Advanced in Maritime Logistics PDF

Advances in Maritime Logistics
and Supply Chain Systems

This page intentionally left blank
Advances in Maritime Logistics
and Supply Chain Systems
editors
Ek Peng Chew
Loo Hay Lee
Loon Ching Tang
National University of Singapore, Singapore
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
ADVANCES IN MARITIME LOGISTICS AND SUPPLY CHAIN SYSTEMS

Copyright © 2011 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN-13 978-981-4329-85-9
ISBN-10 981-4329-85-1
Typeset by Stallion Press

Email: [email protected]
Printed in Singapore.
Steven - Advances in Maritime Logistics.pmd 1 6/14/2011, 2:01 PM

June 14, 2011 11:46 9in x 6in Advances in Maritime Logistics & Supply . . . b1185-fm
CONTENTS
Preface xiii
Part 1: Regional Developments and Performance

Analysis 1
1. Maritime Trade Evolutions and Port City Developments
In Asia 3
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Evolution of Asia’s Maritime Trade . . . . . . . . . . . . 7
2.1. Ancient maritime trade between India and China,
1200–1450 . . . . . . . . . . . . . . . . . . . . . . 8
2.2. Maritime networks in colony times,
1500–1950 . . . . . . . . . . . . . . . . . . . . . . 9
2.3. Asia’s maritime trade under globalization . . . . . 12
3. Asia Port Developments in the 1990s and Beyond . . . . 14
3.1. Northeast Asia . . . . . . . . . . . . . . . . . . . . 15
3.2. Southeast Asia . . . . . . . . . . . . . . . . . . . . 23
3.3. South Asia ports . . . . . . . . . . . . . . . . . . . 30
4. Factors of Port Competitiveness and Development . . . . 32
4.1. Port location . . . . . . . . . . . . . . . . . . . . . 32
4.2. Port efficiency . . . . . . . . . . . . . . . . . . . . 35
4.3. Multimodal network . . . . . . . . . . . . . . . . . 37
4.4. Maritime trade strategy and institutional
settings . . . . . . . . . . . . . . . . . . . . . . . . 39
5. Lessons to be Learnt . . . . . . . . . . . . . . . . . . . . 41
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 44
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 45
v
vi Contents
2. Recent Development of Maritime Logistics 49

1. Development Trends on Global Container Shipping . . . 49
1.1. Global economic condition and industry
perspective . . . . . . . . . . . . . . . . . . . . . . 49
1.2. Recent trends in container shipping industry . . . 53
2. Liner Shipping . . . . . . . . . . . . . . . . . . . . . . . . 54
2.1. Container liners . . . . . . . . . . . . . . . . . . . 54
2.2. Freight rates . . . . . . . . . . . . . . . . . . . . . 57
3. Ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.1. Global container terminal operators . . . . . . . . 58
3.2. Leading terminal operators . . . . . . . . . . . . . 60
3.3. Development in transshipment activities . . . . . 64
3.4. Improvement in port performance . . . . . . . . . 65
3.5. UNCTAD liner shipping connectivity
index 2009 . . . . . . . . . . . . . . . . . . . . . . 65
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 66
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3. Scenario Analysis for Hong Kong Port Development

Under Changing Business Environment 69
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 69
2. Literature Review . . . . . . . . . . . . . . . . . . . . . . 72
3. Changing Business Environment for HKP . . . . . . . . . 74
3.1. Changing business environment
in the PRD region . . . . . . . . . . . . . . . . . . 74
3.2. Potential processing trade relocation trends . . . . 75
4. A MIP Model . . . . . . . . . . . . . . . . . . . . . . . . 78
5. Experimental Design . . . . . . . . . . . . . . . . . . . . 81
5.1. Experimental scenarios . . . . . . . . . . . . . . . 81
5.2. Experimental data . . . . . . . . . . . . . . . . . . 81
6. Results, Analysis and Findings . . . . . . . . . . . . . . . 82
6.1. Modeling results at base scenario . . . . . . . . . 82
6.2. Sensitivity analysis . . . . . . . . . . . . . . . . . 83
6.3. Scenario analysis and findings . . . . . . . . . . . 85
7. Conclusions and Future Work . . . . . . . . . . . . . . . 87
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Contents vii
4. Models for Port Competitive Analysis in Asia-Pacific Region 91

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 92
2. Literature Review . . . . . . . . . . . . . . . . . . . . . . 93
3. Port Benchmarking Models . . . . . . . . . . . . . . . . . 95
3.1. Port efficiency . . . . . . . . . . . . . . . . . . . . 96
3.2. Port connectivity . . . . . . . . . . . . . . . . . . 100
3.3. Impact of factors on individual ports . . . . . . . 105
4. Conclusion and Discussion . . . . . . . . . . . . . . . . . 111
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5. Is Port Throughput a Port Output? 117

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 117
2. Port Production Function . . . . . . . . . . . . . . . . . . 118
3. Port Operating Options . . . . . . . . . . . . . . . . . . . 120
4. Port Resource Function . . . . . . . . . . . . . . . . . . . 120
5. Container Port Output . . . . . . . . . . . . . . . . . . . 121
5.1. TEU Throughput and the port
production function . . . . . . . . . . . . . . . . . 121
5.2. TEU Throughput and port cost functions . . . . . 122
6. Port Interchange Service Measures . . . . . . . . . . . . . 126
6.1. Port revenue . . . . . . . . . . . . . . . . . . . . . 126
6.2. Port throughput ratio . . . . . . . . . . . . . . . . 127
7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6. A Framework for Modelling and Benchmarking

Maritime Clusters: An Application to the Maritime
Cluster of Piraeus 131
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 132
2. Conceptual Definition: Cluster Theory
and Maritime Clusters . . . . . . . . . . . . . . . . . . . 132
3. The Concept of Maritime Clusters . . . . . . . . . . . . . 133
4. Spatial Paradigm: The Greater Area of Piraeus . . . . . 137
4.1. The structure of the Piraeus maritime
cluster . . . . . . . . . . . . . . . . . . . . . . . . 138
4.2. The economic footprint of the maritime industry
in the region . . . . . . . . . . . . . . . . . . . . . 140
4.3. SWOT analysis . . . . . . . . . . . . . . . . . . . 141
viii Contents
5. Methods for Evaluating and Benchmarking Maritime

Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.1. General cluster sizing indicators . . . . . . . . . . 145
5.2. Intra-cluster performance indicators . . . . . . . . 147
5.3 On the use of data and analysis for measuring
performance of maritime clusters . . . . . . . . . . 149
6. Computational Methods for Simulation and Life-Cycle
Management of Maritime Clusters . . . . . . . . . . . . . 151
6.1. Agent-based modeling and simulation . . . . . . . 151
6.2. Modeling case study: the maritime cluster
of Piraeus . . . . . . . . . . . . . . . . . . . . . . 152
6.3. Agent-based modeling toolkit . . . . . . . . . . . 154
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 154
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7. A Performance Evaluation Strategy Towards Dealers

in the Automotive Supply Chain 157
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 157
2. Problems of Dealer Evaluation . . . . . . . . . . . . . . . 158
3. Indicators Definition for Dealers’
Performance Evaluation . . . . . . . . . . . . . . . . . . . 160
3.1. Balanced scorecard method . . . . . . . . . . . . . 160
3.2. Evaluation indicators definition . . . . . . . . . . 160
4. Dealers’ Performance Evaluation via ANP . . . . . . . . 162
4.1. Analytic network process (ANP) . . . . . . . . . . 162
4.2. Enabling factors of dealers’ performance . . . . . 162
4.3. Procedure of dealers’ performance evaluation . . . 163
4.4. Method for dealers’ performance evaluation . . . . 167
5. Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . 168
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 169
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Part 2: Ports and liners operations 171

8. A Yard Allocation Strategy for Export Containers Via
Simulation and Optimization 173
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 173
2. Related Work . . . . . . . . . . . . . . . . . . . . . . . . 174
Contents ix
2.1. Yard allocation for import containers . . . . . . . 174

2.2. Yard allocation for export containers . . . . . . . 175
2.3. Combined yard allocation . . . . . . . . . . . . . . 175
3. Yard Allocation Modeling for Export Containers . . . . . 176
3.1. Problem description . . . . . . . . . . . . . . . . . 176
3.2. Yard allocation modeling . . . . . . . . . . . . . . 176
4. Yard Allocation Algorithm for Export Containers . . . . 181
4.1. Heuristic algorithm for feasible solution . . . . . . 182
4.2. Procedure of genetic algorithm . . . . . . . . . . . 183
5. Simulation Model . . . . . . . . . . . . . . . . . . . . . . 185
5.1. Simulation framework . . . . . . . . . . . . . . . . 185
5.2. Input parameters . . . . . . . . . . . . . . . . . . 186
5.3. Simulation process . . . . . . . . . . . . . . . . . . 186
5.4. Statistical simulation indices . . . . . . . . . . . . 187
6. Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . 188
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 190
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 191
9. Integration of AGVS in Intermodal Rail Operations

at Deep Sea Terminals 193
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 193
2. Earlier Attempts to Address the Problem . . . . . . . . . 195
2.1. Fixed rail mounted gantry cranes linking ship
to shore crane with stacking area and hinterland
modes rail and road . . . . . . . . . . . . . . . . . 195
2.2. NOELL — an approach by K.-P. FRANKE . . . 195
3. The AGV-solution to Integrate Railway Operations
in Deep Sea Terminals . . . . . . . . . . . . . . . . . . . 198
4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 200
10. On the Ongoing Increase of Containership Size 201

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 201
2. Economies of Ship Size . . . . . . . . . . . . . . . . . . . 203
2.1. Modeling ship size economies . . . . . . . . . . . . 203
2.2. Capital related costs . . . . . . . . . . . . . . . . 203
2.3. Labour related costs . . . . . . . . . . . . . . . . . 207
x Contents
3. Shipping Costs of Post Panamax Containerships . . . . . 209

3.1. Fixed annual costs . . . . . . . . . . . . . . . . . . 210
3.2. Fuel costs . . . . . . . . . . . . . . . . . . . . . . . 210
3.3. Shipping costs per roundtrip . . . . . . . . . . . . 211
4. The Ongoing Increase of Containership Size . . . . . . . 216
4.1. Development of ship size and trade . . . . . . . . 216
4.2. Factors limiting economies of ship size . . . . . . 218
4.3. A balance between user and producer costs . . . . 221
4.4. Factors affecting development of user costs . . . . 223
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 225
6. References . . . . . . . . . . . . . . . . . . . . . . . . . . 227
11. A Linearized Approach for Liner Ship Fleet Planning

with Demand Uncertainty 229
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 229
1.1. Background . . . . . . . . . . . . . . . . . . . . . 229
1.2. Literature review . . . . . . . . . . . . . . . . . . 230
1.3. Randomness . . . . . . . . . . . . . . . . . . . . . 232
1.4. Contributions . . . . . . . . . . . . . . . . . . . . 233
2. Problem Description, Assumptions and Notations . . . . 234
2.1. Itinerary . . . . . . . . . . . . . . . . . . . . . . . 234
2.2. Charter strategies . . . . . . . . . . . . . . . . . . 235
2.3. Chance constraints . . . . . . . . . . . . . . . . . 236
2.4. Notations . . . . . . . . . . . . . . . . . . . . . . . 238
3. A Mixed Integer Nonlinear Programming Model
with Chance Constraints . . . . . . . . . . . . . . . . . . 239
4. A Linearized Approach . . . . . . . . . . . . . . . . . . . 243
5. Numerical Example . . . . . . . . . . . . . . . . . . . . . 247
6. Summary and Conclusion . . . . . . . . . . . . . . . . . . 254
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12. Ship Emissions, Costs and Their Tradeoffs 257

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 257
2. Background . . . . . . . . . . . . . . . . . . . . . . . . . 260
3. Some Basics: Algebra of Emissions and Fuel Cost . . . . 267
4. A Simple Logistical Scenario: Factors and Tradeoffs . . . 268
5. The Cost to Avert One Tonne of CO2 . . . . . . . . . . . 281
6. The Port Time Factor . . . . . . . . . . . . . . . . . . . . 283
7. Speed Reduction at SECAs . . . . . . . . . . . . . . . . . 287
Contents xi
8. SECAs Continued: Effect on Modal Split . . . . . . . . . 290

9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 292
10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 293
13. Exploring Tanker Market Elasticity with Respect to Oil

Production Using Foresim 297
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 297
2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 300
3. Simulation Results . . . . . . . . . . . . . . . . . . . . . . 308
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 311
5. References . . . . . . . . . . . . . . . . . . . . . . . . . . 313

PREFACE
Over the recent years, maritime logistics and supply chains have witnessed
tremendous growth rates around the world, notwithstanding the recent
economic downturn. Maritime transportation accounts for the majority of
international trade and it has become a vital factor for the economic health
of many nations. In emrging economies, more new ports have also been
developed to tap into the global maritime logistics network. The global
landscape of the martime industy is changing rapidly and this has generated
many issues which are worthy of more in-depth research. In particular,
topics related to maritime logistics and supply chains have been drawing
immense attention of both academia and industry.
The objective of this book is to reflect the recent developments in
maritime logistics and supply chains, and to examine some research issues
concerned with quantitative analysis on port competitiveness and decision
support for maritime logistics and supply chain systems. Twelve papers
have been selected for publication after a thorough peer review. The papers
are categorized into two main areas: regional developments and performance
analysis; and, ports and liners operations.
Regional Developments and Performance Analysis

The first paper by Xue-Jing Yang, Joyce M.W. Low and Loon Ching Tang,
tracks maritime trade evolution in Asia from the thirteenth centuries to the
post-World War II, followed by an examination on the contemporary devel-
opment of some major Asia ports. Some factors affecting port competition
and development are identified and reviewed. Their study concludes with
their beliefs that maritime trade industry in Asia is promising and positive
future economy trends will continue despite recent concerns over rising oil
prices.
xiii
xiv Preface
The paper by L. H. Lee, E. P. Chew, L. Zhen, C. C. Gan, and J. Shao

presents the recent development in Maritime Logistics during the recent
economic crisis. The recent development of container shipping industry has
shown the following trends: (1) The size of the largest container vessel and
the average vessel size are both increasing. (2) Transshipment handling
has become more and more significant globally. (3) Global container
terminal operators are increasing their market share. (4) Liner companies
are adopting more rigorous measures to reduce cost and stabilize freight
rate.
Hong Kong port had been the world’s busiest container port during the
1990s and early 2000s. However, in recent years, its growth slowed down due
to rising competition from mainland ports. Abraham Zhang and George Q.
Huang perform some scenario analysis for Hong Kong port development
under changing business environment, so as to understand the relationships
between business environment factors and potential relocation trends by
using a mixed integer programming model.
A study on port benchmarking in Asia-Pacific region is performed
by Ek Peng Chew, Loo Hay Lee, Jianlin Jiang and Chee Chun Gan.
Some models for port competitive analysis are proposed from three
perspectives: port efficiency, port connectivity, and the impact of various
factors on individual ports. The authors examine the three perspectives by
presenting a model and a case study for each perspective. Data envelopment
analysis technique, a port connectivity analysis framework, and a network
flow model are proposed and employed to investigate the above three
perspectives respectively.
Wayne K. Talley investigates the question: is port throughput a port
output in port economic production and cost functions? His study disagrees
on this statement. The author proposes the port throughput ratio — the
ratio of cargo interchanged to the total time incurred in interchanging the
cargo, for measuring the output of a port.
The paper by Vassilios K. Zagkas and Dimitrios V. Lyridis proposes a
framework for modeling and benchmarking maritime clusters. The authors
investigate the factors that contribute to the decisions of companies from
key maritime sectors to be established in a specific area that evolves into a
network of firms. Agent-based modeling technology is employed to simulate
the networking process within maritime clusters and managing their life
cycle. This study gives an insight of firm survival strategies within the
cluster, optimum timing for new entrants in the cluster and overall cluster
management.
Preface xv
The paper in this group by Min Chen, Wei Yan and Weijian Mi
propose a performance evaluation strategy for dealers in the automotive
supply chain. The performance evaluation strategy is developed from four
dimensional criteria, i.e., the financial condition, customer satisfaction,
internal processes and self-innovation. The analytic network process (ANP)
technique is employed to analyze the surveyed data. By comparing with
traditional performance evaluation strategies, their approach can eliminate
such disadvantages as time delaying and benefit orientation.
Ports and Liners Operations

Container yard management is essential for the efficiency of terminal
operations. A yard allocation strategy is proposed in the paper by Wei
Yan, Junliang He, Daofang Chang. By using the objective programming,
the proposed model is based on a rolling-horizon strategy, which aims at
allocating export containers into yard. For solving the model, a hybrid
algorithm by using heuristic rules and genetic algorithm is employed.
A simulation model, which embeds the yard allocation model and algorithm,
is also developed to evaluate the proposed system.
The paper by Bernd H. Kortschak examines the integration of railway
links with other functions in a deep sea terminal. Currently, railway links are
often operated separately which incurs additional costs when transferring
containers and hinders the competitive strength of rail versus road links.
An integrated AGV system is proposed to improve productivity and enable
faster transshipment times.
In the paper by Simme Veldman, the author conducts a statistical
analysis of economies of ship size. This study shows that economies of ship
size, expressed as the elasticity of costs as a function of ship size, differ only
slightly from those of ships up to Panamax. For avoiding too high cost for
users, the increase in size has to be in balance with the combined increase
in trade volumes and the number of port pairs between coast lines to be
connected. The study draws a conclusion that the ongoing increase in ship
size will continue.
In the paper by Qiang Meng, Tingsong Wang and Shahin Gelareh, a
linearized approach is proposed for liner ship fleet planning under demand
uncertainty. The authors develop a mixed integer nonlinear programming
model for the problem; then the fuel consumption cost of a ship is
approximated by a linear function with respect to its cruising speed. Hence
a mixed integer linear programming model can be built to approximate the
xvi Preface
originally proposed nonlinear programming model. Numerical examples are

performed to assess the linearized approach.
The paper by Harilaos N. Psaraftis and Christos A. Kontovas studies
ship emissions, costs and their tradeoffs. The authors investigate various
tradeoffs that may impact the cost-effectiveness of the logistical supply
chain, and propose some models that can be used to evaluate these tradeoffs.
Their study validates that speed reduction can lead to a lower fuel bill and
lower emissions, even if the number of ships is increased to meet demand
throughput. In addition, cleaner fuel at SECAs may result in a reverse cargo
shift from sea to land that has the potential to produce more emissions on
land than those saved at sea.
By using a simulation tool named by FORESIM, P.G. Zacharioudakis
and D.V. Lyridis study the future tanker market freight levels in relation
to current market fundamentals and future values of demand drivers.
The authors follow a systems analysis seeking for internal and external
parameters that affect market levels. By using the proposed methodology,
decision makers can measure the behavior of future market as long as twelve
months ahead with very encouraging results. The output information is
potentially useful in all aspects of risk analysis and decision making in
shipping markets.
Concluding Remarks
This book has greatly benefited from the cooperation among the authors,
reviewers and editors. We would like to express our sincere thanks to them.
E.P. Chew
L.H. Lee
L.C. Tang
December 2010, Singapore

June 28, 2011 16:39 9in x 6in Advances in Maritime Logistics & Supply . . . b1185-ch01
PART I
REGIONAL DEVELOPMENTS
AND PERFORMANCE ANALYSIS
1
2
CHAPTER 1
MARITIME TRADE
EVOLUTIONS AND PORT CITY
DEVELOPMENTS IN ASIA
X.J. Yang, Joyce M.W. Low and Loon Ching Tang
Department of Industrial & Systems Engineering
National University of Singapore
1 Engineering Drive 2, Singapore 117576
[email protected]
Historically, almost all goods transported worldwide have been carried by sea
with the current estimate stands at approximately 90 percent by volume and 70
percent by worth. Maritime industry is an important economic sector as it has
a direct impact on the prosperity of a region and/or city. This chapter presents
a review on maritime trade evolution in Asia from the thirteenth centuries
to the post-World War II, followed by an examination on the contemporary
development of some major Asia ports. From the extant port literature, a list of
factors affecting port competition and development is identified and reviewed.
The chapter concludes with brief discussions on fruits for thought, future trend,
challenges and opportunities facing the Asia maritime trade industry.
1. Introduction
Maritime shipping represents the most ancient global transportation,
holding an irreplaceable role in geography discovery, culture communication
and economy development in history. As early as the fifteenth century,
Chinese admiralty Zheng He had visited Indian ports and east Africa
seven times through Southeast Asia. His famous voyages marked the
beginning of the international maritime trade in Asia. Subsequently,
international and global voyages in the following centuries had led the
European admiralties to discover the North America and opened Africa
continents, as well as, the principle of earth’s as a spheroid planet. To some
degrees, these illustrious maritime explorations had shaped the world’s
3
4 X. J. Yang et al.
history. Particularly, international maritime trade has played a significant

role in spreading civilization to many parts of the world and promoting
communications that aid economic development.
In this modern era, maritime trade industry has gained unsurpassed
vigor with more high-value cargo transported by seaborne shipping than
before. Between 1990 and 2003, the value of maritime trade to the
US economy ballooned from 434 billion dollars to 800 billion dollars.
[DeGaspari, 2005] indicated that approximately 75 container ships cross
the Pacific each week, making the trans-Pacific an economic engine between
North America and Asia in 2005. In Europe, container port throughput has
increased by some 58 percent to 69 percent, reaching 49.5 million Twenty
foot equivalent unit (TEU) in 2001; and is expected to hit 53.0 million TEU
in 2010. The transshipment market, despite volatility, has also grown by an
average 9.7 percent per annum since 1995 in North Europe. In Asia, some
major ports have replaced New York, Tokyo and Rotterdam ports as the
top container ports in 2005 [Lee et al., 2008]. These ports have continued to
grow and by 2008, Singapore, Hong Kong, Shanghai, Shenzhen and Busan
have registered impressive container traffic of 29,918 thousand TEUs, 24,494
thousand TEUs, 27,980 thousand TEUs, 21,414 thousand TEUs and 13,453
thousand TEUs respectively.
Continual globalization of world economy, coupled with recent energy
crisis, will propel maritime trade to a higher level in the near future. In
Asia, maritime industry occupies a central position, not only for economies
with large transshipment ports like Singapore and Hong Kong, but much
more for China and Japan who have an increasing demand for oil. At
present, China is the world’s manufacturing center and ranked as the third
largest economy with huge domestic market. This newly industrialized giant
economy represents an exploding demand for oil and the Malacca Straits
is the only major way for China to receive oil from Middle East and India
currently. [Chua et al., 2000] substantiated the claim in noting that this
trade route delivered over 100,000 oil and cargo vessels each year, 3.23
million barrels of crude oil each day. In addition, the Malacca Straits is
also an important shipping route for productions from Chinese and Japan
to European and African market. Fig. 1 shows that the Singapore and the
Malacca Straits are significant transshipment point and route for cargo
originating from, or destined for, the European, East/Northeast Asian and
Australasian markets in the world container flow. The Asia-Europe Route
overtook the transpacific route as the largest containerized trading lane,
with lane totaled 27.7 million TEUs in 2007.
Maritime Trade Evolutions and Port City Developments in Asia 5
Flow of container traffic in 2007 (Millions of TEUs), source from [UNCTAD, 2008].
Fig. 1
6 X. J. Yang et al.
According to [Klink and van den Berg, 1998] and [Heilling and
Poister, 2000], ports are the most significant elements in maritime trade.
The authors highlighted that role of ports as gateways to domestic and
international trade, connecting the region as well as intra-region to the
world is pivotal in global logistical network. Furthermore, the importance of
ports to an economy cannot be underestimated, recognizing that the impact
of having a competitive port is far reaching beyond the immediate benefits
such as higher operating efficiency, profitability, competitive exports and
employment opportunities. Being a vital link in the overall trading chain
and consequently, port performances, to a large extent, determine a nation’s
international competitiveness. Prior to this, [Yabe, 1990] emphasized ports
not only function as junctions of marine and land transportation but also
as nucleus areas for industrial activities and citiesa since ancient times.
Nonetheless, developments in the technological infrastructure have
significantly altered the landscape of the port industry over these recent
decades. Among these, logistical and communication advancements have
led to hinterland expansions and overlaps and ports are no longer captive of
their hinterlands. As carriers deploy larger vessels for higher cost efficiencies,
the accompanying reduction of port calls made by these carriers increases
the potent impact of a move of a carrier and fuels greater competition in the
port industry. In order to insure that their ports will remain attractive in the
heightened competition, port authorities will need to have a clear picture of
the changing playing field of the port and maritime industry. These include
an understanding on the port selection criteria adopted by carriers and
the underlying factors of competitiveness. Ports are required to continually
assess its performance relative to the rest of the world so that appropriate
strategies can be devised to meet the ever increasing and more demanding
needs of port users as well as maintain continuing competitiveness of their
ports and economy [Tang et al., 2008].
In this chapter, we review previous research on maritime trade evolution
in Asia and trace the development of major Asian ports, so as to validate
previous findings and provide further insights on the development of port
a [Fujita and Mori, 1996] observed that most East Asian countries (like Indonesia,
Philippines and Thailand) had experienced a disproportionate share of population and

manufacturing industries that are concentrated in their primal cities located at ports.
The same goes for Shanghai, Hong Kong and Singapore. One key reason is because
ports represent the most convenient location for exports and imports. For some large
cities such as Chicago and Paris, even though ports do not play an important role today,
their primal growth had been initiated by the ports.
cities. Specifically, the long tale of maritime trade evolution in Asia and
development of Asian port cities seeks to unveil the relationship between
maritime industry and the economic progress of a city. Recognizing that
port forms a major pillarb in the development of a port city, we identify
some of the key success factors and challenges faced by major ports in Asia.
These identifications are subsequently verified with findings in the extant
literature. It is hoped that such historical documentations supported by
contemporary case studies will provide invaluable guidance for the future
development in Asia port and maritime industry.
The remainder of this chapter is organized as follows: Section 1.2
reviews of maritime trade evolution in Asia from thirteenth century to
post World War II. Section 1.3 documents the development of major
Asian ports. Section 1.4 discusses some of the important factors that
are found to influence the performances of ports in the classic literature.
Section 1.5 highlights some insights and fruits for thoughts in terms of
future development. Section 1.6 summarizes and concludes the chapter.
2. Evolution of Asia’s Maritime Trade

Maritime trade and ports evolution are affected by revolutions in the trans-
port sector and the industries, as well as, the globalization of the economy.
In ancient times, international maritime is the major transport mode for
geographical discoveries, and maritime trade plays an important role in
civilizations of mankind. Following the industrial revolution, British traders
extracted raw materials at low cost from their colonies, while dumping
industry goods into these colonies markets. This was the key driving force
behind maritime trade. After World War II, trade liberalization led to
increased participations from developed and developing countries in inter-
national trade and fueled the growth of maritime activities. Subsequently,
globalization in the 1990s had brought about a large expansion of world
trade and shipping, of which, maritime trade has played an increasingly
important role in stimulating economic growth.c
b Therefore, factors contributing to a successful port are important not only for port
marketing strategy, but also for the strategies management for the spatial economic
development of port cities.
c [Irwin and Tervio, 2002] have proven one of the most fundamental propositions of
international trade theory, which advocates that trade allows a country to achieve a
higher real income than would otherwise be possible.
8 X. J. Yang et al.
In the following sub-sections, we explore the evolution of Asia’s mari-

time trade classified into three important distinct periods: ancient maritime
trade between India and China prior to the fifteen century, maritime
networks during times of colonization and Asia’s maritime trade under
globalization.
2.1. Ancient maritime trade between India and China,

1200–1450
Folk traders had established China and India maritime links from the first
century BC. More notably, 1200–1450 marked a distinct milestone in the
history of China and southern Asia maritime relations with the forming of
the government maritime network. From the end of Song Dynasty to Ming
Dynasty, Chinese government organized many fleets to southern India and
even east Africa as parts of the Indian commercial zones. It was said in
[Liu, 1988] that traders involved in these maritime exchanges are primarily
the Persians, other Middle Easterners, South and Southeast Asians. The
Malacca Straits, controlled by the port kingdom of Malacca at that time,
was then already a critical trade route linking the Indian Ocean to the South
China Sea and Pacific Ocean. Silk yarn entered India and was shipped to
Rome through Indian ports, while coral and glass from Roman reached
Chinese markets through Indian ports [Lin, 1998]. Hence, ports in southern
Asia were important transit points for Chinese traders to Persian Gulf, and
were also transition centers for Chinese and Roman goods. [Sen, 2006] noted
that these maritime trades had grew so rapidly thereafter that the ports-
of-trade in Southeast Asia and a Muslim trading network were formed in
the eighth and ninth centuries respectively.
After Qubilai (the King of Yuan Dynasty) took control the ports in
China, he executed an aggressive maritime policy with a desire to expand
the military and political influences to the southern coastal region. Large
Chinese ships,d which were more than thirty meters in length with capacity
over hundred tons and staffed with at least sixty crews, were deployed
to carry out the maritime trade in the early twelfth century. These ships
sailed around Chinese sea and were capable of reaching the land of Korea
and Japan by today’s standards. [Sen, 2006] noted that significant fiscal
revenue was derived through maritime commerce, which supported the
d Marco Polo, who visited China at Yuan Dynasty, described the ships that were
transporting goods between China and India as ships having nailed hulls and multiple
masts and cabins and were able to carry 1860 tons load.
world military expansion of Yuan Dynasty. Particularly, the maritime route

through the southern Indian coasts was crucial for commerce between China
and Persian Gulf. The associated significance of the south Indian ports to
maritime trade during the thirteenth and fourteenth century is described
in many history literatures, including [Grant, 2002].
Unlike Yuan Dynasty, the objective of the Ming court for developing
maritime links was not to profit from the commercial exchanges between
China and India coastal regions. Rather, the Ming court desired to use its
naval power to spread its rhetoric civilizing, maintaining peace and order,
and economic prosperity across regions. From 1405 to 1433, Ming court
supported Zheng He’s seven voyages to southern Asia, Red Sea, Indian coast
and east Africa [Sen, 2006]. [Lin, 1998] asserted that the mission of Zheng
He’s first two voyages, in 1405 and 1407 respectively, was a solicitation of
tributary. The Ming ruler, in turn, invited the envoys from Calicut and
other foreign representatives to banquets, conferred titles and returned
gifts. Meanwhile, the expeditions of Zheng He also increased the maritime
exchanges between China and the kingdoms along the Indian coast sharply.
Owing to Ming court’s prohibition to private overseas trades, many
Chinese merchants resided at foreign ports. These Chinese communities in
Southeast Asia had found it easier and more profitable to operate at Java,
Malacca, or other ports in the region for their convenient access to India
and South China. In addition, they benefited from participating in the
tributary system of Ming court, and avoided trade competition with other
foreign traders in the Indian ports. These Chinese traders travelled from the
Southeast Asian ports with the northeastern winds between December and
March, and returned with southwestern winds between April and August
[Grant, 2002].
Notwithstanding the differing motivations, the influences of Yuan
and Ming courts in the Indian Ocean world were far-reaching and were
recognized by officials in the coastal regions of India. Based the analysis of
[Sen, 2006], with City University of New York, the formation of maritime
networks to Indian coast by private and official Chinese traders had
augmented the domain knowledge on Indian geography, coastal kingdoms
and commercial prospects under the ruling of both dynasties.
2.2. Maritime networks in colony times, 1500–1950

In 1511, Portugal captured Malacca for its strategic importance. In 1641,
the Dutch occupied Jakarta, and established Dutch East India Company
10 X. J. Yang et al.
to control the trade in the Straits from the seventeenth to eighteenth

century [Chua et al., 2000]. After the industrial revolutions, the British
also recognized that a safe passing of British cargos into the Chinese market
could be ensured with a control over the Malacca Straits. Thus, in 1819,
British established a colony in Singapore, and agreed to open the Straits for
other friendly nations, which ended the long-standing dispute with Dutch.
Singapore, Hong Kong and Calcutta ports had become colonial maritime
centers serving a key global trading route. In the nineteenth century, these
port cities were integrally linked by the East India Company in Asia. As
the colonial port cities, Singapore, Hong Kong and Calcutta enjoyed rapid
economic growth, physical transformation, ascending population numbers,
and dense maritime networks connected by international trade [Tan, 2007].
Their ports played a crucial role not only as trading places, but also
as centers for technological transfer and culture communication during
the period of colonization. The following paragraphs shall focus on the
discussions of each of these port cities in turn.
Singapore occupies a strategic location between the Indian Ocean and
Pacific Ocean, and at the southernmost tip of Asian landmass. Owing
to its geographical position, transshipments made up a large proportion
of Singapore trade in the nineteenth century. Being part of a trading
environment, the economic, social and cultural conditions of Singapore
were determined by the flow of its maritime networks. During the colonial
times, the extended commercial networks of Singapore were integrated by
a full range of maritime vessels and formed by physical connections, mar-
itime routes, functional inter-dependence such as trade, labor, commodity
exchange and capital flows [Harper, 2002]. The colonial trading pattern
promoted Singapore to be a critical node in the whole maritime network.
It connected the Persian Gulf and India to the west with China to the
east for centuries, which was named Maritime Silk Road. By the middle
of nineteenth century, Singapore had become a congregation of multiple
communities as Indians, Chinese, Malays and European trading together
in the market.
In the late nineteenth century, the speed of globalization (generated by
trade and imperialism) was accelerated by the advent of steam vessels and
telegraph. During this time, ancient trans-national connections stretching
from the Arab lands to the south Chinese coast were revitalized. Singapore
became not only the key economic node but also the heart of intellectual
world of Asia. The island housed a dynamic mixture of diverse classes
and culture, triggering innovation with local experience and improving

relationships with regional and international communities. With the open-
ing of Malayan peninsula at the end of nineteenth century, Singapore
acquired a physical hinterland from which the Malayan’s agriculture
productions were exported. These export activities had become a major
driving force of the port traffic and Singapore economy by the end of
nineteenth century.
Similar to Singapore, Hong Kong was a fishing coastal villages con-
sisting of a hundred dwellers before the intervention of external powers.
While advantageous locations and nautical accessibilities conferred both
economies their strategic importance as British colonies, the main reason
for becoming colonial city ports differed. According to [Lee et al., 2008], it
was Hong Kong’s potential as a gateway to China that motivated the British
Empire to establish and start trade negotiations in Hong Kong. However,
the failure to dominate market in China had lead to the development of
Hong Kong to inevitably parallel that of Singapore.
In contrast to Singapore and Hong Kong, Calcutta’s early success
as an international hub port city was owed to its ability to transform
its immediate hinterland in northern India into an international market
and not relying mainly on transshipment. This hinterland, which spanned
from the Gangetic plains to the west and the Brahmaputra valley in the
northeast of India, had provided Calcutta with large volumes of trade and
labors. As all important offices moved from Murshidabad to Calcutta in
1772, Calcutta became the capital of British India [Tan, 2007]. Through the
connection of the Indian hinterland to world market (especially China and
Southeast Asia), Calcutta port delivered half of Indian’s export of cotton,
tea, coal, sugar and saltpeter in the late 18th century. For the whole 19th
century, Calcutta was a centre of commerce, culture and administration.
The transformation of Calcutta’s hinterland into an international market
was greatly aided by the opening of the Suez Canal and its special location
on a navigable river that expanded its trade-related commerce, as well as,
its extensive rail and road network that provided it with a large number
of laborers and immigrants. Until the early 20th century, Calcutta was
an international port and the center of colonial trading, serving the vast
business created by the East Indian Company. However, from the 20th
century, the British moved its capital of Indian Empire from Calcutta
to New Delhi, and Calcutta gradually lost its hinterland market and its
position as the empire port city.
2.3. Asia’s maritime trade under globalization

After the World War II, a trading regime GATT (General Agreement on
Trade Regime 1947) was established to govern trade between industrial
countries. According to [Francois and Wooton, 2001], the Uruguay Round
of World Trade Organization (WTO) in 1993 marked a new shift in the
maritime trade system by involving a commitment from the developing
countries to participate in the multilateral trading system. Thereafter,
tariff barriers had further assuaged with the formation of more trade
liberalization districts under the agreements of WTO. In addition, other
shipping conferences are organized to set rates, analyze market conditions,
assess other development such as fuel prices and port charges. One of these
major conference agreements is the Transpacific Stabilization Agreement,
which controlled about 86 percent of US maritime trade with Asia in 1998.
Between the 1950s and 1990s, the Asia economy saw the industrializa-
tions of large economic giants and the globalization of port cities. During
the 1950s and 1960s, Japan embarked on an aggressive industrialization of
its economy which was then followed by Korea some twenty years later. As
part of these industrialization efforts, the Japanese and Korean governments
gave strong support to lead industries that were tailored for exports.
These include the shipbuilding, motor vehicles and electronics industries
etc. Together with the huge amount of maritime imports of raw material
and energy (such as metal ores and coal), the trade in Japan and Korea
increased dramatically and stimulated traffic growths in the, Kobe, Osaka
Tokyo, Yokohama and Busan ports. Particularly, [Yabe, 1991] noted that
the emphasis on the roles of ports to the industries in Japan during the
early industrialization period. Extension of wharves and large landfills for
industrial areas were carried out in many ports and harbors, in response
to the rapid increase in production and distribution. Many ports have
been developed seaward so that sufficient water depths for larger ships and
adequate areas for cargo handling can be created. Comparatively, urban
life was neglected resulting in the occurrences of various problems such as
water pollution, traffic congestion and loss of access to the waterfront for the
people living in the city. In recognition of these, Japanese port development
policy has been drastically changed to take into considerations of the effect
on the city since 1985.
Subsequently, with the trend towards globalization, ports in Asia
delineated free trade zones to boost their attractiveness and competitiveness
as logistics hubs within world maritime trade networks. As ports continue to
play an important role in economic development, many saw an expansion in

capacity and upgrading of technology, aimed at improving port operations
efficiency to attract more direct carrier servicese and keep pace with
the booming maritime trade caused by globalization. Through a series
of advanced modernization and urban expansions, Singapore and Hong
Kong had become two hub port cities, connecting Europe and North
America with China and Southeast Asia. [Cullinane et al., 2007] had
proclaimed that Singapore posited itself as a global city-state, defining
itself as a city with global orientations, and entrenched itself as hub for
global international manufacturing, commerce, communications and finance
networks. Together, the post-industrialization, globalization and China’s
Open Door Policy had provided new changes for Singapore and Hong Kong.
Nevertheless, emphases on port productivity and efficiency improvements,
urban attractiveness and a total port-city separation were not relaxed.
In summary, the Asia’s maritime landscape had undergone three
prominent phases of trade evolution (i.e., ancient official and private trade,
colonial economic trade and globalization maritime trade) that produced
profound impacts on the development of ports in Asia (see Fig. 2). Likewise,
Globalization Maritime Trade

Global hub port cities: Singapore and Hong Kong, 1950-present
Colonial Economic Trade

British colonial maritime trade through Malacca Strait, 1500–1950
Ancient Official and Private Trade

Ancient maritime trade between India and China through
Southeast Asia, 1200–1500
Fig. 2 Evolution of maritime trade in Asia.
e Directcarrier services augment the connectivity of a port by increasing the number of

global destinations can be reached by shippers at the port and speeding transiting times.
Fig. 3 Evolution of ports in Asia, modified from [Lee et al., 2008].
the development of port city in Asia had followed a progression path from
a fishing coastal village to colonial city port, to entrepot city port, to free
trade port city and finally emerged as a hub port city, as illustrated in
Fig. 3.
3. Asia Port Developments in the 1990s and Beyond

Since the 1990s, Asia has experienced rapid economic growth. Compared to
the world gross domestic product (GDP) that is growing at an estimated
rate of 3.1 percent in real terms, the aggregate economy of the East Asia
maintains its upward momentum with a 7.0 percent growth rate in 2008.
Despite the contraction of the Japanese economy,f China and India have
shown remarkable growth of 9.6 percent and 7.4 percent respectively while
South Korea grew by 2.2 percent in 2008.
During the same period, the world container port throughput grows
by 4.06 percent to over 507 million TEUs in 2008 (UNCTAD, 2009).
Meanwhile, some ports in Asia have reported double-digit gains. These
ports include Ningbo (19.94 percent), Tianjin (19.67 percent), Guangzhou
(19.58 percent), Port Klang (11.96 percent), and Dubai (11.02 percent). On
the average, the mainland Chinese ports grew by 9.42 percent.
f Japan has experienced a negative GDP growth of −0.7% in 2008.

Fig. 4 Ports in Northeast Asia.
3.1. Northeast Asia

The following subsection describes the key advantages facilitating the
growth of some of the major container ports in China (including Hong
Kong and Taiwan), South Korea and Japan, as well as, development plans
to counteract the challenges facing these ports in their future growths.
Figure 4 illustrates the major ports in Northeast Asia.
3.1.1. Greater China region

The Hong Kong port is the world’s 3rd busiest ports. Located on the north
shore of the South China Sea at the mouth of the Pearl River Delta, the
port of Hong Kong is the leading container port for the mainland of China
and a major hub port for intra-Asia trade. [Wang, 1998] advocated that
Hong Kong’s proximity to underdeveloped Chinese ports is one of the prime
reasons that had allowed Hong Kong to achieve its load center status in a
very short period of time. [Cullinane et al., 2004] added that highly educated
workforce in Hong Kong is another factor that promoted the Hong Kong
Port’s international status as a major hub port in Asia. Being at the centre
of the Asia — Pacific Basin and strategically placed on the Far East trade
routes, the port has also been a key factor in the development of the area.
The Hong Kong port is well-regarded as a highly efficient international

container port in the world. It possesses one of the most perfect natural
harbours in the world and operates in a business-friendly environment
with world-class infrastructure, The port, handling 24,248,000 twenty-foot
equivalent units (TEUs) of containers in 2008, is served by 192,000 vessels
from some 80 international shipping lines that provide over 450 container
liner services per week connecting to over 500 destinations worldwide.
However, Wang indicated that Hong Kong port faced two dimensions
regarding the space problems: first, the lack of stacking space within the
port; second, the lack of stacking, parking, and repairing space outside the
port. Furthermore, [Loo, 2002] observed that the abundance of labor in
China has led to a large-scale relocation of labor-intensive and export-
oriented industries into China, which spurred the growth of ports in
South China. As operations in the Chinese ports improve, the differential
advantage in terms of efficiency at Hong Kong port will be gradually eroded.
While Hong Kong is still the leader in terms of value-added trade services
such as consolidation, forwarding and financing, the cost advantage of its
adjacent Shenzhen port and other ports in Southern China represents a
constant threat. Statistics reveal that the Hong Kong port has lost as much
as 40 percent of its monopolized traffic from the region in the 1990’s to
ports in Southern China. In response to these challenges, Hong Kong has
taken some measures to further enhance port productivity and efficiency
as well as setting up high technical logistics centers and open space (OS)
zones.
Spacious water areas, developed hinterland and convenient land trans-
port links are some of the advantages that had contributed to the early
development of Kaohsiung port. Situated in the South-Western part of
Taiwan at the nexus of main Asia Pacific trade routes, the naturally deep-
water port derives 52.2 percent of its volume from transshipment and
enjoys low tidal variance. The port also has ample space for expansion
and provides one of the world’s largest ship scrapping facilities. However,
[Haynes et al., 1997] noted that the total cargo and containerized cargo
growth in Kaohsiung (Taiwan’s largest port) has been lagging behind Hong
Kong and Singapore due to customers’ dissatisfactions with service such as
cumbersome custom clearances, high costs and poor management. In 2008,
the port of Kaohsiung received 36,000 vessel calls and handled 9,677,000
TEUs which puts the port in the 9th and 12th position respectively in the
ranking of Asia’s and world’s ports. The most recent statistics reveal that
the port served 8,102 incoming container vessels, of which, 1,584 are over
60,000 tons in 2009. In line with the Taiwanese government’s strategy to

promote Kaohsiung into an Asia Pacific operation headquarter, a privatized
Free Trade Zone has been established and operated since 2006. The Free
Trade Zone has vast area of adjoining land, which serves as an offshore
shipping center or logistics center that supports a combination of repacking;
processing and other value-added functions.
The port of Keelung, another major Taiwanese port, lies on the
northern part of Taiwan, 40 km away from the capital city of Taipei.
A series of dredging programs have been undertaken to serve the increasing
number of larger vessels and attract liners to call upon Keelung Port.
After the completion of the first two phases of the intended dredging
programs in 2001, the approximate depth of main channel and the diameter
of turning basin are 15.5 meters and 650 meters, respectively. Currently,
these dimensions are sufficient to meet the berthing needs of 60,000
tonnage container vessels. In addition, the Keelung port also establishes
affiliations with the ports of Oakland, Los Angeles, Bellingham and San
Francisco in the United States and the Port of Southampton in the United
Kingdom as “sister ports” for purpose of promoting international friendship
and strengthening the exchange of technology and experiences on port
developments. These port development efforts are seen to have paid off
from the recorded container traffic of 2,128,000 TEUs in 2008. With an
average annual port call of 9,200 vessels, the port has shipping routes linking
globally with all the other major container ports.
Among the main container ports of Taiwan, Taichung is the closest
to mainland China. The port of Taichung, which is called upon by 5,950
vessels, is located on the west coast of Taiwan with a total length of
12.5 km and widths between 2.5 and 4.5 km. While container traffic is
on a much smaller scale (i.e., approximately 1,200,000 TEUs), it has
grown the most rapidly among all the Taiwanese ports. Owing to the
fast increase in business volume, a construction plan to expand the port
was carried out in cooperation with Taiwan Construction. Given the good
investment environment, fully automated warehouse operations and high
service efficiency, the port of Taichung is potentially the main contender
for direct trading links between Taiwan and mainland China.
China ports can be divided into Northeast China ports, east China
ports and south China ports. As a new active economic driver in Asia
and the world’s largest container generator since its trade liberalizations,
China has dramatically increase its financial investment to improve the
infrastructure and superstructure in many of the Chinese container ports.
In 2007, Chinese ports alone accounted for 139.1 million TEU, which
represent 28.4% of world container port throughput [UNCTAD, 2008].
Because nearly 50 percent of all foreign investments into China are
devoted into Shanghai and its surrounding, the port of Shanghai alone takes
nearly half of the total container traffic through all ports in China making
it the leading Chinese container port. The port of Shanghai is located at
the mouth of Changjiang (Yangtze) on the apex of a vast hinterland with
inter-modal waterways, rail and road links running inland to central China.
Containerization is high at Shanghai port, reaching above 55 percent, and
the port is attracting an increasing number of direct, deep-sea vessel calls.
The 2008 statistics show that a total of 27,980,000 TEUs from 55,000
vessels passed through the port of Shanghai, and this impressive volume
has placed Shanghai as the second busiest port in the world. Alongside,
Shanghai and its surrounding provinces have become the driving engine of
China’s economy growth. Particularly, Shanghai functions as the economic,
financial and shipping centre of China, with its surrounding provinces as
the centre of manufacturing.
To some extent, the rapid ascend of the Shanghai port on the world
rankings can be attributed to a series of aggressive port development efforts
(particularly, with regards to the sustained investment in new terminals)
embarked by its port authority. For instances, the second phase of the
modernization of Yangshan Port (offshore of Shanghai port) has seen an
installation of 13 double-decker conveyorsg that can handle two 20 or
40 feet containers at the same time. The completion of the third Phase
that involves the development of a deep water port at Yangshan in 2009,
helps to partly overcome the problem of shallow draught in Shanghai port
that has previously limited the size of the vessels calling at the Shanghai
port and volume of cargo they ship. Currently, the bulk of Shanghai’s cargo
originates in or travels to the conurbation and neighboring provinces of
Jiangsu and Zhejiang. A coastal and inland container hub is being developed
at Longwugang in Shanghai Harbor to extend the port’s hinterland. On
the softer aspects, the port authority has introduced world-class port
management practices into the port of Shanghai. These include but not
limited to the simplifications of custom procedures, implementation of
computer linkage between the port, customs and other related agencies etc.
g Since
the double-decker conveyors are implemented in 2008, the port’s load or unload
efficiency is 850.53 TEU/hour, and the conveying speed of a single conveyor is 123.16
TEU/hour.
The next major port in China is the Shenzhen port, which handles
21,414,000 TEUs in 2008. Ranking as the fourth busiest port in the world,
Shenzhen port is an agglomeration of several ports including Yantian,
Shekou, Chiwan and other smaller ports in Southern China’s Guangdong
province. Of these, Yantian is biggest and sited in the sheltered waters
on Dapeng Bay just 20 nautical miles north of Hong Kong. The port of
Yantian is opened in July 1994 as an alternative access point to Southern
China. Since its opening, this deep-water container port has lifted its
throughput to over 1 million TEU in just 4 years and over 8 million by
2008. The main bulk of Yantian’s cargoes originated largely from Shenzhen,
Dongguan, Guangzhou, Huizhou and other Pearl River locations. The port
is equipped with advanced port facilities and is served by a sophisticated rail
and road network. Meanwhile, Chiwan and Shekou international container
terminals have constantly improved the efficiency of custom procedures.
More international shipping companies are choosing the Shenzhen port for
transshipment due to its merits such as lower costsh and simplified customs
procedures. By offering shorter time and lower cost of transport (including
handling charges) between Hong Kong and the rest of China, the port
of Shenzhen port has now become the 2nd largest port on the Chinese
mainland in terms of handling international transshipment goods.
Other Chinese ports with impressive traffic performances are Ningbo
(11,226,000 TEUs), Guangzhou (11,001,000), Qingdao (10,320,000) and
Tianjin (8,500,000), which ranked 7th, 8th, 10th and 14th respectively in
the world. Another newly developed port is Qinzhou in Guangxi province.
Qinzhou port is the 6th bonded port in China. The port has 22 berths and a
capacity of 15 million tons. Among the 22 berths, 11 can handle ships above
10,000 tons. Currently, another 9 berths providing an additional 25 million
tons, are under construction. With all these recent developments, earlier
observations from [Cullinane et al., 2004] and others that Chinese ports
were under-provisions of physical infrastructure resulting long waiting time
is no longer valid. With logistics infrastructure and management knowhow,
the future of the Chinese ports is optimistic.
3.1.2. Korea
The Busan port is by far the most important container port in South Korea,
accounting for more than 90 percent of the nation’s container throughput.
h The Shenzhen port’s loading and unloading charges are low, nearly half of those at
ports in neighboring Hong Kong.
Located at the eastern tip of the Korean Peninsula, the gateway on

transpacific route, Busan port includes four branches: North Harbour,
South Harbour, Gamcheon Harbour and Dadepo Harbour. As of 2008,
Busan port is the 5th busiest port in the world with a container throughput
of 13,425,000 TEUs and 83,547 vessel calls. Transshipment cargo accounts
for some 43.2 percent of container throughput, while inbound and outbound
cargo accounts for 28.5 and 28.3 percent respectively. A number of factors
have contributed to the growth of Busan port. Firstly, Busan is a natural
deep water harbor which allows the berthing of big vessels. Secondly, Busan
is at the cross road of Northeast China, Japan and Western Russia and thus
has potential to be the regional hub. Third, Busan is an attractive relay
centre for minor Japanese ports because it is cheaper and thus able to
undercut major Japanese ports. Thus, many shippers have been sending
their cargos through Busan for transshipment to/from regional Japanese
ports. Busan is planning build a total of 30 new berths for 50,000 ton ships
by 2011 as competitions from Chinese ports (such as Shanghai and Dalian)
intensify. With the addition of the new berths, the annual handling capacity
of Busan is expected to reach 8 million TEUs.
Other than Busan, the Port of Incheon has contributed greatly to the
development of the economy and industries in South Korea. Located on the
mid-western coast of the Korean Peninsula, Port of Incheon is a gateway to
Seoul. As an artificial port with the world’s largest and most advanced lock
gate (wet dock) facilities that overcome a tidal difference of 10 meters and
permit vessels up to 50,000 DWT to berth directly in the inner closed harbor
basin, the port is also equipped with various modernized harbor facilities
for trade promotion with the main ports of the world. Nonetheless, the
container traffic at the Port of Incheon is merely over 10 percent of that in
Busan (i.e., 1,655,500 TEUs).
Going down south, Port of Gwangyang is situated on the south coast
of South Korea above the Gwangyang Ha River of Yosu. The port of
Gwangyang is equipped with an annual capacity of 5 million TEU and
is the fastest expanding port in Korea. Since 1998, the port has been
operating three branches — West, East and Yulchon Harbour. The port
is connected to land through four eastern and western container driveways,
and directly to a 2.5 km railroad. Yeosu Airport, which is near the Port
of Gwangyang, is currently under expansion. Thus, a systematical network
that enables fast commuting in every direction to and fro the port is formed
with the integration of railroad, highway and other private airports. In
terms of future developments, the port of Gwangyang is scheduled to be
developed into a 33-berth super-scale container port by 2011 that is capable

of handling 9.3 million TEUs annually.
However, [Bong, 2009] warned that growth rate of cargo traffic through
Korean ports is drastically slowing down from 13.8 percent in the 1990s to
2.2 percent in 2008. Bong advocated that several causes of slowdown in
cargo traffic through ports in Korea are: (i) a rise in operating cost, (ii)
a slowdown in foreign trade due to change of industrial structure (i.e.,
increased portion of service industry), (iii) loss of competitive advantage
in traditional labor intensive light industries, and (iv) failure to induce
newly emerging high technology capital intensive industries. The author
also suggested that Korea should capitalize on the underdeveloped logistics
sector, so as to create more value-add through implementations of multi-
modal transportation systems, technology and management know-how that
drives the growth of ports.
3.1.3. Japan
Kobe, Osaka, Tokyo and Yokohama ports represent four of the major
ports in Japan. Specifically, Kobe Osaka, Tokyo and Yokohama ports have
experienced container traffic of 2,432,000 TEUs, 1,725,500 TEUs, 4,271,000
TEUs and 3,490,000 TEUs, respectively in 2008.
Kobe port is located in the central part of the Japanese Archipelago.
Originating as a hub of trading between Japan and the Chinese continent
and Korean peninsula during ancient and medieval times, Kobe port has a
hinterland that covers the whole of western Japan. The geography location
and topology have conferred Kobe port several unique advantages that
makes it the principal foreign trade port of Japan. Firstly, Kobe port lies
on the main routes of world marine-transportation networks. Secondly,
Kobe port is accessible from various directions as it stretches from east to
west. Thirdly, expensive dredging is unnecessary owing to favorable natural
conditions that include some deep-waters berth and no seasonal winds and
rivers flow into the port. Fourthly, the port is also ideal for mooring since
it has little variation in tides. In terms of connectivity, the Kobe port is
served by many regular carrier service lines, including North American,
European, Southeast Asian, and Chinese lines that linked the port with
500 ports in 130 countries. 2006 marks the opening of Kobe Airport with
the provision of Kobe-Kanku Bay Shuttle that provides a ferry service
between Kobe Port and Kansai International Airport. Together with the
existing expressway networks, domestic feeder services, and ferry services,
intermodal transportation efficiency is secured. As part of a continual

improvement process, Kobe seeks to constantly enhance its services for user
convenience and friendliness by reducing port facility charges, simplifying
various port procedures, computerizing operations using EDI (electronic
data interchange) system for submitting various application. In order to
provide greater flexibility to carriers, domestic container feeders are also
permitted to use overseas berths.
The port of Osaka is located in the western part of the city of Osaka.
Similar to Kobe port, the port of Osaka is directly connected to the main
area of the country through an advanced network of expressways and other
main roads as well as a feeder network. It is also directly linked up with
Kansai International airport. The port possesses wharf facilities, 11 berths
and over 3000 meters of quay with draft between 10 and 14 meters in the
Sakishima District. Meanwhile, the terminals in the Yumeshima District
offer 3 berths and over 1100 meters of quay 15 meters with draft of 15
meters. Port of Osaka is called by more than 7,000 ocean-going vessels per
year, of which, more than 5,000 are container carriers. Over the years, the
port enjoys increasing volume of container cargo, particularly with Asia,
as a result of its constant attempt to promote user-friendly services to the
users.
The Port of Tokyo is located on the west coast of Honshu in area
between the estuaries of the Arakawa and Tamagawa Rivers. In 2007,
the port served 31,332 incoming vessels and handled 87.63 million ton of
cargos. The port plays an important role in the distribution of essential
commodities such as sundry goods, foodstuffs, paper products, building
materials and so forth throughout the Tokyo Metropolitan area (Shinetsu
and southern Tohoku) for its industrial activities and 40 million citizens.
Hence, the port has taken early actions to enhance the accessibility and
functionality of its terminals for container, ferry and specialized cargo
use. For examples, warehouses and distribution centers have been set up
in the reclamation areas behind each terminal to complement terminal
functions. Arterial routes, rail and other roadways are developed to
facilitate distribution activities. There are also plans to construct new
container terminals in the Outer Central Breakwater Reclamation Area
to serve the key routes from Asia.
Close to the port of Tokyo, the Port of Yokohama is located on the
northwestern edge of Tokyo Bay, 30 km from Tokyo. The Port of Yokohama
is a naturally blessed port with a spacious water area of ample depth on
the eastern side and undulated hills on the northern, western and southern
sides. The port operates 24 hours daily and has been equipped with various
facilities such as inner and outer breakwaters to protect the port from the
effects of winds and tides. As part of the future plans, Japanese government
aspires to develop the Port of Yokohama into a major container hub port,
with separate facilities for intercontinental and Asian container traffic. It is
estimated that the total cargo volume in 2015 will reach 150 million tons
and the number of cargos handled will exceed 4 million TEUs.
Operating in a country that is technologically and economically
advanced, all Japanese ports are able to offer efficient services through
extensive use of sophisticated state-of-the-art facilities and implementation
of modern management practices. However, apart from the generally more
expensive labor and associated operations cost, [Imai et al., 2001] pointed
out that charges in Japan’s ports have been consistently higher than those
in other major hubs owing to overcapitalization of the port for relatively
small cargo volume.
3.2. Southeast Asia

The following subsection describes the key advantages and challenges facing
some of the major ports in Singapore, Malaysia, Indonesia, Philippines and
Thailand. Figure 5 illustrates major ports in Southeast Asia.
3.2.1. Singapore
The Singapore port is strategically positioned to participate as a tranship-
ment hub for South East Asia and contribute significantly to the country’s
growthi process into one of the core global cities in Asia. Specifically, the
Singapore port is located at the crossroads of international trading in sea
routes in the Asia-Pacific where the geographical topology endows the
port with a naturally deep harbour. The port represents an active feeder
shipping spot in Asia, with a network service ranging from short to long
routes. Other than being highly efficient, the port offer full range of service,
including fuel, pilotage and towage, cargo, vessel repairs, warehousing,
banking, insurance, communications, entertainment, training and education
in port operation and management, logistics and distribution management
and other transport studies.
i The maritime industry comprises more that 5,000 establishments, employs around
100,000 people and contributes more than 7 percent of the Singapore’s GDP in 2008.
Fig. 5 Ports in Southeast Asia.
Throughout many decades, the Singapore port has retained her position
as one of the world’s busiest ports in terms of vessel arrivals, bunker
sales, cargo tonnage handled and container throughput [Cullinane et al.,
2006]. The ‘secrets’ of Singapore port’s success are well-documented in
commentaries and academic studies. Among these, [Zhu et al., 2002] argued
that conductive Singapore’s business environments and well-developed
infrastructures are the main factors attracting MNCs investments. The
traffic at the Singapore port is further augmented through the port-
related industries, which are located in dense and compact districts and
high technical logistic centers as a response to global and local forces
that promotes in and outward multi-national operations. Other reasons
for Singapore port’s success can be attributed primarily to the resident
port and maritime-related community which provide competitive products
and top service standards in world-class to meet the requirements of port
customers.
Today, Singapore port has achieved an impressive container throughput
of 27,900,000 TEUs and become a focal point for 174,620 vessels of some
200 shipping lines with links to more than 600 ports in over 120 countries
worldwide. Singapore port has 49 berths serving container ships, which
can handle up to 26.1 million twenty-foot-equivalent units (TEUs). The
construction of 16 berths has begun in October 2007, when completed in
2013, the port will have an annual handling capacity of 14 million standard
containers (which is an increase of more than 50 percent). In large part,
Singapore’s historical importance was due to its geographic position in
relation to the Straits of Malacca, one of the world’s busiest sea-lanes.
While the port of Singapore continues to serve as an important link for
goods shipped between Asia and Europe, the port has been faced with
stiff competition as an international transportation hub from neighboring
Malaysia in these recent years. Malaysia started taking away Singapore’s
container trade business with the opening of its Port of Tanjung Pelepas
and immediately secured two of Singapore’s biggest shipping clients. [Lam
and Yap, 2006] conducted a comparison on the cost competitiveness among
the port of Singapore, Port Kelang and Tanjung Pelapse port.
3.2.2. Malaysia
Port Kelang and Port of Tanjung Pelepas are two major ports in Malaysia.
Port Kelang is the one of the most established ports in Malaysia, with a
container traffic of 7,970,000 TEUs that ranks 15th in the world and 11th
in Asia. Situated on the west coast of Peninsular Malaysia (40 km from
the capital Kuala Lumpur), Port Klang’s proximity to the greater Kelang
Valleyj makes it a premier port in Malaysia. The port has trade connections
with over 120 countries and dealings with more than 500 ports around
the world. It serves as the nation’s load centre and regional transshipment
centre, and is called upon by 17,000 vessels annually. Port efficiency is
ensured through modern infrastructure facilities, computer information
systems (including EDI), pre-clearance and advanced pre-clearance on
Customs, Health and Immigration formalities.
The major thrust of Port Kelang’s developments will be more
industrial-based dealing with very large consignments, which are in line with
the economic growth in the central region of the country and its identity as a
regional transshipment base. Currently, the port authority is constructing
additional facilities as part of its supply-driven policies. When Westport
is completed, the facilities at Port Kelang will be sufficient to handle the
projected cargo throughput 130.5 million tonnes (i.e., 8.4 million TEUs)
at the end of 2010. As part of the future development plan, Port Kelang
j The Kelang Valley is the commercial and industrial hub of Malaysia as well as the
country’s most populous region.

will see further expansion of port facilities south of Port Kelang between
Tanjung Rhu and Batu Laut (30 km from Port Kelang).
Port of Tanjung Pelepas (PTP) starts operations in October 1999 and
aspires to be the region’s premier transshipment hub. The port is located
at the confluence of major shipping routes at the southern tip of Johor
West in Malaysia. Being only 45 minutes from the confluence of the world’s
busiest shipping lanes, the PTP has steadily attracted the world’s leading
main shipping lines which include Maersk Sealand in 2000 and Evergreen
Marine Corporation in 2002. Port traffic statistics shows that 3,368 vessels
had stopped at the PTP and brought 5,600,000 TEUs of container traffick
to the port in 2008. This put the port into the 18th place in the world; and
12th place in Asia just behind Port Kelang. Factors that have contributed to
rapid growth in the PTP are its excellent port facilities and infrastructure,
supported by a state of the art integrated information technology systems
and highly trained staff, which enabled high efficiency and productivity to
be achieved. The 15 meters naturally sheltered deep-water port also boosts
of its excellent connectivity via road, rail, air or sea. PTP currently has
12 berths and a terminal-handling capacity of 10 million TEUs. Under the
existing expansion plan, the port would build eight new berths and include
land reclamation and dredging. The long-term plan is to have 95 berths
such that capacity will reach 150 million TEUs.
3.2.3. Indonesia
Indonesia has two principal ports, namely the Tanjung Priok and Tanjung
Perak ports. Tanjung Priok port (also known as Jakarta’s port) is located
in western Java 13 km from the city centre of Jakarta. Tanjung Priok port
is the main port for the major manufacturing region around Jakarta and
west Java, and deals with both coastal and international trade. The port
is constructed after the independence of the Indonesia Republic with the
main purpose of ships’ loading/unloading among the islands on recognition
that the existing Sunda Kelapa Port was unable to be further developed
to accommodate increasing trade ships brought about by the opening of
Suez Canal. The Tanjung Priok port is well protected by breakwaters, with
k This figure translates into 8 percent of South-East Asia’s total port market. Of the
5,600,000 TEUs, 95 percent are transshipment and 5 percent hinterland (i.e., local cargo).
The port hopes to increase the latter to 20 percent in the short- to medium-term.
facilities for all types of cargoes. However, the port ownerl noted that the
growth of the port has been hampered by limited capacity and inefficient
operations, including poor road access. Its current capacity is only 600
containers per hour, about two-thirds of the volume projected for next year.
Currently, the Tanjung Priok port ranks 26th in the world and 18th
in Asia by handling 3,984,000 TEUs (which is about 30 percent of total
freight handled by Tanjung Priok Port). In terms of vessel traffic, the
port registered a total of 7,150 calls in 2008. The government has set an
ambitious goal of developing an international-standard regional port. In
order to fulfill this mission, the state-owned port operator has announced
its plans to invest US$286.2 million to improve infrastructure (i.e., the
purchase of new equipment and stronger cranes, and redesigning the docks)
at Tanjung Priok port in Feb 2010. The investment is part of a five-year
plan to modernize Tanjung Priok, cut costs in half and reduce the ship
docking-time from three or four days to two days.
Meanwhile, the Port of Tanjung Perak (also known as the port of
Surabaya City) is located on the northern coast of the island of Eastern
Java, opposite Madura. The port serves 4,700 vessels as one of the main
gateway ports to Indonesia. Being the principal port in East Java with an
annual container traffic of 1,000,000 TEUs, the port also functions as a main
cargo collection and distribution center for both the Province of East Java,
and the whole eastern archipelago of Indonesia. The Port of Tanjung Perak,
equipped to accommodate tankers, general cargo vessels and container
vessels, has undergone continual physical development with modification
of existing berths, and provision of additional berths specifically designed
for container handling operations. The Port Authority, in its efforts to
encourage development of the associated port industries and construction
of the passenger terminal, continues to upgrade and improve both port
facilities and services to meet demand.
3.2.4. Philippines
The two primary container ports in Philippines are Manila and Davao.
Manila port, situated at the East end of Manila Bay, is the most significant
port in Philippines. The port handles 4,062,000 TEUs, which accounts for
over 90 percent of the nation’s international cargoes. With regards to its
l The port operator JICT is jointly owned by Hong Kong’s Hutchison Port Holdings and
state-owned port operator PT Pelabuhan Indonesia II.

global and regional standings, Manila port is the 25th busiest in the world
and 17th seventeenth in Asia. The port of Manila has significantly benefited
from its topology that bestows it a shoreline of 2 km and protected by 3050m
of rock barriers, while enclosing approximately 600 hectares of anchorage.
The port of Manila is presently equipped with an annual capacity of 1.5
millions TEUs. Press release on April 2009 said that the Philippine govern-
ment is forging ahead with plans to establish some 70 modular ro-ro ports
over a period of four years to the tune of US$248.13 million. The proposed
facilities call for the use of prefabricated steel modules for the port super-
structure — pier, causeway, mooring platform, ramp dolphin and terminal
with solar-powered utilities. Negotiations are under way for the requisite
loan package, estimated to cover nearly 88 percent of total project cost.
The Davao port, situated on the southeastern coast of Mindanao Island,
is the second largest port in Philippines with container traffic of 72,000
TEUs. The Port of Davao, otherwise known as Sasa Wharf, holds the
distinction as the premier export and import hub in Mindanao. In addition,
Davao port also functions as the front-line port in the exchange of commerce
and trade between provinces and other parts of the country, as well as, the
principal seaport for most commodities produced along the Davao Gulf. In
terms of topological advantage, the natural islands of Samal and Talikod
along Pakiputan Strait of Davao Gulf bound the port in the east. Hence,
Davao port is relatively protected by landmasses on all sides except at the
South. In December 2008, Davao port has completed a rehabilitation project
involving the construction of a new 42.35 m × 18 m quay, an expansion of the
3,179 square meter Reinforced Concrete Wharf, 13,180 square meter back-
up area, mooring and fendering area, drainage system, and the installations
of port lighting and rockworks. The port can now accommodate eight ships
at the same time.
3.2.5. Thailand
Bangkok and Laem Chabang are two of the leading ports in Thailand.
Bangkok Port (also known as Krung Thep and Klong Toey) is located on the
left side of the Chao Phraya River between 26.5 and 28.5 km from Klongtoey
District, Bangkok. The port serves 2,800 vessels and handles 1,480,000
TEUs in 2008. Bangkok Port is well-connected with road and rail systems,
which enable fast and economical transport of cargoes between the port
and its hinterland. It is also equipped with a bonded warehouse that offers
several value-added services such as online inventory account reporting,

more equipment for lifting and moving goods, and expansion of storage
areas. To provide a more comprehensive support for port-related activities
and ensure optimal resource utilization, four zones are set out for future
development. Specifically, Zone 1 will be developed into a Maritime Business
Centre to accommodate shipping agents, freight forwarders, a Maritime
Training Centre, an Educational and Exhibition Center, a Business Center
catering to banks and financial institutions, and an Integrated IT/EDI
Document Center; Zone 2 is set aside for the construction of high-rise
logistics center, cargo consolidation/distribution center, tax free zone,
general/bonded warehouses, and a Truck Terminal; Zone 3 will serve
as a modern market that complies with sanitation and environmental
standards; and Zone 4 will house a modern office building and other relevant
activities.
Laem Chabang port is located on eastern Thailand in the Sriracha
district, about 130 km south of Bangkok and Thailand’s industrial heart-
land. The deep water harbors of the Laem Chabang port is opened in
1999. The port provides a comprehensive range of 24-7 services to exporters
and importers. Further improvement to transport links has increased the
accessibility of the port and the port has witnessed steadily rising traffic
volume since its opening. Today, approximately 5,134,000 TEUs on 4,650
vessels go through Laem Chabang port making it the 21st busiest in the
world and 15th in Asia. The current plan is to develop Laem Chabang
into an e-Port. With a primary objective of relieving traffic bottleneck,
the main features of this e-port are: (i) RFID-enabled payment system
to inspect vehicles passing through e-Gate Control, verify data on the
number of containers and fee collection; and (ii) Electronic Data Center
for Real Time exchange of e-manifest, Container List and other data
between Laem Chabang Port, dock operators, Customs and Immigration
Department. Also, in Feb 2010, the National news bureau of Thailand
publically announced that The Port Authority of Thailand will push
forward Laem Chabang Port to become the main port for international
and Mekong Sub Region trade, as well as, the center of trade links, business
traffic, merchant marine. A four years development plan has been drawn
up to increase the capacity and energy efficiency of the port so as to
accommodate the rapid increase of the economy, trade and the growth
of Thailand’s international seaborne trade. When completed in 2016, Laem
Chabang port will provide a capacity of up to 7.2 million TEUs a year.
Fig. 6 Ports in South Asia.
3.3. South Asia ports

Compared to ports in Northeast and Southeast Asia, ports in South Asia
are generally less developed and handle less container traffic. Exceptions are
the major Indian and Sri Lankan ports. The following subsection gives brief
accounts of the Jawaharlal Nehru, Chennai and Colombo ports. Figure 6
illustrates major ports in South Asia.
3.3.1. India ports

Jawaharlal Nehru and Chennai are major Indian ports. Jawaharlal Nehru
port, also known as Mumbai port, is the biggest and most environmental
friendly port in India. Commissioned in 1989, the port handles 3,953,000
TEUs (which translates into 55 to 60 percent of the nation’s total
containerized cargo) in 2008. This volume of container throughput places
Jawaharlal Nehru port on 27th and 19th positions in worldwide and Asia.
Jawaharlal Nehru port is well connected to the national extensive network
of railways. The port operates 24 hours per day, possesses modern handling
facilities and adopts up-to-date customs EDI and vessel traffic management
system. In order to cope with the expected growth in container traffic

and container vessel sizes, projects to expand ports facilities and further
improve its rail and road connectivity are underway. More specially, these
include: (i) the development of a standalone container handling facility with
a quay length of 330 m towards North at JNPT; (ii) development of Fourth
Terminal and Marine Chemical Terminal; (iii) deepening and widening of
main harbor channel and JN Port channel; (iv) doubling of rail track; (v)
development of an integrated and centralized tractor parking zone; and
(vi) upgrading of existing roads and yards. On and all, when completed in
2011, the expanded port will assume an additional capacity of 0.8 million
TEUs per annum from the optimum utilizations of the water front area and
accommodate up to 14 m draught, while providing a safer movement for
cargo and avoiding difficulties in container stacking which together enable
faster turnaround time.
At the Coromandel Coast in South-East India, the Chennai Port
(previously known as Madras) operates on a much smaller scale with
container throughput totaling to 1,128,000 TEUs. Since 2009, the port
operator has started the construction of a Second Container Terminal which
features a 832 m long berth with 15.5 meters alongside depth and a back
up area 35 hectares. Capacity of Terminal is 1 million TEUs per annum.
Other efforts to modernize Chennai Port are the realignment of rail and road
network inside the harbor in progress, mechanization of the coal conveyor
system, and the deepening of channels, basins and berths. In addition,
Chennai Port is contemplating to carry out the construction of a dedicated
Elevated Expressway from Chennai Port to Maduravoyal, runs for a length
of 19.01 kilometers along the riverbank and followed by the NH 4 road. This
project is likely to be commissioned in February 2012. Through these series
of continuous modernization that enable the provision of cost-effective and
efficient services, as well as, the implementation of simple and integrated
procedures, and user-friendly approach, the port seeks to achieve greater
heights in the near future.
However, the growths of the Indian ports have generally lagged behind
other ports in Asia. [De Monie, 1995] and [Haralambides and Behrens,
2000] cited poor physical configurations of ports, proximity to urban
development, outdated port facilities, insufficient equipment maintenance,
backward cargo handling techniques, inadequate accountability for cargo
handling, bureaucratic administration, regulations and weak coordination
between departments as possible reasons for poor performance of Indian
ports. [De and Ghosh, 2003] confirmed the hypothesis that if an Indian port
can perform better by improving its operational and asset performances,

then it is likely to get higher traffic but the reverse is untrue.
3.3.2. Sri Lanka ports

The port of Colombo lies on Sri Lanka’s southwestern shores on the Kelani
River. Apart from being a major port in the Indian Ocean, the Port
of Colombo also makes it presence felt as the world’s 28th and Asia’s
20th busiest container ports with a registered traffic of 3,687,000 TEUs.
The port handles most of the Sri Lankan’s foreign trade, including the
manufacturing exports of processed raw materials. Other leading industries
in the Port of Colombo include jewelry, chemicals, glass, textiles, leather
goods, cement, and furniture. As the commercial center of Sri Lanka, the
Port of Colombo also contains head offices of both foreign and local banks,
insurance companies, government offices, and brokerage houses.
With one of the world’s biggest artificial harbors, the Port of Colombo
is made up of three terminals and offers a total of 6245 square meters
of bonded warehouse including 125 square meters of cool room. These
warehouses are equipped to accept all types of goods (except dangerous or
perishable goods), with 24 hr security service provided, and small processing
services for re-export cargoes available at Warehouse BQII. The Port
of Colombo is presently undergoing an expansion on west of its current
southwest breakwater. According to the Sri Lanka Ports authority, the
South Harbor Development Project involves four terminals. Each of this
four terminals will be over 1200 meters long with alongside depths of
18 meters and covers about 600 hectares in total. There are future plans to
deepen the berths in the Port of Colombo to 23 meters for deep-draft vessels.
The South Harbor channel will be 560 meters long with a depth of 20 meters
and a harbor basin depth of 18 meters with a 600-meter turning circle.
4. Factors of Port Competitiveness and Development

In this section, we compile a list of port characteristics that past studies
have found to be significant in influencing port traffic performance as well
as the various measures of port performances.
4.1. Port location

The location aspects of ports have received considerable attention. Before
the introduction and development of containerization of highly efficient
intermodal transportation system, a port’s site clearly defined its hinter-

land.m [Weigend, 1958] suggested that an ideal site for a port should have
sufficient space for its operation, easy entrance, deep water, a small tidal
range and a climate that will not hamper port operations at any time of
the year.
The roles of a port can be distinguished as that of a feeder port, regional
or global hub port. A study by [Fleming and Hayuth, 1994] identified
centrality, intermediacy and proximity as three key location attributes that
confer competitive advantage to some ports and allow them to become
hub. Relative to feeder ports, [Sutcliffe and Ratcliffe, 1995] pointed out
that a successful (transshipment) hub port must be situated at a location
where there is a minimum diversion from the main shipping lanes for the
line haul vessels and distance to the markets served is short. In addition,
[Cargo System, 1998] suggested a port that is ideally located with respect
to the main axial truck routes and either the rich hinterland or the feeder
connections, and supported with appropriate services is more likely to be
chosen by carriers to be their hubs.
Location also affects the way a port should compete. Noting that ports
in small island economies may be at a disadvantage compared to ports
that are natural gateways to rich hinterlands, [Robinson, 2002] advocated
that the former should position themselves in such ways to achieve cost
leadership (economies of scale) or service differentiation (economies of
scope) in order to attain growth in the fast changing and highly competitive
environment. Meanwhile, some other researchers at that time reported the
key success factors for the positioning of small-island seaports to achieve
competitive advantage. By conducting two case studies with Bahamas
Freeport and Malta Freeport, their findings suggested that small-island
ports with no direct hinterlands must first focus on cost leadership and
then develop value-added services after cargo has been attracted.
While ports with locations that are natural gateways to rich hinterlands
are evidently in a better position to develop the sea-to-land interface and
inland transport services, what was once a secure arean for a port to draw
traffic from is no longer the case with the advent of double-stacking of
containers on rail-cars and the establishment of inland intermodal hubs.
Containers can now be shipped long distances across continents to make
m Ports situated at good geographical locations benefit from the advantages of a large
local market and an opportunity to capture transshipment cargo at the intersection of
major sea routes.
n This area is determined based on land distance of the area from the port.
connections with ports. Moreover, as larger and larger ships operating in

alliances seek economies of scale, fewer ports will be served directly by
the larger transoceanic vessels. These load-centered ports are able to send
or draw their traffic far and wide, by water mode or rail or truck, thus
expanding the hinterland of the port. Consequently, ports are not just
in competition with ports in their local area and along their immediate
seaboard, they also are in competition with ports on distant seaboards
attempting to serve the same inland areas [Haynes et al., 1997] [Fleming
and Baird, 1999].
As competitions among distant seaports intensify, there is a growing
role of intermediacy in a port’s traffic structure that contributes to shrinking
captive hinterlands. For load centers that depend on their intermediacy for
cargo traffic, their positions are highly dependent on the ever changing
service networks of shipping lines service network. In view of this, [Heaver
et al., 2001] asserted that global carriers and forwarders challenge a port’s
capacity to influence goods flow and that ports can no longer expect to
attract cargo simply because that are natural gateways to rich hinterlands.
This point had also been noted earlier on by [Slack et al., 1985] and
subsequently by [Carbone and De Martino, 2003].
With regards to the challenge from peripheral ports, spatial expansion
on new sites may be brought about by two seemingly different factors.
Obsolescence of older facilities ties is behind the Bird’s ‘Anyport’–based
explanations. [Bird, 1963 and 1971] postulated that a shift in activity
comes about as a result of a search for new sites that could offer space for
mechanized terminal operations and/or sites adjacent to deepwater channel
that would allow access for ever larger ships. [Frankel, 1987] and [Baird,
1996] supported Bird in their argument that the deeper waterso and the
ocean locations of downstream ports allow the shipping lines to deploy
their largest vessels and save sailing times and improve port turnaround
time. Compared to river or upstream ports, ships can avoid the need for
transiting through the long, narrow, inland waterways. The [Hayuth, 1981]
model,p on the other hand, implies the congestion and diseconomies at
o [Notteboom et al., 1997] disputed the argument, noting that the increase in vessel
capacity is accounted more by an increase in the beam of ships rather than a draught.
The authors cited the continued success of Hamburg and Antwerp as evidence for the
persistence of upstream ports.
p In the light of Hayuth’s five-stage load-center model, [Wang, 1998] examined the
development of the Hong Kong container port in a regional context. The port-hinterland
relationship between Hong Kong and China is found to be unique as the hub. Later,
Fig. 7 Location factors affecting port attractiveness and competitions.
established terminals as the causes of competition. Other factors that have

been put forward include access to shipping lanes, the search for deep-water
sites, labor cost differentials, and environmental restrictions. According to
Hayuth, factors that are important for the development of a load center
port are the large-scale local market, high accessibility to inland markets,
advantageous site and location, early adoption of new technological and
social systems, aggressiveness of port management and the economic and
political incentives of ports. The geographical location factors affecting port
competition may be summarized in Fig. 7.
4.2. Port efficiency

Port efficiency is an important determinant of shipping costs, especially
with the ever increasing vessel sizes that increase the unproductive cost of
vessels waiting for services at ports. According to [Voorde, 2005], improving
port efficiency from the 25th to the 75th percentile can reduce shipping costs
by 12 percent. Besides, inefficient ports may equal to be 60 percent farther
away from markets for the average country.
Since its inception, containerization has gained popularity and has
become an essential component of a unit-load-concept in international sea
[Slack and Wang, 2002] examined the concept of the peripheral port challenge from
an Asian perspective by focusing on the local and regional competition faced by Ports
of Hong Kong, Singapore and Shanghai. The authors confirmed that these ports are
subjected to challenges from peripheral ports in accordance with established models
of port development but also suggested that the causes go beyond the challenge those
postulated by the Hayuth’s models.
freight transportation today. Hence, many researchers have examined the

issue of achieving port efficiency via operations optimization in the context
of container terminals. Among these studies, the adequate provision and
effective allocation of berthing facilities are seen to have received much
attention due to the fact that berthing (including waiting time of ships to
load and unload their cargo) accounts for a significant portion of vessel
time at port and inefficient berthing entails unnecessary productivity loss.
[Steenken et al., 2004] provided a detailed container terminal management
and optimization review from more than 200 papers. His review described
and classified the operations process in container terminals, and presented a
survey of methods employed in the optimizations of quayside transport, the
landside transport and crane transport, as well as, storage and staking logis-
tics. [Vis et al., 2003] also presented an overview survey on transshipment of
containers at a container terminal, organized in accordance to the processes
at container terminals: arrival of the ship, uploading and loading of the
ship, transport of containers from ship to stack and vice versa, stacking of
containers and inter-terminal transport and other modes of transportation.
Apart from the specifics in managing operations, it should be noted
that the provision of up-to-date facilities, equipments and information
technology (IT) infrastructures play an important leading role in enhancing
the overall port efficiency. Nonetheless, investment in modern equipments
to increase capacity is more complicated in practice as [Song, 2002]
demonstrated the value of intelligent facilities investment in a port’s success.
Figure 8 is a summary of significant infrastructural factors affecting port
operations efficiency.
Fig. 8 Infrastructural factors affecting port efficiency.

4.3. Multimodal network

Ports often exist as part of a multimodal network, where various modes
of transport (road, rail, water) are utilized in the successive movements
of goods in an identical loading unit without any handling of the goods
themselves during the transfers between modes. [Fleming and Hayuth,
1994], [Haynes et al., 1997], [Klink and van den Berg, 1998] and [Helling and
Poister, 2000] suggested that the ability of the port to support intermodal
services could be a major factor in the container liners’ choice of port of call.
As such, port choice becomes more a function of network costs. The ports
that are being chosen are those that will help to minimize the sum of sea,
port and inland costs. By supplying intermodal services, ports can also open
new markets beyond their traditional hinterland and create a competitive
advantage over other ports without access to intermodal system.
The intermodal (or multimodal) networks in EU and US have been
explored in [Caramia and Guerriero, 2009]. However, these models devel-
oped in the contexts of the EU or US may not effective in Asia because the
regional patterns of market concentrations and/or logistic chains in Asia
are fundamentally different from EU and US. In North America, markets
are concentrated in east and west coast districts. These, the coastal cities,
which are port cities at the same time, are connected by highway. In Western
Europe, the Europe continent forms the hinterland, where markets are
concentrated and linked to ports cities around the boundaries. In Southeast
Asia, the concentrated markets are located at the port cities which are
separated from one another.
Notwithstanding the unique characteristics of the EU, US and Asia
continents, network optimization that has far-reaching implications on
the development of ports remains as a substantial problem for any large
scale of intermodal transportations. Broadly speaking, network operations
take into account of infrastructure planning, service schedules, routing
and pricing of services, location of international intermodal terminals, and
their associated daily operations. Because of the inherent complexity that
arises in international intermodal transportation, most of these network
operation issues have been explored in the perspective of uni-modal
transportation. Putting peripheral concerns aside, in comparison to local
uni-modal transportation, international intermodal transportation faces a
most strategic problem that is frequently characterized by (i) multiple
(or conflicting) objectives such as minimization of cost and/or transport
time and maximization transport capability; (ii) scheduled transportation
modes; and (iii) time window constrains. Figure 9 is the multimodal
Multimodal transport network.
Fig. 9
transport network of one origin, two destinations and four transport modes,
constructed in the contexts of South and Southeast Asia.
Literature on the route optimization of intermodal network is rel-
atively sparse. Based on the international intermodal transportation in
Asia-Pacific region, [Min, 1990] developed a goal programming model with
chance-constraint to choose the most effective intermodal route which min-
imizes cost and risk while satisfying various on-time service requirements.
The author tested this model by optimizing the intermodal routes between
suppliers in Japan and manufactures in New York. Later, [Bookbinder and
Fox, 1998] contributed an intermodal routing study for Canada-Mexico
shipments under North American Free Trade Agreement (NAFTA). In
their research, a network is constructed between five Canadian origins and
three Mexican destinations. A shortest algorithm is proposed to calculate
routes which minimize cost and throughout time. Most recently, [Chang,
2008] formulated a multi-objective, multimodal and multi-commodity flow
problem with time windows and concave costs. The author optimized the
route from three LCD suppliers in Taiwan to PC manufacture located in
Denver, USA.
For the regional intermodal transportation network problems, [Modesti
and Scimachen, 1998] presented a model to find shortest paths between vari-
ous origin and destination pairs within the urban intermodal transportation
networks of Genoa. Specific to the regional maritime transportation
network, [Al-Khayyal and Hwang, 2007] formulated a model for finding a
minimum cost routing in a network for ships engaged in pickup and delivery
of liquid products. Their research analyzed the trade-off between transport
costs and inventory costs in maritime routing across the whole supply chain
within the Korean maritime networks. For the regional intermodal network
of rail and road, [Caramia and Guerriero, 2009] studied a long-haul freight
transportation problem with multiple objective functions and a focus on
the implementation of a service network design that best to satisfy specific
customer requests.
In a nutshell, the two major factors affecting multimodal network
are (1) proximity/integration to trucking, rail and air transport; (2)
competitiveness of complementary and substitutable trucking, rail and air
transport.
4.4. Maritime trade strategy and institutional settings

The existing economic foundation of maritime and supporting industries,
to a fair extent, determines the degree of aggressiveness in the pursuit
of expansionary maritime trade strategies. At the same time, maritime

trade strategies set the direction and lay the economic foundation for the
developments of port and maritime industry and the nation as a whole.
The effect of the undertaken maritime trade strategy may be reflected on
the export production price and import price, import and export volume,
open or protective trade mode.
[Francois and Wooton, 2001] proposed several maritime strategy models
for different market structures to examine the implications of liberalization
for profits, trade, and national gain from maritime trade. Using data from
South Asia, Latin America and Southern Africa, Francois and Wooton
confirmed that trade liberalization can lead shipping firms to capture a
significant proportion of the benefits in multilateral trade concessions,
and General Agreement on Trade in Service (GATS) negotiations in
this area have important implication for multilateral efforts aimed more
broadly at trade liberalization. By means of theoretical explicative and
simulation modeling, [Coto-Millan et al., 2005] determined national and
world income, prices of imports, exports and maritime transport services,
and the utilizations degree of the productive capacity as the key variables
that explain the behavior of maritime imports and exports for a particular
economy. Quite recently, [Li and Cheng, 2007] collected data from 30
maritime WTO-membership nations and concluded that maritime policy
is firmly based on economic conditions rather than the result of a rational
analysis of policy makers in their relationship exploratory research.
An appropriate maritime strategy devised under considerations of the
institutional settings will aid port development. As [Loo and Hook, 2002]
commented, the evolution and competitive position of a container port
needs to be understood as the interplay of international, national and
local factors. [Fung, 2001] attempted to provide a systematic treatment
for the interactions between the ports of Singapore and Hong Kong, and
to investigate how the rise of South China ports affects the demand
for Hong Kong container handling services. By including the Shenzhen
port’s throughput volume and various external trade variables as exogenous
variables, the study also allowed for sensitivity analysis of their possible
effects on the conditional growth path. Later, [Cullinane et al., 2004]
analyzed the port of Shenzhen using Robinson’s criteria for hub port
development in an attempt to discern whether it will take over the role of
Hong Kong to become the dominant regional hub. Their study concluded
that despite Shenzhen’s current competitive advantages, Hong Kong would,
in all probability, retain its dominant role owing to other institutional
Table 1 Port characteristics on performances.
Port efficiencies Frequency of Port Calls

Port Berthing Time Length
Labor Problems
Economies of Scale (e.g., Cargo Volume/Size of Port)
Diseconomies of Scale (e.g., Cargo Crowding Out Effect/Port
Congestions)
Material Handling Efficiency
Goods Loss and Damage
Flexibility of Operations Process — Large/Odd size Freight,
Large Volume Shipment, Special Handling
Port productivity Rate of Container Movement
Number of Cranes Moves per hour
Port service Working (or Port Operations) Hours
Shipment Information
Provides Assistance in Claims Handling
Offers Convenient Pickup and Delivery Times
Port Service Coverage — Routes
Reliability
Port management Management Expertise and Aggressiveness
Number of Port Operators
Privilege Contracts to Shippers/Carriers
Government Taxes and Incentives
Bureaucracy, Custom Administration and Regulations
Coordination Between Departments
Port-induced cost Port Charges
Handling Charges
Loading/Discharging Rate
Inland Freight Rates
Others Port Reputation
Port Security (Port Safety/Terminal Security)
concerns. Other than direct head-on competition, [Song, 2002 and 2003]
examined the possibility of co-operation between adjacent container ports
in Hong Kong and South China as another maritime strategy.
Section 1.4 has explored into the various factors such as port location,
port efficiency, multimodal network and maritime strategy that may have
a bearing on port development. Table 1 provides further descriptions of the
port characteristics that promote port performances in various dimensions
depicted in Table 2.
5. Lessons to be Learnt
Evolutions of maritime trade over the centuries have revealed that there
exists a close relationship between the development of a maritime industry
Table 2 Measures of port performances.
Competitive performance Volume of TEUs handled

Market Share
Annual Growth
Operational performance Ship Turnaround time
Pre-berthing waiting time
Percentage of idle time at berth to time at working berth
Asset performance Output per ship berth day
Berth throughput rate
Berth occupancy rate
Idle rate of equipment
Financial performance Operating Surplus per ton of cargo handled
Rate of return on turnover
Proportion of General Cargo to Total tonnage handled
and its environment. Every emergent of a hub port city or prosperous mar-
itime trade region is seen to astutely leverage on its own advantages, which
may be a strategic location in proximity to major trading axes and affluent
markets, advanced transportation system, superior services and/or modern
management practices, and catches on the bandwagon during the special
bull period. In turn, a budding maritime industry spurs port developments
just as the opening of Suez Canal promotes the growths of many ports in
Italy, Greece, India and Malacca Strait along the Asia-Europe line. The pace
and degree of port development can be further facilitated by intelligent hard
and soft investment in port infrastructure and management, made under
careful considerations of its intrinsic characteristic relative to its peers.
In this section, we present some fruits for thought related to oppor-
tunities and threats for future port and maritime development in Asia.
First of all, the associations between maritime trade and its business
environment will irrevocably continue to hold. With the progressive global
integration process, world trade is expected to expand and increase the
demand for cargo shipping. More reductions of tariff in maritime trade
will be reached through forthcoming conference agreements with lower
trade barriers resulting to benefit both the producers and customers. As
trade liberalizations across regions continue and invigorate their district
economies, more long haul and short sea shipping will be induced and
subsequently be developed into highway transportation on sea. At the
meantime, China’s Open Door Policy and its membership of the World
Trade Organization (WTO) will provide further opportunities for ports that
are based in China’s manufacture centers (such as Shanghai, Shenzhen and
Tianjin), and also for Singapore and Hong Kong ports that are situated on
the crucial route providing China its needed energy and resources.
Second, maritime trade is a part of world economy which directly

reflects the sentiments of exports and imports conditions. It has been
observed that there is an obvious pronounced change in maritime trade
industry after every major downturn or upturn of economy. During the
economy recession, consuming markets will shrink, and the volume of
world’s import and export dramatically decreases. Contrarily, every boom
of economy brings along a boom in maritime trade, as seen from the
dramatic developments of both maritime trade and economy in Japan,
Korea, Singapore and Taiwan in 1970s. Therefore, in addition to the
specifics of port investment, prevailing and forecasted future economic
conditions would be another element to be considered when deciding on
the timing such investments.
Thirdly, mergers and partnerships in the liner shipping companies have
allowed the deployment of ever larger and fewer vessels, with a primary
motivation to reduce operations cost through economies of scale. However,
this concentration within the liner shipping industry has increased the
potent impact of a move by a major port user on the port’s traffic. Coupled
with advances in logistical systems that lead to the overlapping hinterlands
of ports, the port industry faces heighten competition that entails the
need for greater efficiency in ports (particularly, landside technological
improvement will be required to reduce the waste or relieve bottlenecks at
landside operations) and an ability to meet the changing demands of liner
shipping companies. Nonetheless, [Low et al., 2009] noted that the hub-
and-spoke systems of operations in the liner shipping companies has also
opened up co-operation opportunities for port to engage in collaborative
efforts. Besides the port and maritime industry, the economy at large
will also benefit from such scale increases and collaborative efforts in the
process.
Fourthly, with the advancements of communication and transportation
technologies that transcend time and space, the business competition will
be in the global market in the coming decades. Despite the intensification
of competition, more benefits can expected to be reaped from the scale
of global industry that includes product design, manufacturing, orders
processing, transportation, retailing and other miscellaneous service sectors.
At the same time, greater emphasis will be placed on maritime trade
strategy that optimizes the routing decisions of the entire value chain,
spanning across geographical boundaries and comprising multiple transport
modes. Naturally, ports as a component of the logistics chain will also have
a more significant role to play a key role in export competitiveness, exports
and import prices, as well as, the overall competitiveness of an economy.
Last but not least, port city development may cause land-use conflic-
tion, environment pollution and transportation congestion. We see that
unrestrained investment may not be a good strategy as seen in the case
of the Japanese ports, where there are insufficient traffic and excessive
capacity. Instead of being overly port centered, a lot of other factors
will need to be considered as well. For example, Hong Kong, having
a high population density, high labor cost, limited inland space and
traffic capacity, may face problems related to (noise and air) pollutions,
congestions and financial impediments in its future development. In Japan,
private transport is very expensive and public transportation is often too
crowded. Nevertheless, the impact of the latter in terms of pollution and
traffic congestion is less severe on a per head basis. In Korea, traffic
congestions have resulted in long hours of jams during peak period. For
the developing countries like India, [Pacione, 2006] highlighted problems
such as the inadequacy of infrastructure (i.e., safe drinking water, hygienic
sewage and low cost housing), inability to handle natural disasters (i.e.,
flooding) and control environmental pollution (gaseous emissions). It may
be fortunate for China, who has benefited from the experiences of these
developed countries, to construct an artificial port totally on the sea far
away from the Shanghai city. Such decision not only helps to alleviate
problems related to congestions and pollution, it also effectively solves the
problem that the water near Shanghai city is not deep enough for large
vessels. Another alternative to reduce environmental pollution in the port
city is the development of dry port. The dry port concept is based on a
seaport directly connected by rail with inland intermodal terminals where
shippers can leave and/or collect their goods in intermodal loading units as
if directly at the seaport [Roso, 2007].
6. Conclusions
This chapter tracks maritime trade evolution in Asia from the thirteenth
centuries to the post-World War II, with an analysis of the recent
developments in some of the Asian major ports. Through an extensive
review of the extant literature, this chapter identifies factors that have been
found to affect the port’s standings in various dimensions of operating and
financial performances. Particularly, the impact of port location, inter-port
competition, multimodal network, maritime strategy and institutional set-
tings on how a port should compete are discussed in depth with references to
the classic and contemporary theories. Additionally, the chapter highlights
some possible opportunities and threats facing the future developments of

the port and maritime industry in Asia. We conclude the chapter with
our beliefs that growths in the Asia’s economy and its maritime trade are
promising.
References
1. Al-Khayyal, F. and Hwang, S. (2007). Inventory constrained maritime
routing and scheduling for multi-commodity liquid bulk, Part I: Applications
and model, European Journal of Operational Research, 176, pp. 106–130.
2. Bird, J. (1963). The major seaports of United Kingdom. (Hutchison,
London).
3. Bird, J. (1971). Seaports and Seaport Terminals. (Hutchison, London).
4. Bong, M.J. (2009). Challenges and Opportunities of Port Logistics Industry
in Korea. (Korea Maritime Institute).
5. Bookbinder, J.H. and Fox, N.S. (1998). Intermodal routing of Canada-
Mexico shipments under NAFTA, Transportation Research Part E, 34,
pp. 289–303.
6. Caramia, M. and Guerriero, F. (2009). A heuristic approach to long-
haul freight transportation with multiple objective functions, OMEGA, 37,
pp. 600–614.
7. Carbone, V. and De Martino, M. (2003). The changing role of ports in
supply-chain management — an empirical analysis, Maritime Policy and
Management, 30, pp. 305–320.
8. Cargo Systems (1998). Opportunities for Container Ports. (London IIR
Publication).
9. Chang, T. (2008). Best routes selection in international intermodal networks,
Computers and Operations Research, 35, pp. 2877–2891.
10. Chua, T., Ingrid, R.L., Gorre, S., Adrian Ross, Stella Regina Bermad,
Bresilda Gervacio and M. Corazon Ebarvia (2000). The Malacca Straits,
Marine Pollution Bulletin, 41, pp. 160–178.
11. Coto-Millán, P., Carrera-Gómez, G., Baños-Pino, J. and López de
Sabando, V. (2005). Rate of return regulation: the case of Spanish Ports,
International Advances in Economic Research, 11, pp. 191–200.
12. Cullinane, K., Fei, W.T. and Cullinane, S. (2004). Container terminal
development in Mainland China and its impact on the competitiveness of
the Port of Hong Kong, Transport Reviews, 24, pp. 33–56.
13. Cullinane, K., Yap, W.Y. and Lam, J.S.L. (2006). The port of Singapore
and its governance structure, Research in Transportation Economics, 17,
pp. 285–310.
14. De Monie, G. (1995). The problems faced by Indian Ports today, Maritime
Policy and Management, 22, pp. 235–238.
15. De, P. and Ghosh, B. (2003). Causality between performance and traffic —
an investigation with Indian Ports, Maritime Policy and Management, 30,
pp. 5–27.
16. DeGaspari, J. (2005). Ports look outward: seaports, shippers, and the
government are linking their efforts to maintain a critical infrastructure,
Mechanical Engineering, 127, pp. 30–35.
17. Fleming, D.K. and Hayuth, Y. (1994). Spatial characteristics of transporta-
tion hubs: centrality and intermediacy, Journal of Transport Geography, 2,
pp. 3–18.
18. Fleming, D.K. and Baird, A.J. (1999). Comment — some reflections on port
competition in the United State and Western Europe, Maritime Policy and
19. Francois, J.F. and Wooton, I. (2001). Trade in international transport
services: the role of competition, Review of International Economics, 9,
pp. 249–261.
20. Frankel, E.G. (1998). China’s maritime developments, Maritime Policy and
21. Fujita, M. and Mori, T. (1996). The role of ports in the making of
major cities: Self-agglomerations and hub-effect, Journal of Development
Economics, 49, pp. 93–120.
22. Fung, K.F. (2001). Competition between the Ports of Hong Kong and
Singapore — a structural vector error correction model to forecast the
demand for container handling service, Maritime Policy and Management,
28, pp. 3–22.
23. Grant, P. (2002). Ex oriente luxuria: Indian commodities and Roman
experience, Journal of the Economic and Social History of the Orient, 45,
pp. 40–95.
24. Haralambides, H.E. and Behrens, R. (2000). Port restructuring in a
global economy: an Indian perspective, International Journal of Transport
Economics, 27, pp. 19–39.
25. Harper, T. (2002). Globalism and pursuit of authenticity: the making of
a diasporic public spere in Singapore, Sojourn, 12, pp. 261–292.
26. Haynes, K.E., Hsing, Y.M. and Stouch, R.R. (1997). Regional port dynamics
in the global economy: the case of Kaohsiung, Maritime Policy and Manage-
ment, 24, pp. 93–113.
27. Hayuth, Y. (1981). Containerisation and the load centre concept, Economic
Geography, 57, pp. 160–176.
28. Heaver, T., Meersman, H. and van der Voorde, E. (2001). Cooperation
and competition in international container transport: strategies for ports,
Maritime Policy and Management, 28, pp. 293–305.
29. Helling, A. and Poister, T.H. (2000). U.S. maritime ports — trends, policy
implications and research needs, Economic Development Quarterly, 14,
pp. 298–315.
30. Imai, A., Nishimura, E. and Papadimitriou, S. (2001). The dynamic berth
allocation problem for a container port, Transportation Research Part B, 35,
pp. 401–417.
31. Iris, F.A. Vis and Rene, de Koster (2003). Transshipment of containers
at a container terminal: an overview, European Journal of Operational
Research, 147, pp. 1–16.
32. Irwin, D.A. and Tervio, M. (2002). Does trade raise income? evidence from
the twentieth century, Journal of International Economics, 58, pp. 1–18.
33. Klink, H.A., Van and Berg, G.C. (1998). Gateways and Intermodalism,
Journal of Transport Geography, 6, pp. 1–9.
34. Lam, J.S.L. and Yap, W.Y. (2006). A measurement and comparison of cost
competitiveness of container ports in Southeast Asia, Transportation, 33,
pp. 641–654.
35. Lee, S., Song, D. and Ducruet, C. (2008). A tale of Asia’s world ports: the
spatial evolution in global hub port cities, Geoforum, 39, pp. 372–385.
36. Li, K.X. and Cheng, L. (2007). The determinants of maritime policy,
37. Lin, M. (1998). Maritime contacts between China and Rome, The Western
Regions and the Chinese Civilization during the Han and Tang, pp. 307–321.
38. Liu, X. (1988). Ancient India and Ancient China: Trade and Religious
Exchanges, AD 1-600, (Oxford University).
39. Loo, B and Hook, B. (2002). Interplay of international, national and local
factors in shaping container port development: a case study of Hong Kong,
Transport Reviews, 22, pp. 219–245.
40. Low, J.M.W., Lam, S.W. and Tang, L.C. (2009). Assessment of hub status
among Asian ports from a network perspective, Transportation Research
Part A, 43, pp. 593–606.
41. Min, H. (1990). International intermodal choices via chance-constrained goal
programming, Transportation Research Part A, 25, pp. 351–362.
42. Modesti, P. and Sciomachen, A. (1998). A utility measure for finding
multi-objective shortest paths in urban multimodal transportation networks,
European Journal of Operational Research, 111, pp. 495–508.
43. Notteboom, T.E. (1997). Concentration and load centre development in the
European container port system, Journal of Transport Geography, 5, pp. 99–
115.
44. Pacione, M. (2006). City profile — Mumbai, Cities, 23, pp. 229–238.
45. Robinson, R. (2002). Ports as elements in value-driven Chain systems — the
new paradigm, Maritime Policy and Management, 29, pp. 241–255.
46. Roso, V. (2007). Evaluation of dry port concept from environmental
perspective: a note, Transportation Research Part D, 12, pp. 523–527.
47. Sen, T. (2006). The formation of Chinese maritime networks to Southern
Asia, 1200–1450, Journal of the Economic and Social History of the Orient,
49, pp. 421–453.
48. Slack, B. (1985). Containerization, inter-port competition and port selection,
49. Slack, B. and Wang, J. (2002). The challenge of peripheral ports: an Asian
perspective, GeoJournal, 56, pp. 159–166.
50. Song, D.W. (2003). Port co-opetition in concept and practice, Maritime
Policy and Management, 30, pp. 29–44.
51. Song, D.W. (2002). Regional container port competition and co-operation —
the case of Hong Kong and South China, Journal of Transport Geography,
10, pp. 99–110.
52. Steenken, D., Vob, S. and Stahlbock, R. (2004). Container terminal operation
and operations research — a classification and literature review, OR
Spectrum, 26, pp. 3–49.
53. Sutcliffe, P. and Ratcliffe, B. (1995). The battle for med hub role, Container-
ization International, 28, pp. 95–99.
54. Tan, T. (2007). Port cities and hinterlands: a comparative study of Singapore
and Calcutta, Political Geography, 26, pp. 851–865.
55. Tang, L.C., Low, J.M.W. and Lam, S.W. (2008). Understanding port
choice behavior — a network perspective, Network and Spatial Economics,
10.1007/s11067-008-9081-8.
56. United Nations Conference on Trade and Development Secretariat (2008).
Review of Maritime Transport 2008.
57. Voorde, E.E.M.V.D. (2005). What future the maritime sector? Some con-
siderations on globalization, co-operation and market power, Research in
Transportation Economics, 13, pp. 253–277.
58. Wang, J.J. (1998). A container load with a developing hinterland — the case
of Hong Kong, Journal of Transport Geography, 6, pp. 187–201.
59. Weigend, G.G. (1958). Some elements in the study of port geography,
Geographical Review, 48, pp. 185–200.
60. Yabe, Y. (1991). Major characteristics of urban waterfront redevelopment in
Japan, Marine Pollution Bulletin, 23, pp. 397–401.
61. Zhu, J., Lean, H. and Ying, S. (2002). The third-party logistics services
and globalization of manufacturing, International Planning Studies, 7,
pp. 89–104.
CHAPTER 2
RECENT DEVELOPMENT OF
MARITIME LOGISTICS
Loo Hay Lee, Ek Peng Chew, Lu Zhen, Chee Chun Gan, and Jijun Shao
Department of Industrial & Systems Engineering,
National University of Singapore
1 Engineering Drive 2, Singapore 117576
[email protected]
The economic crisis from the end of 2007 to early 2010 has greatly affected
the maritime industry. Some of the biggest problems include oversupply of
tonnage, sudden drop in demand, and declining freight rates.
Global containerized trade enjoyed a year-on-year increase of 4.6 percent
in 2008. The liner shipping market accounts for 16 percent of the world’s trade
goods loaded in terms of tons. One of the most significant trends in liner
development is the increase in vessel size. There are notable increases in both
the size of the largest vessel and the average vessel size. The productivity of
most container lines, however, decreased in 2008, partly due to the various cost
control measures adopted by carriers, such as slow steaming, to mitigate the
effect of the crisis. Freight rates also suffered in 2008 and 2009. With demand
recovering in 2010, freight rates have increased, and have even reached pre-crisis
levels in some trade routes.
For container terminals around the world, global operators are still the
market leaders and this trend is likely to continue. Container transshipment
activities have enjoyed rapid growth during the 1990s, and the share of
transshipment activities in total container volume has reached equilibrium from
2000 onwards. There were some notable improvements to port performance in
2008. The United Nations’ Liner Shipping Connectivity Index 2009 has clearly
shown the effect of the crisis in the industry.
1. Development Trends on Global Container Shipping

1.1. Global economic condition and industry perspective
Global container shipping has been growing in the past 20 years before
the current economic crisis. Figure 1 shows the growth of maritime
49
50 L. H. Lee et al.
Fig. 1 GDP, trade and maritime containerised transport. 1985–2007.

Source: Sánchez, Ricardo J.(1)
containerized trade and world gross domestic product (GDP). The growth
rate for seaborne containerized trade is higher than the growth rate of
world GDP and total seaborne trade. There has been sustained growth
even during years of past recessions.
The economic crisis of 2007–2010, also known as the Great Recession,
has seriously damaged the global economy. In 2009, the world GDP declined
by 1%, for the first time in the post-World War II era. The recession has
also hit global trade, with trade volumes declining by as much as 25% in
2009 from 2008’s level, the largest single year drop since World War II.
Lower commodity prices, tighter credit, and increasing protectionism all
contributed to the drop in trade demand during the recession.(2)
The container shipping industry has been badly hit by the recession.
Throughout year 2009, container volumes have declined, liners and terminal
operators have suffered from diminishing profits or even losses, new
container ship ordering has almost halted, vessels have been laid up by
ship owners, and freight rates and charter rates have reached record low
levels. The crisis has quickly caused a shake up in the container shipping
industry.
There are quite a few reasons for the quick spread of the downturn
to the maritime industry. One of the major reasons is that the demand
Recent Development of Maritime Logistics 51
for maritime transport services, especially container shipping services, is

derived from economic growth and trade volume. As the economic condition
is bad during the recession, the demand for trade, as well as the need
for maritime transport services to carry goods from manufacturers to
customers, has diminished. Another side effect of such strong dependency is
that the global seaborne trade and container shipping is likely to rebound
and recover from the recession slower than other sectors.(3) Another factor
that affects the industry is the ever increasing gap between supply and
demand for container shipping. The recession has caused a sudden and
unexpected drop in the demand side. However, most of the orders for new
ships were placed during a period of rapid growth for both the container
shipping industry and the demand for shipping services. As these orders
are approaching their delivery date, the growth in the supply side of the
container shipping industry will outpace the growth, if not contraction, in
demand. In 2009, the total container fleet is expected to grow at 9.6 percent
while the demand is expected to go down by 9.1 percent. The growth rates
of supply and demand in global container shipping from 2000 to 2009 are
shown in Fig. 2. We can see that before the current crisis, the growth
rates of demand and supply closely matches each other. However, in 2009,
there is a huge gap between the growth rates for demand and supply. The
gap in the growth rates would likely lead to the gap in the actual figure
Fig. 2 Growth rate of supply and demand in global container shipping.

Source: the UNCTAD secretariat(3)
52 L. H. Lee et al.
Fig. 3 Expected container fleet.

Source: Sánchez, Ricardo J.(1)
of demand and supply, which will likely cause huge overcapacity in the
industry. Furthermore, as new vessels typically take years from point of
order to delivery, the existing orders are likely to keep the growth rate
high on the supply side for several years and thus enlarge the overcapacity
problem until they are fully delivered in 2011. Figure 3 illustrates the
expected total container fleet at the end of each year to 2011, based on
current orders. We can see from the chart that the speed of the capacity
expansion is quite fast, and the problem of overcapacity would likely to
remain in years to come. This overcapacity problem has been recognized by
many industry experts, and carriers have adopted several measures to deal
with the problem. Orders are being cancelled or delayed by carriers, and
excessive capacity is being removed by laying up idle ships and demolishing
older vessels.
Many industry experts feel that the unavailability of trade financing
during the downturn has also made a significant contribution to the rapid
spread of the downturn in the maritime industry. Since the recession started
as a financial crisis and banks are taking extra caution on lending, it is
much harder for carriers, shippers and port operators to obtain sufficient
financing for projects. According to the UNCTAD Secretariat, the shortage
of financing is most severe in developing economies, with unmet demand
estimated to range between $100 billion and $300 billion annually.(3)
The recovery from the recession started at the end of 2009 and
continued into the first half of 2010. The global economic condition has
become better, and the recession has been considered to have ended in the
first half of 2010. However, there is still substantial risk in the shipping
industry in 2010. In a survey conducted in September 2009 by AMR
Research, “the recovery cycle” has been identified as the biggest risk in
2010. The main reasons given in the survey include “potential commodity
price increases, limited internal skills after work force reductions, and
problems meeting new demand with constrained capacity, low inventory
and transportation constraints.”(4) Europe-based shipping industry analyst
group Seabury has forecasted that global seaborne container trade will
grow about 11 percent this year over 2009.(5) There is much uncertainty
in the container shipping industry as well as the global economic condition
during the recovery cycle in 2010. Many carriers remain cautious about
the economic prospect, and some shipping executives still feel that the
downturn has not brought enough needed consolidation and efficiency to
the container shipping industry and overcapacity may still be a big problem
during the recovery.(6)
1.2. Recent trends in container shipping industry

The world container ship fleet continues to expand in 2008 despite the
economic crisis. Table 1 Development in World Container Ship Fleet shows
the development in global container ship fleet for the past decade. We can
see that there were 4,638 ships by the beginning of 2009, with a total
capacity of 12.14 million TEUs. This represents an increase of 8.5 percent
in the number of vessels from 2008 to 2009. In terms of TEU capacity, there
is a 12.9 percent increase. Another interesting trend in the world container
fleet is that the average vessel size increased from 2516 TEU in 2008 to 2618
TEU in 2009, representing a 4 percent growth. For new container ships, the
Table 1 Development in world container ship fleet.
World Growth
total 1987 1997 2007 2008 2009 2009/2008
Number of 1 052 1 954 3 904 4 276 4 638 8.47

vessels
TEU 1215 215 3 089 682 9 436 377 10 760 173 12 142 444 12.85
capacity
Average 1 155 1 581 2 417 2 516 2 618 4.04
vessel
size
Source: the UNCTAD secretariat(3)

54 L. H. Lee et al.
trend towards larger vessels is much clearer. In 2008, the average size of new
container ships entering service is 3489 TEU, an increase from 3291 TEU
in 2007. On 31 October 2009, there were 218 new 2009-built fully cellular
container ships in service with average carrying capacity of 4,125 TEU.(3)
Another trend in the world cellular fleet is the trend towards more
gearless vessels. Among 2008-built container ships, nearly 80 percent of
vessels and 90 percent of TEU capacity are gearless, comparing to only
about half of the vessels built 10 years ago. This is also because of the
development in port facilities, since more and more ports are equipped
with modern handling equipment, especially specialized gantry cranes. The
trend towards gearless vessels is likely to encourage ports to invest further
in port handling equipment.(3)
2. Liner Shipping
2.1. Container liners
The container shipping market has grown rapidly over the past 20 years.
In 2008, global containerized trade was estimated at 1.3 billion tons, which
represents a year-on-year increase of 4.6 percent. The liner shipping market
accounts for 16 percent of the world’s trade goods loaded in terms of tons.(3)
Liner shipping companies are integral components of the global con-
tainer shipping network, and their decisions and policies obviously have large
effects on the liner industry. The past few years have been a difficult time for
most liners and they have had to make many adjustments in order to stay
afloat in the face of sharp reductions in volume and severe overcapacity.
2.1.1. Liner market developments

The decade from 2000 to 2010 saw further concentration of the container
market in the hands of the top 10 liner operators. The capacity share of the
top 10 operators grew from 49 percent to 58 percent. The majority of the
increase can be accounted for by the growth from Maersk, MSC and CMA-
CGM. Maersk’s and CMA-CGM’s growth arose mainly from mergers and
acquisitions, whereas MSC’s increase in capacity share came from organic
growth.
As of 1 January 2010, AXS-Alphaliner’s Top 100 Report on the fleet
capacity share of the top 20 liner operators around the world includes
13 liners from Asia, reinforcing Asia’s position as the leading region for
containerized trade. However, several Asian carriers have also significantly
reduced the size of their container fleet over the past year. Most notably, the
three Japanese carriers NYK, MOL and “K” Line have decided to reduce
their exposure to the volatile liner trade as part of a long term corporate
strategy shift. According to AXS Alphaliner, seven major Asian operators
have disposed of a combined 165 vessels totalling 282,000 TEU in the past
15 months. This includes 155,000 TEU sent for scrap and 127,000 TEU in
secondhand sales.
Maersk Line maintained its dominance as the market leader in Jan
2010, but lost ground to second-ranked MSC in market share in terms of
fleet capacity. Maersk’s share dropped from 16 percent in 2007 to 15 percent
in 2010 while MSC’s share increased from 10 to 11 percent in the same
period. However, almost all of the top 20 liner companies experienced mas-
sive reductions in their profits, with Hapag-Lloyd being the only exception.
2.1.2. Increase in liner ship size

The size of container ships has been increasing steadily over the past three
decades, as shown in Fig. 4. In 1975, the largest container ship had a
capacity of 3,000 TEU. In 1991, that increased to 4,000 TEU and 6,800
TEU in early 2000. Most recently in 2009, container ships with capacity
of up to 14,000 TEU have been pushed into service. The vast majority
of these ships are used on the Asia-North America and the Asia-Europe
routes, which offer the optimum combination of high volumes, long voyages
and deep, efficient ports. The construction of wider locks at the Panama
Fig. 4 Trend in Largest Container Vessels.

Source: Historical series compiled from Containerisation International, various years
56 L. H. Lee et al.
Canal, which is expected to be complete by 2014, will provide carriers with

additional options for deploying their super-sized container ships. The wider
locks will be able to accommodate “New Panamax” ships of about 12,500
TEU, compared with the current Panamax capacity of about 5,000 TEU.
High-volume container ships can provide significant cost savings to liner
companies because of the economies of scale in major trade lanes. A 12,000-
TEU ship can be operated with the same 13 or 14 crewmembers required by
a ship with half the capacity. Per-unit costs of capital investment and fuel
consumption are substantially less than for two vessels of half the size. It is
thus no surprise that large container ships also dominate the order books,
with ships of capacity over 7,000 TEU making up 58 percent of the current
capacity on order. 30 percent of the order book is comprised of 128 vessels
of 12,000 TEU and above. In combination with the growth in the number of
container ships worldwide, this trend represents a very significant increase
in global container shipping capacity.
The increase in size of container ships has also lead to more hub-and-
spoke structures in the global shipping network. This is because of the
fact that only several major trade lanes have sufficient demand to support
these mega vessels. Furthermore, the mega vessels can only dock in a small
number of ports because of their deep drafts. It is thus natural for carriers
to adopt a hub-and-spoke network structure in order to take advantage of
the economies of scale by consolidating the demand from smaller trade lanes
and ports. Consequently, the demand for transshipment operation has also
increased due to the shift towards larger vessels.
Not all liner companies, however, are convinced that the trend towards
increasing container ship sizes is healthy for the market. Evergreen has
decided not to include any of these mega-containerships in their recent
order for 100 new container vessels. Evergreen’s order enables them to
take advantage of the drop in newbuilding prices amidst the current
oversupply in shipping capacity, but Evergreen chairman Chang Yung-Fa
had previously expressed reservations on the market impact of such high-
volume vessels, which are unable to take advantage of their economies of
scale unless they are constantly full. With the current depressed global
economy, it remains to be seen whether the trend of liners increasing their
ship sizes will continue.
2.1.3. Liner productivity

Due to the high cost of assets such as container ships, liner companies are
naturally concerned about liner productivity, which measures the utilization
of such expensive assets. Unfortunately, the productivity of the world’s

shipping fleet in terms of ton-miles per deadweight-ton (dwt) has been
decreasing since 2005 and is expected to decline further. By 2008, liner
productivity had fallen to levels only 6 percent higher than those in
1990.
The decrease is due to a combination of many factors, most notably the
oversupply of tonnage coupled with the decrease in trade growth. In 2009,
the size of the world’s container fleet was estimated to grow by 9.6% in
contrast with a drop in demand of 9.1%. The cargo volumes carried by the
world residual fleet, including container ships and general cargo carriers,
dropped from 10.84 to 10.4 tons per dwt.(3)
Furthermore, liner companies have introduced various cost saving mea-
sures which further decrease liner productivity. Previously, liner companies
introduced slow steaming in order to reduce fuel costs. With the recent
decline in fuel prices, liner companies have also taken to choosing longer
but cheaper routes, which has also resulted in fewer ton-miles per dwt.
On the other hand, these longer routes, e.g. bypassing the Suez Canal
via the Cape of Good Hope, can result in significant cost savings (up
to $300,000 in some cases for the largest ship) even after considering
additional fuel and crew costs, mainly due to savings on canal transit fees.
Such rerouting also bypasses the hotspot for piracy near Somalia, which
presents additional cost savings for insurance. It is also a good measure to
absorb additional capacity, since such rerouting will increase the average
sailing time by approximately 7 days. However, the viability of the reroute
strategy relies on the current relatively low bunker prices. The long-term
applicability of such strategies could come into question when bunker prices
increase.
2.2. Freight rates

Liner companies have been hit hard by the sharp decline in freight rates
since the start of the economic crisis. Freight rates for most vessels
plummeted due to the decrease in shipping volumes, with April 2009 rates
dipping below the levels experienced in 2000. Container volumes on the
Asia-Europe route fell by around 15 percent in 2008. This contributed to a
severe drop in freight rates in early 2009 to about $300 per TEU, a decline
of 80% from the peak in 2007.(3)
Furthermore, in 2008 the European Union repealed the block exemption
that had previously been granted to liner conferences with regards to price
58 L. H. Lee et al.
and capacity setting. This resulted in the former members of the Far East
Freight Conference having to set their own tariffs and surcharges since
18 October 2008, presenting an additional burden to shippers in having
to keep track of the multitude of different rates. The first quarter of
2009 saw a corresponding decrease in trade volumes on major routes to
and from Europe. The Europe-Asia route declined by 22 percent on the
westbound route from Asia, and the eastbound route from Asia experienced
a 17 percent decline at the same time. The transatlantic westbound route to
North America also experienced a 17 percent decline, while the eastbound
route from North America to Europe dropped by 30 percent.
However, great improvements in freight rates can be seen recently due
to recovering demand with the peak season and great efforts from liners in
maintaining freight rate increases despite heavy competition. Transpacific
freight volumes have increased 13% year-on-year in the first quarter of 2010,
with corresponding freight rate increases to $2,607 per FEU compared to
pre-crisis levels of $2,000 per FEU.
Such recovery does not come without cost, however. Shippers are
becoming increasingly unhappy about sharp increases in rates, in addition
to service changes such as slow-steaming which increase shipping time.
Furthermore, it is unclear if the recovery in shipping volumes is sustainable
rather than simply a case of retailers restocking inventory. A deterioration
in volumes, combined with an expected wave of new containership deliveries
(410,000 TEU in Q2 2010), may undermine the recovery potential in freight
rates. Some shippers have proposed holding regular pricing discussions with
liners in an effort to prevent seeing such sharp swings in freight rates,
which also add uncertainty to shippers’ supply costs. Liner companies, on
the other hand, have also expressed concerns about regular communication
and cooperation with shippers to reduce costs and conflicts on both sides.
3. Ports
3.1. Global container terminal operators
Global container terminal operators are companies with activities in more
than one geographical region. Most global terminal operators are private
organizations, while PSA, DP World, HHLA, COSCO Pacific and China
Shipping are majority owned by elements attached to a state or public
entity. Global container terminal operators have the highest market share
in South East Asia and North Europe, controlling 76.6% and 75.5%
of throughput respectively in these two regions. Global operators are
least significant in Eastern Europe (30.3%), South America (32.2%) and

Australasia (32.8%). Global operator throughput capacity is expected to
rise slightly above the average by 2.7% annually between 2008 and 2014. By
2014, the combined market share controlled by global operators as measured
by capacity will be 55.7%.
According to Drewry, global container terminal operators can be
classified into three categories:
(1) Global stevedores: these are companies whose core business is termi-
nal operation. They view terminals as profit centers. As a result, global steve-
dores often focus on achieving greater efficiency by implementing common
systems across a terminal network. HPH, PSA and DP World are the leading
stevedore-based global operators. This type of terminal operator accounts for
57.2% of the total throughput by global terminal operators in 2008.
(2) Global carriers: these are companies for which container shipping
is the prime focus, and whose terminal networks support shipping activity.
Terminals are often run as cost centers and efficiency gains are achieved
through the integration of the terminals with a global shipping service
network. Evergreen, Hanjin, K Line, China Shipping and MSC are the
leading carrier-based global operators. This type of terminal operator
accounts for 12.2% of the total throughput by global terminal operators
in 2008.
(3) Global hybrids: these are companies focusing on both shipping
and terminal operation. The main activity of such companies, or that of
their parent groups, is container shipping, but separate terminal-operating
divisions have been established. These terminal-operating divisions are
expected to handle a significant amount of third-party traffic besides serving
the core liner shipping business of the parent companies. They are often
designed to operate as independent profit centers. For example, APM is the
biggest and longest-established hybrid terminal operator. CMA, COSCO
and APL/NOL are also the examples of this type of operator; their terminal
operating divisions are Terminal Link, COSCO Pacific and APL Terminals
respectively. This type of terminal operator accounts for 30.5% of the total
throughput by global terminal operators in 2008.
There are some big stevedore-based terminal operators which are not
considered as global terminal operators as they only focus on one country.
The Shanghai International Ports Group and Gulftainer are the examples
of these operators. They have expressed interest in expansion overseas.
Thus more companies will be added to the existing list of global terminal
operators in the future.
60 L. H. Lee et al.
Some liners may step up investments in their sectors of terminal

operations. For example, Hamburg Sud is reportedly participating in
a terminal development in Santa Catarina, Brazil; UASC has taken a
shareholding in a new development in the Egyptian port of Damietta.
However, the uncertainty involved in current business climate may imply
that such plans are put on hold for the immediate future.(7)
The financial crisis is likely to seriously curtail the investments in
terminals in the short to medium term. Some carriers may be forced to
dispose of their terminal assets. Bankruptcies of carriers and mergers and
acquisitions between carriers are anticipated. It may create some larger
liner-based global terminal operators over the next few years. In the first
half of 2010, as the economic conditions improve, operators are once again
considering future investment opportunities.
3.2. Leading terminal operators

Drewry ranks container terminal operators by two different forms of
measurement:
Measurement 1: the summation of total throughput of all terminals in which
the global operator has a stake of more than 10%.
Measurement 2: the weighted summation of terminal throughput based on
the equity share of the operator in the terminal, which could correctly reflect
the amount of interest of the operator in the terminal and eliminate double
counting on terminals with multiple interests.
The league table of global operators according to the above two forms
of measurement are listed as follows in Table 2:
The big four terminal operators (PSA, HPH, APMT, DPW) are cur-
rently the dominant players in the global container terminal sector, by both
forms of measurement. There are quite a few challengers, like Cosco, MSC,
and perhaps China’s SIPG (Shanghai International Port Group), but their
positions appear to be secure for the foreseeable future. By Measurement 1,
which is the total throughput measure, these top four market leaders enjoyed
a 45.6% share of world throughput in 2008. This compares with 45.1% in
2007, underlining the fact that the influence of these four companies within
the container terminal market continues to strengthen. The fastest-growing
operator by Measurement 1 is the Marseilles-based shipping group CMA
CGM, which jumped from 15th (in 2007) to 11th (in 2008). The biggest
rate of decline among the global operators was APL, which dropped from
Table 2 The league table of global operators’ throughput in 2008.
Measurement 1 Measurement 2
Million % Million %
Rank Operator TEU Share Rank Operator TEU Share
1 HPH 67.6 13.0 1 PSA 50.4 9.6

2 APMT 64.4 12.3 2 HPH 34.4 6.6
3 PSA 59.7 11.4 3 APMT 33.8 6.5
4 DPW 46.2 8.9 4 DPW 32.9 6.3
5 Cosco 32.0 6.1 5 Cosco 11.1 2.1
6 MSC 16.2 3.1 6 Evergreen 8.9 1.7
7 Eurogate 13.2 2.5 7 MSC 7.9 1.5
8 Evergreen 10.3 2.0 8 Eurogate 7.4 1.4
9 HHLA 7.4 1.4 9 HHLA 6.7 1.3
10 SSA Marine 7.4 1.4 10 SSA Marine 4.6 0.9
11 CMA-CGM 7.0 1.3 11 APL 4.2 0.8
12 Hanjin 5.7 1.1 12 CMA-CGM 4.1 0.8
13 Dragados 5.5 1.1 13 NYK Line 3.8 0.7
14 NYK Line 5.5 1.1 14 ICTSI 3.7 0.7
15 APL 5.4 1.0 15 Dragados 3.7 0.7
16 OOCL 3.9 0.7 16 Hanjin 3.6 0.7
17 ICTSL 3.8 0.7 17 K Line 2.6 0.5
18 K Line 3.4 0.6 18 MOL 2.4 0.5
19 MOL 3.2 0.6 19 GrupTCB 2.4 0.5
20 Grup TCB 3.2 0.6 20 OOCL 2.1 0.4
21 Yang Ming 2.0 0.4 21 Yang Ming 1.3 0.3
22 Hyundai 1.1 0.2 22 Hyundai 1.2 0.2
Total 374.1 71.6 233.1 44.6
Source: Drewry Shipping Consultant(7)
11th (in 2007) to 15th (in 2008). PSA of Singapore is the leading global
container terminal operator by a huge margin according to Measurement
2. Its 50.4million equity TEU throughput in 2008 is more than 16 million
above the second place, i.e., HPH.
Table 3 places the operators in order of the total available capacity at
all terminals in 2008.
The company with the biggest capacity at the end of 2008 was APM
Terminals. HPH and PSA also achieved solid increases in available capacity
of around 5.4% and 7.6% respectively, allowing them to retain their second
and third rankings in this capacity table. DP World and Cosco both
achieved double-digit capacity growth in 2008. Among the top 20 global
operators, 8 companies increased capacity by more than 10% between 2007
and 2008. The highest growth rates were recorded by OOCL, SSA Marine,
NYK and ICTSI.
62 L. H. Lee et al.
Table 3 The league table of global operators’ capacity in

2008.
Rank Operator million TEU % Change 07-08
1 APM Terminals 98.4 3.8

2 HPH 84.8 5.4
3 PSA 84.6 7.6
4 DP World 63.7 10.6
5 Cosco 61.9 13.1
6 MSC 22.2 13.2
7 Eurogate 20.7 16.4
8 Evergreen 16.3 1.2
9 SSA Marine 13.6 24.4
10 Hanjin 11.2 −0.6
11 NYK 10.2 22.0
12 CMA CGM 10.1 7.7
13 HHLA 9.8 4.9
14 Dragados 9.1 −9.0
15 APL 7.8 0.0
16 ICTSI 6.6 19.3
17 K Line 5.2 −10.3
18 OOCL 5.0 25.0
19 MOL 4.7 −6.9
20 Hyundai 4.3 2.7
21 TCB 4.2 −0.1
22 Yang Ming 3.6 −5.0
Despite the uncertain short term prognosis for the world’s container
trades, the future outlook beyond this is still positive, with China,
India and Vietnam in particular expected to push container traffic levels
higher. Consequently, there is an underlying need for further expansion of
container terminal capacity in the medium and long term. Indeed there
are several regions where, once economic recovery is underway, container
terminal bottlenecks could quite quickly strangle growth unless investment
is sustained. Actually, during the first half of 2010, such trends have already
been noted in some ports.
DP World is one the most acquisitive operators in recent years and
continues this pattern in 2008. Notable developments include the purchase
of a 90% shareholding in the operator of the Egyptian port of Sokhna, a 60%
shareholding in a terminal operator in Tarragona, Spain, and so on. PSA
invested in International Trade Logistics of Argentina and consequently
obtained a stake in the Buenos Aires operator Exolgan. PSA also purchased
substantial shareholdings in terminals in Kandla and Kolkata in India, and
entered into a joint venture with PIL to develop and operate a terminal
in Singapore.
PSA and HPH remain heavily dependent on traffic generated by
terminals in the Far East and South East Asia. These two regions account
for more than 81% of the PSA’s equity TEU throughput, and 59% of HPH.
Both PSA and HPH also have a significant presence in Europe, deriving
respectively 16.8% and 25.5% of their equity TEU from European terminals.
APM Terminals and DP World have the most geographically balanced
spread of container terminal activities of all leading global container
terminal operators.
The container terminal business is highly competitive and competition
has further intensified since the economic downturn. The competition
between global operators does exist within many of the world’s leading
container ports, including, for example, Busan, Kaohsiung, Hong Kong,
Rotterdam, Antwerp, Hamburg, Long Beach and Los Angeles. In fact, of
the world’s top ten container ports, competition between global container
terminal operator is absent only in Dubai, Singapore and China’s Shanghai
and Shenzhen ports.
Global container terminal operators are increasingly willing to coop-
erate where it is in their mutual interest to do so. A large number of the
joint ventures in which two or more global operators have shareholdings are
partnerships between a stevedore-based global operator and one or more
carrier-based operators.
Last but the most important, the financial performance of the global
container terminal operators is listed as follows:
As shown in Table 4, HPH and DP World achieved the highest turnover
and earnings from their global container terminal operations. The third
Table 4 The financial performance of the selected global operators in 2008.
Terminal operator Throughput Turnover (million USD) Earnings (million USD)
Eurogate 14.2 996.97 282.59

HPH 67.6 5109.08 1707.80
ICTSI 3.8 435.52 184.73
NYK Line 6.4 1447.78 65.81
PSA 63.1 3025.09 901.60
APMT 76.9 3119.00 573.00
Cosco Pacific 45.9 unknown 128.20
DPW 46.6 3283.00 1340.00
HHLA 7.4 1104.74 525.38
APL 5.4 577.00 93.00

64 L. H. Lee et al.
most profitable global operator is PSA. Among the top five global terminal
operators, DP World achieved the best earnings per TEU, with $28.74 per
TEU, followed by HPH with $25.25 per TEU, and PSA with $14.30 per
TEU. APM Terminals and Cosco fared less well with earnings per total
TEU of $7.45 and $2.79 respectively.
3.3. Development in transshipment activities

The container transshipment volume has been growing steadily. The
proportion of transshipment volume in total container volume rose from
18 percent in 1990 to 25 percent in 2005.(8) The following figure illustrates
the growth of the percentage share of transshipment activities in total port
volume from 1990 to 2006.
The major driver of transshipment activities is the use of hub-and-
spoke system by liners. The hub-and-spoke system can consolidate demand
from small ports to hub ports, so as to achieve economies of scale. However,
the hub-and-spoke system may also incur extra costs in longer transit time,
extra handling at the hub port, etc. From Fig. 5 Growth in Transshipment
activities (Source: Transport and Tourism Division in UNESCAP(8) we
can see that the percentage share of transshipment activities increased
rapidly during the 1990s, and at a much slower rate from 2000 onwards. The
percentage share of transshipment activities has reached equilibrium, which
Fig. 5 Growth in Transshipment activities.

Source: Transport and Tourism Division in UNESCAP(8)
Fig. 6 Trends in LSCI Indicators.
is likely caused by stabilization in the hub-and-spoke network structure of

major carriers.
According to UNESCAP estimates, global container transhipment vol-
ume will increase from around 85 million TEU in 2005 to 184 million TEU in
2015 at an average growth rate of 7.6 percent per annum. This growth rate is
on par with the expected growth rate for total world container volume.
3.4. Improvement in port performance

Improvements on cargo handling in container terminals would greatly help
the global supply chain. The most notable improvement to port perfor-
mance in 2008, as mentioned by the Review of Maritime Transport 2009,
is the increase in number of ports achieving higher crane productivity.(3)
Following the trend of increasing vessel size, there is also much pressure on
the efficiency of terminal handling operation. With the help of new training
programmes, yard improvements, and the deployment of new equipment,
many ports have achieved great improvement on crane moves per hour. In
recent years, tandem-lift, triple-lift and even quad-lifts cranes have been
installed on more and more terminals. However, these new cranes have not
revolutionized the industry.(3)
3.5. UNCTAD liner shipping connectivity index 2009

The UNCTAD Liner Shipping Connectivity Index, or the LSCI, is a famous
measure used by the United Nations to assess different countries’ connection
66 L. H. Lee et al.
to world markets and their import and export activities. Figure 6 shows the
trends in LSCI indicators from 2004 to 2009.
There are 5 indicators in the LSCI: the TEU capacity of the largest
container vessels, the number of liner companies per country, the TEU
capacity deployed per country, the number of services per country, and the
average number of vessels per country. We can see that the ecomonic crisis
has caused a significant drop in the last three indicators. For the number
of liner companies, we can see that there is a decreasing trend from 2005,
indicating a large number of merger and acquisition activities.(3)
4. Conclusions
The container shipping industry has enjoyed unprecedented growth in
the past 20 years. The economic crisis of 2007 to 2010 has caused
substantial contraction in world containerized trade and, at the mean-
time, unveiled several problems in container shipping industry, including
oversupply of tonnage, sudden drop in demand, and declining freight
rates.
The recent development of container shipping industry has shown the
following trends:
(1) The size of the largest container vessel and the average vessel size
are both increasing. Many carriers are investing in mega-containerships with
carrying capacity over 12,000 TEU. This trend causes ports to also investing
in advanced equipments to cater for the mega-vessels. Furthermore, in
order to achieve higher load factor for the mega-vessels, as well as to take
advantage of economies of scale, many carriers are adopting more of a hub-
and-spoke structure in their shipping network.
(2) Transshipment handling has become more and more significant
globally. The container transshipment volume has been growing steadily
to about 25% of total container handling activities. As a result, many ports
are investing in port infrastructure and productivity improvement projects
to attract more transshipment demand.
(3) Global container terminal operators are increasing their market
share. It is often believed that global operators have higher expertise and
better equipments and capital than smaller terminal operators.
(4) Liner companies are adopting more rigorous measures to reduce cost
and stabilize freight rate. These measures include slow steaming, rerouting,
and holding regular meeting with customers.
During the recovery after the crisis, the above trends will continue to
affect the development of the container shipping industry. There are both
challenges and opportunities for liner companies and terminal operators in
the post-crisis era.
References
1. Sánchez, Ricardo J. The Crisis: Long Run Perspective, Signals of Over-
Tonnage, and Crazy Pricing. Maritime Bulletin. 2009. Vol. 36.
2. The World Fact Book. Central Intelligence Agency Web site. [Online] June
24, 2010. [Cited: July 20, 2010.] https://www.cia.gov/library/publications/
the-world-factbook/geos/xx.html.
3. The UNCTAD Secretariat. Review of Maritime Transport. Geneva: s.n. 2009.
4. Field, Alan M. Uploading a Recovery. The Journal of Commerce Magazine.
Janurary 11, 2010.
5. JOC Staff. Analyst Forecasts 11 Percent Ocean Trade Growth. The Journal
of Commerce Online. [Online] April 13, 2010. [Cited: July 26, 2010.] http://
www.joc.com/maritime/analyst-forecasts-11-percent-ocean-trade-growth.
6. — Downturn Not Deep Enough, Says Shipping Executive. The Journal of
Commerce Online. [Online] April 13, 2010. [Cited: July 26, 2010.] http://
www.joc.com/maritime/Downturn-not-deep-enough-says-shipping-executive.
7. Drewry Shipping Consultants Ltd. Annual Review of Global Container
Terminal Operators. s.l.: Drewry Publishing, 2009.
8. Transport and Tourism Division in UNESCAP. Regional Shipping and Port
Development: Container Traffic Forecast 2007 Update. New York: United
Nations, 2007.

CHAPTER 3
SCENARIO ANALYSIS FOR HONG

KONG PORT DEVELOPMENT
UNDER CHANGING BUSINESS
ENVIRONMENT
Abraham Zhang∗ † and George Q. Huang
∗Departmentof Industrial and Manufacturing Systems Engineering,
The University of Hong Kong, Pokfulam Road, Hong Kong
†[email protected]
Hong Kong port (HKP) had been the world’s busiest container port during the
1990s and early 2000s. However, in recent years, its growth slowed down due
to rising competition from mainland ports. This paper identifies that potential
relocation of its key cargo source, processing trade enterprises, may also have
fundamental impacts on HKP development. Scenario analysis is conducted
to understand the relationships between business environment factors and
potential relocation trends using a mixed integer programming (MIP) model.
It suggests production operations are likely to move to Pan-PRD and lower-
cost areas of Guangdong if business environment does not deteriorate much
further. However, continual appreciation of Chinese currency RMB and further
reduction of value-added tax (VAT) rebate in China will make Asian lower-
cost countries more competitive. Very low oil prices will favor Inland of China.
Very high oil prices will cause global manufacturing move near major markets.
Processing trade relocation to western Guangdong will be slightly favorable to
HKP development, while other relocation scenarios will adversely affect HKP
development.
1. Introduction
Hong Kong port (HKP) had been the world’s busiest container port during
1990s and early 2000s. However, dramatic changes happened in recent years.
In 2005, Singapore overtook Hong Kong as the world’s busiest container
port. In 2007, Shanghai surpassed Hong Kong for the first time in terms
69
70 A. Zhang and G. Q. Huang
Table 1 Port statistics on container throughput (Mil-

lion TEU).
Port 2003 2008 AAGR(%)
Singapore 18.4 29.9 10.2

Shanghai 11.3 28.0 19.9
Hong Kong 20.4 24.5 3.7
Shenzhen 10.7 21.4 15.0
Busan 10.4 13.4 5.2
Dubai 5.2 11.8 18.0
Guangzhou 2.8 11.0 31.8
Ningbo-Zhoushan 2.8 10.8 31.3
Rotterdam 7.2 10.8 8.4
Qingdao 4.2 10.4 19.9
Source: Marine Department, Hong Kong SAR (2009).
of container throughout. In terms of annual average growth rate (AAGR),

Table 1 shows that HKP has had the lowest growth rate among the world’s
top 10 container ports from 2003 to 2008. If the trend continues, HKP will
soon be overtaken by Shenzhen port, becoming the world’s third busiest
container port. In addition, Guangzhou port, another neighboring port,
has also been on a fast ascending track. HKP is now under unprecedented
challenges to keep its role as a regional shipping hub.
The slow-down on HKP development has direct impacts on the
economy of the city. Ocean shipping between seaports has been the major
transportation mode in global trade. The port, thus, has been a key vehicle
for the operations of the trading and logistics industries in Hong Kong.
These two industries have been one of the four pillar industries of the
city, driving its economic development, creating employment positions and
providing growth opportunities for other industries. In terms of economic
contribution, trading and logistics services account for 25.8% of the GDP
in 2007. The number of employees in the two sectors was 842,200, or 24.2%
of total employment (Census and Statistics Department 2009).
The classical port development theory concludes that “port growth is a
function of the production outcomes of firms in the port’s adjacent space —
or of that space to which it is linked, either in landward space or in areas
linked across water or ocean” (Robinson, 1998). Indeed, HKP development
has been largely determined by the growth of processing trade industries in
the hinterland it serves, that is the Pearl River Delta (PRD) region adjacent
to Hong Kong. This region has grown to become the “World’s Factory” of
many labor-intensive products. The prosperity of processing trade in the
Scenario Analysis for Hong Kong Port Development 71
region boosted the HKP development. In 2008, over 70% of container traffic
handled in Hong Kong is related to the PRD region and its adjacent areas
(Hong Kong SAR Government 2009). Most of such traffic was generated by
the import and export of processing trade (GPRD Business Council 2007a).
Processing trade activities in the PRD region have become vital for HKP.
In recent years, an alarming phenomenon has emerged for HKP
development. In 2007 and 2008, thousands of factories have ceased their
operations in the PRD region due to the pressure of rising operating
cost. Local governments in the PRD region also announced industry
restructuring program to discourage processing trade operations in the
region (GPRD Business Council 2007b). It appears to be just a matter
of time that processing trade industries will relocate. If major relocation
destinations are far from Hong Kong, its port development may be
devastated by the loss of cargoes. Coupled with the rising competition from
Shenzhen and Guangzhou ports serving the same territory, HKP faces the
risk of being marginalized over long term. This new phenomenon of the
relocation of processing trade out of the PRD region has not been studied
thoroughly by researchers. There is also a lack of scholarly research on its
implications on HKP development. This paper intends to narrow the gap
by answering the following research questions:
What are the key business environment factors driving processing trade
to relocate out of the PRD region, which will in turn endanger HKP
development?
What are the potential relocation destinations of processing trade?
What further development in business environment may cause each
relocation trend dominant?
What are the implications of each relocation trend on HKP development?
This paper surveys government and industrial reports to answer the
first and second research questions. The third research question is addressed
by scenario analysis through a Mixed Integer Programming (MIP) model.
Sensitivity analysis is performed for key factors so as to understand their
impacts on relocation destinations. Insights on potential relocation trends
are drawn from a case study on a representative processing trade enterprise
(PTE) in the PRD region. Based on findings from scenario analysis,
the fourth research question is answered by discussing implications from
potential relocation trends.
The rest of this paper is organized as follows. Section 2 reviews
literature on HKP development and MIP models. Section 3 describes key
business environment factors in the PRD region and potential relocation
trends. Section 4 presents a MIP model. Section 5 explains experimental

design. Section 6 gives modeling results, conducts scenario analysis and
summarizes findings on HKP development. Section 7 concludes the paper
and identifies areas for further study.
2. Literature Review
Research studies on HKP development can be categorized by their per-
spectives. One of the key focuses is on port competition. An early work
among them was performed by Wong et al. (2001). A fuzzy number-based
distribution model was built to predict the cargo distribution among the
three main seaports in Southern China. Modeling results suggested that
HKP is most attractive and its position is less likely to be overtaken.
Cullinane et al. (2004) reviewed the recent rapid development of mainland
ports and analyzed its impact on the competitiveness of HKP. In particular,
competition from Shenzhen port is discussed in details. It was suggested
that HKP would retain its dominant role in the region.
However, most recent studies suggested a less optimistic future for
HKP. Yap et al. (2006) studied the port competition on hub status in East
Asia. Hong Kong, Busan and Kaohsiung ports have been dominant from
the 1980s. Competition intensified as Shanghai, Shenzhen and Qingdao
also emerged as hub candidates in recent years. Comparative studies were
conducted between traditional hub ports and emerging hub candidates.
Evidence shows that HKP has become less attractive for cargoes related
with China. Yap and Lam (2006) examined the competition dynamics
between the major container ports in East Asia. Results suggested that
Hong Kong and Pusan are beneficiaries from inter-port competition in the
region from the 1970 to 2001. However, port competition in the region would
intensify as the centre of gravity of cargo volume shifts to mainland China.
Mainland Chinese ports have observed rapid growth on cargo throughput
and they have already attracted many direct calls by shipping lines.
Several scholars made forecast on the throughput for HKP. Fung (2002)
and Hui et al. (2004) constructed error-correction models. Their forecast
results are different from those by government. Seabrooke et al. (2003)
forecasted the cargo growth for HKP through regression analysis. It was
suggested that cargo movement would continue to increase, though at
a slower pace. Neighboring ports will divert cargo from HKP, but the
continuous growth of the total cargo pool in Southern China would be more
influential. Lam et al. (2004) proposed neural network models to predict
cargo throughput. It was suggested neural network models are more reliable
and accurate than traditional regression analysis models.
Several research studies discussed factors affecting HKP development.
Loo and Hook (2002) examined the interaction among international,
national and local factors. It suggested that changing policies of Hong Kong
government have not been consistent with the need of a more integrated
port-inland distribution system. Fung et al. (2003) identified that the
separation of terminal handling charges (THC) from ocean freight rates
since 1991 adversely affected the throughput of HKP. Cheung et al. (2008)
used HKP for a case study on drayage services between a cargo terminal and
the origin or destination of cargo. Challenges and issues in the cross-border
logistics flows were highlighted. Results from an attribute-decision model
suggested that relaxing policy regulations could be very beneficial. All these
papers unanimously acknowledged that the manufacturing industries in the
PRD region are the key cargo source for HKP. However, none of them
addressed the impacts from potential relocation of manufacturing activities
as a result of changing business environment.
Location decisions of global manufacturing operations often employ
MIP techniques (Wilhelm et al., 2005). Several researchers reviewed
location decision models (Vidal and Goetschalckx, 1997; Pontrandolfo and
Okogbaa, 1999; Meixell and Gargeya, 2005; Melo et al., 2009). Several MIP
models discussed explicitly the impacts of changes in factors.
An early MIP model was formulated by Hodder and Jucker (1985).
Their model incorporated mean and variance of prices and exchange rates
to address uncertainties. Modeling results suggested that it is important to
address uncertainties in the international context.
Another early MIP model was developed by Cohen and Lee (1989) for
a personal computer manufacturer. Its objective function is to maximize
after-tax profit. It incorporated international factors including tariffs, duties
and transfer pricing. Sensitivity analysis was performed to understand
impacts of uncertainties in the foreign exchange rate and market demand.
Vidal and Goetschalckx (2000) analyzed impacts of uncertainties on
global supply chains through a MIP model and sensitivity analysis. The
paper reviewed mathematical modeling methods and suggested sensitivity
analysis as the best way to analyze system variations. Uncertainties in
four factors were addressed, including exchange rate fluctuation, demand
change, supplier reliability and raw material shipping lead time variance.
Modeling results suggested that uncertainties significantly affect optimal
global supply chain configurations.
Mohamed (1999), and Mohamed and Youssef (2004) built MIP models
to address impacts of exchange rates and initial capacity levels on produc-
tion, distribution and investment decisions and operating profit. Modeling
results were obtained for several experimental scenarios and comparative
analysis suggested that both factors have significant effect.
Wilhelm et al. (2005) built a MIP model for a hypothetical laptop
computer assembly operations in the Texas-Mexico environment under
North America Free Trade Agreement (NAFTA). Impacts of business
decisions, exchange rates and government investment incentives are assessed
through experimental cases and sensitivity analysis.
In general, the above MIP models concluded significant impacts of
various factors on global manufacturing location decisions. Sensitivity
analysis is effective and beneficial for the analysis. Unfortunately, these
models cannot be adapted easily for the relocation decisions of PTEs in
the PRD region. None of the above MIP models was for the business
environment of a developing country like China. Their relevant business
models were also distinctly different from processing trade, which is
dominant in the PRD region.
3. Changing Business Environment for HKP

3.1. Changing business environment in the PRD region
In the past five years, business environment has deteriorated rapidly for
PTEs in the PRD region. Market changes have driven up business operating
costs dramatically. First, Chinese currency RMB appreciated more than
20% against USD from 2005 to 2008. This appreciation has directly
impacted the cost competitiveness of products made in the PRD region.
Second, labor cost in the PRD region has been at sharp rise. From 2004 to
2008, the minimum wage standard was revised three times and the lowest
salary for production workers increased by about 70%. In early 2008, the
implementation of a new labor contract law pushed up labor cost further by
an average of 23.5%, as China government mandated better social security
and job security for employees. Third, oil price has been highly volatile
since 2004. In July 2008, oil price hit its historical high level of about
$150 per barrel, while that before 2003 was about only $30 per barrel.
High oil prices led to high transportation costs, which discourages the
intensive global trade activities of PTEs. Fourth, prices of many industrial
raw materials have soared in correlation with that of oil price, as they are
by-products of oil refinement, or incur high energy consumption during
production processes. Last, utility costs have gone up too, as they are also
affected by the rise of oil price (HKTDC 2008b).
Industrial policies have also changed to be in less favor of PTEs. At cen-
tral government level, China reduced export value-added tax (VAT) rebates
in both 2006 and 2007. It affected a very large category of labor-intensive
products. “Prohibited” category has expanded for processing trade. For
the production of these products, PTEs now have to make full payment of
tariffs and VAT for imported raw materials and parts. “Restricted” cate-
gory has also expanded. For products in the category, PTEs are subject to
the customs duty deposit system. Their cash flows are affected (HKTDC
2007a). At local government level, Guangdong province started its industrial
restructuring program as an attempt to upgrade its industries toward heavy,
chemical and high-tech industries. Further set-up of labor-intensive and
energy-intensive PTEs is discouraged in the PRD region (GPRD Business
Council 2006). With the implementation of more strict environmental pro-
tection standard, sewage treatment charges more than tripled from RMB
0.25 per tonne to RMB 0.8 per tonne (HKTDC 2007a).
Among the above changes, four factors have been most influential on
production costs and business profits. These four factors are RMB currency
exchange rate, labor cost in the PRD region, oil price and export VAT
rebate in China (HKTDC 2007a, HKTDC 2008b). As PTEs are now under
tremendous pressure to relocate, HKP development is at risk due to its
heavy dependence on cargoes from PTEs. If the traditional processing trade
industries relocate far away from the PRD region, it is less likely that they
will continue to use HKP for their import and export. On the other hand,
the port and logistics facilities of Hong Kong were primarily developed for
containers. They will not be able to provide the logistics support needed
by the heavy and chemical industries promoted in Guangdong. Coupled
with the more fierce competition from neighboring ports, HKP faces a very
challenging future (GPRD Business Council 2006).
3.2. Potential processing trade relocation trends

Relocation involves various business considerations. Low labor costs are
crucial for the production of labor-intensive products. Availability of skilled
labors is also important. Exchange rate risks should be minimized. Ade-
quate infrastructure is essential for manufacturing and logistics operations.
Conducive business environment and investment incentives are also valued
by global manufacturers. In addition, companies often locate within existing
industrial clusters where necessary raw materials and services are readily
available. Using footwear products serving the US market as an example,
this research identifies five popular relocation destinations:
Guangdong (Qingyuan, Guangdong, China): The location represents
lower cost areas in Guangdong province out of the PRD region (GPRD
Business Council 2007b). The PRD region is in the central part of Guang-
dong and consists of nine counties. They are Zhaoqing, Jiangmen, Zhuhai,
Foshan, Zhongshan, Guangzhou, Dongguan, Shenzhen and Huizhou. Their
locations can be seen in Fig. 1. The north, west and east parts of Guangdong
are less developed. Their labor costs are lower in comparison with that of
the PRD region. Most of these areas are within 3 hours driving distance
to the seaports in the PRD region. Relocation to these areas maintains the
connectivity with existing industrial clusters in the PRD region. Many local
governments offer tax and land incentives.
Pan-PRD (Ganzhou, Jiangxi, China): The location is within Pan-
PRD region and 450 km away from the PRD region. The location of
Jiangxi can be seen in Fig. 2. It represents lower cost areas in Pan-PRD
(GPRD Business Council 2007b, HKTDC 2007b, HKTDC 2008a). The local
government offer tax and land incentives.
Fig. 1 Map of Guangdong province of China.

Fig. 2 Map of Pan-PRD of China.
Inland (Bishan, Chongqing, China): The location is about 2,500 km

away from the east coast of mainland China. Its labor cost is among the
lowest in China. It represents other inland provinces of China (GPRD Busi-
ness Council 2007b). Several large footwear plants have been established in
the region to serve domestic market. Skilled labors and raw materials are
available.
India (Chennai, Tamil Nadu, India): The location has easy access to a
container port. It represents lower cost countries in Asia (HKTDC 2008a).
The location hosts one of the largest footwear clusters in the country. Its
local government offers very attractive investment incentives for export
oriented manufacturing operations.
Mexico (Leon, BC, Mexico): The location is about 600 km away
from the US border. It represents low cost countries which are near
major international markets. A footwear cluster exists in the region. The
location incurs lowest transportation costs and zero tariffs to serve the
US market. It is least vulnerable to risks from high oil prices (Rubin and
Tal, 2008).
PTEs’ choices of relocation destinations are expected to have funda-

mental impacts on HKP development as they serve as the key cargo source.
If PTEs relocate to Asian lower cost countries or near major markets, as
represented by India and Mexico respectively in the example, HKP will be
in a very unfavorable position. If PTEs relocate within China, their choices
among Guangdong, Pan-PRD and Inland will cause different implications on
HKP development. The routing patterns of cargo are influenced by several
key factors, including port connectivity, transportation modes, transport
distance, shipping cost and port competition. Generally speaking, better
connectivity, shorter transport distance and lower shipping cost attract more
cargo. In terms of transportation modes, barging incurs substantially lower
cost than land transportation modes, and thus is preferred as long as water-
ways are available. HKP possesses geographical advantage for cargoes from
its west via barging (Lin, 2008). If the Hong Kong-Zhuhai-Macau Bridge is
completed in the next few years as planned, HKP will extend its advantage
to cargoes from its west via trucking. Though the port of Gaolan in Zhuhai is
nearer to cargoes from the west side, it is far less established and may not be
able to develop into a competitive mainline port. However, for cargoes from
its north and east, the nearest seaports are well-established Guangzhou and
Shenzhen ports respectively. For both waterway and land transportation,
these two ports are in more convenient locations than HKP. In addition,
high THCs, high trucking cost and the low efficiency of border-crossing have
been causing cargo diversion from HKP to Guangzhou and Shenzhen ports
(Seabrooke et al., 2003; Cullinane et al., 2004).
Currently, it is unclear which relocation destination will attract most
PTEs. As different relocation trends cause varying effects on HKP develop-
ment, it is essential to understand which trend would become dominant and
which factor plays a most influential role. To answer this research question,
scenario analysis is conducted in the following three sections through MIP
modeling and sensitivity analysis.
4. A MIP Model
This research builds a MIP model (Zhang and Huang, 2009) to analyze
potential relocation trends. As PTEs in PRD competes mainly on cost
in international markets, the model assumes that PTEs stay or move to
the locations leading to lowest landed cost (LC) in markets. In general,
labor-intensive manufacturing in the PRD region is not capital-intensive,
and thus only major variable production costs are considered in the model.
Transportation costs are assumed to be in liner correlation with oil prices.

Raw materials are assumed to be equally priced and available at all
candidate locations. The PTE maintains a same net profit margin no matter
where it may relocates. The model is formulated by using the parameters
and variables as defined below. After presentation of the model formulation,
model components are described and discussed.
Parameters
M Markets {1, 2, . . . m, . . . M}
F Facilities {1, 2, . . . f, . . . F}
Dm Yearly demand for the market m supplied by the PTE
DFC f Base domestic freight cost per unit from facility f to port
IFC f m Base international freight cost per unit from facility f to
market m
RMC f Raw material cost per unit of production at facility f
CRC f Unit capacity retaining cost at facility f
RLC f Unit capacity relocation cost at facility f
DLC f Hourly direct labor cost at facility f
Hr f Man hour per unit of production at facility f
OP Oil price
OPB Base oil price
ef Exchange rate to USD from the currency used for facility f
Ctax f Corporate income tax rate at facility f
VAT f Realized VAT rate at facility f with consideration of VAT
rebate
NPAT f Net profit margin after tax at facility f
IDM f Days of inventory accountable to facility f
IDT f m Days of inventory in transit from facility f to market m
Tariff f m Tariff rate imposed by market m for the product imported
from facility f
φ Correlation factor between freight costs and oil price
µ Inventory holding cost coefficient
L A sufficiently large constant
Decision variables and convenience variables

πf Binary value for facility f (1 if open; 0 if not open)
qf m Yearly production quantity at facility f for market m
SP f Unit selling price from facility f at EXW term
MFGC f Unit manufacturing cost at facility f
DFCC f Domestic freight cost per unit from facility f to port

IFCC f m International freight cost per unit from facility f to market m
CTAXC f Corporate income tax per unit at facility f
VATC f VAT per unit at facility f
Objective function
Minimize:

qf m SPf (1 + IDTf m µ)
f m
(1)
+ qf m (DF CCf + IF CCf m + Tariff f m SPf m )
f m
Constraints
M F GCf = (CRCf + RLCf + DLCf Hrf )ef
(2)
+ RM Cf (1 + IDMf µ) ∀ f ∈ F
DF CCf = DF Cf (1 + φ(OP/OP B − 1))ef ∀f ∈ F (3)
IF CCf m = IF Cf m (1 + φ(OP/OP B − 1)) ∀ f ∈ F (4)
CT AXCf = (SPf − M F GCf − V AT Cf )Ctaxf ∀f ∈ F (5)
V AT Cf = (SPf − RM Cf )V ATf ∀f ∈ F (6)
SPf − M F GCf − V AT Cf − CT AXCf = N P ATf SPf ∀f ∈ F (7)

qf m = Dm ∀m ∈ M (8)
f

qf m ≥ πf ∀f ∈ F (9)
m

qf m ≤ πf L ∀f ∈ F (10)
m
qf m is an integer ∀ f ∈ F and m ∈ M (11)
πf ∈ {0, 1} ∀ f ∈ F (12)
All variables ≥ 0 (13)

The objective function (1) minimizes yearly LC in all international mar-

kets. LC components include selling prices of the manufacturer, freight costs
and tariff costs. Equations (2)–(6) are convenience constraints to simplify
model formulations while defining individual cost components, including
manufacturing cost, domestic freight cost, international freight cost, corpo-
rate income tax and VAT tax. Equation (7) is a constraint on net profit margin
after tax. Equation (8) is a demand constraint. Constraint (9) ensures that
no facility is open without any production. Constraint (10) is a binary force
constraint. Constraints (11)–(13) define the characteristics of variables.
5. Experimental Design
5.1. Experimental scenarios
The modeling work focuses on impacts of the four key factors as identified
in Sec. 3.1. These four factors are RMB currency exchange rate, labor cost
in the PRD region, oil price and export VAT rebate in China. Other factors
are assumed to be stable as in 2008 constant price.
An experimental scenario is defined as a unique combination of
parameter values to represent a certain business environment. Base scenario
represents the best prediction on the future business environment for the
planning horizon of 10 years. Table 2 shows parameter values at base
scenario, which closely represents business conditions in the PRD region
in early 2008. RMB to USD exchange rate is 0.143. Its labor cost is based
on that of 2008. Export VAT rebate in China is 13%. Average oil price is
projected to be at $75 per barrel. Scenarios for sensitivity analysis are also
defined in Table 2.
5.2. Experimental data

The MIP model is applied through a case study with a labor-intensive
PTE in the PRD region. The company produces low to middle-end sports
Table 2 Definition of experimental scenarios.
Scenarios for sensitivity analysis

(absolute values or in relation to
Factor Base scenario base scenario)
RMB to USD exchange rate 0.143 90%; 110%; 120%

Labor cost in the PRD region As in 2008 80%; 120%; 140%
Oil price ($ per barrel) 75 37.5; 150; 225; 300
Export VAT rebate in China 13% 8%; 11%; 15%; 17%
Table 3 Manufacturing cost data.
CRCf RLCf DLCf Hrf IDMf IDTf m RM Cf
PRD 6.85 0 12.77 0.8 35 28 $2.35

Guangdong 5.01 1.17 10.23 0.8 36 28 $2.35
Pan-PRD 4.81 1.33 9.81 0.8 37 28 $2.35
Inland 4.38 1.67 7.36 0.8 42 28 $2.35
India $0.87 $0.50 $0.85 0.9 42 35 $2.35
Mexico $1.19 $0.58 $2.17 1.0 35 7 $2.35
Table 4 Prevailing tax, tariff rates and freight costs.
Ctax f VAT f Tariff f m DFC f IFC f m
PRD 15.0% 4% 20.0% 0.288 $0.593

Guangdong 4.0% 4% 20.0% 0.552 $0.593
Pan-PRD 4.0% 4% 20.0% 1.000 $0.593
Inland 4.0% 4% 20.0% 4.000 $0.593
India 4.0% 0% 20.0% $0.049 $0.752
Mexico 5.8% 0% 0% $0 $0.208
shoes. Its business model and material flow represent most PTEs in the
region.
Tables 3 and 4 present various input data in relation to the US market.
The first rows give parameters corresponding to definitions in Sec. 4. Labor
costs at all locations are valid as in 2008. All transportation modes use 40FT
full containers or equivalent, which contains 4,800 units of finished products.
Data for costs in India and Mexico are given in the USD. Major data sources
are Hong Kong Trade Development Council (HKTDC), Investment Com-
mission of India, ProMexico, Werner International, Shenzhen Container
Trailer Association and four logistics service providers.
Demand of the US market for the manufacturer is 6,000,000 units per
year. Correlation factor between freight costs and oil price (φ) is 40.4% in
relation to the base oil price (Rubin and Tal, 2008). Inventory holding cost
coefficient (µ) is set as 25% per year to account for the cost of holding
inventory in transit or at manufacturing sites.
6. Results, Analysis and Findings

6.1. Modeling results at base scenario
As the numerical example considers only the US market, only a single
manufacturing facility is suggested to be in operation at all scenarios. Values
Table 5 Landed cost comparison at base scenario.
Manufacturing location Landed cost per unit Difference to PRD
PRD $7.06 —
Guangdong $6.54 −7.46%
Pan-PRD $6.53 −7.52%
Inland $6.58 −6.90%
India $6.68 −5.50%
Mexico $6.94 −1.70%
of decision variable qf m at all scenarios equal to the market demand. Table 5

shows modeling results of LC per unit of product at candidate locations at
base scenario. It suggests highest LC savings of 7.52% can be achieved by
relocation to Pan-PRD, followed by 7.46% for Guangdong and 6.90% for
Inland. Inland has the lowest labor cost in China, but its freight cost is
much higher than that of PRD and Guangdong. Relocation to India brings
LC savings by 5.50%, which is less competitive than the three candidate
locations in China. The advantage of lower labor cost in India is offset by
its higher logistics cost. It is surprising that relocation to Mexico brings LC
reduction of 1.70% though its labor cost is much higher than that of PRD.
This is mainly due to its much lower logistics cost to the US and zero tariff
benefit under NAFTA.
6.2. Sensitivity analysis

Figure 3 consists of four graphs to show the results of sensitivity analysis.
Graphs 3(a) and (b) employee comparative values of factors in relation to
base scenario, while graphs 3(c) and (d) use absolute values of factors.
The comparison of graphs (a), (b), (c) and (d) suggests that impacts
of RMB exchange rates and labor costs in the PRD region are most
significant. For the current manufacturing operation in the PRD region,
10% RMB appreciation against base scenario would raise LC from $7.06 to
$7.40. Labor cost increase by 20% would push up LC from $7.06 to $7.46.
In comparison, impacts of oil prices and VAT rebates in China are less
significant.
Impacts of factors on optimum relocation destinations could be dra-
matic. Graph 3(a) suggests that RMB appreciation of 10% against base
scenario would cause all three locations in China to lose advantage to India.
The cost advantage of manufacturing in China has become very slim after
RMB appreciation of over 20% since 2005. Further RMB appreciation would
quickly erode its cost competitiveness over lower cost countries in Asia.
$8.0
$7.5
Landed cost in U.S.
$7.0
$6.5
$6.0
0%
0%
0%
0%
0%
0%
50
25
00
%
%
%
%
5
8%
5
$7
7.
90
80
11
13
15
17
$1
$2
$3
10
11
12
10
12
14
$3
(a) RMB Exch (b) DLC in PRD (c) Oil price (d) VAT rebate in China
PRD Guangdong Pan-PRD Inland India Mexico
Fig. 3 Results of sensitivity analysis.
Graph 3(b) suggests that the labor cost in the PRD region has become too
high for the PTE to stay competitive. Even if its labor cost decreases to 80%
of the level as in base scenario, it would still not become most competitive.
Continual rise of labor cost in the PRD region would further undermine its
competitive position. Graph 3(c) shows that high oil prices adversely affect
manufacturing locations with higher logistics costs, for example, Inland
of China and India. Such effects from oil prices could dramatically alter
optimum relocation destinations. If oil price stays at low level of $37.5 per
barrel, Inland of China would be the optimum relocation destination due to
its lowest labor cost. At base oil price of $75 per barrel, Pan-PRD is most
cost competitive. At oil prices of $150 and $225 per barrel, Guangdong is
most cost effective due to its lower freight costs than Pan-PRD. If oil price
rises wildly to the very high level of $300 per barrel, Mexico would be the
best location to serve the US market. High oil prices amplify the logistics
advantage of production in Mexico. Graph 3(d) suggests that changes of
VAT rebates in China between 11% and 17% do not cause the shift of
optimum relocation destination. However, if VAT rebate decrease to be as
low as 8%, India will become most cost competitive.
In general, manufacturing in Pan-PRD and Guangdong leads to
minimal difference on LC at all scenarios. In comparison with base scenario
of manufacturing in the PRD region, relocation to either of these two
locations will bring about 20% labor cost savings and only incur slightly
higher freight costs. Besides lowest LC, they will bring least challenges on
supply chain lead time, language, culture difference, or availability of skilled
labors. It is likely that Pan-PRD and Guangdong will attract a large number
of relocating PTEs if business environment does not deviate far from base
scenario.
6.3. Scenario analysis and findings

Table 6 summarizes impacts of factors on the choice of primary relocation
destinations in relation to base scenario. Future scenarios only consider
changes in a single factor. In most scenarios, Guangdong and Pan-PRD
are the most preferred relocation destinations. Exceptions only happen in
four scenarios: if RMB currency appreciates more than 10% against base
scenario, Asian lower cost countries become more preferred; If oil price stays
at very low level of $37.5 per barrel or below, Inland of China become most
cost competitive; If oil price persists at very high level of $300 per barrel
or above, Mexico would become most favorable. If VAT rebate in China
Table 6 Scenario analysis on factors and relocation trends.
Primary relocation
Factor Future scenario destinations
RMB to USD exchange Depreciate Guangdong and

rate Pan-PRD
Stable Guangdong and
Pan-PRD
Appreciate by less than Guangdong and
10% Pan-PRD
Appreciate by more than Asian lower cost countries
10%
Labor cost in the PRD Drop slightly Guangdong and
region (2008 constant Pan-PRD
price)
Stable Guangdong and
Pan-PRD
Rise Guangdong and
Pan-PRD
Oil price per barrel (2008 $37.5 Inland
constant price)
$75; $150; $225 Guangdong and
Pan-PRD
$300 Near major markets
VAT rebate in China Between 11% and 17% Guangdong and
Pan-PRD
8% or below Asian lower cost countries
falls to 8% or below, Asian lower cost countries will gain advantage over
locations in China.
It should be noted that the combined effect of multiple factors may
cause the relocation dynamics more complicated. For example, both RMB
currency appreciation and VAT rebate decrease cause locations in China less
competitive than lower cost countries in Asia. The combined effect of these
two factors would cause changes in critical levels of factors which trigger the
shift of optimum locations. Taking VAT rebate as an example, its changes
between 11% and 17% do not alter optimum manufacturing location if other
factors stay as in base scenario. However, if RMB appreciates another 5%
against base scenario, changes in VAT rebates in the same range would
cause the shift of optimum manufacturing locations. Another example can
be seen on the change of critical oil price causing manufacturing moving
near major markets. If RMB appreciates another 10% against base scenario,
oil price $170 per barrel would cause Mexico more competitive than lower
cost countries in Asia.
Following the above analysis on relocation trends, Table 7 suggests
their implications on HKP development. Relocation of PTEs to other parts
of Guangdong province may bring mixed effect. If PTEs move to western
Guangdong, their cargoes are more likely to be captured by HKP. If PTEs
relocate to northern or eastern Guangdong, HKP will be less favored in
comparison with Guangzhou and Shenzhen ports respectively. If PTEs
move to Pan-PRD, HKP development will be in greater danger as land
transportation will become the dominant mode of cargo transportation.
Besides its disadvantage on cargo transported via trucking, HKP is lack
of railroad linkage. It will lose opportunities to Shenzhen and Guangzhou
ports when railroad becomes most economical for long distance land
transportation. Such an unfavorable position will be seen for HKP if PTEs
Table 7 Implications of relocation trends on HKP development.
Implications on HKP
Relocation destination Nearest mainline ports development
Guangdong West Hong Kong Slightly favorable

North Guangzhou Slightly unfavorable
East Shenzhen Slightly unfavorable
Pan-PRD Shenzhen, Guangzhou Modestly unfavorable
Other Inland areas of China Shanghai, Ningbo-Zhoushan, Most unfavorable
Qingdao, Tianjin
Lower cost countries in Asia Ports in or near the location Most unfavorable
Near major markets Ports in or near the location Most unfavorable
shift their operations to provinces which are far away from Hong Kong,
including west part of Pan-PRD and other inland provinces. If PTEs move
to lower cost countries in Asia, HKP will lose geographical proximity to
cargo source. At most, it will intercept some transshipment cargo from
nearby lower cost countries in Asia. Relocation to near major markets
would happen when oil price stays at very high levels, which will greatly
discourage global trade. HKP development, along with other ports in Asia,
would become very pessimistic in this scenario.
7. Conclusions and Future Work

This research studies impacts of business environment factors in the PRD
region on the port development of Hong Kong. Processing trade activities
in the PRD region have been the primary cargo source for HKP. Their
rapid expansion in the past two decades has boosted the role of Hong Kong
as a regional shipping hub. However, in recent years, business environment
for processing trade has deteriorated dramatically. Chinese currency RMB
appreciated over 20% since 2005. Minimum wage limits in PRD had moved
up by about 70% since 2004. The introduction of a new labor contract
law in China at the start of 2008 pushed up labor cost further. Oil price
has fluctuated wildly at various high levels since 2004, which leads to
high transportation costs in the global trade. China central government
reduced export VAT rebates for a very large category of processing trade
products in both 2006 and 2007. Local Guangdong government also started
industrial restructuring program to discourage processing trade activities in
the PRD region. All these factors are pushing processing trade activities to
relocate. If relocation destinations are far from Hong Kong, its port would
be marginalized over long term.
This paper contributes to the study on HKP development. First, it
points out that potential relocation of processing trade activities out of the
PRD region may have fundamental impacts on HKP development. Relevant
key factors and potential relocation trends are also identified. Second, a MIP
model is constructed for scenario analysis. The relationships between factors
and potential relocation trends are established through sensitivity analysis.
PTEs are likely to move to Pan-PRD and other parts of Guangdong if
business environment does not deteriorate much further from that of early
2008. Lower cost countries in Asia may become more competitive if RMB
appreciates another 10% from base exchange rate of 0.143, or VAT rebate
in China falls to as low as 8% or below. If oil price stays at very low level of
$37.5 per barrel or below, Inland of China may become most cost effective.
If oil price persists at very high level of $300 or above, locations near
major markets will become most preferred. Last, implications of potential
relocation trends on HKP development are discussed. Relocation of PTEs to
western Guangdong will be slightly favorable. Relocation to northern and
eastern Guangdong will be slightly unfavorable. Relocation to Pan-PRD
will be modestly unfavorable. Most unfavorable trends are the relocations
of PTEs to other inland provinces of China, lower cost countries in Asia
and near major markets.
The research is limited in several areas. First, the modeling work uses
a footwear PTE for case study. Modeling results are expected to be valid in
general as footwear is a typical labor-intensive processing trade product in
the PRD region. However, due to differences on cost structures and freight-
intensiveness, other products may have different critical levels in factors
to trigger various relocation trends. Second, the MIP model focuses on
the effect of four key factors. A real life manufacturing relocation decision
usually involves more factors. Third, potential relocation trends are subject
to uncertainties. It is not impossible that unexpected factors would alter
the potential relocation trends. Last, implications on the HKP development
are based on the current situation of the port and neighboring mainland
ports. The changing port competition dynamics in the region may lead to
different prospects.
Future work on the subject is to derive policy responses for the HKP
development. This will be of practical benefits to the port and the economy
of Hong Kong. As the logistics industry of Hong Kong is twinned together
with the trading industry, further study on trading industry of Hong Kong
will be required. Port competition from mainland ports has diverted a
large portion of cargo from the HKP. Future work will include a detailed
comparative study on the advantages and disadvantages of ports in the
territory.
References
1. Census and Statistics Department (2009). The Situation of the Four Key
Industries in the Hong Kong Economy in 2007. Hong Kong Monthly Digest of
Statistics. Hong Kong: Census and Statistics Department, The Government
of Hong Kong SAR, FB0-FB16.
2. Cheung, R., Shi, N., Powell, W. and Simao, H. (2008). An attribute-decision
model for cross-border drayage problem, Transportation Research. Part E,
Logistics & Transportation Review, 44(2), pp. 217–234.
3. Cohen, M.A. and Lee, H.L. (1989). Resource deployment analysis of global
manufacturing and distribution networks, Journal of Manufacturing and
Operations Management, 2, pp. 81–104.
4. Cullinane, K., Fei, W.T. and Cullinane, S. (2004). Container terminal
development in Mainland China and its impact on the competitiveness of
the port of Hong Kong, Transport Reviews, 24(1), pp. 33–56.
5. Fung, M.K. (2002). Forecasting Hong Kong’s container throughput: an error-
correction model, Journal of Forecasting, 21(1), pp. 69–80.
6. Fung, M.K., Cheng, L.K. and Qiu, L.D. (2003). The impact of terminal han-
dling charges on overall shipping charges: an empirical study, Transportation
Research Part A — Policy and Practice, 37(8), pp. 703–716.
7. GPRD Business Council (2006). Report on Guangdong’s industrial restruc-
turing — opportunities and challenges for Hong Kong. Hong Kong: The
Greater Pearl River Delta Business Council.
8. GPRD Business Council (2007a). The Greater Pearl River Delta Business
Council 2006/2007 Annual Report. Hong Kong: The Greater Pearl River
Delta Business Council.
9. GPRD Business Council (2007b). Implications of Mainland Processing Trade
Policy on Hong Kong. Hong Kong: The Greater Pearl River Delta Business
Council.
10. HKTDC (2007a). Cost escalation and trends for export price increase —
A look at the rising production costs in the PRD. Hong Kong: Hong Kong
Trade Development Council.
11. HKTDC (2007b). Relocating processing trade from PRD — An assessment
of alternative destinations on the mainland. Hong Kong: Hong Kong Trade
Development Council.
12. HKTDC (2008a). Latest development and strategies of Hong Kong compa-
nies’ processing trade business in the Pearl River Delta. Hong Kong: Hong
Kong Trade Development Council.
13. HKTDC (2008b). Upward pressure on export prices mounting but com-
petitiveness maintained. Hong Kong: Hong Kong Trade Development
Council.
14. Hodder, J.E. and Jucker, J.V. (1985). International plant location under
price and exchange rate uncertainty. Engineering Costs and Production
Economics, 9(1–3), pp. 225–229.
15. Hong Kong SAR Government (2009). Transport. Hong Kong Yearbook 2008.
Hong Kong: Hong Kong SAR Government.
16. Hui, E.C.M., Seabrooke, W. and Wong, G.K.C. (2004). Forecasting
cargo throughput for the port of Hong Kong: error correction model
approach, Journal of Urban Planning and Development-ASCE, 130(4),
pp. 195–203.
17. Lam, W.H.K., Ng, P.L.P., Seabrooke, W. and Hui, E.C.M. (2004). Forecasts
and reliability analysis of port cargo throughput in Hong Kong, Journal of
Urban Planning and Development, 130(3), pp. 133–144.
18. Lin, W. (2008). Towards the upliftment of the Hong Kong port. Shippers
Today. Hong Kong.
19. Loo, B.P.Y. and Hook, B. (2002). Interplay of international, national and
local factors in shaping container port development: a case study of Hong
Kong, Transport Reviews, 22(2), pp. 219–245.
20. Marine Department (2009). Ranking of Container Ports of the World. Hong
Kong: The Government of the Hong Kong SAR.
21. Meixell, M.J. and Gargeya, V.B. (2005). Global supply chain design: a litera-
ture review and critique, Transportation Research Part E, 41, pp. 531–550.
22. Melo, M.T., Nickel, S. and Saldanha-da-Gama, F. (2009). Facility location
and supply chain management — A review, European Journal of Operational
Research, 196(2), pp. 401–412.
23. Mohamed, Z.M. (1999). An integrated production-distribution model for a
multinational company operating under varying exchange rates, Interna-
tional Journal of Production Economics, 58(1), pp. 81–92.
24. Mohamed, Z.M. and Youssef, M.A. (2004). A production, distribution and
investment model for a multinational company, Journal of Manufacturing
Technology Management, 15(6), pp. 495–510.
25. Pontrandolfo, P. and Okogbaa, O.G. (1999). Global manufacturing: a review
and a framework for planning in a global corporation, International Journal
of Production Research, 37(1), pp. 1–19.
26. Robinson, R. (1998). Asian hub/feeder nets: the dynamics of restructuring,
27. Rubin, J. and Tal, B. (2008). Will Soaring Transport Costs Reverse
Globalization? Toronto: CIBC World Markets Inc.
28. Seabrooke, W., Hui, E.C.M., Lam, W.H.K. and Wong, G.K.C. (2003).
Forecasting cargo growth and regional role of the port of Hong Kong. Cities,
20(1), pp. 51–64.
29. Vidal, C.J. and Goetschalckx, M. (1997). Strategic production-distribution
models: a critical review with emphasis on global supply chain models,
European Journal of Operational Research, 98, pp. 1–18.
30. Vidal, C.J. and Goetschalckx, M. (2000). Modeling the effect of uncertainties
on global logistics systems, Journal of Business Logistics, 21(1), pp. 95–120.
31. Wilhelm, W., Liang, D., Rao, B., Warrier, D., Zhu, X. and Bulusu, S. (2005).
Design of international assembly systems and their supply chains under
NAFTA. Transportation Research Part E, 41, pp. 467–493.
32. Wong, W.G., Han, B.M., Ferreira, L. and Zhu, X.N. (2001). Factors
influencing container transport: a fuzzy number-based distribution model
approach, Transportation Planning and Technology, 24(3), pp. 171–183.
33. Yap, W.Y. and Lam, J.S.L. (2006). Competition dynamics between container
ports in East Asia, Transportation Research Part A — Policy and Practice,
40(1), pp. 35–51.
34. Yap, W.Y., Lam, J.S.L. and Notteboom, T. (2006). Developments in container
port competition in East Asia, Transport Reviews, 26(2), pp. 167–188.
35. Zhang, A. and Huang, G.Q. (2009). A Mixed Integer Programming Model
to Evaluate the Impact of Business Factors on Global Manufacturing
Relocation Decisions. Proceedings of 39th International Conference on
Computers & Industrial Engineering. Troyes, France, pp. 390–395.
CHAPTER 4
MODELS FOR PORT

COMPETITIVE ANALYSIS
IN THE ASIA-PACIFIC REGION
Chew Ek Peng∗ , Lee Loo Hay, Jiang Jianlin and Gan Chee Chun
Department of Industrial & Systems Engineering
National University of Singapore, Singapore, 119260
∗ [email protected]
The trend of increasing globalization of the world’s economies and the relentless
development of international trade links has resulted in a great boom in many
transportation industries, notwithstanding the recent economic downturn. Sea
container traffic accounts for the majority of international trade due to its
cost-effectiveness and as such maritime transportation is vital for the economic
health of many nations. The container transportation industry has therefore
experienced great growth, resulting in intense competition among international
container ports. In such a competitive environment, port benchmarking has
become particularly important for individual ports as it can help them
recognize potential threats and opportunities as well as reveal their strengths
and weaknesses in the face of the ever-changing competitive landscape.
This paper focuses on port benchmarking from the following three
perspectives: port efficiency, port connectivity, and the impact of various
factors on individual ports. First we present a brief review on the study of
port benchmarking in recent years. Next, we examine the aforementioned
three perspectives by presenting a model and a case study for each model.
Efficiency is an important factor in port competitiveness, and by using Data
Envelopment Analysis methodology many port attributes can be evaluated in
terms of their impact on port performance. Port connectivity indicates how well
ports connect to others in the actual maritime transportation network and is
vital for the competitiveness of transshipment services at a port. However, to
date this concept has not yet been clearly defined. A port connectivity analysis
framework is proposed in this paper and under this framework this concept
can be characterized in terms of the impact of transshipment operations at
a port on the overall network. Lastly, many factors can influence a port’s
performance and competitiveness and a given factor can play different roles
for different ports. The complex interplay of factors in port systems can make
91
92 E. P. Chew et al.
it difficult to study the impact of individual factors on the port. A network flow
model is presented to analyze the impact of different factors on different ports.
As Asia-Pacific is the pre-eminent region in terms of world container through-
put, this paper focuses on the benchmarking analysis of major ports in the Asia-
Pacific region in order to provide some insights into the performance of these
ports.
1. Introduction
In recent years, the maritime transportation industry has seen great
growth in many areas due to the globalization of trade, growth in sea
transport, development of business logistics and specialization in production
(Cuadrado et al., 2004). As the major gateways for international trade,
maritime ports are vital components of the maritime transportation
network. Accordingly, their importance as economic drivers in the region
has also led many countries to invest large sums of money in order to lower
operational costs or improve service quality. More and more new ports have
also been developed to tap into the global maritime traffic. As a result, the
competition among maritime ports has also been rapidly increasing. This is
of great concern for individual ports, as their competitive ranking can have
a very large effect on their future well-being in such a fierce competitive
environment. A decline in traffic can potentially lead to irreparable long-
term repercussions. If liner companies shift services away from a port due to
poor connections, high costs or poor handling quality, the loss of connections
at the port may incite additional liner companies to shift away as well. This
can lead to a slippery slope where ports can find it very difficult to recover
their competitive advantage once it is lost.
In such an environment, it becomes vitally important for ports to
conduct benchmarking studies to constantly evaluate their position within
the competitive landscape. According to Bemowski (1991), benchmarking
is “the measurement of a company’s performance in comparison to the
best, determining how those companies achieve superior performance and
using that information as the basis to decide on and implement objectives
and strategies”. By examining the strengths and weaknesses of themselves
and their competitors, ports can quickly detect any facet of operations at
which they are lagging behind. This enables port managers to quickly make
decisions and put in place measures to arrest the slide before too much
damage is done. In this chapter, port benchmarking is used to analyze
port competitiveness from three different perspectives: port efficiency, port
connectivity and the impact of different factors on individual ports.
Models for Port Competitive Analysis 93
2. Literature Review
Numerous studies have been done on port competitiveness, some of which
have involved benchmarking in the port sector, though not necessarily
looking at only the container terminals. Various ways of evaluating or
comparing ports and even container terminals have been applied. For
instance, Porter’s diamond has been used to identify the factors that
determine the competitiveness of the ports (Rugman, 1993) and Strategic
Positioning Analysis (SPA) which focuses on the market share and growth
within a time period has also been proposed and applied (Huybrechts,
2002). Kleywegt et al., (2002) analyzed the competition between the ports
of Singapore and Malaysia. The paper compares various attributes of the
two ports, considers the economic factors behind the transfer of Maersk
Sealand and provokes some thinking on strategies the ports should adopt
in the future. Using the benchmarking analysis, Pardali and Michalopoulos
(2008) give a score for each attribute considered based on how good the
ports are in that attribute and identify the best port as being the one
which has the highest number of maximum scores. The port competitiveness
degree (PCD) is then considered to further test the result. The ranking of all
other ports is obtained through comparison to the best port in the model.
According to the authors, this methodology has the advantage of taking
into consideration an unlimited number of factors.
The factors which can influence a port’s efficiency are complex and
the correlation between those different factors often cannot be determined
exactly. Nevertheless, the Data Envelopment Analysis (DEA) methodology
is an ideal tool to deal with such problems due to its attributes. In fact,
in the last decade, there has been a continuous and steady increase in
the use of DEA for port benchmarking in terms of the efficiency of
individual ports. The reason for such an increase is that the methodological
and computational benefits of DEA make it rather suitable for efficiency
measurement in the complicated port environment. The work of Roll and
Hayuth, (1993) was the first attempt to use DEA to measure port efficiency
and thus it can be regarded as a milestone of DEA application to seaport
efficiency measurement, although it was purely theoretical and did not
use any actual data. Poitras et al., (1996) analyzed the relative efficiency
of 23 Australian and international ports with both the CCR model and
the BCC model. Martinez-Budria et al., (1999) used DEA to measure the
efficiency of 26 Spanish ports. Tongzon, (2001) applied DEA to measure the
relative efficiency of some Australian and other international ports. Both
constant and variable returns to scale models (CCR and ADD) were chosen
to analyze the selected ports and the author concluded that the CCR model
would result in more inefficient ports than the ADD model, and that the
number of the selected ports should not be too small. Valentine and Gray,
(2001) adopted DEA to analyze the relationship between port ownership
structure and efficiency. Barros, (2003) used DEA to analyse 11 ports of
Portugal using panel data between 1990 and 2000. The author concluded
that the most important factors of a port are its dimensions and location
while capital intensity and private ownership offered little significant impact
to port efficiency. Park and De, (2004) presented a four-stage DEA model
to measure port efficiency, in which the overall efficiency is divided into
four stages, namely productivity, profitability, marketability and overall
efficiency. Cullinane et al. (2006) compared two methods for port efficiency
measurement: DEA and SFA. This research finds that the two methods
would yield similar efficiency scores in terms of the ranking of the ports.
Rather recently, Sharma and Yu, (2009) combined data mining and DEA
to present a diagnostic tool to measure the efficient terminals and prescribe
a step-wise projection to reach the frontier according to their maximum
capacity and similar input properties.
Port connectivity has also been examined from a few aspects dur-
ing the last several years. Hoffmann, (2005) combined 10 indicators of
maritime transportation, including fleet assignment, liner services, and
vessel and fleet sizes and so on, to generate an overall liner shipping
connectivity indicator for each country. McCalla et al., (2005) measured the
connectivity for Caribbean shipping networks. Notteboom, (2006) discussed
the time factor in liner services which reflects one port’s connectivity.
Marquez-Ramos et al., (2006) used principle component analysis (PCA)
methodology to build three complex connectivity component variables and
analyze the determinants of maritime transport costs of Spanish exports
and their effect on international trade flows. Wilmsmeier et al., (2006)
investigated maritime trade among 16 Latin-American countries and their
findings revealed that inter-port connectivity has a significant impact on
international maritime transport costs. Wilmsmeier and Hoffmann, (2008)
analyzed the impacts of liner shipping connectivity on intra-Caribbean
freight rates and the relationships between the structure of liner services,
port infrastructure and liner shipping freight rates. Tang et al., (2008) and
Low et al., (2009) proposed a direct measure of port connectivity based on
the number of origin and destination pairs served by individual ports in
real transportation networks.
Port selection criteria of carriers and shippers have been studied for
several decades and the impacts of various factors on ports in different
regions have been investigated in depth. Slack, (1985) analyzed the
containerized traffic between the North American Mid-West and Western
Europe, from which the author concluded that the price and service are the
most important factors. Murphy et al., (1992) developed a framework to
analyze port selection factors for different maritime players. Tiwari et al.,
(2003) used a discrete choice model to simulate port choice behavior from
which they find that the distance to destination, the distance from origin,
port congestion, and shipping line’s fleet size play an important role in port
selection. Malchow and Kanafani, (2004) present an alternative form of the
discrete choice model to analyze the distribution of maritime shipments
among US ports and found that the most significant factor of a port is its
location. Recently, Tongzon and Heng, (2005) summarized eight key deter-
minants of port competitiveness from the existing literature, namely port
operation efficiency level, cargo handling charges, reliability, port selection
preferences of carriers and shippers, the depth of the navigation channel,
adaptability to the changing market environment, landside accessibility,
and product differentiation. Analytic Hierarchy Process (AHP) is a multi-
objective, multi-criteria theory of measurement created by Saaty (1977)
and it has already been employed to determine the predominant factors
in port selection decisions. Lirn et al., (2004) use AHP to analyze and
reveal important service factors for transshipment port selection by global
carriers. Song and Yeo, (2004) employ AHP to identify competitiveness of
container ports in China and provide managerial and strategic implications.
By using AHP, Ugboma et al., (2006) determine the factors that carriers
and shippers consider important. Their findings show that the carriers and
shippers place high emphasis on efficiency, frequency of ship visits and
adequate infrastructure.
3. Port Benchmarking Models

Many effective port benchmarking methodologies exist to evaluate different
aspects of port operations. This chapter focuses on evaluating ports from
the perspective of port efficiency, port connectivity and the impact of factors
on port performance. Accordingly, a brief overview of 3 separate models
is presented. Port efficiency will be evaluated using a Data Envelopment
Analysis (DEA) model. A framework is then proposed for the analysis of
port connectivity from a network perspective. Lastly, a network flow model
based on the work of Lee et al. (2006) is used to evaluate the impacts of
various factors on different ports. To illustrate these methodologies, a case
study is performed on various ports in the Asia Pacific region.
3.1. Port efficiency

Maritime transportation is vital to the health of a country’s economy, as
more than 85% of international trade is transported through seaports (Liu,
2008). Coelli, (1996) indicates that there are mainly two methods used to
measure efficiency: the stochastic frontiers approach (SFA) and the Data
Envelopment Analysis (DEA) approach. SFA utilizes regression analysis to
determine the inefficiency values and the factors which have an impact on
efficiency. This unfortunately requires the functional form of the regression
equation to be known in advance, which can result in biases introduced by
subjective recognition.
The DEA methodology measures the relative efficiency of a set of
homogeneous entities called Decision Making Units (DMUs). DEA can
account for the impact of numerous factors and does not require prior
assumptions such as the relationship between inputs and outputs, or the
weightings for each factor. As DEA is not limited to the operating types of
the objects being evaluated, it can be used in a wide variety of fields. The
DEA methodology is based on generating a frontier which is composed
of the best DMUs, meaning those that have the greatest output/input
ratio. By measuring the distance of each DMU from the efficient frontier,
an efficiency value between zero and one can be assigned. DEA can also
identify a set of weights that maximizes the efficiency score of a DMU
while keeping the efficiency scores of other DMUs less than or equal to
1 using the aforementioned weights. However, care must be taken in the
selection of the appropriate inputs and outputs as the DEA methodology
does not assign weights based on the actual importance of the attributes
to the final performance. DMUs can be classified as being efficient (weakly
or strongly) if their efficiency score is 1. Otherwise, they are classified as
being inefficient.
The last decade has seen a steady increase in the DEA’s popularity as
a measure of port efficiency. More details about the application of DEA
methodology in port economics can be referred to in the work of Panayides
et al., (2009). In this section we will focus on the performance of the
container terminals of the major Asian ports and DEA will be used to
benchmark the major ports in the Asia Pacific region. Furthermore, the
key attributes that contribute to the success of the ports will be identified.
The model is constructed using n DM Uo (o = 1, . . . n) with m inputs

and s outputs. xo ∈ R+
m
and yo ∈ R+s
represent the input and output vectors
of DM Uo respectively, while X = (xo ) ∈ R+ m×n
and Y = (yo ) ∈ R+ s×n
denote the input and output matrices. Based on X and Y , the DEA model
can be given by the following linear programming problems
Min θ
s.t. Xλ ≤ θxo
Y λ ≥ yo
(1)
(eT λ = 1)
λ ≥ 0,
Max e T s− + e T s+
s.t. s− = θ∗ xo − Xλ
s+ = Y λ − y o
(2)
(eT λ = 1)
s− ≥ 0, s+ ≥ 0, λ ≥ 0.
where θ and λ are the decision variables among which the former is a
scalar while the other is a column vector (λ1 , . . . , λn )T . e is a column vector
in Rn with all elements equal to 1. The optimal solution of (1), denoted
by θ∗ , is employed as the technical efficient score of DM Uo , which is used
in (2).
If DM Uo satisfies (i) θ∗ = 1; (ii) s− ≥ 0 and s+ ≥ 0, then it is called
fully efficient; otherwise, if DM Uo only satisfies (i) then it is called weakly
efficient. In (1) and (2) (eT λ = 1) means this constraint can either be
included or excluded, which corresponds to different DEA models. If it is
excluded then (1) is called the CCR model which stipulates variable returns
∗
to scale and the units satisfying (i) and (ii) are BCC-efficient. Let θCCR
∗
and θBCC be the CCR score and BCC score of a DM Uo , respectively. Then
the scale efficiency is derived by
∗
θCCR
SE = ∗ , (3)
θBCC
∗ ∗
SE is not greater than one since θCCR is always not greater than θBCC .
SE = 1 indicates scale efficiency and SE < 1 indicates scale inefficiency.
Both the CCR and BCC models given by (1) are input-oriented which
means they minimize inputs while satisfying the given output levels. There
is another type of CCR and BCC model, called output-oriented, that
attempts to maximize outputs while using no more than the given input
levels. Besides the CCR and BCC models, there are several other classical
DEA models for efficiency measurement, such as the additive model, slacks-
based measure model, Russell measure model and so on. As we are only
focused on using DEA models in measuring the efficiency of major ports
in the Asia-Pacific region, we will not delve into the details of these other
DEA models.
3.1.1. Case study

The model detailed in the previous section was applied to 9 major ports
in Asia: Singapore, Shanghai, Hong Kong, Shenzhen, Busan, Kaohsiung,
Qingdao, Ningbo, and Tanjung Pelepas. The listed ports were chosen as
they are some of the largest and fastest growing ports in the world, and
thus benchmarking on them should be representative and meaningful. The
port data used was obtained from CI-Online, 2008.
Several scenarios were constructed in order to analyze different aspects
of port operations. All the scenarios have the same input variables (Area of
Container Terminals, Quay Length, No. of Quay Cranes, Storage Capacity).
However, their output variables are different combinations of Derived
Revenue, Port Calls/week and throughput, as shown in Table 1.
The input-oriented CCR DEA model is chosen as our benchmarking
analysis tool and the efficiency scores of the nine major ports in different
scenarios are shown in Table 2:
By comparing the results of Scenario 1 and Scenario 2, we can observe
the effect of excluding Derived Revenue as an input on the resulting
efficiency scores. The ports of Shenzhen and Tanjung Pelepas suffer a
decrease in their efficiency score, which shows that they are more efficient
at bringing in revenue with more throughput or higher cost. In effect, they
are able to have higher throughput or higher cost or both with the same
input or they are able to have the same throughput and cost with less
input. As for Busan and Qingdao, the exclusion of Derived Revenue does
not have an impact on their inefficient scores. This probably indicates that
Table 1 The attributes used in our study.
Scenario 1 Scenario 2 Scenario 3 Scenario 4

√ √
Output Variables Derived Revenue
√ √
Port Calls/week
√
Throughput
their efficiency score in Derived Revenue is exactly the same as their overall
score. The alternative is that the attribute does not have any impact on
their overall score, which is highly unlikely.
Comparing the results of Scenario 1 and Scenario 3 provides insight
on the contributions of the attribute Port Calls/week. It is found that
the efficiency scores of 5 ports (Busan, Kaohsiung, Qingdao, Ningbo and
Tanjung Pelepas) drops because of the exclusion of the attribute Port
Calls/week, which shows that they are doing a better job in bringing in
port calls, i.e., they are able to have higher port calls with the same input
or they are able to have the same port calls with less input. The efficiency
score of the port of Kaohsiung experiences the biggest drop, 86.57%, which
means Kaohsiung is doing the best in bringing in port calls. The next are
Pusan, Ningbo, Qingdao and Tanjung Pelepas.
The attribute of Throughput is often chosen as an output in many DEA
studies, and is sometimes the only output considered. Thus it is interesting
to compare the difference between Derived Revenue and Throughput, which
is carried out by the comparison of Scenario 3 (Derived Revenue is the
product of Throughput and Normalized handling cost) and Scenario 4. Ports
of Singapore, Shanghai, Hong Kong and Shenzhen are efficient in both
scenarios, which shows that they excel at bringing in throughput. Inefficient
ports in Scenario 3 remain inefficient in Scenario 4, but they all experience
an increase in their efficiency scores, which shows they are doing worse in
terms of handling cost when compared with the efficient ports. Kaohsiung
has the greatest increase, followed by Busan, Tanjung Pelepas, Ningbo and
Qingdao.
According to Table 2, it is clear that Busan, Kaohsiung, Qingdao
and Ningbo are more efficient in bringing in port calls while Shenzhen
and Tanjung Pelepas are better in bringing in revenue. The three ports
Table 2 The efficiency score of different scenarios.
Port Scenario 1 score Scenario 2 score Scenario 3 score Scenario 4 score
Singapore 1 1 1 1
Shanghai 1 1 1 1
Hong Kong 1 1 1 1
Shenzhen 1 0.532664 1 1
Busan 0.917816 0.917816 0.352235 0.644344
Kaohsiung 1 1 0.134321 0.84481
Qingdao 0.638416 0.638414 0.545662 0.596209
Ningbo 1 1 0.681831 0.959301
Tanjung Pelepas 0.398707 0.373839 0.395433 0.666486
Table 3 The efficiency score for

turnaround time.
Port Scenario 1 score
Singapore 1
Shanghai 0.638757
Hong Kong 1
Busan 0.51116
Tanjung Pelepas 0.35266
Singapore, Shanghai and Hong Kong remain strongly efficient in all the 4
scenarios. This suggests that the 3 ports are likely to be the most efficient
ones among all the ports studied.
A similar analysis can also be performed using Turnaround Time as
the output, with the inputs normalized using throughput handled in order
to remove the bias against large ports. Due to unavailability of data, the
analysis could only be performed on 5 ports, namely Singapore, Shanghai,
Hong Kong, Busan and Tanjung Pelepas.
The results for Turnaround Time (Table 3) show that Singapore and
Hong Kong are the most efficient, while Shanghai is in 3rd place. This
reinforces the earlier findings that Singapore, Hong Kong and Shanghai are
the most efficient out of those ports included in the study.
3.2. Port connectivity

Transshipment operations have rapidly grown in importance in the past
few decades. According to a report from Drewry Shipping Consultants,
transshipment volume as a proportion of total volume handled has grown
from 18% in 1990 to 28% in 2008. As such, it is no surprise that any
thorough study of port competitiveness should include their transshipment
capability. A major factor influencing transshipment capability is the
concept of port connectivity. Intuitively, port connectivity indicates how
well one port connects to others in a maritime transportation network
and its ease of accessibility by liner services. A port with a high level
of connectivity is likely to have a great advantage for transshipment
operations, so much so that connectivity could be used as a direct proxy
for a port’s competitiveness in terms of transshipment. In general, the
higher connectivity level a port has, the more attractive it will be to
liner companies in the sense of facilitating the transportation of cargo
and reducing transportation cost and time, leading to the port being more
competitive than its peers.
However, it is not immediately evident how to determine a port’s

connectivity as it is a concept that can have different interpretations and
meanings in different cases. To date, this concept of port connectivity has
not been well defined despite several papers on the topic. In this section,
we propose a connectivity analysis framework from a network perspective
that can be used to benchmark the competitiveness of various ports in the
network in terms of the impact of their transshipment operations.
At first glance, a possible way to measure the connectivity of a port
is to simply count the number of direct connections it has to other
ports in the network, i.e., the number of incoming and outgoing shipping
services to/from the port in question. Despite providing a reasonable
estimate, this simplistic method can leave out some important details.
Three such important factors are discussed as follows for an origin port
A and destination port B:
Responsiveness: The average waiting time that a supplier has to wait
for a service to ship his goods from port A to port B. This factor is related
to the frequency of shipping between port A and port B. The lower the
frequency, the longer the expected waiting time for the supplier. Vice versa,
the higher the frequency, the more trips from A to B is made per time period
and thus the shorter the expected waiting time.
Capacity: Even if there are frequent services between port A and port
B, it could be that the ships serving port A and port B have low capacities,
which limit the amount of cargo that the supplier can ship at any time.
On the other hand, if services between port A and port B have low frequen-
cies but high capacities, a large amount of cargo can be shipped in a single
shipment even though the average waiting time might be long. Therefore,
capacity should also be considered as a factor in port connectivity.
Network structure: Besides direct links, there may also be services
connecting ports A and B that comprise of multiple stops. In this case, the
structure of the transportation network, such as the existence of bottlenecks
or the ease of accessibility to large hubs, can also play a very large part in the
viability of transshipment services and hence affect a port’s connectivity.
In order to account for these factors, a new definition of connectivity is
proposed as follows. Considering that transshipment has become a major
port service, the connectivity of a port can be defined as the impact on
the transportation network’s performance when transshipment services are
not available at this port. Intuitively, such impact is proportional to the
connectivity of the corresponding port. This methodology also accounts
for any network effects that may influence services between ports that are
not directly connected. If a port has a high degree of connectivity, then

many carriers will come to this port to transship their cargoes. If said
port’s transshipment capabilities are now disabled, then many of these
carriers will have no choice but to select other ports at which they can
transship their cargoes, which will likely result in greater transportation
cost or longer shipping time. Thus, ports with a high degree of connectivity
will result in a greater impact on the transportation network’s performance
than those with low connectivity when transshipment services are not
available.
Such an impact on the network can take many forms. For example,
the model can measure the impact on the transportation flow capacity of
the system, the impact on the transportation time of cargoes and so on. It
is necessary to examine the results from different perspectives in order to
obtain a thorough and comprehensive understanding of this concept and
provide meaningful benchmarking results.
3.2.1. Case study

This subsection focuses on the effects of transshipment on container
throughput and shipping time of major ports in the Asia Pacific region. By
comparing various scenarios in which transshipment is disabled for certain
ports against a base case, we can rank the ports analyzed according to
the network benefits provided by transshipment services. The network was
constructed using data from the top 10 largest liner companies in 2008
according to CI-Online.
When considering the throughput model, the disabling of transship-
ment services at Singapore, Shanghai, Dalian and Qingdao has the greatest
impact in descending order (shown in Table 4). Singapore is the largest by
a wide margin as it serves as the primary transshipment hub for all traffic
to Oceania and between Asia and Africa/Europe. Shanghai is close behind
due to its share of transshipment traffic to the USA, while Dalian and
Qingdao serve as gateway ports to the North China region. The disabling
of transshipment services at Qingdao and Dalian has a significant impact
on the flow of traffic to Tianjin and the Beijing area. In comparison,
Hong Kong and Shenzhen have a relatively small impact on the network,
especially considering that they have a very large number of linking
services.
The waiting time model provides some different insights as seen in
Table 5. Disabling transshipment services at Singapore still has the greatest
Table 4 Impact on throughput.
Port Weekly TEUs per OD Change in throughput % change
Base case 235697.6

Singapore 230484.3 -5213.32 -2.21
Shanghai 232170.3 -3527.29 -1.50
Dalian 232577.5 -3120.17 -1.32
Qingdao 232759.1 -2938.56 -1.25
Yokohama 233312.9 -2384.69 -1.01
Ningbo 233785.5 -1912.16 -0.81
Busan 233826.1 -1871.53 -0.79
Kaohsiung 234298.4 -1399.24 -0.59
Port Klang 234361.1 -1336.51 -0.57
Tanjung Pelepas 234838.7 -858.878 -0.36
Hong Kong 234739.9 -957.743 -0.41
Tianjin 234746.1 -951.491 -0.40
Shenzhen 234770.8 -926.856 -0.39
Guangzhou 234946.6 -751.026 -0.32
Yingkou 235334.8 -362.835 -0.15
Laem Chabang 235463.6 -233.981 -0.10
Table 5 Impact on waiting time.
Port Waiting time per OD (days) Change in waiting time % change
Base case 6.924952

Singapore 7.451377 0.526424 7.60
Busan 7.329748 0.404796 5.85
Yokohama 7.22962 0.304668 4.40
Qingdao 7.107357 0.182405 2.63
HongKong 7.04775 0.122798 1.77
Shanghai 7.029669 0.104717 1.51
Shenzhen 6.996822 0.071869 1.04
PortKlang 6.991358 0.066406 0.96
Kaohsiung 6.983623 0.05867 0.85
LaemChabang 6.980875 0.055923 0.81
Dalian 6.958633 0.033681 0.49
Ningbo 6.957301 0.032349 0.47
TanjungPelepas 6.945159 0.020207 0.29
Guangzhou 6.929884 0.004932 0.07
Yingkou 6.925413 0.00046 0.01
Tianjin 6.924952 0 0.00
impact by a wide margin due to the very large number of transshipment

services available. Singapore also serves as a consolidation hub for South
East Asia ports to major markets such as the USA and Europe, thus there
is a large increase in waiting time when these smaller services are not able
to transship onto larger international liner services. Busan, Yokohama and

Qingdao are ranked second, third and fourth after Singapore in terms of
impact on waiting time when transshipment is disabled. This is likely due
to their geographical location as gateway ports to the North China region,
forming a bottleneck through which all services must pass. Tianjin has no
impact on waiting time due to its position deep in the bay of Bohai. This
means that services will not be routed via Tianjin as there will be some
backtracking and hence time wasted.
The results provide an indication of the relative impact of these ports
on the transportation network as a whole, which can be seen as a reflection
of their connectivity. Further analysis can be performed using the same
framework to obtain different insights, which can then be combined as part
of the benchmarking process. For example, the results from the throughput
model and the waiting time model can be integrated using the Pareto graph
in Fig. 1 below.
The grouping of ports shows that Singapore clearly has the highest
connectivity in terms of both impact on throughput and impact on waiting
time. Busan, Yokohama, Qingdao and Shanghai form a second group of
ports that have similar connectivity rankings, with some tradeoffs between
throughput and waiting time within the group. Dalian has a large impact
on throughput, but a small impact on waiting time, which pulls down its
overall connectivity ranking.
Fig. 1 Pareto analysis of throughput vs. waiting time.

3.3. Impact of factors on individual ports

The degree of complexity inherent in a port system can often make it very
difficult to determine the influence that any single factor can have on the
port’s performance. For example, carriers and shippers can base their port
selection on any number of factors such as cargo handling charges, reliability
or land accessibility. The complexity is compounded by the fact that a given
factor can play very different roles for different ports. In any benchmarking
study, it is useful to examine each port’s strengths and weaknesses, as well
as the factors contributing to their success or failure.
In order to analyze the impact of these factors on individual ports, this
section will utilize a network flow model (MCNFM) which is similar to the
model presented in Lee et al., (2006). The model attempts to derive optimal
shipping decisions by balancing the actual cost of transportation and
opportunity cost of transportation while satisfying constraints on supply
and demand, port capacity and link capacity. Thus, the model is able to
account for the trade-offs between transportation cost and shipping time.
The model is based on several assumptions, which are presented below.
Assumptions:
(i) The volume of trade demand between countries is known and deter-
ministic. Trade data are obtained in terms of dollar volume of trades,
from the United Nations COMTRADE database.
(ii) Each port is assumed to be the sole sea link gateway for that country.
For China, there are five ports in the study, namely Hong Kong,
Shenzhen, Qingdao, Ningbo and Shanghai. As it is difficult to model
demand and supply operations for 5 different ports under the same
country, they are grouped according to their geographical proximity
and linked into three regions. These are: Hong Kong-Shenzhen, for
Southern China; Shanghai-Ningbo, for Middle China; and Qingdao,
for Northern China.
(iii) For the segregation of Chinese ports and regions, it is assumed that
the port for each specific region handles all of the demand and supply
for the region. In effect, each region acts as a separate country with
their own gateway ports.
(iv) As the study focuses on the connectivity of maritime ports and does not
consider land links, each port-country/region pair will be considered as
one node. For example, Kaohsiung-Taiwan will be considered as one
node and Hong Kong-Shenzhen will be considered as another node
serving the southern part of China.
(v) As per Lee et al. ‘s model, the rationality of shippers is assumed, and
informed choices are made on the basis of perfect information.
Parameters:
N: Set of nodes (port-country/region)
O: Set of origin nodes
D: Set of destination nodes
L: Set of links
M: Set of ports
αij : The freight cost of transporting one TEU of cargo from node i to
node j
βij : The time incurred of transporting one TEU of cargo from node i to
node j
ρ: Nominal cost coefficient
λ: Average monetary value of one TEU for the single commodity
Kk : the capacity of port k
Skod : Denotes the amount of supply or demand for the commodity at
node k for origin-destination pair o-d. Skod is non-zero only if node
k either the destination or the origin. If node k is the origin, then
Skod is either 0 or negative (supply). If node k is the destination,
then Skod is either 0 or positive (demand).
MCNFM:

Min (αij · xod
ij ) + λ · ρ (βij · xod
ij ) (4)
o d ij o d ij

ik −
xod xod ∀o ∈ O, d ∈ D, k ∈ N,
od
s.t. kj = Sk , (5)
i∈N j∈N

ij ≤ Uij ,
xod ∀(i, j) ∈ L, (6)
o d

xod od
ik + δk Sk ≤ Kk , ∀k ∈ N (7)
o d i∈N
ij ≥ 0,
xod (8)
where xod
ij are the decision variables which indicate the number of container
flow in TEU shipped from origin o to destination d through link (i, j). The
objective value of the linear program minimizes the sum of two terms, of
which the first is the cost of transportation from node i to j and the second
is the opportunity cost of transporting goods from i to j. Equation (6)

is the conservation of flow constraint, which states that the amount of
goods going into port k for all origin — destination pairs must be equal
to the sum of the demand or supply and the amount of goods going out
of port k. Equation (7) states that the total amount of goods utilizing a
particular link must be within the link capacity. Equation (8) limits the
total goods passing through a port to within its annual capacity, where an
index is adopted which indicates whether port k is a supply port or demand
port, i.e.,

1, if port k is a supply port,
δk = (9)
0, otherwise.
The last equation (9) sets the decision variables to be non-negative.

The model is implemented using ILog OPL Development Studio 5.5
based on a CPLEX solver, and generates throughput figures for all origin-
destination pairs. By varying certain parameters such as link capacities and
frequencies, we can analyze the interactions among the major Asia Pacific
ports to help understand the competitive landscape in the region.
3.3.1. Case study

This subsection focuses on the analysis of the impact of some factors
on the nine major ports mentioned in Subsection 3.1.1. In our analysis,
each port is assumed to be the sole sea link gateway for that country.
For example, Kaohsiung is assumed to be the only port serving Taiwan,
although in actual fact Keelung is also a major port in Taiwan. Data from
three major liner shipping companies was used in our analysis, namely
Maersk, Evergreen and Mediterranean Shipping Company. Detailed data
for other shipping companies can be added in future studies to obtain more
comprehensive results. In our analysis, we focus on the major shipping
partners with the Asia Pacific region: the United States of America and
Europe. The model requires some calibration to ensure consistency with
actual throughput figures as only selected ports, liner companies and
regions are included in the study, which is based on trade data from 2002.
However, the relative proportion of throughput between the ports is fairly
accurate.
The model is implemented using ILog OPL Development Studio 5.5
based on a CPLEX solver. In total, it has 6571 decision variables and 826
Table 6 Comparison of model throughput and actual throughput.
% of total % of total
Throughout generated Throughput actual
Port (MODEL) throughput (ACTUAL) throughout
Singapore 28975954 19.51 29918200 19.45

Shanghai — Ningbo 39412267 26.53 39206000 25.49
Busan 21648917 14.57 13425000 8.73
Qingdao 5155761 3.47 10320000 6.71
Hong Kong - Shenzhen 35114351 23.64 45661888 29.69
Tanjung Pelepas 15000000 10.10 5600000 3.64
Kaohsiung 3241739 2.18 9676554 6.29
Total 14764783 100.00 153807642 100.00
linear constraints. Table 6 below compares the generated throughput from

the model with actual port throughput data.
The overall generated throughput for the model is lower than the actual
throughput. This is to be expected as only selected ports are included in
the study, and thus the results are not exhaustive. Furthermore, regions
such as Australia, Middle East, Africa and South America are not included
in the study. However, the generated throughput is still relatively accurate,
which can be attributed to the fact that Europe and the United States
are the main trading partners of Asia-Pacific countries and thus form the
majority of the container throughput for Asia-Pacific container ports. For
the purposes of this study, which focuses on the topic of port competition,
the absolute throughput is not as important as the relative proportion
of throughput between the ports and also the proportionate changes in
throughput when certain variables are changed. Ranking ports in terms of
proportions, the relative magnitude of the throughput figures are accurate
except for Tanjung Pelepas and Shanghai-Ningbo. This can be attributed
to the aggregation of Hong Kong and Ningbo into a single port and also
the approximation of certain data.
Moving towards the comparison of throughput at the individual port
level, the throughputs of Singapore and Shanghai-Ningbo ports reflected
as a proportion of the total are relatively close to the actual throughput,
contributing a combined percentage of 46.04%, which is almost half of the
overall throughput. All the other ports, with the exception of Qingdao,
have throughput that is less than the actual throughput. This might
be attributed to the same reason mentioned previously (not all trading
partners considered) and the assumption that only a single port is used
to represent each country. In actual fact there can be numerous ports in a
single country, for example Tanjung Pelepas, Port Klang in Malaysia, and
Keelung, Kaohsiung in Taiwan.
After developing a model that can adequately describe and approximate
the actual container shipping system in Asia-Pacific, sensitivity analysis can
be carried out to examine the impact of different factors on major ports.
As an example of such analysis, we consider the impact of varying link
capacity on Singapore port and Busan port. The rationale for choosing
these two ports is that they are major container shipping hubs, Singapore
being a global transshipment hub and Busan being a key transshipment
hub for the North Asia region, particularly for the transpacific trade route
serving North America. For these two transshipment hubs, link capacities
are more important than other ports, as they transfer cargo more frequently
than other ports. The ability to receive and send out large amount of
cargo is essential to transshipment ports. Sensitivity analysis for these two
transshipment hubs might provide insights on the impact on other ports in
the region if their link capacities are altered.
Singapore port
Figure 2 shows the variation in throughput (measured in thousands of
TEUs) of the various ports in the study when the overall link capacities of
Singapore port are varied. Specifically, the cases where the link capacities
are reduced to 50% and 30% are considered.
Fig. 2 Effects of reducing Singapore port’s link capacities.

It was found that the traffic flow for Singapore takes a much bigger
dip when the link capacity is reduced from 50% to 30%. This concurs with
the observation that link capacities are very important to the success of a
major transshipment hub such as Singapore. At 50% of the link capacity,
we see increased traffic flow to Qingdao, Shanghai-Ningbo and Hong Kong-
Shenzhen, in decreasing order.
When Singapore’s link capacity is reduced to 30%, traffic flow increases
greatly to Shanghai-Ningbo, Hong Kong-Shenzhen and Qingdao (in decreas-
ing order), which all lie on the coastline of China. This is due to con-
tainer flow that normally passes through Busan and Kaohsiung being first
diverted through the three China ports due to insufficient capacity on the
Singapore-Busan/Kaohsiung direct links with the reduction in Singapore’s
link capacity. These three ports can share the burden of a reduced capacity
on Singapore’s side. Shanghai-Ningbo’s throughput experiences the greatest
increase as it is geographically advantageous for both direct shipments to the
USA or transshipment to USA via Busan (it is the next closest port to Busan
while being nearer to Singapore than Qingdao). Both these possibilities can
raise the throughput of Shanghai-Ningbo port by a large amount, leading to
the observation from the model. This suggests Shanghai-Ningbo’s strategic
importance as a port in Asia Pacific, being well endowed with a superb
geographical location which can provide alternatives for Singapore’s USA-
bound cargo which traditionally is shipped via Busan.
Tanjung Pelepas, being the closest port to Singapore in this study,
surprisingly did not see any change in traffic when Singapore’s link
capacities are reduced. One possible reason for this is that the port is
already operating at its maximum capacity in the model, which indicates
an area for possible future refinement.
Busan port
The analysis is done on Busan in a similar manner, with link capacities
reduced to 70% and 50%. The results are summarized in Fig. 3.
At 70% of the link capacity, increased traffic flow to almost all other
ports can be observed, with the exception of Tanjung Pelepas. The ports
of Shanghai-Ningbo, Hong Kong-Shenzhen, Kaohsiung and Qingdao, in
decreasing order, see the greatest increases in traffic. These ports are within
the geographical vicinity of Busan. This reinforces the previous findings
that Shanghai-Ningbo is Busan’s greatest competitor, and that a port’s
link capacity has an impact on neighboring ports. In addition, with the
exception of Qingdao, the increase in throughput again seems to be related
Fig. 3 Effects of reducing Busan port’s link capacities.
to the geographical location. Shanghai-Ningbo is further north than Hong

Kong-Shenzhen, which is roughly location in the same region as Kaohsiung.
A preliminary hypothesis is that the nearer the port, the greater the gain
in traffic of that neighbor port when the target port’s link capacity is
decreased. That is, port competition is localized.
In the 50% case, the above hypothesis is strengthened. With Singapore’s
throughput remaining at roughly the same levels, the ports with the great-
est increase in throughput are Shanghai-Ningbo and Qingdao, which are the
ports nearest to Busan. However, the overall improvement in neighboring
ports is not as great compared to the 70% case. This might imply that as
the link capacities for Busan port decreased, the overall throughput of the
region is being hampered. This suggests inter-dependencies between ports
in the same region, even though there is competition between them.
The sensitivity analysis performed on both Singapore and Busan reveal
that Shanghai-Ningbo has the potential to be a major competitor in the
region, at least when considering link capacities. Shanghai-Ningbo stands to
gain the most from any reduction in link capacities through both Singapore
and Busan and could become a key port for the Asia-Pacific region in
the future.
4. Conclusion and Discussion

Port competition has grown more intense with the growth of international
trade and global economies. Asia is the pre-eminent region in the world
container throughput and the competition among Asian container ports

is even more intense than in other regions of the world, highlighting the
need for effective benchmarking methodologies. This paper focused on
benchmarking major ports in the Asia-Pacific region from three aspects:
port efficiency, port connectivity and impacts of factors on individual ports.
Using DEA models, it was found that Singapore, Hong Kong and Shanghai
rank as the most efficient. When considering a network framework to
evaluate connectivity in terms of throughput capacity and waiting time,
Singapore ranks as the most well connected, followed by Busan, Yokohama,
Qingdao and Shanghai. Lastly, the sensitivity of ports towards the impact
of various factors can be measured and benchmarked using a network
flow model, which revealed the importance of Shanghai-Ningbo as major
competitors to Singapore and Busan.
Some facts should be mentioned as follows:
• Most of the data used in the paper are collected from CI-online and some
other resources. However, some of the data could not be found and they
had to be estimated which will lead to inaccuracy of data. In addition,
some data used may be outdated or inaccurate as there is no reliable
universal source. So, further study may be required.
• Only nine ports and three liner shipping companies are included in this
study for port efficiency and impact of factors on individual ports. The
small number of ports and liner companies may not be representative
of the whole maritime transportation network of Asia-Pacific region and
this could likely lead to some inaccurate conclusions. More ports and
liner shipping companies should be involved in future studies.
• Port connectivity in this paper is still a preliminary study and as
mentioned earlier it is closely related to some other factors such as waiting
time, responsiveness and cost. Thus, it will be more meaningful if we
integrate these factors together in our connectivity analysis.
References
1. Barros, C.P. (2003). The measurement of efficiency of Portuguese sea port
authorities with DEA. International Journal of Transport Economics, 30(3):
335–354.
2. Bemowski, M.T. (1991). The competitive benchmarking wagon. Quality
Progress, January, pp. 11–16.
3. Coelli, T.J. (1996). A guide to DEAP version 2.1: a data envelopment
analysis (computer) program, unpublished manuscript, University of New
England at Armidale.
4. Cuadrado, M., Frasquet, M. and Cervera, A. (2004). Benchmarking the

port services: a customer oriented proposal. Benchmarking: An International
Journal, 11(3): 320–330.
5. Cullinane, K.P.B., Wang, T.F., Song, D.W. and Ji, P. (2006). The tech-
nical efficiency of container ports: comparing data envelopment analy-
sis and stochastic frontier analysis. Transportation Research A, 40(4):
354–374.
6. Goss, R. (1990). Ecomomic policies and seaports: 4. Strategies for port
authorities. Maritime Policy and Management, 17(4): 273–287.
7. Hoffmann, J. (2005). Liner shipping connectivity. UNCTAD Transport
Newsletter, 27: 4–12.
8. Huybrechts, M., Meersman, H., Van De Voorde, E., Verebeke, A., Van Hooy-
donk, E. and Winkelmans, W. (2002). Port competitiveness, an economic
and legal analysis of the factors determining the competitiveness of seaports,
p. 18.
9. Kleywegt, A., Goh, M.L., Wu, G.Y. and Zhang, H.W. (2002). Competition
between the Ports of Singapore and Malaysia. The Logistics Institute —
Asia Pacific, National University of Singapore, Singapore.
10. Lee, L.H., Chew, E.P. and Lee, L.S. (2006). Multicommodity network flow
model for Asia’s container ports. Maritime Policy & Management, 33(4):
387–402.
11. Lirn, T.C., Thanopoulou, H.A., Beynon, M.J. and Beresford, A.K.C. (2004).
An application of AHP on transhipment port selection: a global perspective.
Maritime Economics and Logistics, 6: 70–91.
12. Liu, C.C. (2008). Evaluating the operational efficiency of major ports in the
Asia-Pacific region using data envelopment analysis. Applied Economics, 40:
1737–1743.
13. Low, J.M.W., Lam, S.W. and Tang, L.C. (2009). Assessment of hub status
among Asian ports from a network perspective. Transportation Research
Part A: Policy and Practice, 43: 593–606.
14. Malchow, M. and Kanafani, A. (2004) A disaggregate analysis of port
selection. Transportation Research Part E: Logistics and Transportation
Review, 40: 317–337.
15. Marquez-Ramos, L., Martinez-Zarzoso, I., Perez-Garcia, E. and Wilmsmeier,
G. (2006). Determinants of maritime transport costs. Importance of connec-
tivity measures. Presented to International Trade and Logistics, Corporate
Strategies and the Global Economy Congress Proceedings, September,
Le Havre.
16. Martinez-Budria, E., Diaz-Armas, R., Navarro-Ibanez, M. and Ravelo-Mesa,
T. (1999). A study of the efficiency of Spanish port authorities using data
envelopment analysis, International Journal of Transport Economics, 26(2):
237–253.
17. McCalla, R., Slack, B. and Comtois, C. (2005). The Caribbean basin: adjust-
ing to global trends in containerization. Maritime Policy and Management,
32(3): 245–261.
18. Murphy, P.R., Daley, J.M. and Dalenberg, D.R. (1992). Port selection
criteria: an application of a transportation. Logistics and Transportation
Review, 28: 237–254.
19. Notteboom, T. (2006). The time factor in liner shipping services. Maritime
Economics and Logistics, 8: 19–39.
20. Panayides, P.M., Maxoulis, C.N., Wang, T.F. and Ng, K.Y.A. (2009).
A critical analysis of DEA applications to seaport economic efficiency
measurement. Transport Reviews, 29(2): 183–206.
21. Pardali, A. and Michalopoulos, V. (2008). Determining the position of
container handling ports, using the benchmarking analysis: the case of the
Port of Piraeus. Maritime Policy & Management, 35(3): 271–284.
22. Park, K.R. and De, P. (2004). An alternative approach to efficiency
measurement of seaports. Maritime Economics & Logistics, 6(1): 53–69.
23. Poitras, G., Tongzon, J. and Li, H. (1996). Measuring port efficiency:
an application of data envelopment analysis. Working paper (Singapore:
National University of Singapore).
24. Roll, Y. and Hayuth, Y. (1993). Port performance comparison applying data
envelopment analysis. Maritime Policy & Management, 20(2): 153–161.
25. Rugman, A.M. and Verbeke, A. (1993). How to operationalize Porter’s
diamond of international competitiveness. The International Executive,
35(4): 17–39.
26. Saaty, T.L. (1977). A scaling method for priorities in hierarchical structures.
Journal of Mathematical Psychology, 15: 234–281.
27. Sharmaa, M.J. and Yu, S.J (2009). Performance based stratification and
clustering for benchmarking of container terminals. Expert Systems with
Applications, 36(3): 5016–5022.
28. Slack, B. (1985). Containerisation, inter-port competition and port selection.
Maritime Policy and Management, 12(4): 293–303.
29. Song, D.W. and Yeo, K.T. (2004). A competitive Analysis of Chinese
Container Ports using the Analytic Hierarchy Process. Maritime Economics
and Logistics, 6: 34–52.
30. Tang, L.C., Low, J.M.W. and Lam, S.W. (2008). Understanding Port
Choice Behavior¡a A Network Perspective. Networks and Spatial Economics,
published online.
31. Tiwari, P., Itoh, H. and Dio, M. (2003). Shippers’ port and carrier selection
behaviour in China: a discrete choice analysis.Maritime Economics and
Logistics, 5: 23–39.
32. Tongzon, J. (2001). Efficiency measurement of selected Australian and
other international ports using data envelopment analysis. Transportation
Research A, 35(2): 113–128.
33. Tongzon, J. and Heng, W. (2005). Port privatization, efficiency and com-
petitiveness: some empirical evidence from container ports (terminals).
Transportation Research Part A: Policy and Practice, 39: 405–424.
34. Ugboma, C.H., Ugboma, O. and Ogwude, I. (2006). An analytic hierarchy
process (AHP) approach to port selection decisions — Empirical evidence
from Nigerian ports. Maritime Economics and Logistics, 8: 251–266.
35. Valentine, V.F. and Gray, R. (2001). The measurement of port efficiency
using data envelopment analysis, in: Proceedings of the 9th World Conference
on Transport Research, 22–27 July (Seoul, South Korea).
36. Wilmsmeier, G., Hoffmann, J. and Sanchez, R.J. (2006). The impact of
port characteristics on international maritime transport costs. Research in
Transportation Economics, 16: 117–140.
37. Wilmsmeier, G. and Hoffmann, J. (2008). Liner shipping connectivity and
port infrastructure as determinants of freight rates in the Caribbean.
Maritime Economics and Logistics, 10: 130–151.

CHAPTER 5
IS PORT THROUGHPUT
A PORT OUTPUT?
Wayne K. Talley
Department of Economics, Old Dominion University,
Norfolk, Virginia, 23529 U.S.A.
[email protected]
This paper investigates the question: is port throughput a port output in port
economic production and cost functions? The paper concludes that the answer
to this question is no. Port throughput is the amount of cargo received from
carriers that passes through the port. A port does not produce throughput
but rather provides interchange service for the cargo that it receives. By using
port throughput as a measure of port output (or the dependent variable) in
a port’s interchange service economic production function: (1) the interchange
service economic production function no longer exists and (2) the port’s
long-run total cost and short-run variable economic cost functions cannot
be derived. A measure of port interchange service that is consistent with
measures used heretofore by ports for evaluating their performance is the port
throughput ratio — the ratio of cargo interchanged to the total time incurred
in interchanging the cargo.
1. Introduction
Port cargo throughput is the amount of cargo that passes through a
port from one transport carrier to another. The technical (or productive)
efficiency of cargo ports has been investigated in the literature under the
assumption that the cargo output of ports is cargo throughput. A port
is technically efficient if its output is the maximum obtainable output in
the use of given levels of resources. A port’s economic production function
represents this functional relationship. Port cargo throughput as a measure
of port output is found in empirical port economic production studies
by Chang,1 Kim and Sachish2 Bendall and Stent,3 Dowd and Leschine4
and Liu.5 Container port output measured as TEU (twenty-foot equivalent
117
118 W. K. Talley
unit) throughput has been used in empirical Data Envelopment Analysis

(DEA) and stochastic frontier analysis models for investigating the relative
technical efficiency among container ports by Tongzon,6 Cullinane, Song,
Ji, and Wang,7 and Cullinane and Song.8,q
The cost efficiency of cargo ports has also been investigated in
the literature under the assumption that the cargo output of ports is
cargo throughput. A port is cost efficient if its output is provided at
minimum costs. A port’s economic cost function represents this functional
relationship. Port cargo throughput as a measure of port output is found
in empirical port economic cost studies by Kim and Sachish,2 Martinez-
Budria,9 Jara-Diaz, Martinez-Budria, Cortes, and Vargas,10 Jara-Diaz,
Martinez-Budria, Cortes, and Basso11 and Jara-Diaz, Tovar and Truillo.12,r
This paper investigates the question: is port throughput a port output
in port economic production and cost functions? The paper concludes that
the answer to this question is no. The output of ports is interchange service,
i.e., ports use its resources to interchange cargo provided by carriers between
arriving and departing carrier vessels and vehicles. The amount of cargo
that is interchanged is the port’s throughput. The paper also investigates
the impact of measuring port output as port throughput rather than as
port interchange service in the investigation of port technical and cost
efficiencies.
In the next section, a port interchange service economic production
function is presented, followed by discussions of port operating options and
the port resource function in Section 3 and 4, respectively. In Section 5, the
impacts of using port throughput as a measure of output in port economic
production and cost functions are discussed. In Section 6, measures of
port interchange service are presented. Finally, conclusions are found in
Section 7.
2. Port Production Function

For the sake of simplicity, only the throughput of a container port is used
in the following discussion. However, the discussion can be applied to any
q For further discussion of technical efficiency and ports, see Cheon, Dowall and Song18,
Song, Cullinane and Roe17, Talley14, Talley15, Wang, Cullinane and Song19 and Yan,
Sun and Liu.20
r A review of these cost studies is found in Tovar, Jara-Diaz and Trujillo.21 An estimate of
an input distance function for investigating the technical inefficiency of ports, where tons
of port cargo throughput are used as measures of port outputs, is found in Rodriguez-
Alvarez, Tovar and Trujillo.22
Is Port Throughput a Port Output? 119
type of port cargo. It is important to note that a port does not produce
throughput, but provides interchange service to the cargo that it receives.
For example, a container port does not produce TEUs (twenty-foot equiva-
lent units) or 20 foot containers but rather provides interchange service for
the TEUs that it receives, i.e., import TEUs received from ocean vessels
are interchanged with domestic vessels and land vehicles and vice versus
for export TEUs; transshipment TEUs are interchanged between vessels.
In order for a container port to provide TEU interchange service, at
least two parties must be in agreement. If either party is not in agreement,
no container port TEU interchange service will occur. Specifically, trans-
portation carriers must be willing to transport containers to and from a port
and the port must be willing to interchange containers that are received. If
the port is unwilling to accept containers, even though carriers are willing
to provide it with containers, no port TEU interchange service will occur.
If carriers are unwilling to provide the port with containers, even though
the port is willing to accept containers, no port TEU interchange service
will occur. The port can not force carriers to provide it with containers and
carriers can not force the port to accept its containers.cs
From the above discussion, a container port interchange service eco-
nomic production function is a function that relates the maximum amount
of TEU interchange service provided by the port to the levels of resources
utilized in providing interchange service and the number of TEUs provided
by carriers, i.e.,
TEU Interchange Service

= f (Ri ; TEUs Provided by Carriers) i = 1, 2 . . . M (1)
where, TEU interchange service is the service of interchanging TEUs

between arriving and departing carrier vessels and vehicles at the port and
Ri is the ith resource.
If the port adheres to its interchange service economic production
function in the provision of TEU interchange service, the port is technically
efficient in the provision of this service. Further, since the dependent
variable of a firm’s economic production function is the amount of output
provided by the firm, it thus follows that the dependent variable of economic
production function (1), TEU Interchange Service, is the amount of output
provided by the container port.
s c. The exception is when the port and carrier are owned or leased by the same firm.
120 W. K. Talley
3. Port Operating Options

A firm that produces a tangible product seeks to produce homogenous units
of the product, i.e., it does not seek to produce a product that varies in
quality. Alternatively, a port that provides a service (rather than producing
a tangible product) does seek to provide a service that varies in quality, e.g.,
to improve the quality of its interchange service in order to attract more
cargo from carriers. The means by which a port can differentiate (or vary)
the quality of its interchange service have been referred to in the literature
as the port’s operating options.13,14,15
Operating options for container ports, for example, include14 : (1)
loading and unloading service rates, i.e., TEUs loaded and unloaded per
unit time to and from vessels, vehicles and port equipment; (2) entrance
gate reliability (percent of time that the port’s entrance gate is open for
vehicles); (3) departure gate reliability (percent of time that the port’s
departure gate is open for vehicles); (4) harbor waterway reliability (percent
of time that the port’s harbor waterway is open to navigation); (5) berth
reliability (percent of time that the port’s berth is open to the berthing
of vessels); (6) berth accessibility (percent of time that the port’s berth
adheres to authorized depth and width dimensions); (7) harbor waterway
accessibility (percent of time that the port’s harbor waterway adheres to
authorized depth and width dimensions); (8) the monetary loss of cargo,
vessel property, inland carrier property and port equipment property in
port to theft; and (9) the monetary damage to cargo, vessels, inland carrier
vehicles and port equipment in port.
4. Port Resource Function

A port resource function for the ith resource (Ri ) relates the minimum
amount of this resource to be employed by the port to the port’s levels
of operating options and the mount of cargo provided by carriers (see
Talley14 ). If the cargo is TEUs, then the port resource function may be
expressed as:
Ri = Ri (OPTION1 , OPTION2 , OPTIONn . . . OPTIONN ;

TEUs Provided by Carriers) i = 1, 2 . . . M (2)
where, OPTIONn is the nth operating option of the port.

If no excess capacity exists for the port’s ith resource, then a change
in the port’s nth operating option for the purpose of increasing the quality
of its interchange service will require that an additional amount of the

ith resource be employed by the port. If an increase in the quality of
service requires an increase in the nth operating option, e.g., an increase
in the vessel loading service rate, an increase in Ri will also occur, i.e.,
∂Ri /∂OPTIONn > 0. If an increase in the quality of service requires a
decrease in the operating option, e.g., a decrease in the lost of port cargo
to theft, an increase in Ri will also occur, i.e., ∂Ri /∂OPTIONn < 0. If no
excess capacity exists for resource Ri , an increase in TEUs Provided by
Carriers to the port will result in an increase in the amount of resource Ri
used by the port, i.e., ∂Ri /∂ TEUs Provided by Carriers > 0.
5. Container Port Output

Rather than using TEU interchange service as the amount of output
provided by a container port, the literature heretofore has instead used TEU
throughput. However, what are the impacts in doing so on the estimation
of container port’s economic production and cost functions for investigating
container port technical and cost inefficiencies?
5.1. TEU Throughput and the port production function

Theoretical economic production functions utilized heretofore in the esti-
mation (via econometric and frontier analysis procedures) of port economic
production functions are those that have appeared in the literature in the
estimation of economic production functions of firms that produce a tan-
gible product rather than provide a service. Specifically, these production
functions relate the maximum amount of a product (or output) produced
by a firm to the levels of resources utilized, i.e.,
Output = g1(Ri) i = 1, 2, . . . M (3)
Port studies where estimates of economic production function (3)

are found include, for example, studies by Chang,1 Bendall and Stent,3
Liu,5 Dowd and Leschine4 and Cullinane and Song.8 In the estimation of
economic production function (3) for container ports, output is measured
as TEU Throughput.
Criticisms in using function (3), where output is measured as TEU
Throughput, to investigate the technical efficiency of container ports
include: a) container ports provide a service as opposed to producing a
tangible product, b) economic production function (3) unlike interchange
service economic production function (1) does not capture the fact that
122 W. K. Talley
the number of TEUs received by ports are not determined by the ports
but by carriers through their port-choice decision making process and c)
interchange service economic production function (1) no longer exists if
interchange service is measured as TEU Throughput (see Proposition 1
below).
Proposition 1: Interchange service economic production function (1) no

longer exists if interchange service is measured as TEU Throughput.
Proof. If TEU Interchange Service in function (1) were replaced with

TEU Throughput, a problem would arise, i.e., from the fact that TEU
Throughput = TEUs Provided by Carrierst or TEUs interchanged by the
port are the same TEUs provided to the port by carriers. Port output
(a dependent variable) cannot be a function of itself and the function be
a port economic production function. A function by definition denotes the
relationship between a dependent variable and one or more independent
variables. A function does not denote the relationship between a dependent
variable and a set of variables that includes the same dependent variable.
Alternatively, a variable cannot be both a dependent and an independent
variable in the same relationship and the relationship be a function.
5.2. TEU Throughput and port cost functions

A port will be cost efficient in the provision of its output if it provides
its output at the least cost, given the prices that it pays for the resources
employed. The relationship between the least or minimum costs and the
amount of output provided is the port’s economic cost function. In order to
be cost efficient, it is necessary that the port be technically efficient in the
provision of this output. Otherwise the port could provide the same amount
of output with at least a lesser amount of one resource and therefore incur
lower cost in providing this output (for given resource prices).
5.2.1. Long-run total cost function

The long-run is a time period that is sufficiently long enough so that
amounts of all resources employed can be varied. To be cost efficient in
t In practice, i.e., once a time period is specified (say a year), this equality may not hold
exactly. For example, containers that are received by a port at the end of the year may
not depart from the port (and therefore become port throughput) until several days into
the next year.
providing output in the long-run time period, a port will minimize its
long-run total cost (LTC) in the provision of technically efficient output
(represented by the interchange service economic production function (1)),
i.e., minimize LTC subject to interchange service economic production
function (1). Further, LTC will be minimized subject to resource function
(2), since the latter represents the minimum amount of a given resource to
be utilized the port given the amount of TEUs provided by carriers and
the levels of the port’s operating options. LTC is the sum of the products
of the amounts of resources (Ri ) utilized by the port and the prices (Pi )
incurred in their utilization, where Pi is the price incurred by the port for
the ith resource.
From the above discussion, the port’s long-run total cost function will
be derived by:

Minimizing LTC = Pi Ri
subject to
TEU Interchange Service = f (Ri ; TEUs Provided by Carriers) (4)
TEUs Provided by Carriers) i = 1, 2, . . . M
The choice variables in the optimization of equation (4) are the port’s
operating options. That is to say, the port would select the levels of its
operating options (or quality of service). This selection in turn determines,
given the TEUs provided by carriers, the port’s minimum levels of resources
to be employed — which are also the levels to be employed that minimize
the port’s long-run costs (LTC). From this optimization, the port’s long-
run economic total cost function can also be derived, where LTC is a
function of the prices of the resources employed by the port, the port’s
TEU Interchange Service and TEUs Provided by Carriers, i.e.,
LTC = LTC(Pi , TEU Interchange Service,

TEUs Provided by Carriers) i = 1, 2, 3 . . . M (5)
Proposition 2: Port long-run economic total cost function (5) cannot be

derived if TEU Interchange Service is measured by TEU Throughput.
Proof. By Proposition 1, interchange service economic production func-

tion (1) does not exist if interchange service is measured as TEU
124 W. K. Talley
Throughput. Hence, Eq. (4) cannot be optimized to obtain long-run

economic total cost function (5).u
5.2.2. Short-run variable cost function

To be cost efficient in providing output in the short-run (a time period that
is sufficiently short so that the amounts of one or more resources employed
cannot be varied, i.e., are fixed), a port will minimize its short-run variable
cost (SRVC) subject to interchange service economic production function
(1), RM amount of fixed resource M and resource function (2), i.e.,

Minimize SRVC = Pi Ri
subject to
TEU Interchange Service = f (Ri ; RM , TEUs Provided by Carriers) (6)
TEUs Provided by Carriers) i = 1, 2 . . . M − 1
As for Eq. (4), the choice variables in the optimization of Eq. (6) are
the port’s operating options. From the optimization, the port’s short-run
economic variable cost function can be derived, where SRVC is a function
of the prices of the variable resources (1 through M − 1) employed by the
port, the port’s TEU Interchange Service, TEUs Provided by Carriers and
RM amount of fixed resource M , i.e.,
SRVC = SRVC(Pi , TEU Interchange Service,

TEUs Provided by Carriers, RM ) i = 1, 2, 3 . . . M − 1 (7)
u Criticsof the use of port aggregated cargo throughput (i.e., the aggregation of the
throughput of two or more different types of cargo) as a measure of port output in the
estimation of port economic production and cost functions note that port aggregated
cargo throughput generates aggregation-bias estimates of technical and cost efficiencies
for ports that handle more than one type of cargo, e.g., solid bulk, liquid bulk, general
non-containerized and containerized cargoes. To avoid aggregation bias in such studies,
Kim and Sachish2 recommend that the port throughput of each type of cargo be used
as port output measures rather than using a single aggregate throughput measure. In a
port cost function estimation study by Martinez-Budria,9 aggregate throughput is used
as the measure of port output. However, the study makes the restrictive assumption that
the cost share of a ton of any type of cargo is independent of the activity where it is
handled.
Proposition 3: Port short-run economic variable cost function (7) cannot

be derived if TEU Interchange Service is measured by TEU Throughput.
Proof. By Proposition 1, interchange service economic production func-

tion (1) does not exist if interchange service is measured as TEU Through-
put. Hence, Eq. (6) cannot be optimized to obtain short-run economic
variable cost function (7).
5.2.3. Port cost function estimates

Estimates of long-run economic total cost functions for various types of
ports are found in studies by Kim and Sachish,2 Martinez-Budria,9 Jara-
Diaz, Martinez-Budria, Cortes and Vargas,10 Jara-Diaz, Martinez-Budria,
Cortes and Basso11 and Jara-Diaz, Tovar and Trujillo.12 Port long-run total
costs in these studies were expressed as functions of port throughput and
the prices of the resources employed by ports — but not functions of port.
Interchange service.v Measures of port throughput found in these
studies include tons of: cargo, container general cargo, non-container general
cargo, liquid bulk cargo, dry bulk cargo and roll-on roll-off (Ro-Ro) cargo.
As noted in Proposition 1, a port’s TEU throughput is also the TEUs
provided by carriers to the port. This is also true for any type of port
throughput — i.e., a port’s throughput of any type is also the amount
of cargo of any type provided by carriers to the port (to be interchanged
between arriving and departing carrier vessels and vehicles). Hence, the
port long-run economic total cost functions used in the above cost studies
may be alternatively stated as functions, where port long-run economic
total cost is expressed as a function of the amount of cargo provided by
carriers to the port and the prices of the resources employed by the port.
However, the derived long-run economic total cost function for a container
port (function 5) has long-run total cost being a function of not only of
TEUs provided by carriers and the prices of resources employed, but also
of the port’s TEU interchange service. For a port handling any type of
cargo, function (5) would be rewritten as port long-run total cost being a
function of cargo provided by carriers, prices of resources employed and the
level of cargo interchange service.
v The one possible exception is the study by Jara-Diaz, Martinez-Budria, Cortes and
Basso11 that includes the explanatory variable, index added of other activities that use
part of the port infrastructure.
126 W. K. Talley
Since the above port cost studies do not include the explanatory
variable, cargo interchange service, in obtaining their long-run total cost
function estimates, the studies’ estimated parameters are statistically
biased due to specification error, i.e., from a relevant explanatory variable
having been omitted from the estimation (see Pindyck and Rubinfeld.16,w
The studies conclude that ports exhibit economies of scale, i.e., port long-
run total cost does not increase in proportion to the increase in port
throughput. Would this conclusion hold if the explanatory variable, cargo
interchange service, had been used in obtaining the long-run total cost
function estimates?
6. Port Interchange Service Measures

6.1. Port revenue
In a study by Song, Cullinane and Roe,17 a container port’s output was
initially described as an interchange service. Specifically, the output was
defined as the container turnover derived from the delivery of container
terminal services and measured by the revenue received for these services.
However, problems arose in the use of revenue as a measure of container
port output. Port revenue not related to interchange service, e.g., from
property sales, had to be excluded. The wide diversity of accounting
systems used by port terminals in the study resulted in an intractable
problem of separating out revenue attributable to different sources. In
the end, the study chose instead the output measure of annual container
throughput TEUs.2
Even if revenue received from port interchange service could be mea-
sured without error, port revenue would not be an appropriate measure of
port interchange service. That is to say, port revenue from port interchange
service and the amount of port interchange service provided are not
necessarily positively related. Port revenue may be increasing, while port
interchange service is decreasing over time. Port revenue is the product of
port price and port interchange service. Port revenue will increase over time
even though the amount of port interchange service demanded is decreasing
over time if the demand for port interchange service is price inelastic, all
else held constant.
wA parameter estimate is biased when the expected value of the estimated parameter
does not equal to the population parameter. Since the bias will not disappear as the size
of the sample increases, the parameter estimate is also inconsistent.
6.2. Port throughput ratio

A feasible measure of port interchange service is the port throughput
ratio — i.e., the ratio of port throughput to the time that the throughput
is in port. This ratio would be determined by dividing the amount of cargo
provided by carriers to and passing through a port (or interchanged) by
the total time that the cargo is in port. This total time is a measure of
the utilization of resources by a port in the interchange of cargo. Note
that if a port’s throughput is increasing faster over time than the time
it incurs in interchanging throughput, the port’s throughput ratio will be
increasing over time. Alternatively, it can be stated that the port’s technical
inefficiency is decreasing over time, since the port’s utilization of resources
is not increasing in proportion to the increase throughput.
The throughput ratio is consistent with measures used heretofore by
ports for evaluating the performance of their interchange service, e.g., vessel
loading and unloading services rates — the amount of cargo loaded to and
from vessels per unit of time. Also, ports seek to reduce the dwell time of
cargo, e.g., storage time per TEU stored or increase the reciprocal, TEUs
stored per unit of storage time.
7. Conclusion
This paper investigated the question: is port throughput a port output in
port economic production and cost functions? The paper concludes that
the answer to this question is no.
A port provides interchange service for cargoes provided by transport
carriers, i.e., a port uses its resources to interchange cargoes between
arriving and departing carrier vessels and vehicles. A port’s interchange
service economic production function relates the maximum amount of cargo
interchange service provided by the port to the levels of resources utilized in
providing interchange service and the amount of cargo provided by carriers.
The amount of cargo that is interchanged is the port’s throughput.
A port’s long-run economic total cost function relates the minimum
long-run total cost to be incurred by the port to the prices of the resources
employed by the port, the port’s cargo interchange service and the cargo
provided by carriers. A port’s short-run economic variable cost function
relates the minimum short-run variable cost to be incurred by the port to
the prices of the variable resources employed by the port, the port’s cargo
interchange service, the cargo provided by carriers and the amount(s) of
the fixed resource(s).
128 W. K. Talley
By using port throughput as a measure of port output (or the

dependent variable) in a port’s interchange service economic production
function the: (1) interchange service economic production function no
longer exists and (2) port’s long-run economic total cost and short-run
economic variable cost functions cannot be derived. A measure of port
interchange service that is consistent with measures used heretofore by
ports for evaluating the performance of their interchange service is the port
throughput ratio — the ratio of cargo interchanged (or throughput) to the
total time incurred in interchanging the cargo.
References
1. Chang, S. (1978). Maritime Policy and Management, 5.
2. Kim, M. and Sachish, A. (1986). Journal of Industrial Economics, 35.
3. Bendall, H. and Stent, A. (1987). Maritime Policy and Management, 14.
4. Dowd, T. and Leschine, T. (1990). Maritime Policy and Management, 17.
5. Liu, Z. (1995). Journal of Transport Economics and Policy, 29.
6. Tongzon, J. (2001). Transportation Research Part A, 34.
7. Cullinane, K., Song, D.-W., Ji, P. and Wang, T.-F. (2004). In Review of
Network Economics (Special Issue), Ed. Talley, W.K.
8. Cullinane, K. and Song, D.-W. (2006). In Port Economics: Research in
Transportation Economics, Eds. Cullinane, K. and Talley, W.K. (Elsevier,
Amsterdam).
9. Martinez-Budria, E. (1996). Revista Asturiana de Economia, 6.
10. Jara-Diaz, S., Martinez-Budria, E., Cortes, C. and Vargas, A. (1997). 25th
European Transport Forum: Proceedings Seminar L (PTRC, London).
11. Jara-Diaz, S., Martinez-Budria, E., Cortes, C. and Basso, L. (2002).
Transportation, 29.
12. Jara-Diaz, S., Tovar, B. and Trujillo, L. (2005). Transportation, 32.
13. Talley, W.K. (1994). Logistics and Transportation Review, 30.
14. Talley, W.K. (2006). In Port Economics: Research in Transportation Eco-
nomics, Eds. Cullinane, K. and Talley, W.K. (Elsevier, Amsterdam).
15. Talley, W.K. (2009). Port Economics (Routledge, Abingdon, United
Kingdom).
16. Pindyck, R.S. and Rubinfeld, D.L. (1981). Econometric Models and Eco-
nomicForecasts (McGraw-Hill Book Company, New York).
17. Song, D.-W., Cullinane, K. and Roe, M. (2001). The Productive Efficiency
of Container Terminals: An Application to Korea and the UK. (Ashgate,
London).
18. Cheon, S., Dowall, D.E. and Song, D.-W. Transportation Research Part E,
(forthcoming).
19. Wang, T.-F., Cullinane, K. and Song, D.-W. (2005). Container Port Produc-
tion and Economic Efficiency (Palgrave-Macmillan, Basingstoke).
20. Yan, J., Sun, S. and Liu, J. (2009). Transportation Research Part B, 43.
21. Tovar, B., Jara-Diaz, S. and Trujillo, L. (2007). Maritime Policy and
Management, 34.
22. Rodriguez-Alvarez, A., Tovar, B. and Trujillo, L. (2007). International
Journal of Production Economics, 109.

CHAPTER 6
A FRAMEWORK FOR
MODELLING AND
BENCHMARKING MARITIME
CLUSTERS: AN APPLICATION
TO THE MARITIME CLUSTER
OF PIRAEUS
Vassilios K. Zagkas and Dimitrios V. Lyridis
National Technical University of Athens School
of Naval Architecture & Marine Engineering
Laboratory for Maritime Transport, Athens, Greece
[email protected]
[email protected]
This paper investigates the factors that contribute to the decisions of companies
from key maritime sectors to be established in a specific area that evolves
into a network of firms. It is also the scope of this paper to investigate and
benchmark the circumstances under which a network of firms is transformed
into a competitive Maritime Cluster. In this framework we present methods for
developing and evaluating possible models for the Cluster creation and devel-
opment, addressing more specifically the case of Piraeus. The concentration of
the research on the Greater Area of Piraeus as opposed to the whole country is
considered given that the country’s major port is located in the Greater Area of
Piraeus which constitutes a very active maritime community. Furthermore, the
paper will give a short introduction into new computational methods such as
Agent Based Modelling for simulating the networking process within maritime
clusters and managing their life cycle. This will give an insight of firm survival
strategies within the cluster, optimum timing for new entrants in the cluster
and overall cluster management.
131
132 V. K. Zagkas and D. V. Lyridis
1. Introduction
This paper addresses the role of regional entrepreneurial networks and
their evolution into dynamic cluster formations through the emergence
of competitive advantages. Several theories have been applied in the
study of clusters; such theories are agglomeration economics, industrial
districts, spatial economics, and economic geography- all of them being
useful tools. However, the competitiveness theory as developed by Michael
Porter in the 1990’s is the most well-known theory on clusters and their
economic behavior. The integration of Porter’s theory with the maritime
context can give a pragmatic approach to Maritime Clusters. Planning and
structuring the maritime cluster can be considered as a cyclical process
consisting of iterative cycles infatuated by governmental or private initia-
tives. However, managing the maritime cluster, retaining and enhancing
its competitive advantages in the context of international competition are
complex issues related to dynamic systems and complexity theory. In the
framework of this research a sophisticated computational model such as
Agent Based Modelling is employed in order to simulate the actions and
interactions of firms that act as autonomous individuals in the maritime
cluster, with a view to assessing their effects on the cluster system as
a whole.
2. Conceptual Definition: Cluster Theory

and Maritime Clusters
The increasing interest in cluster structures and their valuable outcomes
have led us into their investigation and tempted us into their application
on the Maritime sector. The development of clusters is by some seen
as the only way to overcome the risk of being outperformed in the
global economy (Lagendijk, 2000). Cluster theory was first introduced
almost a century ago by Alfred Marshall under the term of industrial
districts. In Principles of Economics he described the phenomenon as ‘the
concentration of specialized industries in particular localities’ (Marshall,
1922). The concept of knowledge spill-over and externalities were crucial
elements on Marshall’s theory, as he elegantly states: ‘The mysteries of
the trade become no mysteries, but are as it were in the air’ (Marshall,
1922). Industrial districts enjoy the same economies of scale that only
giant companies normally get. Specialized suppliers arrive. Skilled workers
know where to come to ply their trade. And everyone involved benefits
from the spill-over of specialized knowledge (Surowiecki, 2000). Later on,
A Framework for Modelling and Benchmarking Maritime Clusters 133
the competitiveness theory as developed by Michael Porter presented in

his 1990’s book, ‘The Competitive Advantage of Nations’, is the most
well known theory on clusters and their economic behavior. Economic
sciences hesitate to get involved in subjects where the use of numbers
and quantities is limited. However, thanks to Porters approach, it became
widely known that cluster has a very good impact on the economy and
many papers like this one struggle to define how much good that is.
The main argument in Porter’s theory is that firms and not nations
compete in international markets and the presence of competing clusters
is a key dynamic factor to nation competitiveness. Porter’s Diamond
can be better comprehended through the crucial question: ‘Why do
firms based in particular nations achieve international success in distinct
segments and industries?’ (Porter, 1990). According to Porter the answer
lies in four broad attributes of a nation that shape the environment in
which local firms compete and promote competitive advantage. Those
four elements are: Factor Conditions, Demand Conditions, Related and
Supporting Industries, Firm strategy, structure and rivalry (Porter, 1990).
The integration of the four elements in Porter’s theory with the maritime
context can give a pragmatic approach to Maritime Clusters. However,
it is very difficult to speak eloquently about the cluster of firms or
the competitive advantage of some regions or cities without explicitly
taking into consideration the ‘space’. Over the years, economists have
neglected spatial issues, due to the difficulty of modeling increasing
returns and imperfect competition. Thus, the study of economic geography
and space was pushed to the periphery of economic theories (Krugman,
1991).
This overview shows that there is an old and strong theoretical
background for clusters addressing both their economic and spatial matters,
originating from industrial districts and they are related to into agglom-
eration economics, spatial economics and the competitiveness theory. This
paper addresses the need for a theory integrating the four factors of Porter’s
theory with economic values and spatial development; a system that will
reveal the effect the micro level has on the macro level — the whole
cluster.
3. The Concept of Maritime Clusters

Experience around the world has shown that the concept of clustering
suits particularly well to maritime businesses. There are numerous benefits,
ranging from specialized labor to targeted training, from increased mar-

ket awareness to connections with R&D institutes and from strategic
co-operations to inter-related maritime activities (Wijnolst, 2009). Despite
the large maritime industry in Europe and worldwide, we have little sys-
tematic information concerning the degree of interaction between maritime
firms. The European network of maritime clusters is one of the pioneering
initiatives concerning the cross-country maritime cluster of Europe. Several
country reports, as Norway’s’ and Netherlands’, have been published there
revealing the structure and some quantitative data of their maritime cluster
(Wijnolst, 2006).
The need for a flexible theory to base our research on, has directed us
towards a bottom-up approach to the maritime cluster concept. Therefore,
the first task was the conceptual definition of the maritime cluster, hence,
maritime cluster as per se in our research can be defined as: ‘The outcome
of one or more spatial consolidations, of cooperating — competing firms and
institutions within all sectors, sub-sectors and economic activities directly or
indirectly linked to the shipping industry, maritime transport and generally
the utilization of the sea’.
Based on the above definition, it is then necessary to define the sector
and sub-sectors that make up a maritime cluster. The European commission
has identified the following traditional maritime sectors in Europe (E.C.
Report, 2009), as shown in Table 1.
However, many differences exist per country and maritime cluster
regarding the scope of the maritime industry and its specialization. The
European Network of maritime clusters give us a more narrow or more
pragmatic perspective on the sectors of European maritime cluster. Here,
we have eight sectors: Shipping, Shipbuilding, Marine equipment, Seaports,
Table 1 Traditional maritime sector according to EC

study (E.C. Report, 2009).
Traditional Maritime Sectors (EC study)
Shipping Scrapping
Shipbuilding Offshore supply
Ports & Related Services Cable & Submarine telecom
Classification Societies Inland Shipping
Repair & conversion Naval Shipbuilding
R&D and Education Dredging & Maritime works
Equipment Manufacturing Recreational Vessels
Support Services Fishing & Aquaculture
Maritime services, Yacht building, Offshore services and Fishing (E.C.

Report, 2009).
Much of the literature on clusters has ignored the issue of market
structure. On the other hand, literature on maritime clusters is dedicated
on replicating the market networking structure that exists without inves-
tigating in depth the reasoning for such networking. In the framework of
approaching the concept of maritime clusters, there is the need to analyze
the structure of the network and identify the key relationships that control
the supply and demand in the maritime sector. The study on the maritime
service sector in London (Grammenos, 1992) gave us an insight on how
different firms in the shipping industry are interconnected. The study came
to a model with a core centre of three main sectors: Charterers, Owners
and Brokers around whom ancillary services revolve. The model proposed
here is based on one fundamental value that shall govern the behavior of
the cluster; there is supply and demand of knowledge between the different
categories that live in the cluster and among the firms that populate each
nod in a micro perspective. In practice, the demand and supply of knowledge
can be translated into exchange of services and goods among the firms.
The triangle of charterers, owners and brokers (Grammenos, 1992) has
been replaced by one triangle incorporating Ship-owners, Ship-managers
and Charterers and another inverted triangle included in the previous
one, presenting brokers as the intermediary activity between supply and
demand. The following scheme presents the idea of the double inverted
triangle.
Fig. 1 Double inverted triangle, market modeling.

As Fig. 1 above implies, the demand for shipping is expressed in the

market by charterers who seek vessels for the movement of their cargoes
while on the other side the supply is expressed by Ship-owners and Ship-
managers who offer ships to charterers for their needs. The most usual
case is that the supply side and the demand side employ brokers to match
their needs. This is the reason why brokers absorb demand information
from Charterers with the one side of the triangle and supply information
from Ship-owners and Ship-managers with the other two sides respectively.
The purpose of this model is only to present the core of shipping activity
around which satellite services revolve and they constitute as a whole a
universe of maritime activities — the maritime cluster. For the moment,
this model is simplified and relieved from complexity issues that certainly
exist in the market. Further on, the addition of satellite services around
the core activity creates a network that can be considered as a maritime
cluster, shown in Fig. 2 below. In order to harmonize the above model with
the definition of maritime clusters we should also consider the factor of
localization. It is therefore essential to enhance the model with the spatial
dimension, meaning that the players of the core activity with the firms that
offer the satellite services are co-locating in a region that can be therefore
characterized as a Maritime Cluster.
Fig. 2 Double inverted triangle, market modeling.

4. Spatial Paradigm: The Greater Area of Piraeus

The greater area of Piraeus is the shelter of a significant number of
companies and organisations that participate in the maritime activity of
the country. This significant number can be roughly estimated at over
1.000 shipping companies, a vast number of technical companies, many
banks with shipping departments and more than 1.200 marine services
companies. Greek shipping has been considered among the most successful
industries in the world accounting for about 4,392 ships, a figure translated
in 8.7% of the world fleet being controlled by Greek owners (GSCC, 2008).
Having in mind that Greece is one of the small countries in the world
with a population of around 11 million people and ranking 96th in total
area out of 231 countries worldwide (Wikipedia, 2009), some important
observations arise. Such observations lead us to assume that there is a
high density of maritime related services in Greece and that there shall
be significant factors that create competitive advantages for the Greek
shipping industry. The high density of maritime activity in a small country
has directed our theoretical framework towards spatial theory. Mapping
maritime related businesses in Greece has resulted in identifying four key
regions that include the vast majority of firms with maritime activity.
The most significant region that will also serve as our case study is the
Greater Area of Piraeus, consisting of the city of Piraeus and seven of its
adjacent suburbs. The other areas mentioned in line of importance are the
Northern Suburbs of Athens, the Southern Suburbs of Athens and the City
of Athens.
The cluster population of the greater area of Piraeus can be analyzed
with the use of firm statistics created from the ‘Greek-Cypriot Maritime
Guide’ (MIS, 2008). All registered firms are included in this dataset. Some of
the firms in the dataset are members of larger groups or subsidiaries created
for economic reasons. Therefore, the number of ‘real’ firms is overestimated.
This shortcoming is not important for the purposes of this study, because
the general picture of Piraeus’s cluster is fairly reliable. The figure below
identifies the importance of each area by the number of companies in each
key sector for the maritime cluster.
The sector of maritime services in the greater area of Piraeus can be
characterized as large and dynamic. It can compete with other key exporters
of maritime services as London, whilst lacking in maturity and organization.
The sector demonstrates cluster behaviour with geographic concentrations
and interconnected companies, while the cluster forces that shall hold it
together, seem weak and vague. It is vital that these forces be strengthened
Chart 1 Sector fragmentation by region.
in the face of international competitive pressures that will try to pull the
cluster apart. The axis of world economic activity is moving eastwards
and competing centres in the Far East are expected to gain in stature
(Lagendijk, 2000).
4.1. The structure of the Piraeus maritime cluster

The maritime industry in the greater area of Piraeus constitutes a complete
cluster. It is composed of three main bodies: shipping, maritime services and
maritime industry. The cluster is also surrounded by research & educational
institutions, governmental bodies, the port & port authority and some
maritime associations.
Figure 3 suggests that the three core segments in the cluster consisting
of services that are directly connected to each other. This network of services
is not abandoned in the marketplace, but it acts in the framework of big co-
operating institutions that are concerned with the quality and the quality of
the services provided. The maritime associations, port authority, research
&educational institutions and governmental bodies not only are part of
the cluster but they also surround it since they can contribute into policy
making. The most important competitive advantages of the cluster are
described and explained in textbox 1 below. Over the years, a variety of
factors has affected the structure of the cluster as described above. However,
there are four significant variables recognized: the agglomeration effects,
internal competition, cluster barriers and heterogeneity (Langen, 2004).
These variables will be later on specifically discussed for our case study.
Fig. 3 Structure of the Piraeus cluster.
Textbox 1. The core segments of the Piraeus maritime cluster.
Shipping : Shipping is the core of this cluster and it is constituted

by owners and operators of all kinds of vessels, e.g., bulk carriers, oil
tankers, container ships, general cargo, gas carriers, reefer ships, fishing
vessels, cruise ships and ferries. The shipping segment is considered to
be the most important in the cluster not only because it is the largest
network of companies, but also because shipping companies are the
most international and fundamental ones in the internationalisation of
the cluster (Wijnolst, 2009). Shipping companies create the excessive
demand in services of high quality, hence stimulating innovation and
creativity in the whole cluster. According to our survey 2009, there are
608 shipping firms in the greater area of Piraeus.
(Continued)
Textbox 1. (Continued)
Shipbrokers: The role of shipbrokers as explained earlier is crucial for

the shipping market and for the cluster. The greater area of Piraeus
hosts approximately 290 companies, composed from small, medium even
large firms with international reputation and branch offices in important
maritime centres.
Marine Consultants (Naval Architects — Surveyors): There
is a massive concentration of technical offices or individual brand
firms specializing in technical consultancy, ship design and surveying in
Piraeus. According to our survey, there are 168 marine consulting firms
active in the greater area of Piraeus, mostly addressing the demand
created by shipping companies located in the area.
Spare Part Suppliers: Firms specializing in spare part supplies
dealing with repairs in the shipping industry of Piraeus. More than 400
firms support the most demanding fleet of the world constantly.
Machinery & Engine Repairs: This segment is constituted of 160
companies, specializing in low cost repairs of machinery and engines; it
is a crucial service for the shipping companies located in the area.
Legal Services: There are a large number of lawyers specializing in
maritime law and consulting in the greater area of Piraeus, this segment
consists of both big firms and individual lawyers, counting over 100
lawyers in the core area of Piraeus.
Banking & Financial Services: Another strong segment of the
cluster. There are over 210 institutions, banks and firms specializing
in financial services for the maritime sector in Piraeus. This includes
local banks and firms as well as representatives of famous international
institutions.
4.2. The economic footprint of the maritime

industry in the region
There are several ways to assess the economic importance of an industry.
Such are employment, profitability, productivity and knowledge externali-
ties. The maritime industry in Greece is large, internationally competitive
and geographically concentrated. These characteristics make it a very
important asset of the Greek economy. The geographical concentration of
the industry, as indicated above, tempts us to assess its economic footprint

on the corresponding region.
The shipping sector, only one segment of the maritime cluster con-
tributes strongly to the Greek Economy. Only for the year 2007, the net
income from shipping was $17 billion, meaning 7% of the GDP covering
28% of the trade balance deficit (World Bank, 2007). The added value of
the maritime sector in Greece according to a report form ‘Policy Research
Corporation’ is 6400 million, that is 3.24% of the GDP of the country.
Added Value is the net output of a sector after adding up all outputs and
subtracting intermediate inputs.
Concerning the employment factor, the maritime industry in Greece
is a substantial employer. There around 76,200 people employed in the
sectors of the maritime cluster. The concentration of the cluster in Attica
and Piraeus represents 43.3% of the total maritime employment in Attica
and 55% of that in the Greater Area of Piraeus.
Source: Policy Research Corporation, Report on Greek Maritime Industry.
4.3. SWOT analysis

The SWOT analysis is a summary of the results coming from a preliminary
survey undertaken as well as a comparison with other prominent clusters
and a review on the perspective of sector experts. On one hand, strengths
(S) and weaknesses (W) can reveal the internal conditions of the cluster
and its current position while on the other hand, opportunities (O) and
threats (T) focus on future growth and suggestions (Wijnolst, 2009).
Fig. 4 Multi-scale cluster organization for ABMS modeling.
• There may certainly be additional weaknesses, threats, etc. other than

those mentioned in Figure 4 such as:
• Deindustrialization of Greece and gradual relocation of industries from
Western Europe to emerging countries with cheaper labor force.
• Relatively expensive Greek labor force compared to emerging countries
wage levels.
• The Mediterranean Sea no longer plays a central role in international
trade as it used to, since it is faraway from BRIC countries.
• Major shipyards are nowadays located in the Far East, very faraway from
Greece.
• Countries such as China are beginning to develop commercial fleet as
well.
• The new international industrial center is in the Far East.
• London is Europe’s major trade center and represents direct threat to
Piraeus, as it attracts so many Greek shipping companies.
Therefore Table 2 must not be considered as comprehensive.

The basic conclusion from the analysis is that Greece is a leading
maritime center, but there are a number of actions and initiatives that
should take place in order to become a maritime centre of excellence, pro-
viding quality maritime services for international customers. The identified
opportunities should be encountered in policy strategies and the threats and
weaknesses shall be balanced out by direct public and private initiatives.
Table 2 Summarized SWOT analysis based on survey & interviews.
SWOT-Strengths
Large number of leading international companies in shipping
The largest fleet of the World
Concentration of diversified maritime services at a specific region
Very high gross earning from shipping & high added value from the entire maritime
industry
Strong networking, trust and family ties between the member of the maritime
community
Competitive taxation for shipping companies
Concentrated knowledge and experience in maritime matters
International reputation and long maritime tradition
SWOT-Weaknesses
Declining Seafarer’s labor force
Lack of R&D and Private Maritime Institutions
Limited number of institutions for maritime education
Lack of awareness of the significance of the maritime industry in the Greek Economy
Lack of Maritime Cluster Organization backed from public and private initiatives
Lack of regulatory framework and initiatives for high capaciy activities such as: Marine
Arbitration, Greek P&I Clubs, Greek Stock Market open to shipping, New financing
tools for shipping
SWOT-Opportunities
Attract more shipping companies and other maritime activities in Greece
Greater government efforts to promote and support Greek maritime interests
internationally
Further support and development strategies for the small enterprises of the Greek
maritime community
Creation of an official Greek Maritime Cluster organization
Support the shipbuilding industry and increase shipbuilding capacity
Increase Research & Development expenditures on maritime matters
Greater focus on quality of maritime services and international competition
Increase networking and intitiatives for collaboration
SWOT-Threats
Competition from countries with high innovation index and advanced R&D activity in
the maritime sector
Insufficient flow of skilled labor into the sector
Lack of integrated maritime policy and strategic plan
Lack of unified safety culture and environmental awareness
5. Methods for Evaluating and Benchmarking

Maritime Clusters
There is an increasing request for a more fact-based input to Cluster
formulation and development. Good practice methods have not yet been
identified in evaluating and benchmarking clusters since the field is still
considered to be relatively new. One of the most recent and robust
methodologies in the literature is the one developed by the European
Cluster Observatory (www.clusterobservatory.eu). This European initiative

has created a Cluster Mapping database that is built in the intersection
of regions and sectors in Europe. The European Cluster Observatory has
developed the so-called 3-Star evaluation method to determine the strength
of clusters. This method utilizes three factors which are the following:
Size of the Cluster (represented by employment share in relation to
total European employment),
Specialization (represented by the proportion of employment in a
cluster category over total employment in the region),
Focus (represented by the share of cluster category in the regional
economy). Further to the classification of cluster strength the European
Cluster Observatory uses two additional cluster performance indicators:
innovation index (based on the based on the Regional Innovation Score-
board 2006) and world export share.
Another very important factor in the development and performance of
clusters is the human capital. Many researchers in the area of economic
growth believe that the growth of human capital is the most important
factor in developing knowledge-based economies and innovative processes.
Padmore and Gibson also emphasize the role of human capital in the devel-
opment of clusters. Their GEM model (Groundings-Enterprises-Markets)
includes factors very similar to Porter’s diamond (Padmore et al., 1998).
The GEM model was created to assess the potency and effectiveness of
a given cluster by scoring each of six determinants from the following
major categories: supply determinants, structural determinants, market
determinants.
There is a wide variety of performance indicators that can be utilized
for the benchmarking of entrepreneurial systems; however for our research
on maritime clusters we have divided them into general cluster sizing
indicators and intra-cluster indicators. The general cluster sizing indicators
are predominately used to exhibit the existence of the cluster in the
area under study and also to rank the cluster in comparison with other
competing ones in terms of size and influence on the local and national
economy. However the general cluster sizing indicators, such as the size of
the cluster in terms of the number of firms or the percentage share in GDP
are not only useful for determining the size of the cluster or for ranking
it against others, they are also essential in order to determine whether
the cluster has reached the specialized critical mass to develop positive
spill-overs and linkages. On the other hand intra-cluster indicators aim to
reveal the performance of internal cluster dynamics that contribute to the

existence and the development of the cluster.
Starting with the general cluster sizing indicators we have excluded
some possible general performance indicators for being simplistic and
non-proper for the aim of our research. These are: productivity and
foreign Direct Investment. Productivity refers to metrics and measures
of output from production processes, per unit of input. It is therefore
a non-applicable measure for dynamic — evolving systems as clusters,
since it does not facilitate measures for capturing the changes of cluster
networking population (Maillat, 1998). Last, Foreign Direct Investment is
not sufficient since clusters are dominated by local players focusing on
outward investments; investments that can be measured by added value
and employment indicators. On the other hand, inward investments can be
more helpful on performance but still insufficient.
After identifying and reviewing the indicators that are not satisfactory
for our study, the general performance indicators, below, are analyzed that
are more suitable for maritime clusters.
5.1. General cluster sizing indicators

The dynamics of clusters depend strongly on structural, economic and social
performance indicators, such as:
Cluster Structure Indicators: The structure of the cluster is
a fundamental element of cluster’s strength. The type and number of
maritime sectors that exist in the cluster play a significant role. The broader
is the cluster in terms of sectors, the greater is the probability of high
networking and cluster effects. There is also a distinction between sectors
in the cluster. Not all sectors have the same importance; for example,
shipping is the most important sector, highly contributing to the added
value generated and also pulling the demand for services within the cluster.
Therefore, maritime sectors are attributed with weights of importance, thus
representing their effect on the cluster in a more realistic way.
Another structure indicator is the population of the cluster and its
sectors. It is critical to enter population barriers for sectors in order to
qualify as members of the cluster. It is believed that a linear relationship
exists between the population of firms in the cluster and the strength of
the cluster.
Economic Performance Indicators: The use of standard economic
performance indicators as used in all markets and economies is reasonable
since these indicators can be used as tools for benchmarking and comparison
against other clusters. Therefore, the following indicators in Textbox 2 are
suggested for measuring the economic performance of maritime clusters.
Textbox 2. Economic performance indicators for maritime clusters.
Direct/Indirect Added Value: Added Value refers to the additional

value of a commodity over the cost of commodities used to produce
it from the previous stage of production. For the delivery of maritime
services, the value added consists mainly of labor expenses, depreciation
and profit before tax. Since this indicator cannot be directly calculated,
for this research, added value is directly linked to employment data and
the added value per person statistic from Eurostat (Policy Research
2008).
% Share in GDP: Gross domestic product (GDP) is defined as the
“value of all final goods and services produced in a country in one year;
it is more simply the total output of a region. Therefore the share that
the Cluster has in the total output of the country is strongly indicating
the importance of clustering, especially if it monitored over a time series,
while the cluster matures.
Growth Rate: Economic growth is the increase in the amount of the
goods and services produced by the Cluster over time. This can be
conventionally measured for our purposed as the percent rate increase
in share in the GDP.
Employment: This is the most stable and significant indicator for the
performance of any business activity. Existing employment data are
assessed and their correlation with the cluster performance and its added
value are evaluated.
Risk Tolerance: It is a measure of how much a company will risk
in order to gain a specified return. This paper strongly supports the
use of risk tolerance as performance indicator for the cluster since we
expect it to be proportional with the level of clustering and geographical
concentration. Clustering is exposed to the effect of risk aversion,
hence seeking greater returns that can substantially increase the firms’
perceived utility.
(Continued)
Textbox 2. (Continued)
Profitability : Profitability can be problematic for measuring the

performance of the cluster since clustering does not necessarily lead to
higher profits of firms in the cluster (Langen, 2004). However all firms are
ultimately driven by profitability and Clusters that their firms gradually
improve the profitability can seen to be stronger and more stable. This
indicator shows the efficiency of the company’s activity.
5.2. Intra-cluster performance indicators

The amount and quality of knowledge circulating and spilling over between
firms located in a cluster is dependent upon the cluster’s size (which is
determined by the indicators described above) the cluster structure and the
intra-cluster dynamics. The later are processes as well as expected outcomes
of the clustering effect. The main intra-cluster indicators are innovation
(which can be measured with many different variants and perspectives),
knowledge diffusion, intra-firm linkages and intra-cluster initiatives (new
entrants-firms, institutions, collaboration schemes).
Innovation & Research Indicators: Innovation is a key factor to
determine productivity growth. The importance of the cluster’s structure
is also present here. The existence of strong maritime services and marine
equipment sectors indicates increased research activity and innovative
spirit. According to many scholars, the more innovative the individual
sectors are, the stronger the cluster becomes as a whole. Furthermore, the
existence of leader firms significantly drives the innovation cycles within
sectors, since they can signal demand and lead SMEs to an integrated
research strategy. For the purpose of our research, we have defined a simple
Innovation Efficiency Index (IEI) that is defined as the ratio of innovation
outputs over innovation inputs. Inputs and outputs concerning innovation
are the sum of specific sub-categories as defined in Table 3.
Calculating the innovation efficiency index is a challenging effort. How-
ever, there are several difficulties that arise from the use of the above factors.
The above data are very demanding due to their rarity. The numbers used
are based on existing quantitative databases, as Eurostat, reinforced by
qualitative data received from questionnaires and interviews from sector
experts. Producing the index requires that all data are normalized and then
summed up, in order to calculate the desired index. The construction of a
synthetic index requires comparability of data (Hollanders, Esser, 2007).
The innovations indicators are incommensurate with each other as several
Table 3 Maritime cluster innovation inputs and outputs. Relevant data retrieved from
(Hollanders, Esser, 2007).
INNOVATION INPUTS IN THE MARITME CLUSTER

Education Attainment Level
(% share of population in the Maritime Cluster aged 20–28 with upper secondary
education in Maritime matters)
Participation in Maritime Seminars and Education
(% share of Maritime Cluster population)
Population having attained Msc or Phd on Maritime Education
(% share of Maritime Cluster population)
Public R&D expenditure for Maritime matters
(% share of Clusters’ Added Value)
Maritime Businesses R&D expenditure for development
R&D expenditures for high technology & manufacturing
Firms that have developed in-house R&D and Innovation
(% share of Maritime Cluster firm population)
Firms co-operating in Innovative products and R&D
Total Expenditures in Innovation for the Maritime Sector
INNOVATION OUTPUTS FROM THE MARITME CLUSTER
Employment in high-tech services
(% share of total workforce in the Maritime Cluster)
Sales of new-to-market products
Sales of new B2B products
Number of scientific publications
(Measured per number of educated employees in the Maritime Cluster)
Number of patents
(Measured per number of employees in the Maritime Cluster)
New entries of companies
(% share of firm population in the Cluster measure over a time span of 5 years)
Firms that have developed in-house R&D and Innovation
of them have different units of measurement. R&D expenditure indicators

e.g., are expressed as a percentage of value added in Maritime Cluster while
other indicators are expressed as share in population of firms or workforce.
There are a number of normalization methods available. In this research we
predict that the use of the two most common methods, standardization (or
z-scores) and re-scaling, shall be the most suitable. The Innovation Index
in therefore is computed as a weighted sum of its normalized component
indicators:
Where Q is the number of innovation indicators.
Q I
q=1 [wq (outq )]
IEI = Q
.I
q=1 [wq (inq )]
Knowledge Diffusion: This is the process of communicating

research, innovations and or knowledge to individuals, groups, firms or
organizations. When firms of the same or adjacent sector are located in
clusters they share a common set of values and knowledge. The significance
of such values and knowledge form a cultural environment in which these
firms co-exist. The nature of this cultural environment can either encourage
or discourage the exchanged of knowledge and problems. The existing liter-
ature (Schrader, 1991) suggests that knowledge diffusion through informal
channels happens as information trading. This refers to informal exchange
of information between employees working for different and even competing
firms. The exchanged knowledge can either be specific and of high value or
general and of low value. To attain data on knowledge diffusion of the
case study cluster we have used interviews and questionnaires. These are
designed to obtain information on the knowledge base of the firms and also
information on the value of transferred knowledge to and from collocating
firms. Preliminary results of our above research demonstrate that leader
firms of the cluster tend to have more developed knowledge base and
stronger knowledge flow with other leader firms in the cluster. After normal-
izing and retrieving all data from interviews and questionnaires the pattern
of knowledge diffusion will also be verified by a simulation with the aid of
agent based modeling which is described in section 4 of this paper. The basic
idea behind the knowledge diffusion simulation is that a sample population
of the cluster is represented by nodes; each node is of equal importance
and has knowledge to supply to one or more of the other nodes. As the
simulation runs the weight of importance of each node changes in relation
to the knowledge demand and knowledge supply functions attached to it.
5.3. On the use of data and analysis for measuring

performance of maritime clusters
Numerous international studies have presented results which indicate that
clusters have a positive impact on innovation and economic growth,
therefore many organizations and even countries and regions have embraced
the concept of clusters and try to develop them through specific initiatives.
Facing the need to back-up the analysis of the maritime cluster of Piraeus, a
range of tools was employed after being adapted to the needs of a maritime
cluster. The methodology as described below is used in order to tackle
the problem of cluster performance analysis regarding our selected case
study.
Cluster Mapping : The first step of this project was the mapping of
the cluster. In a very practical way, mapping of the cluster is classifying the
firms by sector and marking them on the map. The patterns of clustering
are thus very easy to identify. This is visual evidence that the cluster
exists. Geographical proximity is then proved by numbers for each sector
region and sub-region. This task is essential in cluster modeling in order to
understand the practical structure of the maritime community and identify
the regions of competence.
Cluster Database: The following step is the creation of a cluster
database. The database carries information for nearly all the firms in the
cluster. The maritime cluster of Piraeus consists of around 3,000 firms
diversified in various maritime activities. A variety of data is contained,
as location, number of employees, number of ships and number of new-
building orders (in the case of shipping companies), annual turnover (where
available), market share in the sector, patents, publications, education of
employees, average wage etc. Those data are normalized and statistically
analyzed in order to be used as performance indicators for the cluster as a
whole.
Survey/Interviews: The most comprehensive tool, that can shed
light into the dark corners of firms networking patterns, is the use of survey
and much more the use of interviews with sector experts. For the purposes
of our research, the second step after identifying cluster sectors is to identify
the leading firms in the cluster for every sector and proceed with identifying
the experts. All experts are invited to an interview so as to collect personal
opinions about the structure of the maritime community in Piraeus. Except
from firm experts’ interviews applied to experts from organizations, classi-
fication societies, governmental bodies and educational institutes are inter-
viewed as well. The results of the survey and interviews used as qualitative
data are processed by SPSS software in order to identify correlation between
key factors that drive the development of the cluster. The outcome is then
used to feed the computational model and it is analyzed in the following sec-
tion. The use of surveys and interviews is efficient for the localized maritime
cluster of Piraeus, however it would be very difficult for these methods in

order to apply multi-cluster comparison. Nevertheless, surveys and inter-
view are considered an essential tool when studying and modeling a cluster,
since it is vital to identify the consciousness of the major stakeholders.
Concluding cluster analysis can be achieved with several tools and
enables accurate and effective policy and management intervention. It is
essential to have a good understanding of a cluster’s internal workings —
components, structures, processes, routines and development pathways.
6. Computational Methods for Simulation and Life-Cycle

Management of Maritime Clusters
Managing the maritime cluster, retaining and enhancing its competitive
advantages in the context of international competition are complex matters
that lend themselves to dynamic systems and complexity theory. In the
framework of this research, a sophisticated computational model such as
Agent Based Modelling is employed in order to simulate the actions and
interactions of firms that act as autonomous individuals in the maritime
cluster, with a view to assessing their effects on the cluster system as
a whole. The model intends to simulate the simultaneous operations of
multiple agents-firms, in an attempt to re-create and predict the actions
of complex phenomena such as the maritime business environment. Agent–
Based modeling has connections to many other fields; its historical roots
can be traced in the study of complex systems (CAS) and has thereon
extended into techniques and theories such as Cellular Automata, Swarm
Intelligence, Network Science and Social Simulation.
6.1. Agent-based modeling and simulation

The increasing complexity of the world and its systems, calls for man-
agement tools that must be able to capture the whole lattice of their
complexity. Industrial and governmental organizations frequently base their
research and decision–making on fine data organized in the form of analyt-
ical databases. However, there are no robust tools for revealing emerging
patterns from this data. Competitive advantage can be missed without
the use of sophisticated tools. Simulation and Agent–Based modeling
can contribute into assembling patterns from the chaotic interactions of
firms. Agent–Based modeling is used to increase the capabilities of experts
to grasp micro-level behavior and to relate this behavior to macro-level
outcomes (North, Macal, 2007). This technique is based on the notion that
unique rules, parts and components of a system are represented in the form
of individual agents. Agents have varying influence and none of them can
solely determine the ultimate outcome of the system. On the other hand,
every agent contributes to the results in some way.
Implementing computational agents is the next step. Agents are
the decision–making components in complex adaptive systems. They are
attributed with sets of rules or behavior patterns that allow them to
receive information, process them and then reflect them in the environment.
Another characteristic of agent-through information processing is adaption
and learning. Before modeling agents, it is important to understand their
structure as units. Agents are individuals with a set of attributes and
behavioral characteristics. These are explained in textbox 3.
Textbox 3. Carrying characteristics of agents.
Agent Attributes: There are various agent attributes. These are

essentially some key characteristics of the agents that are ascribed by
the user, in order to measure the outcome of the simulation. In an agent-
based simulation, attributes are carried by each agent and can evolve or
change over time as a function of each agent’s learning experiences.
Agent Behaviors: Agents have behavior features that can vary from
agent to agent in order to reflect pragmatic situations. There are two
levels of rules. The first level specifies how the agent will react to
routine events and the second level provides rules for the adaption
of changing routines. Generally, agent behaviors follow three steps:
1. Agents evaluate their current state and determine their actions,
2. Agents execute the actions that they have chosen 3. Agents evaluate
the results of their actions and adjust their rules.
In this case the firms within the cluster are agents. When seeking a
detailed simulation the result is a multi-scale model of cluster behavior
with the smaller scale firm interactions combined to produce the larger-
scale activities of the cluster as a whole.
6.2. Modeling case study: the maritime cluster

of Piraeus
The complex nature of our research, has directed us towards computer sim-
ulation with the use of an agent-based model. After identifying previously
some characteristic of agent-based model, we need to define our model for

the maritime cluster of Piraeus. Firstly, some principal assumptions need to
be considered. Agents in this model represent only firms, organizations and
institutions. Every agent belongs to a sector in the maritime cluster. For the
purposes of the simulation, the population of firms in each sector is in not
full scale; a sample of companies is attributed to each sector. The problem
that this model addresses is to determine the emergence of competitive
advantages for each individual firm within a cluster. In a knowledge based
economy the source of competitive advantage for firms is no more limited to
cost and differentiation advantages but it is linked to resources-competences
that firms possess and their capability to create knowledge (Carbonara N
et al., 2006). The model seeks to investigate if the emergence of knowledge
externalities drives the development of clusters and determines the factors
that control some critical performance indicators for clusters.
All agents-firms have attributes and behaviors. These can change over
time and by sector. After a detailed survey on sector experts, here are the
selected attributes and behaviors for our model (Table 4).
Explaining each attribute, Size: The size of each firm is measured in
accordance with the number of employees that have attained an educational
degree, Knowledge: This is the heart of the model. As explained before,
the long-term growth of firms and regions depend on their ability to
continually develop and produce innovative products and services that are
directly linked to knowledge. Services that are provided and acquired in the
market are here modeled as demand and supply of knowledge. Knowledge is
therefore exchanged within the cluster, with different rate of accumulation
for its firm. Measuring the accumulation of that knowledge can present the
emergence of competitive advantage in firms, Innovation: Is critical to
measure the innovative capacity of each firm and sector. This is a derivative
Table 4 Attribute and behaviors of each agent–firm.
Firm
Attributes Behaviors
Size Knowledge Demand

Knowledge Stock Knowledge Supply
Innovation Learning
Growth Rate Moving in new positions
Risk Tolerance
Market Share Targets
Position on the grid
of knowledge as described above, Position: This attribute indicated the

position of the firm in a dimensional grid. The grid contains all the firms
and the agent by calculating the maximization of his competitive advantage
that depends on the knowledge stock and market share he can acquire; takes
the decision to move or not on a more competitive position in the grid. The
rest of the attributes are described before as performance indicators, that
when attached to each agent they can derive valuable information. The first
experimental stage of the simulation uses a sample population of firms from
all sectors, assuming that they all position in a dimensional grid, having all
the same knowledge capacity but different weight; something that depends
on the firms’ size. Starting the simulation, knowledge is circulated according
to demand and supply. Then, firms try to locate where networking favors
their competitive advantage, from this routine geographical concentrations
arise and clusters of firms are developed. The results from this simulation
are then validated against realistic data from the existing structure of the
maritime community, in Piraeus. This confirms that the assumption of the
initial model was pragmatic, that indeed, in reality, knowledge externalities
drive clustering and that clustering of firms maximizes the performance
indicators chosen. A multi-scale cluster model, as perceived, is shown below,
with firms as subagents, sectors and relating institutions and bodies that
are agents as well.
6.3. Agent-based modeling toolkit

There are a number of toolkits available for implementing agent-based
modeling. Thanks to substantial public and private research, many com-
putational environments have been developed and are now available for
business use without any charge. The software environment for this research
project is Repast (the REcursive Porous Agent Simulation Toolkit) and it
is a leading open-source large scale ABMS toolkit. Repast was developed
in order to support the development of extremely flexible models of agents
focusing on social and economic simulation (North et al., 2007). Repast’s
goal is to represent agents as discrete entities that act as social actors and
are mutually defined with recombinant motives. The broader scope of the
toolkit is to replay cases with altered assumptions (ROAD, 2004).
7. Conclusions
The traditional dynamic of Greek shipping companies and services that
accompany them, together with special circumstances, contribute into
making our era a unique opportunity for strengthening the development of

the Piraeus & Greater Area maritime cluster. This emerging competitive
advantage of the region must be nourished and encouraged.
Nowadays, there are significant opportunities to defend the existing
Greek Maritime cluster formation and organise it, against both cost
pressures and competition. However, in order to utilise such opportunities,
it is essential that all stakeholders act with collective response on a cluster
level basis. Talking about stakeholders, it is essential to identify them
and assign their role and response to the cluster movement. According to
the subject research, one of the major stakeholders is the Public sector
and more specifically the Central and Local Government. Results from
other cluster surveys have shown that the public sector has a major
role in cluster formations. In fact, a supportive government is one of the
most important criteria for the competitiveness of the cluster. Central
government must develop enhanced understanding of the cluster and offer
increased priority and support. This is also implemented in the agent-based
model.
The awakening of the private sector is also essential. The behaviour
of the private sector in Greece, as we know it today, must significantly
change. Companies shall incorporate in their strategies the managerial
theory of the 20th century. Cooperation amongst companies is a must
for improved competitiveness and collective behaviour. Companies can be
more efficient by developing a cross-selling culture in order to grow business
across the cluster as a whole. All stakeholders shall develop a philosophy
of partnership. The public and the private sector shall learn to work in the
framework of a strong funded cluster organisation, pursuing the promotion
of Piraeus as a global maritime services centre. Cluster initiatives and
projects shall be pursued, both by the government and companies. The
maritime identity of Piraeus shall be promoted worldwide and it should
develop an image of offering cost-effective office space for smaller firms and
associations, and opportunities for co-location to maximise cluster factors.
Synergies shall be exploited with other services clusters.
Concluding, the efforts of the central government in these first crit-
ical steps of cluster development shall be based on supporting research
and projects around the cluster and its built up. The results of this
research should then form the basis for structuring public policies and
financial proposals, as tax relaxations and land use for services localisation,
which will favour the emergence of Piraeus as a global maritime services
centre.
References
1. Carbonara, N., Albino, V. and Giannoccaro, I. (2006). Knowledge External-
ities in geographical clusters: An agent — based simulation study. EIASM
Workshop on Complexity and Management, Oxford.
2. European Commission Report (2009). DG Fisheries and Maritime Affairs
studies: “Employment trends in all sectors related to the sea or using sea
resources” and “Employment in the fisheries sector”, http://ec.europa.eu/
maritimeaffairs/study employment en.html, Accessed on 06/02/09.
3. Fisher Associates, (2004). ‘The Future of London’s Maritime services cluster:
A call for Action’, www.fisherassoc.co.uk, London.
4. Greek — Cypriot Maritime Guide, Marine Information Systems, 2009.
5. Greek Shipping Cooperation Committee/Lloyd’s Register — Fairplay,
February 2008 ‘Greek Controlled fleet annually by number of ships’.
6. Hollanders, H. and Esser, F.C. (2007). Measuring innovation efficiency
INNO-Metrics Thematic Paper, PRO INNO Europe.
7. International Standard Industrial Classification, United Nations, 2009,
http://unstats.un.org
8. Krugman, P. (1991). Geography and Trade, The MIT Press, Oxford.
9. Langen, P. (2004). The Performance of Seaport Clusters; a framework to
analyze cluster performance and an application to the seaport clusters
in Durban, Rotterdam, and the lower Mississippi, Erasmus University
Rotterdam.
10. Lagendijk, A. (2000). Learning in Non-core regions: Towards ‘Intelligent
Clusters’: Addressing Business and Regional needs. In Boekerna, F., Morgan,
K., Bakkers, S. and Rutten, R. (Eds.). Knowledge, Innovation and Economic
Growth: The Theories and Practice of Learning Regions (pp. 165–191),
Edward Elgar, Cheltenham.
11. Marshall, A. (1922). Principles of Economics, London.
12. North, M.J. and Macal, C.M. (2007). Managing Business Complexity,
Discovering Strategic lutions with Agent — Based Modeling and Simulation.
Oxford Press, New York.
13. Porter, M. (1990). Competitive Advantage of Nations. MacMillan, London.
14. ROAD (2004). Repast homepage: http://repast.sourceforge.net/ (Accessed
on 08/02/09).
15. Wijnolst, N. (Ed.) (2006). Dynamic European maritime clusters, Rotterdam:
Dutch Maritime Network.
16. Wijnolst, N. and Wergeland, T. (2009). Shipping Innovation, IOS Press,
Amsterdam.
17. Wikipedia (2009). http://en.wikipedia.org/wiki/List of countries and
outlying territories by total ara (Accessed on 09/02/09).
18. World Bank Organization, Data & Statistics, http://web.worldbank.org/
(Accessed on 10/02/09).
19. Padmore, T. and Gibson H. (1998). Modelling systems of innovation: II.
A framework forindustrial cluster analysis in region, Research Policy 26.
20. Schrader, S. (1991). Informal technology transfer between firms: cooperation
through informal trading. Research Policy 20: 153–70.
CHAPTER 7
A PERFORMANCE
EVALUATION STRATEGY
TOWARDS DEALERS IN THE
AUTOMOTIVE SUPPLY CHAIN
Min Chen, Wei Yan and Weijian Mi
Logistics Engineering School, Shanghai Maritime University
1550 Pudong Avenue, Shanghai 200135, P.R. China
[email protected]
Owing to the paradigm shift from the make-to-stock (MTS) to the make-to-
order (MTO), it was imperative to integrate the front-end market information
for automotive supply chain. As a linkage between manufacturers and con-
sumers, the performance of automobile dealers secured a crucial role in the
automotive market-place. Accordingly in this study, a novel performance
evaluation strategy has been developed for the automotive supply chain
regarding four dimensional criteria, i.e., the financial condition, customer
satisfaction, internal processes and self-innovation. More specifically, the
balanced scorecard method was initially applied to evaluate the dealers’ per-
formance. Subsequently, a survey was conducted for the next-step evaluation.
Consequently, the analytic network process (ANP) technique was employed
to analyze the surveyed data. To this end, this strategy assisted automobile
dealers in achieving both short-term and long-term objectives. Compared with
traditional performance evaluation strategies, this approach could eliminate
such disadvantages as time delaying and benefit orientation.
1. Introduction
In China, automobile manufacturers primarily rely on dealers to sell their
automobiles and parts as well as to provide after-sale services to customers.
This forms a partnership between the manufacturers and dealers. Although
the dealers in China operate the 4S processes in a similar way like
those in the Western, their self-fulfilled scale spanning from management
157
158 M. Chen, W. Yan and W. Mi
capability, sale power to public relationship is still lacking. In the domestic

automobile sales system, the vital player in the automotive supply chain
is the manufacturers rather than other partners. As a result, they should
balance and control regional market, establish comprehensive and in-depth
sales network through which they secure the brand reputation and capitalize
the profit (Zhang, 2003).
Pertaining to tougher competitions in today’s market-place, together
with additional pressures from current financial crisis, the paradigm of
automotive industry was shifted from make-to-stock (MTS) to make-to-
order (MTO). This could reduce the inventory and cost among the auto-
motive supply chain. Based on this notion, it was imperative to integrate
the front-end market information for automotive supply chain (Ma, 2007).
As a linkage between manufacturers and consumers, the performance of
automobile dealers secured a crucial position in the automotive market-
place. As such, the performance evaluation became critical towards dealers
(Dreyer, 2000; Chen et al., 2008).
2. Problems of Dealer Evaluation

Dealers secure a crucial position between manufacturers and end customers
for selling the products and providing customer services. Figs. 1 and 2
represent the sales and services process in the automobile supply chain.
In particular, the ‘dealer market information integration’ refers to an
integration of the ‘sales’ and ‘services’ functions in a supply chain, whereas
Automobile Manufacture
Dealers
Enterprise strategy Enterprise strategy
Purchase market Purchase market

production
Automobile sale
inventory Automobil
distribution
inventory es
Human resource Human resource
Fig. 1 Sales process in the automobile supply chain.

A Performance Evaluation Strategy Towards Dealers 159
Automobile Manufacture Dealers

f
Enterprise strategy Enterprise strategy
market market
Parts production Purchase
parts inventory sale parts

Purchase
distribution
inventory
Human resource
Human resource
Fig. 2 Services process in the automotive supply chain.
the dealers’ performance evaluation by manufacturers is aimed at enhancing

their sales and service capabilities.
Currently, the performance evaluation mainly focuses on the accounting
results of dealers, considering fairly about longer-term competence improve-
ment of dealers. This impairs the automobile supply chain as a whole.
There exist several impacts on the dealers’ performance evaluation. In
details,
1. The dealers’ performance evaluation was conventionally emphasized on

financial indices, thus these indices tended to be time-delaying and could
not dynamically reflect the market information;
2. The internal business processes were usually lacking of evaluation and
could not be objectively assessed on the supply chain operation;
3. The self-competence development of dealers was not well addressed
because the competitive edge of entire supply chain was rarely
considered.
These might result in following consequences.
1. Only focusing on the sales quantity rather than the potential dealers’
development and royalty;
2. Inaccurately reporting on client order number, which might affect the
production planning of automotive manufacturers;
3. Intensively emphasizing on new clients rather than existing clients, which
might influence the dealers’ sustainable sales capability.
3. Indicators Definition for Dealers’

Performance Evaluation
To properly address the existing problems above-mentioned, the indicators
definition is conducted by manufacturers. With regard to integrating sales
and service processes in a supply chain, the balanced scorecard method is
developed for dealers’ performance evaluation.
3.1. Balanced scorecard method

In this paper, the balanced scorecard method, which was first coined
by Kaplan and Norton (Kaplan and Norton, 1992) is used for perfor-
mance management. It is related to a full-covering strategic evaluation
indicator system, comprising both financial and non-financial indicators.
More specifically, it connects the organizational performance evaluation
with its long-term vision, mission statement and development strategy;
it converts the organizational mission and tactics into tangible tar-
gets and assessment indicators, so as to link organizational plan with
performance.
3.2. Evaluation indicators definition

By employing the theory of balanced scorecard method (Brewer and Speh,
2000; Ma, 2002; Shi and Cai, 2003), it is revealed that, while strategic
objectives of manufacturers and dealers are defined, their development
strategies should be further decomposed and analysed. In particular,
this is aimed at promoting he profit, responding dealers’ demands, and
improving the sales and services capability. Based on this notion, the
evaluation indicators are postulated from four aspects, namely financial
issue, customer satisfaction, internal process, and research and innovation
(see Fig. 3).
For the purpose of evaluation indicators definition, only the indepen-
dency of dealer’s operation is considered. In addition, it is found that the
overall optimization on the supply chain should also be taken into account
(Brewer and Speh, 2000; Yang et al., 2008; Xiong et al., 2006). In details,
relevant indicators are provided as follows.
1. Financial issues. Due to the independent nature of dealer’s operation,

financial issue cannot be replied much on dealers’ revenue and profit-
related data. From the manufacturers’ viewpoint, the major indicators
Increase
Automobile profit of the
Manufacture manufacture
Strategy
Improve
Improve
service
Dealers
sales capability
Strategy
Single vehicle Sales level

Financial indicators sale profit rate rate
Data retention Customer Customer

Customs indicator test drive rate turndown maintenan
rate
Internal process On-time On-time care On-time

indicators interest follow follow up rate complain
Self- innovation sales service

indicators consultant consultant
Fig. 3 Decomposition of supply chain strategies and evaluation indictors.
for financial issues are associated with the single vehicle sale profit and
sales level rates during dealer’s performance evaluation.
(a) The single vehicle sale profit rate is used to avoid vicious compe-
titions among dealers in lowering the prices, so as to maintain the
market stability.
(b) The sales level rate is used to ensure the dealers’ fulfilment on
the sales plan from manufacturers, so as to accurately make the
production plan and to reduce the inventory of manufacturers.
2. Customer satisfaction. In essence, supply chain management is aimed to
lower the cost of production and servicing by the means of information
and resource sharing among the organizations and participants, so as
to meet the increasingly-changing customer demands on high-quality
products and services. The detailed indicators include the data retention,
test drive, customer turndown and customer maintenance rates.
(a) The data retention and test drive rates are used to measure dealers’
ability to attract new customers and to complete successful sales.
(b) The customer turndown and maintenance rates are used to measure
dealers’ ability to keep old customers and to maintain existing
relationships.
3. Internal process. To meet the expectations of stakeholders and

customers, the organization should implement a process that creates
customer values. Corresponding indicators include on-time interest
follow-up rate, on-time care follow-up and on-time complaint handling
rates.
(a) The on-time interest and care follow-up rates are used to improve the
dealers’ ability for proactive sales and services.
(b) The on-time complaint handling rate is used to urge dealers to settle
customer complaints and to resolve customer problems in time, so as
keep old customers.
4. Research and innovation. It is mainly focused on the organization’s long-

term development. The primary indicators include sales and services
consultant turnover rates.
The aforementioned four categories of indicators are interacted with each

other to ensure the achievement of manufacturers’ and dealers’ strategic
objectives.
4. Dealers’ Performance Evaluation via ANP

4.1. Analytic network process (ANP)
The dealers’ performance evaluation involves the operation of the entire
automobile supply chain. Due to the indicators are inter-dependant with
one another, an effective evaluation method should be investigated and
employed to resolve this problem (Yan et al., 2009a, b). Accordingly, the
analytic network process (ANP) (Qi and Ding, 2006) is applied to analyze
the inter-dependant indicators for final performance evaluation.
4.2. Enabling factors of dealers’ performance

As above-mentioned, from a holistic viewpoint of manufacturers, the
enabling factors of dealers’ performance are related to sales capability and
service capability (see Fig. 4). From the evaluation indicators determined
by balanced scorecard method, four aspects of dealers’ performance can
be obtained, i.e. the financial condition, customer satisfaction, process
efficiency and self-innovation. Each of them is composed of the sub-
indicators, as listed in following table (Table 1).
sales services
• initiatively sale • initiatively service
• meet the customers Dealers’ • Solve the customer
demand in time performance complain in time
• inquire the customers • Meet the customer
demand desire of service in
• feed back the market time
requirement to • Initiatively care about
automobile requirement the customers
Fig. 4 Services process in the automotive supply chain.
Table 1 Sub-indicators of dealers’ performance.
Sales Service
Financial condition Sales level rate

Single vehicle sale profit rate
Customer satisfaction Customer maintenance rate
Customer turndown rate
Test drive rate
Data retention rate
Process efficiency On-time complain handling rate
On-time care follow up rate
On-time interest follow up rate
Self innovation Service consultant turnover rate
Sales consultant turnover rate
4.3. Procedure of dealers’ performance evaluation

The procedure of dealers’ performance evaluation is presented as follows.
Step 1: Establish the judgement matrix. The judgement matrix is formed

to compare the relative importance of lower-lever indicators
against upper-lever indictors. In particular, the ANP adopts Satty
ratings (Qi and Ding, 2006), as shown in Table 2.
1. Weights of two enabling factors. Compare the weights of sales
and services capabilities (denoted by F , viz., F = (f1 , f2 ))
(shown in Table 3).
2. Weight of the four aspects. The weights of four aspects (denoted
by M , viz., M = (m1 , m2 , m3 , m4 )) (shown in Table 4).
Table 2 Definition of judgement matrix.
Scale Definition
1 Same importance
3 Former is slightly more important than latter
5 Former is obviously more important than latter
7 Former is very much more important than latter
9 Former is extremely more important than latter
2,4,6,8 Middle value between above adjacent values
Reciprocal Relative importance of latter against former
Table 3 Weights of two enabling factors.
Sales capability Service capability Relatively weight
Sales capability (f1 ) 1 1/2 0.3333

Service capability (f2 ) 2 1 0.6667
Table 4 Weights of four aspects.
Financial Customer Process Self- Relatively

Enabling factor issue satisfaction efficiency innovation weight
Financial (m1 ) 1 1/7 1/5 1/3 0.0666

Customer satisfaction (m2 ) 7 1 2 2 0.4369
Process efficiency(m3 ) 5 1/2 1 2 0.2979
Self-innovation (m4 ) 3 1/2 1/2 1 0.1986
Step 2: Define the relative weights of indicators. Supposed the weights

of indicators for each aspect be the financial condition, customer
satisfaction, process efficiency and self-innovation in terms of the
sales and services capabilities, respectively. They are represented
as akij regarding two enabling factors and four aspects (where
k = 1, 2; i = 1, 2, 3, 4; j = 1, 2). For example of sales factor,
the weights of four indictors under customer satisfaction aspect
could be represented as a121 , a122 , a123 , a124 , respectively. Similarly
to service factor, the weights of the same indicators could be
represented as a221 , a222 , a223 , a224 .
The relative weights of indicators for each enabling factor can
also be calculated in that manner. For instance, Tables 5 and 6
list the customer satisfaction weights of indicators against sales
and services enabling factors, respectively.
Table 5 Customer satisfaction weights of indicators against sales capability.
Data Test Customer

retention drive turndown Maintenance Relatively
Sales capability rate rate rate rate weight
Data retention rate 1 3 5 9 0.4511

Test drive rate 1/3 1 7 9 0.4344
Customer turndown rate 1/5 1/7 1 1 0.0587
Maintenance rate 1/9 1/9 1 1 0.0557
Table 6 Customer satisfaction weights of indicators against services capability.
Data Test Customer

retention drive turndown Maintenance Relatively
Sales capability rate rate rate rate weight
Data retention rate 1 1/2 2 1/6 0.1161

Test drive rate 2 1 1/4 1/6 0.1082
Customer turndown rate 4 4 1 1/2 0.3008
Maintenance rate 6 6 2 1 0.4749
Table 7 Definition of judgement matrix.
Adjusted
Aspect Indicator indicator

weight (mi ) Indicator weight (akij ) weight (akij )
0.0666 Single vehicle sale profit rate 0.2000 0.0133

0.0666 Sales level rate 0.8000 0.0533
0.4369 Data retention rate 0.4511 0.1971
0.4369 Test drive rate 0.4344 0.1898
0.4369 Customer turndown rate 0.0587 0.0256
0.4369 Maintenance rate 0.0557 0.0243
0.2979 On-time complain handling rate 0.1905 0.0567
0.2979 On-time care follow up rate 0.4762 0.1419
0.2979 On-time interest follow up rate 0.3333 0.0993
0.1986 Sales consultant turnover rate 0.6667 0.1324
0.1986 Service consultant turnover rate 0.3333 0.0662
Step 3: Calculate weights of indicators against each aspect. Upon com-

pletion of calculating the relative weights of each indicator, the
weights of indicators are adjusted regarding different aspects.

Supposed the adjusted weights of indicators be akij , akij = akij .mi .
The results are shown in Tables 7 and 8.
Table 8 Weights of indicators against services capability.
Adjusted
Aspect Indicator indicator

weight (mi ) Indicator weight (akij ) weight (akij )
0.0666 Single vehicle sale profit rate 0.5 0.0333

0.0666 Sales level rate 0.5 0.0333
0.4369 Data retention rate 0.1161 0.0507
0.4369 test drive rate 0.1082 0.0473
0.4369 Customer turndown rate 0.3008 0.1314
0.4369 maintenance rate 0.4749 0.2075
0.2979 On-time complain handling rate 0.5300 0.1579
0.2979 On-time care follow up rate 0.4038 0.1203
0.2979 On-time interest follow up rate 0.0662 0.0197
0.1986 sales consultant turnover rate 0.3333 0.0662
0.1986 service consultant turnover rate 0.6667 0.1324
Table 9 Synthetic weights of indicators.
Aspect weight Indicator Indicator weight Sort
Financial issue Single vehicle sale profit rate 0.0266 11

Sales level rate 0.0400 10
Customer satisfaction Data retention rate 0.0995 5
Test drive rate 0.0948 7
Customer turndown rate 0.0961 6
Maintenance rate 0.1464 1
Process efficiency On-time complain handling rate 0.1242 3
On-time care follow up rate 0.1275 2
On-time interest follow up rate 0.0462 9
Self-innovation Sales consultant turnover rate 0.0883 8
Service consultant turnover rate 0.1103 4
Step 4: Elaborate the impacts of sales and services capabilities on the

enabling factors. By calculating the synthetic weights of different
indicators, it is the sum of weights of all enabling factors multiply
1 ∗ 2 ∗
weights of adjusted indictors, viz., wi = aij f1 + aij f2 (also see
Table 9).
Step 5: Sort the evaluation indicators for dealers’ performance. By apply-
ing the ANP technique, further analysis on indicators should be
conducted using the future business areas for dealers, that is,
improving the proactive follow-up post-sale services to ensure cus-
tomers to choose dealers as their maintenance providers; quickly
handling customer complaints and problems; promoting customer
satisfactions; enhancing the post-sale service team; enabling the
service team to provide customers with consistent and high-quality

services; frequently communicating with potential customers by
actively contacting with them and urging them on test drive so as
to increase the deal rate.
4.4. Method for dealers’ performance evaluation

The bespoke indictors are quantitatively calculated based on actual business
data from dealers. Pertaining to the practical situations, the target values
or the up-limits are defined for each indicator (referring to Table 10).
Table 10 Target rates of indicators.
Indicator Assess target

Aspect weight Indicator weight rate % Definition
Financial issue Single vehicle sale 0.0266 15 the higher the better,
profit rate reach the up limit is
full mark
Sales level rate 0.0400 100 he higher the better,
reach the up limit is
full mark
Customer Data retention rate 0.0995 60 he higher the better,
satisfaction reach the up limit is
full mark
Test drive rate 0.0948 40 he higher the better,
full mark
Customer 0.0961 30 the lower the better, the
turndown rate higher rate , the lower
mark
Maintenance rate 0.1464 60 the higher the better,
full mark
Process On-time complain 0.1242 100 the higher the better,
efficiency handling rate reach the up limit is
full mark
On-time care 0.1275 100 the higher the better,
follow up rate reach the up limit is
full mark
On-time interest 0.0462 100 he higher the better,
follow up rate reach the up limit is
full mark
Self-innovation Sales consultant 0.0883 40 the lower the better, the
turnover rate higher rate
Service consultant 0.1103 30 the lower the better, the
turnover rate higher rate
The full-mark for each indicator is 100. In case of higher marking,

indictor score = indicator rate/assessment target rate ∗ 100, whereas for
lower marking, indicator score = (100% – indicator rate/assessment target

rate) ∗ 100. The final score for dealers’ performance evaluation =
(indicator score ∗ indicator weight).
5. Case Study
By applying the proposed evaluation method, a case study was conducted
based on three large-dimensional dealers for a specific automobile company.
In order to promptly detect the dealers’ operation problems and to consider
the dealers’ business turnover cycle, the evaluation period was set to be
1 month, December 2008 in particular. Table 11 shows the results from
performance evaluation for the bespoke three dealers.
From the results obtained, it could be found that Dealers A
and C performed well, whereas Dealer B lagged behind due to low
scoring, especially in internal process efficiency and employee develop-
ment. This indicated that Dealer B should improve such performances
as pre-sale customer interaction, follow-up post-sale service, and team
management.
Table 11 Results from performance evaluation for three dealers.
Indicator Dealer A Dealer B Dealer C
Financial issue Single vehicle sale profit 2.13 2.02 2.29

rate
Sales level rate 4.00 3.80 3.80
Customer satisfaction Data retention rate 7.96 7.46 7.96
test drive rate 8.15 8.25 8.34
Customer turndown rate 7.50 5.77 7.59
maintenance rate 12.59 11.71 12.00
Process efficiency On-time complain 11.80 10.56 11.67
handling rate
On-time care follow up 12.24 7.65 11.86
rate
On-time interest follow 4.30 3.51 4.57
up rate
Self- innovation Sales consultant turnover 7.06 6.53 7.95
rate
Service consultant 9.82 8.82 10.15
turnover rate
The end performance evaluation result 87.54 76.09 88.19
6. Conclusions
In this study, a front market information integration model was established
for manufacturers’ and dealers’ business. Based on this notion, the dealers’
performance was evaluated from the automotive supply chain viewpoint.
By using the balanced scorecard method, a breaking-down analysis was
first conducted for the strategic objectives of an automotive supply chain.
Subsequently, the evaluation indicators for dealers’ performance were
speculated in terms of four aspects, i.e., the financial condition, customer
satisfaction, process efficiency and self-innovation. Next, the ANP technique
was investigated and employed to obtain the indicator weights for dealers’
performance evaluation. Consequently, the importance of indicators was
sorted. To this end, the final results revealed that this approach assisted
manufacturers to implement a more objective and effective assessment
towards their dealers.
It could be found in this study that
1. compared with other dealers’ performance evaluation methods, this

approach was oriented from a manufacturer point of view;
2. this approach involved a more comprehensive method that concerned
both short-term and long-term goal advancements of dealers; and
3. compared with traditional performance evaluation methods, such as time
lag and short-term profit hunting, this approach focused intensively on
the dealer’ internal performance evaluation.
Acknowledgments
This work was supported by funding from Shanghai Science & Technology
Committee (08ZR1409200), Shanghai Education Committee (09ZZ163,
J50604), and Shanghai Maritime University (20080459).
References
1. Brewer, P. and Speh, T. (2000). Journal of Business Logistics, 75.
2. Cao, J.X., Shi, Q.X. and Lee, D.H. (2008). Tsinghua Science & Technology,
211.
3. Chen, C.H., Yan, W. and Chen, K. (2008). International Journal of
Computer Application in Technology, 298.
4. Dreyer, D.E. (2000). Supply Chain Management Review, 30.
5. Kaplan, R. and Norton, D. (1992). Harvard Business Review, 71.
6. Ma, S.H., Li, H.Y. and Lin, Y. (2002). Industrial Engineering and
Management, 5.
7. Ma, S.H. and Liu, X.G. (2007). Automotive Engineering, 174.

8. Qi, E.S. and Ding, X. (2006). Journal of Harbin University of Commerce, 11.
9. Shi, L.P. and Cai, X. (2003). Journal of Harbin University of Commerce
(Social Science Edition), 84.
10. Xiong, S.G., Yi, S.P., Tang, P., Ping and Liu, F. (2006). Journal of Chongqing
University (Natural Science Edition), 4.
11. Yan, W. Chen, C.H. Chang, D.F. and Chong, Y.T. (2009). Advanced
Engineering Informatics, 201.
12. Yan, W. Chen, C.H. Huang, Y.F. and Mi, W.J. (2009). Computers in Indus-
try, 21.
13. Yang, H.J., Sun, L.Y. and Gao, J. (2008). Industrial Engineering and
Management, 96.
14. Zhang, Y. (2003). Journal of Southwest University for Nationalities (Natural
Science Edition), 199.
PART II
PORTS AND LINERS

OPERATIONS
171
172
CHAPTER 8
A YARD ALLOCATION
STRATEGY FOR EXPORT
CONTAINERS VIA SIMULATION
AND OPTIMIZATION
Wei Yan, Junliang He and Daofang Chang
Logistics Engineering School, Shanghai Maritime University
1550 Pudong Avenue, Shanghai 200135, P.R. China
[email protected]
Container terminals, including shipping and land transportation, secure a

crucial position in container transportation. In particular, container yard
management, which involves diverse operational services, significantly affects
the operational efficiency of the entire container terminal. Therefore, it is
imperative to attain an efficient strategy to support the yard allocation
for export containers. In this paper, a yard allocation model via objective
programming was initially postulated based on a rolling-horizon strategy,
which aims at allotting export containers into yard. Accordingly, the model’s
objective functions were subject to the minimum total distance of container
transportation between storage blocks and berthing locations. This could
balance the workloads amongst blocks. To resolve the NP-hard problem
regarding the yard allocation model, a hybrid algorithm, which applies heuristic
rules and genetic algorithm (GA), was then employed. Afterwards, a simulation
model, which embeds the yard allocation model and algorithm, was developed
to evaluate the proposed system. Subsequently, a case study was used for
system illustration and simulation. Consequently, it could be found from
computational results that this approach paves a venue for resolving the
container yard allocation problem.
1. Introduction
As a hinge of global economy and trade, container terminals play
an important role in a worldwide competition environment. It is well
known that it becomes an imperative to improve operational efficiency of
173
174 W. Yan, J. He and D. Chang
existing container terminals. The temporary storage of import and export

containers, i.e., the storage space allocation problem, is critical to the
container terminal services. The yard allocation for export containers at
blocks affects directly on the moving distance of equipments and turnaround
time of vessels. The container storing or retrieving process is associated with
the time to arrange yard cranes, position and load/unload containers. As a
container should be allotted to or picked up from somewhere in a block,
other containers might be re-marshaled. In case when the export containers
for the same vessel are placed intensively, it is likely more yard cranes are
needed for this purpose. Thus, the workload balance between blocks is
problematic. This results in a higher operating time and handling cost for
yard cranes. Therefore, the rational yard allocation for export containers is
imperative to the stowage planning and equipment scheduling.
Based on these understandings, a yard allocation model for export
containers was developed based on a rolling-horizon strategy of objective
programming. This is aimed at minimizing the total horizontal trans-
portation distance and balancing the workloads among blocks. In order
to resolve the NP-hard problem regarding the yard allocation problem, a
hybrid algorithm, which integrates heuristic rules and genetic algorithm
(GA), was employed. In this respect, the heuristic rules were postulated for
generating feasible solutions, whereas the GA was applied for optimizing
these solutions. Furthermore, a simulation model was deployed for system
evaluation. Finally, a case study on a specific container storage yard was
used for system illustration.
2. Related Work
Up to present, a number of researchers attempted to deal with the problems
concerning the yard allocation problems for a container terminal. Current
research work was focused on import containers, export containers, and a
combination of import and export containers.
2.1. Yard allocation for import containers

Sgouridis et al. (2003) focused on the simulation of import containers that
were transported by trucks. This approach was involved in a medium-
size terminal in terms of an ‘all-straddle-carrier’ system. The simulated
system was proposed for short- or medium-term planning using a process
improvement strategy. Meanwhile, Bish (2003) proposed a strategy for
determining a storage location for each import container, dispatching
A Yard Allocation Strategy for Export Containers 175
vehicles to containers, and scheduling the loading and unloading operations

for cranes. This was aimed at minimizing the maximum time for vessel
serving. Further studied, Wang (2007) postulated a plan-rolling model to
resolve the mixed stocking for stochastic import containers in a container
yard. To balance the travel and queuing time of containers, Cao et al.
(2008) proposed a storage allocation strategy by integrating the integer
programming, genetic algorithm and greedy heuristic algorithm. Similarly,
Chang et al. (2008) synthesized the dynamic berth allocation and yard
planning for import containers, which combined the heuristics algorithm
and simulation optimization.
2.2. Yard allocation for export containers

To enhance the space and loading efficiency of export containers, Kim &
Park (2003) developed a mixed integer programming model based on the
sub-gradient optimization. Alternatively, Kim & Lee (2006) maximized the
efficiency of yard trucks and transfer cranes. Kang & Ryu (2006) postulated
a stacking strategy using the simulated annealing search. Furthermore,
Zhang et al. (2007) proposed an optimization model of intra-bay relocation
for export containers to minimize the total number of re-marshals and
ensure the number of re-marshals within the average range. Meanwhile, Yan
et al. (2008) combined yard crane scheduling and yard planning for export
containers allocation. In this study, the hill-climbing algorithm, together
with the best-first-search algorithm, was employed to resolve the NP-hard
problem.
2.3. Combined yard allocation

Preston & Kozan (2001) minimized the turnaround time of container
vessels. Zhang et al. (2003) proposed a yard allocation model for import
and export containers to balance workloads amongst blocks. This study
minimized the total transportation distance between yard and berth, which
involved such terminal resources as quay and yard cranes, yard space
and internal trucks. Based on these notions, Chen et al. (2004) narrowed
the storage problem down and focused on the central allocation process
to guarantee space efficiency. For the same purpose, Lim & Xu (2005)
speculated a meta-heuristic procedure for yard allocation, i.e., critical-
shaking neighborhood search. Moreover, Bazzazi et al. (2007) deployed a
genetic algorithm (GA) strategy to resolve an extended storage space allo-
cation problem (SSAP) simultaneously for import and export containers.
Similarly, Fu (2007) employed heuristic algorithms for space allocation

through various time, quantity and container size.
Although intensive researches have been investigated into yard allo-
cation, the yard allocation problem has not been well addressed for
export containers rather than import containers. Compared with import
containers, a more complicated problem-solving occurred pertaining to
export containers. Therefore, it is imperative to develop a more effective
yard allocation strategy for export containers. Thus, a comprehensive
approach, along with multiple enabling objectives and constraints, is still
lacking. As a result, a hybrid algorithm, which applies heuristic rules and
genetic algorithm (DGA), was explored in this study.
3. Yard Allocation Modeling for Export Containers

3.1. Problem description
The turnaround time of vessels consists of loading and unloading time
for containers. In order to reduce loading time, the storage locations for
export containers should be selected for loading onto vessels efficiently.
However, rational locations for export containers are cohesively associated
with effective yard planning. In this paper, the number of containers for
each vessel was determined for each block. To minimize turnaround time
and handling cost of vessels, the workloads among blocks were balanced
for each vessel, and the total distance of container transportation between
the storage blocks and vessel berthing locations was minimized. In this
regard, yard cranes in blocks were simultaneously served for vessels, and
the berthing time of vessels was related to the maximal processing time of
yard cranes. In general, the workload balancing on each block for vessels
could reduce the completion time of vessels, and eliminate the traffic jam
of equipments. In other words, the transportation distance between storage
blocks and berthing locations affected directly on the turnaround time of
vessels. If the transportation distance was shorter, the turnaround time
would be less.
3.2. Yard allocation modeling

Owing to the uncertainty information of vessel arriving at port, a decision-
making strategy, which synthetically considers all arriving vessels in 4 days,
is developed using rolling-horizon approach (Zhang et al., 2003). Meanwhile,
a planning horizon of 4 days, each day divided into two 12-hour periods, is
set. At the beginning of the first period, a storage space allocation plan is
Decision-making cycle of planning

period2
P1 P2 P4 P6 P8 P10 P12 P14

day1 day2 day3 day4 day5 day6 day7
Decision-making cycle of planning

period1
Fig. 1 Rolling-horizon strategy.
formed for the 8 periods within days 1–4. Only the plan of the first period
is executed and a new 4-day plan is formed at the end of the first period.
The details are shown in Fig. 1.
(i) Assumptions
The yard allocation model for export containers is developed based on the
following assumptions:
1. The berth, berthing time and departing time of all arriving vessels in
decision-making cycle are known;
2. It can be predicted according to statistics that the number, type and
weight distribution of containers loading/unloading;
3. The number of quay crane scheduled for every vessel is estimated;
4. It accords with historical statistics that the number of export containers
into yard, the retrieved number of containers and the retrieved time.
(ii) Notation
TP The total number of planning periods in a decision-making
cycle.
TP = 8 Planning period is denoted t. Only the plan of the first period
is executed;
NA The total number of blocks in the yard;
P The currently decision-making cycle;
V Pt The set of all vessels needing yard planning in decision-making
cycle P ;
V Pjt Vessel j in period t;
NVP ti The set of vessels which have been allocated in block i before
period t;
Bjt The berthing place of vessel VP jt ;
dij Distance between block i and the berthing place of vessel VP jt
in period t;
N 2jt The number of 20-foot export containers of vessel VP jt in

period t;
N 4jt The number of 40-foot export containers of vessel VP jt in
period t;
Ri The lanes of block i;
Ti The tiers of block i;
OPLj The estimated number of quay crane scheduled for vessel
VP jt in period t Kjt . Up to period t, the periods that
export containers of vessel VP jt having arrived at
container terminal;
STH it The set of start times for all vessels in block i in period t.
STH it = {STH it1 , STHit2 , . . . , STHitn };
ETH it The set of end times for all vessels in block i in period t.
ETH it = {ETH it1 , ETHit2 , . . . , ETHitn };
ST tj The loading start time of vessel VP jt in period t;
ET tj The loading end time of vessel VP jt in period t;
N 2ijtk The total number of 20-foot export containers allocated in
block i that arrive at the container terminal in period
t − k;
N 4ijtk The total number of 20-foot export containers allocated in
block i that arrive at the container terminal in period
t − k;
NU i(t−1) The number of empty bay at the end of period t − 1;
NUB i(t−1) The number of empty 40-foot bay (Two adjacent bays) at
the end of period t − 1;
λ Expansion coefficient.
(iii) Decision Variables



 1, Block i is allocated for vessel V Pjt

 in period t
Aijt =

 0, Block i is not allocated for vessel V Pjt


in period t


 1, Block i is allocated for vessel V Pjt

 in period t − k
Hijtk =

 0, Block i is not allocated for vessel V Pjt


in period t − k
N 2ijt –The total number of 20-foot export containers allocated in block

i that arriving at the container terminal in period t;
N 4ijt –The total number of 20-foot export containers allocated in block
i that arriving at the container terminal in period t.
(iv) Mathematical Models

The decision-making objectives are presented as follows.
1. Minimizing the total distance of all vessels to transport the containers

between their storage blocks and the vessel berthing locations;
2. Balancing the workload among blocks allocated vessel V Pjt in period t;
3. Balancing the workload among all blocks.
The Mathematical models are presented as follows.
TP
NA
f1 = Min (N 2ijt + N 4ijt ) · Aijt · dijt , (1)
t=1 j∈V Pt i=1
Equation 1 is the first objective to minimize the total distance of all

vessels to transport the containers between their storage blocks and the
vessel berthing locations, which synthetically considers all arriving vessels
in 4 days.
 
 Kjt

f2 = Min Max  (N 2ijtk + 2 · N 4ijtk ) · Hijtk
 {i}
k=1

8−Kjt

+ (N 2ijt + 2 · N 4ijt ) · Aijt 
t=1

Kjt

− Min  (N 2ijtk + 2 · N 4ijtk ) · Hijtk
{i}
k=1

Kjt

+ (N 2ijtk + 2 · N 4ijtk ) · Hijtk  , (2)

k=1
Equation 2 is the second objective to balance the workload among

blocks allocated for vessel V Pjt , which minimizes the margin between the
block with the maximum export containers of V Pjt and the block with the
minimum export containers of V Pjt .

  
TP
f3 = Min Max  (N 2ijt + 2 · N 4ijt ) · Aijt 
 {i}
t=1 j∈V Pt
 

− Min  (N 2ijt + 2 · N 4ijt ) · Aijt  , (3)
{i} 
j∈V Pt
Equation 3 is the third objective to balance the workload among all

blocks, which minimizes the margin between the block with the maximum
workloads and the block with the minimum workloads. This is used to avoid
the traffic jam.
Min{ω1∗ f1 , ω2∗ f2 , ω3∗ f3 }, (4)
Equation 4 is a multi-objective function formed via Equations 1, 2 and

3. ω1, ω2 and ω3 are respectively the weights of Equations 1, 2 and 3.
t ∈ T P, j ∈ V Pt , (5)
Equation 5 is a constraint to ensure that the planning period is in

the decision-making cycle and the vessel is needed for yard planning in
period t.
NA

N 2jt = A∗ijt N 2ijt , (6)
i=1
Equation 6 is a constraint to ensure that the total number of 20-

foot export containers allocated in blocks i is the sum of these containers
assigned to all the blocks in period t.
NA

N 4jt = A∗ijt N 4ijt , (7)
i=1
Equation 7 is a constraint ensure that the total number of 40-foot

export containers allocated in blocks i is the sum of these containers
assigned to all the blocks in period t.
Kjt N A 8−Kjt N A

Hijtk + Aijt = 2 · OPLjt , (8)
k=1 i=1 t=1 i=1
Equation 8 is a constraint to ensure that the total number of blocks

allocated to any vessel is two times that of quay cranes scheduled.
∀ t = (1, 2, . . . , 8),
 

λ· N 2jt + 2 · N 4jt  (9)
j∈V Pt j∈V Pt
NA

≤ Aijt · N Ui(t−1) · [Ri · Ti − (Ti − 1)],
i=1
Equation 9 is a constraint to ensure that the allowable stacks of all

blocks will not be less than the total number of export containers of all
vessels in period t. If ∃ j ∈ V Pt and ALijt = 1, the constraint can be
ensured. The constraint is only used for the 20-foot export containers.
As well, a special constraint for the 40-foot export containers should be
given, because the 40-foot containers should be stored in two adjacent bays.

∀ t = (1, 2, . . . , 8), λ · N 4jt
j∈V Pt
NA

≤ Aijt · NUB i(t−1) · [Ri · Ti − (Ti − 1)], (10)
i=1
Equation 10 is a constraint to ensure that the allowable stacks of all

blocks will not be less than the total number of 40-foot export containers
of all vessels in period t.
∀ m ∈ NVP ti , Aijt · [ST Hmt − ETjt ] · [ET Hmt − STjt ] > 0. (11)
Equation 11 is a constraint to ensure that the handling time of

all vessels, for which containers stored in the same block, will not be
overlapped.
4. Yard Allocation Algorithm for Export Containers

Specifically for this study, the heuristic algorithm is used for reducing
feasible solution scale via constraints. Meanwhile, the genetic algorithm
is used for optimizing feasible solution to gain the approximate optimal
solution for yard allocation for export containers.
Meeting the The actual blocks for

Start BLNumjt > 0 vessel V jt before period t
formula (11) N
belong to set RDSet jt
Y N Y
Block i belongs to Randomly selecting

Update the current BLNumjtblocks in
rational block set DSet jt The number of containers
decision-making cycle P. set DSetjt for the selected blocks
belong to set CTNSet jt
N Meeting the
i = i+ 1
formula (6)
Yard planning for all
vessels in period t Y j=j+1
i > NA Meeting the N

N
Gaining every vessel Vjt formula (7)
Y
needing yard allocation in Y j>N
period t and the number N Gaining forecasted N20 jt N
and N40 jt Randomly selected BLNumjt Y
blocks belong to actual allocation
blocks set RDSetjt for vessel V jt
Yard allocation for t=t+1
vessel V jt Randomly planning the
Computing the number of needed
number of containers for
blocks for vessel Vjt
the selected blocks
BLNumjt = 2* OPLj * pt −
t>8
Computing rational block kjt
i for vessel V jt in period t (HBNjt − 2* OPLj * ∑ pm ) N Meeting the Y N Y
m =1 formula (9) and (10)
The end
Fig. 2 Procedure of heuristic algorithm for feasible solution.
4.1. Heuristic algorithm for feasible solution

The heuristic algorithm, which generates feasible solution for yard alloca-
tion for export containers, is shown in Fig. 2.
The heuristic algorithm is involved in five rules provided as follows.
Rule 1: Aijt · [STHmt − ETjt ] · [ET Hmt − STjt ] > 0. This rule ensured
that the handling time of all vessels, from which containers are stored in
the same block, will not overlap.
N A
Rule 2: N 2jt = i=1 Aijt · N 2ijt . This rule ensured that the total
number of 20-foot export containers allocated in block i is the sum of these
containers assigned to all the blocks in period t.
N A
Rule 3: N 4jt = i=1 Aijt · N 4ijt . This rule ensured that the total
number of 40-foot export containers allocated in block i is the sum of these
containers assigned to all the blocks in period t.
N A
Rule 4: λ · ( j∈V Pt N 2jt + 2 · j∈V Pt N 4jt ) ≤ i=1 Aijt · N Ui(t−1) · [Ri ·
Ti − (Ti − 1)]. This rule ensured that the available stacks of all blocks are
not fewer than the total export containers of all vessels in period t.
N A
Rule 5: λ · j∈V Pt N 4jt ≤ i=1 Aijt · N U Bi(t−1) · [Ri · Ti − (Ti − 1)]. This
rule ensured that the available stacks of all blocks are not fewer than the
total 40-foot export containers of all vessels in period t.
4.2. Procedure of genetic algorithm

The procedure of genetic algorithm is presented as follows.
Step 1: Encoding representation. The different planning periods in
decision-making cycle represent chromosomes. The chromosome consists
of four dimensions for the indices regarding vessel, allocated block, num-
ber of 20-foot export containers and number of 40-foot export contain-
ers. The vessel gene is arranged in time sequence arrived at terminal,
i.e., V 1, V 2, . . . , V n. The allocated block gene consists of two numbers,
which are at odd bit and right adjacent even bit, respectively. The genes
of number of 20-foot export containers and number of 40-foot export
containers are composed like the allocated block gene.
Step 2: Population initialization. The first generation population of every
processor is randomly generated with heuristic algorithm (Fig. 2). The
population size n and the number of processors m are set. The sub-
population size of every processor is n/m.
Step 3: Judging individual feasibility (Fig. 2). If the individual is feasible,
it will be held; otherwise, it will be mutated.
Step 4: Fitness evaluation. Supposed that Equations 1, 2 and 3 are unitary.
(i) If Equation 1 is unitary, the minimum and the maximum distance to

transport the containers between their storage blocks and the vessel
berthing locations are set as lmin and lmax , respectively. The unitary
formula is presented as follows.

f1 − j∈V Pt (N 2jt + N 4jt ) · lmin
f1 = ,
j∈V Pt (N 2jt + N 4jt ) · lmax − j∈V Pt (N 2jt + N 4jt ) · lmin
(12)
(ii) If Equation 2 is unitary, the minimum and the maximum imbalance

between the block allocated for some vessel are set as CTN Minj and
CTN Maxj , respectively. The unitary formula is presented as follows.
f2j − CTN Minj
f2 = , (13)
CTN Maxj − CTN Minj
j∈V Pt
(iii) If Equation 3 is unitary, the minimum and the maximum imbalance

among all blocks are set as ALLCTN min and ALLCTN max , respec-
tively. The unitary formula is presented as follows.
f3 − ALLCTN min
f3 = , (14)
ALLCTN max − ALLCTN min
The unitary objective function is defined as f = Min(ω1∗ f1 + ω2∗ f2 + ω3∗ f3 ).
The fitness function of the genetic algorithm is defined as f (k) = 1/(ω1∗ f1 +
ω2∗ f2 + ω3∗ f3 ).
Step 5: Selection strategy. The individual is selected in the mating pool
of parents with ‘roulette wheel’ sampling. The process is presented as
follows.
(i) Calculate the sum of fitness of all individuals in sub-population

(Equation 15). Notation n is the sub-population size.
n

SUM = F (xi ), (15)
i=1
(ii) Calculate the selected probability of individual (Equation 16).
F (xj )
P (xj ) = n i
, (j = 1, 2, . . . , n). (16)
i=1 F (x )
(iii) Calculate the cumulative probability of individual.

(iv) Generate a random number RM between 0 and 1.
(v) If the RM is between the cumulative probability of two individuals, the
individual, which cumulative probability is greater, will be selected.
Step 6: Selection, crossover and mutation are implemented with ‘steady

state’ method. The single-point crossover, which crosses the entire gene
including block, is used for the number of 20-foot export containers and
number of 40-foot export containers. In this paper, a mutation operator,
called gauss mutation, is also used. The number of substitutes is decided
by designating the percentage of replacement to replace each generation.
The selected individual will be brought in the pool of parents for the next
generation.
Step 7: Migration operation. In this paper, the migration operation is
conducted according to the migration number, migration frequency and
topology relation among all populations. The migration frequency and
migration number are set as 3 and 5, respectively. The topology relation

among all populations is set as circularity. The migration individual
will be brought in the pool of parents for the next generation. If the
migration operation is done, every sub-population evolves respectively,
and then goes to Step 3; otherwise, directly go to Step 3. If the stop-
page rules (the maximum elapsed generation) are met, the algorithm
stops.
Step 8: Stoppage rule. The algorithm is terminated regarding the
experimentally-determined maximum elapsed generation.
5. Simulation Model
To evaluate the proposed strategy on yard allocation for export containers,
an effective method is imperative. Thus, it is quite complicated to apply
such mathematical models as operational research. Accordingly, simulation
is an efficient tool to evaluate the performance of complex systems.
Hence, the simulation technology could only evaluate an existing approach
rather than an optimized strategy. Therefore, a combination of simulation
and optimization technologies was proposed in this study, i.e., the yard
allocation model and algorithm.
5.1. Simulation framework

A simulation model was established using eM-PlantTM (Fig. 3). As
illustrated in Fig. 4, this model comprised five interacting modules, namely
Fig. 3 Illustration of yard allocation simulation.

Input module
Control
module
Simulation Output
Key module
framework module
Optimization
module
Evaluation module
Fig. 4 Framework of simulation strategy.
input, control, evaluation, optimization and statistical output modules. In

details, the yard allocation model for export containers was embedded in
the control module, whereas the hybrid algorithm was contained in the
optimization module. Subsequently, the feasible solutions, together with
constraints, were generated via the heuristic algorithm. Consequently, the
feasible solutions were optimized using the simulation-embedded DGA
algorithm. In addition, the evaluation indices were subtracted for the
evaluation module.
5.2. Input parameters

The following parameters should be initialized prior to simulation.
1. The parameters of container terminal layout included the quayside
length, water depth, number of blocks, number of bays, lanes and
tiers of blocks, coordinates of blocks, coordinates and length of tracks,
coordinates of gateways, etc.
2. The parameters of mechanism included the size and velocity parameters.
3. The statistical distribution included the time interval distribution
of arrived vessels, the time interval distribution of external trucks,
distribution of vessel types, distribution of container types, distribution
of containers, etc.
5.3. Simulation process

The simulation process of yard allocation for export containers was
presented in Fig. 5.
System Call time control

initialization procedure
It is time to yard
Wait 6 hours
crane scheduling
N Y
Yard cranes scheduling
Simulation pauses procedure is performed based
on the proposed approach
Simulation goes on
Vessel arrives Inbound / Outbound External trucks

containers are generated are created
Vessel finishes
berthing Yard plans are
generated N
Quay cranes are

deployed
Storage location are Truck with
Y
allocated for contianer container
Quay cranes discharge
containers from vessel
N
Internal trucks come Outbound External trucks

N come into block
into block containers
Y
Yard cranes are deployed
Yard cranes are deployed External trucks according to the scheduling
according to the scheduling come into block results computed by the
results computed by the proposed approach
proposed approach Yard cranes are deployed according
to the scheduling results computed Yard crane retrieves
by the proposed approach container
Yard crane stacks
container onto yard Yard crane stacks
External truck is
container onto yard
N checked at gate
Discharging ends Yard crane stacks

container onto yard
External truck
Y departs
Yard crane
retrieves container
Exist outbound
containers
Quay crane
N loads containers It is time to
Y onto vessel simulation stopping
Quay cranes are
deployed
Loading ends Y
Internal trucks come
into block Y
Simulation ends and
N Vessel departs statistic begins
Fig. 5 Simulation process of yard allocation.
5.4. Statistical simulation indices

The following statistical indices should also be obtained during simulation.
1. The loading time of export containers onto vessels.

2. The total horizontal transportation distance between storage blocks and
berthing locations for loading.
3. The balance among blocks for export containers, i.e., the gap between
blocks with the maximum and minimum export containers.
6. Case Study
A case study on a specific container terminal was conducted. In details,
a continuous 4-day data was stochastically selected (Table 1) to study
the yard allocation for export containers. In particular, the container
terminal possesses 32 blocks, together with 35 bays, 5 tiers and 7 lanes,
handled by RTGCs, where each bay possesses a storage capacity of 31
containers. Meanwhile, there are 4 berths with quayside length of 1,100
meters.
The yard allocation for export containers was optimized via the simu-
lation model. The comparison between optimized results and actual results
was listed in Table 2, while the optimization process was shown in Fig. 5.
In this case, some indices, i.e., the population size, sub-population size,
crossover probability, Gaussian mutation probability, maximum elapsed
generation and percentage of replacement, were initially set as 100, 50,
0.8, 0.05, 40 and 20%, respectively.
As shown in Table 2, the total vessel loading time of 13 vessels improved
8.10% against the actual operation, the total horizontal transportation
distance of 13 vessels improved 9.13% and imbalance among all blocks
improved 35.61%.
The computational results from yard allocation for export containers
were discussed as follows. The allocated blocks of export containers for
vessels were shown in Table 1. Meanwhile, the computational time of hybrid
algorithm was approximately taken for 21 minutes.
Table 1 Data of arrived vessels in four consecutive days.
Vessel TEU Number of Number of Berth of

name 20-foot. 40-foot. load port weight vessel
v01 283 121 9 4 3

v02 639 274 7 3 4
v03 250 107 6 4 2
v04 112 48 3 2 1
v05 289 124 4 1 3
v06 188 80 5 1 1
v07 255 109 10 5 1
v08 497 213 8 2 2
v09 524 224 8 3 3
v10 1112 476 12 5 3
v11 622 266 11 5 4
v12 522 223 7 4 3
v13 491 210 5 3 4
Table 2 Comparison between optimized and actual results.
Horizontal transportation
Vessel Vessel loading (h) distance (km) Imbalance among blocks
name Optimized Actual Optimized Actual Optimized Actual
v01 3.9 4.3 267.8 294.0 3 5

v02 6.8 7.3 547.3 586.1 5 8
v03 4.0 4.1 250.4 264.9 5 8
v04 3.1 3.3 105.4 116.4 5 7
v05 4.3 4.7 254.4 288.4 5 7
v06 4.4 4.7 151.1 168.6 6 9
v07 4.1 4.6 318.8 364.2 5 8
v08 5.5 5.9 489.9 543.9 6 8
v09 4.9 5.2 573.6 613.7 4 7
v10 7.7 8.5 1252.9 1421.5 6 10
v11 5.4 6.0 665.9 720.8 6 10
v12 5.1 5.6 599.7 653.1 5 7
v13 4.2 4.5 575.5 625.0 5 7
To synthetically evaluate computational results via simulation, com-

parisons were conducted based on the results between the actual, manual
and proposed approaches (shown in Table 2). In details,
1. Against the actual operation, the total loading time onto vessels,
the total horizontal transportation distance for 13 vessels and the
imbalance among blocks was improved by 8.10%, 13.37% and 36.19%,
respectively.
2. Through the simulation on one month’s data of the container terminal,
it could be found from the experimental results that the average time of
vessels staying at port reduces 9.2%.
To verify the effectiveness and reliability of the proposed yard allocation

approach for export containers, the numerical experiments with regard to
diverse yard occupation ratios were completed based on the data from
Table 1. In this regard, the yard occupation ratios were predetermined
as 40%, 60% and 80% for three experiments, respectively. Pertaining to
statistical indices, the results were then compared with those from actual
approach.
Tables 3 listed the comparisons of experiments regarding the yard
occupation ratios of 40%, 60% and 80%, respectively. From the results
in Table 3, all indices of bespoke three experiments obtained from the
proposed approach were better than those obtained from actual approach.
Furthermore, when the yard occupation ratio was lower, the improvement
Table 3 Comparisons among three experiments using yard occupation

ratio.
Yard occupation ratio
40% 60% 80%
Statistical indices Improvement Improvement Improvement
Vessel loading (h) 8.10 16.53 23.58

Horizontal transportation 9.13 16.88 26.20
distance (km)
Imbalance among blocks 35.61 37.38 36.06
of total loading time onto vessels and the total horizontal transportation
distance were higher. When the yard occupation ratio tended to be
higher, which was resulted from fewer blocks available, the improvement
of imbalance among blocks might be reduced.
7. Conclusions
The efficiency of container terminals is significantly dependent on the
resource allocation at diverse operational stages. Hence, the resource of con-
tainer yard is costly and can be the bottleneck in container handling process.
Accordingly, a yard allocation model was established for export containers
based on a rolling-horizon strategy. This was aimed at minimizing the total
horizontal transportation distance and balancing the workloads among all
blocks.
In order to resolve the NP-hard problem regarding the yard allocation
problem, a hybrid algorithm, which integrates heuristic rules and GA, was
investigated. In details, the heuristic rules were developed for generating
feasible solutions, while the GA was applied for optimizing these solutions.
Eventually, a case study on a specific container storage yard was used for
system illustration.
Hence, there still existed a space for improvement, e.g., the impacts on
the berth allocation were not well addressed in this approach. As such, the
possible extensions could be adapted to this study as future directions.
Acknowledgments
This work was supported by funding from Shanghai Science & Technology
Committee (08ZR1409200), Shanghai Education Committee (09ZZ163,
J50604), and Shanghai Maritime University (20080459, 20100130).
References
1. Bazzazi, M., Safaei, N. and Javadian, N. (2007). Computers and Industrial
Engineering, In Press.
2. Bish, E.K. (2003). European Journal of Operational Research, 83.
3. Cao, J.X., Shi, Q.X. and Lee, D.H. (2008). Tsinghua Science & Technology,
211.
4. Chang, D.F., Yan, W., Chen, C.H. and Jiang, Z.H. (2008). International
Journal of Computer Applications in Technology, 272.
5. Cheng, P., Fu, Z., Lim, A. and Rodriques, B. (2004). IEEE Transactions on
Automation Science and Engineering, 26.
6. Fu, Z., Li, Y., Lim, A. and Rodriques, B. (2007). Journal of the Operational
Research Society, 797.
7. Kang, J. and Ryu, K.R. (2006). Journal of Intelligent Manufacturing, 399.
8. Kim, H.K. and Lee, J.S. (2006). International Conference on Computational
Science and Its Applications, 564.
9. Kim, K.H. and Park, K.T. (2003). European Journal of Operational
Research, 92.
10. Lim and Xu, Z. (2006). European Journal of Operational Research, 1247.
11. Preston, P. and Kozan, E. (2001). Computers and Operations Research, 983.
12. Sgouridis, S.P., Makris, D. and Anqelides, D.O. (2003). Journal of Waterway,
Port, Coastal and Ocean Engineering, 178.
13. Wang, B. (2007). System Engineering Theory and Practice, 147.
14. Yan, W., Huang, Y.F. and He, J.L. (2008). International Journal of
Computer Applications in Technology, 254.
15. Zhang, Y.W., Mi, W.J., Chang, D.F., Yan, W. and Shi, L.D. (2007). IEEE
International Conference on Automation and Logistics, 776.

CHAPTER 9
INTEGRATION OF AGVS IN
INTERMODAL RAIL
OPERATIONS AT DEEP SEA
TERMINALS
Bernd H. Kortschak
Business Administration & Logistics,
Faculty Business, Logistics, Transport
University of Applied Sciences Erfurt, Germany
[email protected]
Rail links in sea-terminals are often separated by several means: fences,

barriers, special gates, costly haulage to overcome the interfaces. Some earlier
solutions are dealt with — like to put a quai-track underneath the gantry
crane to get direct transfer of containers. It will be shown in the paper that
this solution is not state of the art for various reasons, the most important is
that the gantry crane loading/unloading a 10,000 TEU vessel is not prepared to
move forward and backward just to put the right container on the right wagon.
The solution has to be found by including shuttle trains in the AGV regime
of the terminal for better productivity and sufficient flexibility. The paper will
outline which criteria to be dealt with to achieve that goal of better economic
performance.
1. Introduction
Deep sea terminals face a multiplicity of criteria on port operations, e.g.,
short turnaround time, short dwelling time of containers in the port,
enough stacking areas for peak demand. Different concepts have been
deployed, even spatial differentiation between the deep sea transshipment
193
194 B. H. Kortschak
and hinterland connectionsx in the Agile Port System in the U.S.y In

this paper, the focus lies on the integration of all functions in a deep sea
terminal, i.e., to link all modes within the terminal area; to integrate the
functions despite of spatial constraints. Until now railway links to deep sea
terminals are operated separately, which incurs additional costs for transfer
and thus hinders competition strength of rail versus road.
To cater for growing container volumes, bigger ships demand for a
strong hinterland connection provided by mass volume transport modes,
such as inland waterways or rail. Whereas inland waterways depend
on natural geographical landscape, rail services can be deployed almost
everywhere in hinterland connections given that the volume of containers
can justify the construction cost. Therefore the aim of this paper is to
answer the question on how rail services can be integrated in terminal
operations so as to increase their operation efficiency as illustrated by
Fig. 1.
Fig. 1 Functional areas in deep sea terminals without rail integration.
x Some intermodal transfers take place at one point, while others involve two or
more locations. The former is much more efficient. Muller, G.: Intermodal Freight
Transportation, 3rd ed., Landsdowne VA 1995, p. 144.
y Vickermann, M.J. (1999). Agile Port Concept, http://www.transystems.com.
Integration of AGVS in Intermodal Rail Operations at Deep Sea Terminals 195
2. Earlier Attempts to Address the Problem

2.1. Fixed rail mounted gantry cranes linking ship
to shore crane with stacking area and hinterland
modes rail and road
This approach has been investigated by Howaldtswerke Deutsche Werft
AGz for the Terminal of the Future which can serve vessels with size of up
to 8000 TEU. It is implemented by overlapping yard crane services in a
stacking corridor to enable the linkages between different transport modes.
2.2. NOELL — an approach by K.-P. FRANKE

The main idea behind this logistical concept is to load and unload large
vessels on a reduced area of land with minimum impact on the inland public
traffic system and the environment. In addition,aa the system is targeted on
Fig. 2 The integration of rail with rail mounted vehicle.
z Howaldtswerke — Deutsche Werft, AG et al. (1997). Container-Transport-systeme, der
Zukunft, Projekt, gefördert vom Bundesminister für Bildung, Wissenschaft, Forschung

und Technologie, Förderkennzeichen 18S0071 A, Kiel.
aa Franke, K.-P. (2008). A technical approach tot he Agile Port System, in: Koenings,
R., Priemus, H. and Nijkamp, P. (eds.): the Future of Intermodal Freight Transport
Operations, Design and Policy, p. 137.
196 B. H. Kortschak
Fig. 3 Noell Design by K.-P. Franke of the Efficient Marine Terminal: Direct handling
of containers between vessel and trains.cc
maximizing port productivity by transshipping boxes directly from vessels

to trains and vice versa at the quay (shown in Fig. 3).bb
It incorporates the following features:
• Trolleys and ship-to-shore cranes are able to unload containers to

a platform in the quayside portal, where the twist locks from deck
containers can be removed
• A conveyor to move containers from the lashing position on the platform
to a second position underneath a rail mounted gantry (RMG) cantilever,
which could be extended to provide additional buffer-space. It was
realized years ago by Matson Terminal, Los Angeles.
bb Franke, K.-P. (2008). A technical approach tot he Agile Port System, in: Koenings,
cc Franke, K.-P. (2008). A technical approach tot he Agile Port System, in: Koenings,
• RMGs operate under the portal of the ship-to-shore cranes; and they
could cover four rail lanes and a three-lane wide box mover.
• There are two extra lanes under the lashing platform for the ship-to-shore
cranes.
The big advantage of this new design concept is that yard transfer
vehicles are not required, which could save a great deal of machinery and
labor. When serving the vessel, one duty of the RMG would be to take
containers from the platforms and place them on the linear-motor-based
transfer system or the rail cars on the shortest possible way. The linear
motor lanes could serve additional RMG loading and unloading along the
trains as well as a buffer stack. The linear motor system would allow boxes
being out of sequence to be held aside and shuffled without interrupting
the ship-to-shore import — export cycle. Five to eight RMGs could serve
five ship-to-shore cranes between them.dd
By considering other functions of a terminal into account, the spatial
scheme is given by Fig. 4.
Fig. 4 The functional scheme of the NOELL solution by K.-P. Franke.
dd Franke,K.-P. (2008). A technical approach tot he Agile Port System, in: Koenings,
198 B. H. Kortschak
3. The AGV-solution to Integrate Railway Operations

in Deep Sea Terminals
In 2006, Duinkerken et al. compared multi-trailer-systems (MTS),
automated-guided-vehicles (AGVs) and automated-lifting-vehicles (ALVs).ee
Although ALVs are found to be superior, because they do not have to wait
for cranes to load or unload,ff they are not considered here, because they
need more space to be operated. The multi-trailer system uses manned
trucks pulling trains of five trailers.gg 20 manned trucks and 130 sets of
trains of five trailers have to be operated at least at Maasvlaktehh compared
to Hamburg’s Container Terminal Altenwerder (CTA) where a single trailer
system is operated by using 12 traction units and about 300 trailers. CTA
uses less resources due to the fact that the distance to be covered at
Maasvlakte amounts up to 6 km whereas the rail link at CTA is only less
than 1 km since ITT is close to the stacking area — less than 1 km distance
has to be covered — even if the transfer to the train takes place at the
opposite site of the rail terminal. See Fig. 5 below:
Operating an AGV system directly with the rail cranes could enhance
crane productivity. The longitudinal movements of the crane for correct
positioning of the containers are avoided; thus the loading and unloading
activities can take place with a minimum of transfer time. Furthermore,
the direct access from the rails to the ships and shore cranes enables faster
transshipment times and reduces peak loads at the stacking areas. Hence,
it increases the competitiveness of rails against roads because the duration
for transfer between rails and deep sea ships may be accelerated. The only
difficulty is to find a reserved space for AGVs’ movements to and from
the rail link terminal. If the rail terminal is parallel to the shore besides
the stacking area, one solution might be that AGVs pass by the stacking
area, the dead end of truck, rail terminal area, and move parallel to the
ee Duinkerken,M.B.,Deekker,R.,Kurstjens,S.T.G.L.,Ottjes,J.A.andDellaert,N.P.(2006).
Comparing transportation systems for inter-terminal transport at the Maasvlakte container

terminals, in: OR Spectrum 28, p. 469.
ff Duinkerken, M.B., Deekker, R., Kurstjens, S.T.G.L., Ottjes, J.A. and Dellaert, N.P.
(2006). Comparing transportation systems for inter-terminal transport at the Maasvlakte

container terminals, in: OR Spectrum 28, p. 471.
gg Duinkerken, M.B., Deekker, R., Kurstjens, S.T.G.L., Ottjes, J.A. and Dellaert, N.P.

hh Duinkerken, M.B., Deekker, R., Kurstjens, S.T.G.L., Ottjes, J.A. and Dellaert, N.P.

Fig. 5 Present scheme of operations at container terminal altenwerder source:

Engelhardt, T. and Müller-Elsner, H. (2003). Ein Gigant mit Gehirn, in: GEO, 11,
pp. 114–140, pp. 132–133.
Fig. 6 The integration of rail transfer with AGV’s in deep sea terminals.
Source: Adapted by Kortschak after: Engelhardt, T. and Müller-Elsner, H. (2003). Ein
Gigant mit Gehirn, in: GEO, 11, pp. 114–140, pp. 132–133.
200 B. H. Kortschak
rail terminal area on the outer side of the deep sea terminal. This is one
of possible solutions which could still be deployed at new Weser Jade Deep
Sea Terminal in Wilhelmshaven in Germany.
4. Conclusion
Up to now, the terms of trade between rails and roads in hinterland
traffic are in favor of roads because rail access is not integrated in
deep sea terminals. By extending AGVs to rail mounted gantry cranes
and performing the transshipment of containers on rail flat cars, the
competitive disadvantage of rails could be eliminated. Furthermore, global
warming and climate change require competitive transport chains including
rails for hinterland traffic. Moreover, China plans to build huge railway
infrastructures with links to Singapore. A competitive railway link should
be integrated in the terminal operations. The proposed deployment of AGVs
promises an efficient solution.
References
1. Duinkerken, M.B., Deekker, R., Kurstjens, S.T.G.L., Ottjes, J.A. and Del-
laert, N.P. (2006). Comparing transportation systems for inter-terminal
transport at the Maasvlakte container terminals, in: OR Spectrum 28,
pp. 469–493.
2. Franke, K.-P. (2008). A technical approach tot he Agile Port System, in:
Koenings, R., Priemus, H. and Nijkamp, P. (eds.): The Future of Intermodal
Freight Transport Operations, Design and Policy, 2008, pp. 135–151.
3. Howaldtswerke — Deutsche Werft, A.G. et al. (1997). Container-
Transportsysteme der Zukunft, Projekt, gefördert vom Bundesminister
für Bildung, Wissenschaft, Forschung und Technologie, Förderkennzeichen
18S0071 A, Kiel.
4. Kortschak, B.H. (1992). CARGO NET-Lean Production for Combined Trans-
port, in: CARGO SYSTEMS (ed.) Conference Proceedings, Intermodal 92,
The Hague, pp. 77–78.
5. Muller, G. (1995). Intermodal Freight Transportation, 3rd ed., Landsdowne
VA.
6. Vickermann, M.J. (1999). Agile Port Concept, http://www.transystems.com.
CHAPTER 10
ON THE ONGOING INCREASE

OF CONTAINERSHIP SIZE
Simme Veldman
Ecorys Nederland BV
P.O. Box 4175, 3006 AD Rotterdam, Netherlands
[email protected]
For port planning purposes the development of the size of containerships is

of great importance. Parties involved continuously try to beat competitors by
creating the possibility to accommodate ships bigger than existing ones. With
information available on vessel particulars and new building prices for fully
cellular Post-Panamax container ships reaching up to about 14,000 TEU it
is possible to conduct a proper statistical analysis of economies of ship size.
Results show that economies of ship size, expressed as the elasticity of costs as
a function of ship size, differ only slightly from those of ships up to Panamax.
The assessment of economies of ship size is similar to the one used by
Ryder and Chappel (1979), Jansson and Shneerson (1987) and Cullinane
and Khanna (1999) and (2000) and results show similarities and differences.
Wijnolst et al. (1999) created a push in thinking about the role of much bigger
ships with their Suez-Max and Malacca-Max designs and results are compared.
Economies of ships size act as the motor of demand for bigger ships.
In order not to lead to too high user costs the increase in size has to be in
balance with the combined increase in trade volumes and the number of port
pairs between coast lines to be connected. The conclusion is that the ongoing
increase in ship size will continue.
1. Introduction
The development of the size of containers ships is of great importance
for port planning and the authorities involved try to beat competitors by
creating the possibility to accommodate bigger, more cost effective ships.
Since their introduction the size of containerships was increasing steadily
up to Panamax and beyond. In this development limiting factors with the
201
202 S. Veldman
design of ships such as structural strength, engine capacity, cavitation of

propeller and rudder, cargo handling equipment speed and available depth
in ports were gradually solved and barriers pushed forward; this also applies
to limiting factors in ports and their hinterland with respect to stock
carrying and connecting inland transport.
Costs of capital and energy increase with the size of ship, but less than
proportional and the relation between these costs and ship size can be well
described by multiplicative relationships. The increase in costs as a function
of ship size doubling from 500 TEU to 1000 TEU, then to 2000 TEU and
so on to Panamax size of about 5000 TEU can be well described by these
relationships and in log-linear form tested with regression analysis. This
results in elasticities of ship size per category of costs, measured over a
wide range of sizes of ships, showing remarkable similarities at both the
lower and higher end of the range of ship sizes. See section on economies of
ship size.
Do these economies of ship size also apply to Post Panamax ships,
and to what degree? Results of a statistical analysis show the existence of
economies of ships size for containerships ranging from 800 TEU to 14,000
TEU, of which new-building costs and engine capacity are known and also
for the subsets of ships up to Panamax and Post Panamax separately. See
Section on shipping costs of Post-Panama ships.
Will economies of ship size continue beyond 14,000 TEU? It is argued
here that there are no a priori reasons to believe that the gradual process
will stop with a size of, say, 20,000 TEU. Economies of ship size will keep
the engineers and port planners busy in a stepwise manner, sometimes with
small steps and sometimes with large bold ones, to solve the problems posed
by the limiting factors as they did so far and thereby continue to do so in
the future.
Containerships experience economies of ship size in their hauling
capacity and diseconomies of scale in their handling capacity. To maximise
scale effects it is therefore logic to employ big ships on the longest distance
routes. The combined effect is that that the largest ships can best be
employed on the North Europe — Far East trade route and our analysis
therefore concentrates on the outcome on that route.
With a given volume of trade the employment of bigger ships implies
that less roundtrips are needed to meet demand, leading to a lower
frequency of service and thereby to higher stock carrying costs for shippers
and receivers. This effect has a downward impact on the introduction of
bigger ships.
On the Ongoing Increase of Containership Size 203
An analysis of the introduction of the biggest ships shows that it takes

place at a rate somewhat less than half the rate of the increase of trade,
leaving the other half to an increase in the number of roundtrips. The
introduction is negatively influenced by user costs and by the increase in
ports to be covered per coast line, both requiring more trips to be made.
The introduction is positively influenced by transhipment. See section on
the ongoing increase in container ship size.
2. Economies of Ship Size

2.1. Modeling ship size economies
Ship size economies can be expressed as the shipping costs per unit of
service as a function of ship size. Jansson and Shneerson (1987) developed
an analytical framework based on multiplicative relationships between
shipping cost categories such as capital, labour, energy and ship size and
performance. They brought attention to the existence of economies of ship
size with respect to a containership’s hauling capacity and the diseconomies
of its handling capacity. The ship size economies are based on specific
technical relationships existing for capital related costs (construction,
maintenance and insurance), manning costs and fuel use and fuel costs and
related to the time spend at sea, while the handling capacity relates to the
time spend in port. The statistical analysis hereafter is based on technical
information and historic vessel prices of the existing fleet and order book
of fully cellular containerships from WSE (2008) as it existed per January
2008.
2.2. Capital related costs

Capital related costs are based on the ship’s new-building costs, which are
the basis for ship market prices. Annual capital costs are set equal to the
capital recovery factor (annuity) based on the interest rate and the ship’s
economic lifetime, where the interest rate is the weighed average of the
return on equity and loans. The ship’s price can be expressed as function
of the ship’s size and speed or size only according to:
P = α0 Sα1 (1a)
α1 α2
P = α0 S V (1b)
where the price P is a function of size S, expressed in dwt or in TEU,

and speed V in knots. The Greek letter symbols concern the coefficients of
204 S. Veldman
a multiplicative function and are estimated with regression analysis of the

model in log-linear form. A value of α1 , to be referred to as elasticity, of less
than one indicates that the ship’s price increases less than proportionally
with its size, implying that there are economies of ship size with respect to
the ship’s price and derived from that to capital and capital related costs.
For the regression analysis we use WSE (2008) containership data
containing a great number of price quotations over a longer period of time.
By using vessel prices over different years we have to allow for changes in
time and do so by including yearly dummy variables, where D1 , D2 . . . Dt
refer to a dummy variable for year 1, 2 . . . t compared to an arbitrary base
year 0, for which 2008 is taken:
P = α0 Sα1 Vα2 exp(α3 D1 + α4 D2 . . . αt Dt ) (1c)

In log-linear form this becomes:
Ln(P) = ln(α0 ) + α1 ln(S) + α2 ln(V) + α3 D1 + α4 D2 . . . αt Dt (1d)
A higher design service speed, ceteris paribus, requires a greater engine

capacity resulting in higher building costs. Bigger container ships appear to
have a higher speed and the following multiplicative relationship appears
to do well:
V = β0 Sβ1 (2)
A higher speed implies that there is a correlation between the two

explanatory variables of model (1b), so that the regression analysis might
suffer from multi-collinearity implying that there is a risk of obtaining
erratic values of the estimates.
Ryder and Chappell (1979) did some measurements for containerships
using quotations for a sample of the then existing fleet including contain-
erships ranging from 80 to 580 TEU and published values of 0.48 and 1.38
for the size and speed coefficients respectively according to model (1b). The
value of 0.48 is considerably smaller than one, suggesting the existence of
great scale effects. For model (1a) they present a coefficient of 0.62. They
did not publish goodness of fit indicators such as standard errors or R-
square. The implicit relationship between the speed of ships and ship size
according to model (2) can be assessed by equalling model (1a) and (1b)
and solving for V, which leads to an elasticity of speed with respect to size
according to β1 of 0.19.
Jansson and Shneerson (1987) did a number of regression analyses
for tankers, bulk carriers and containerships. They conclude that the ship
capital cost is proportional to the two-thirds power of the ship size and
apply this value for their analysis of containerships ranging from 240 to
3200 TEU. They did a number of regressions for general cargo ships and
measured values ranging from 0.16 to 0.17 for different samples, which
comes close to the value of β2 of 0.19 as mentioned above.
Cullinane and Khanna (1999) tested model (1a) with 153 observations
of a Fairplay database of container ships ranging from 1000 to 8000 TEU
and estimated a size elasticity of 0.759 with an r-square of 0.93. Ship size
was measured in nominal TEU.
We tested model (1b) for a sample of 1364 ships (to be) delivered in
the period 1991–2012 of fully cellular containerships in excess of 10,000
dwt, which corresponds with about 800 TEU. The ship’s size is measured
in deadweight in order not to be influenced by playing by the owner with
size in TEU. The spread in the size of ships is large enough to do statistical
tests for fully cellular containerships up-to-Panamax and Post Panamax
separately.
The statistical results of the first three lines in Table 1 show that
estimated values of ship size elasticities according to model (1a) are
statistically significant with high t-values for all data sets. For the set of all
ships the value is 0.726 and for up to Panamax and Post-Panamax 0.677 and
0.745 respectively. The r-square values are high and t-values also, except
for the speed variable. For the whole set the speed elasticity is 0.235 with a
t-value of 2.6, meaning that the relation between design service speed and
the ship’s price is positive and significant. This also applies for the subset of
ships up to Panamax with a coefficient value of 0.249 and a t-value of 2.3.
For Post Panamax ships, however, the coefficient value is not significant
Table 1 Statistical test results of container ship price as a function of ship size and
speed.
Ln(dwt) Ln(speed)
Dependent Sample
variable α1 t-value α2 t-value description DF r-square
Ln(Price) 0.726 42.0 0.235 2.6 All ships 1364 0.895

Ln(Price) 0.677 29.8 0.249 2.3 Up to Panamax 1014 0.811
Ln(Price) 0.745 26.3 −0.290 −1.4 Post Panamax 349 0.910
Ln(Price) 0.766 92.4 Not included All ships 1364 0.895
Ln(Price) 0.721 59.1 Not included Up to Panamax 1014 0.810
Ln(Price) 0.733 27.2 Not included Post Panamax 349 0.909
Estimated values of coefficients of yearly dummy variables have been omitted

DF: degrees of freedom
Source: Regression analysis based on WSE (2008) data.
206 S. Veldman
and even shows a negative value. This erratic behaviour may be explained
by the small variation in speed of Post Panamax ships.
If, according to model (1a) we only adopt ship size, the corresponding
size elasticities would be 0.766 for all ships, 0.733 for Post Panamax and
0.721 for ships up to Panamax. The values of the ship size elasticities appear
to be rather close and somewhat smaller for up to Panamax compared to
Post Panamax. All values are close to the size elasticity of 0.759 estimated
by Cullinane and Khanna (1999).
As stated there is a positive relationship between ship size and design
service speed according to model (2). For the total set of containerships
the elasticity is 0.167. For the subset of ships up to Panamax the elasticity
is 0.174. For the subset of Post Panamax ships the elasticity is 0.029. It
can be stated that the speed of ships increases with the size of ships until
about Panamax and starts to be constant from there on. A scatter plot
of size in TEU versus speed in knots is given in Fig. 1 and shows that
the average design service speed of containerships ranges from about 16
knots for 800 TEU ships to 25 knots for up to Panamax ships. From there
on it fluctuates at 25 knots. The fact that the design vessel speed does not
24.0
Speed
20.0
16.0
12.0
2500 5000 7500 10000 12500
TEU
Fig. 1 Scatter diagram design service in knots as a function ship size in TEU based on
WSE (2008) data.
Table 2 Statistical test results of design service speed as a function of ship size.
Ln(dwt)
Dependent variable β1 t-value Sample description DF r-square
Ln(Speed) 0.167 67.2 All ships 1364 0.768

Ln(Speed) 0.174 45.4 Up to Panamax 1014 0.670
Ln(Speed) 0.029 4.9 Post Panamax 349 0.061
Source: Regression analysis based on WSE (2008) data DF: degrees of freedom.
increase any more for Post Panamax ships is reflected in the poor statistical
result for the speed variable for these ships.
For the calculations hereafter on economies of size for Post Panamax
ships we take model (1a) as a basis using the coefficients as estimated
for Post Panamax ships. A ship of 10,000 TEU, which corresponds with
a tonnage of 123,000 dwt, has a price of USD 137 million in prices
of 2008.
Jansson and Shneerson (1987) estimated for three different sets of
general cargo ships with container capacity elasticity values ranging from
0.16 to 0.17. Cullinane and Khanna (1999) estimated an elasticity of 0.192
for a sample of 280 ships with an r-square of 0.90, somewhat higher than
the 0.174 we found for ships up to Panamax.
Cullinane and Khanna (1999) assumed that operational costs such as
costs of repairs and maintenance and ship insurance can be considered to be
a fixed percentage of the ship’s price thereby showing the same economies
of ship size. Hereafter we will apply the same approach.
2.3. Labour related costs

Labour related costs depend on the size of the ship’s crew, the nationality
of the crew and the ship’s voyage patterns. Small ships deployed in coastal
shipping may have smaller crews and thereby lower costs of labour. These
aspects have little to do with a ship’s size. Including small ships Jansson
and Shneerson (1987) measured a coefficient value according to model (3) of
0.03, close to zero; Cullinane and Khanna (1999) took two different amounts
for crew costs for ships up to and in excess of 800 TEU respectively.
L(S) = ε0 Sε1 (3)
In our analysis we concentrate on Post Panamax ships and therefore

can take a fixed amount of USD 400,000 per year based on Drewry (2006).
208 S. Veldman
Fuel related Costs

Fuel costs relate to a ships engine capacity. To attain a certain speed the
horsepower requirement of the engine is less than proportional to ship size,
which advantage is often traded off against higher speed. This effect can
be expressed by a shipbuilding rule of thumb as mentioned by Jansson and
Shneerson (1987) that horsepower (HP) is proportional to the two-third
power of the displacement multiplied by the cube of the design speed. The
equation can be expressed as:
HP = γ0 Sγ1 Vγ2 (4)
where capacity measured in horsepower (HP) is expressed as a function

of ship size S and speed V. Of the set of WSE (2008) data used for the
regression analysis with respect to a ship’s price and speed, in 2% to 4%
of the cases information on horsepower is missing. This small percentage
most probably has little effect on the comparability of the outcome of the
regression analyses on price and engine capacity.
The regression analysis results in Table 3 show that, depending on the
set of ships, the estimated values range from 0.42 to 0.61 for the elasticity
of engine capacity with respect to ship size and from 2.0 to 3.0 for the
elasticity of engine capacity with respect to speed. The r-square value is
the lowest for Post Panamax ships, which as noted before, most likely has
to do with the small variation in speed.
The estimated values of the ship size and speed elasticities are smaller
than those according to above-mentioned shipbuilding rule of thumb with
values of 2/3 and 3. In their study Ryder and Chappell (1979) use values
of 0.48 and 3.13 respectively. Jansson and Shneerson (1978) used a sample
of ships of Zim Navigation Company and estimated an elasticity value of
0.72 for fuel consumption against ship size and did not include speed, as
variations in design speed are minute.
Table 3 Statistical test results of main engine capacity as a function of ship size and
speed.
Ln(dwt) Ln(speed)
Dependent
variable γ1 t-value γ2 t-value Sample description DF r-square
Ln(HP) 0.607 42.6 2.215 29.7 All ships 1330 0.941

Ln(HP) 0.586 33.4 2.008 24.3 Up to Panamax 994 0.902
Ln(HP) 0.417 16.3 2.963 12.8 Post Panamax 335 0.624
Source: Regression analysis based on WSE (2008) data. DF: degrees of freedom.
Cullinane and Khanna (1999) estimated a value of 0.967 for ship size
only. In their sample the speed of ships increases with size according to
an elasticity of 0.192. Correcting for the increase in ship size the elasticity
with respect to ship size comes at 0.78, which is still high compared to the
values in Table 3.
The average consumption of heavy fuel oil (HFO) is 127 gram per brake
horsepower for the 311 ships for with information on both horsepower and
HFO consumption is available.
For the assessment of economies of ship size for Post Panamax ships
we take model (4) as basis by using the estimated values of the coefficients.
A ship of 10,000 TEU, which corresponds with a tonnage of 123,000 dwt
and a design service speed of 25 knots, requires a main engine capacity
of 85,700 HP. We assume that the design service speed can be achieved
with 80% utilisation of the HP capacity. This means that the daily fuel
consumption in tonnes comes at 85,700 × 80% × 24/1000000 = 209 tons.
Cullinane and Khanna (1999) cite a publication of Gilman from 1980 stating
that the daily costs of lubricating oil consumption are about 3% of HFO
consumption. With a fuel price of USD 230 per tonne and 3% of costs of
lubrication oils the daily costs when steaming come at USD 49,500.
Conclusions on Economies of Ship Size for Daily Costs

Comparing the estimated values of daily shipping cost elasticities of Post-
Panamax size ships with ships up to Panamax it can be concluded:
• That for capital related costs the ships size elasticity is slightly higher
or practically equal, meaning that economies of ships size are practically
the same or slightly less.
• That for manning costs the ships size elasticity is also zero, meaning the
economies of ship size are the same.
• That for fuel related costs the ship size elasticity is smaller, meaning that
economies of ships size are greater.
3. Shipping Costs of Post Panamax Containerships

The basic elements of ship size economies relate to building costs, the costs
of energy use and the costs of crews. Other costs such as those related
to the handling of cargo and ships in port and the passage of canals are
considered to be practically neutral with respect to the size of ships. Given
the economies of ship size as estimated with regression analysis shipping
210 S. Veldman
costs are assessed as a function of ship size for ships ranging from 6,000 to
14,000 TEU and extrapolated for ships reaching up to 20,000 TEU.
3.1. Fixed annual costs

The annual capital costs are assessed with a capital recovery factor (CRF)
of 10,19%, which is based on an interest rate of 8% and an economic lifetime
of 20 years. In line with Cullinane and Khanna (1999) costs of maintenance
and repairs, ship insurance and administration are set at 3.5% of the ship’s
price, as it can be argued that these costs experience similar scale effects as
new-building costs. The sum of capital related costs appears to range from
USD 12.9 million for a 6,000 TEU ship to USD 31.3 million for a 20,000
TEU ship. The relation for Post-Panamax ships is based on model (1c),
while the design service speed of ships is expected to be 25 knots for the
whole range.
For Post Panamax ships crew costs do not vary with size and amount
to about USD 400,000 per year. The annual fixed costs by category and
size class are given in Table 4. The daily costs are based on a number of
350 operational days per year.
3.2. Fuel costs

The statistical test of model (4) shows that for Post Panamax ships the
main engine capacity increases with ship size with an elasticity of 0.417.
The design vessel speed of the Post Panamax ships does not increase with
size and is about 25 knots. The engine capacity varies with ship size from
Table 4 Fixed annual costs as a function of ships size in TEU (in USD 1000).
Maintenance, Total Total

insurance capital Manning Total fixed
Size in Capital and admin- related and fixed costs in
TEU Costs istration costs overhead costs USD/day
6,000 9,627 3,308 12,935 400 13,335 38,099

8,000 11,887 4,085 15,971 400 16,371 46,775
10,000 13,999 4,810 18,809 400 19,209 54,884
12,000 16,000 5,498 21,499 400 21,899 62,568
14,000 17,914 6,156 24,070 400 24,470 69,916
16,000 19,757 6,789 26,546 400 26,946 76,987
18,000 21,538 7,401 28,939 400 29,339 83,826
20,000 23,267 7,995 31,263 400 31,663 90,465
Source: Based on outcome regression analysis as described in text.

Table 5 Daily cost at sea and in port (USD 1000).
Main Daily
Daily engine Daily fuel fuel costs Daily fuel Daily Daily
Size in fixed capacity consumption main costs cost in cost at
TEU costs in HP tons engine auxiliaries port sea
6,000 38.1 69,267 169 40.0 1.20 40,4 78.1

8,000 46.8 78,096 190 45.1 1.35 49,5 91.9
10,000 54.9 85,712 209 49.5 1.49 58,0 104.4
12,000 62.6 92,482 226 53.4 1.60 66,0 116.0
14,000 69.9 98,622 240 57.0 1.71 73,7 126.9
16,000 77.0 104,270 254 60.2 1.81 81,1 137.2
18,000 83.8 109,519 267 63.3 1.90 88,2 147.1
20,000 90.5 114,438 279 66.1 1.98 95,2 156.6
69,300 HP for a 6,000 dwt ship to 114,400 HP for a 20,000 TEU ship.
The corresponding fuel consumption with 80% utilisation of the capacity
ranges from 40 tonnes per day for the 6,000 TEU ship to 66.1 tonnes for a
20,000 TEU ship. With a fuel price of USD 230 per tonne the daily costs,
including a 3% allowance for lubricating oil, range from USD 40,400 for the
6,000 TEU ship to USD 95,200 for the 20,000 TEU ship.
3.3. Shipping costs per roundtrip

Containerships experience economies of ship size in their hauling capacity
and diseconomies of scale in their handling capacity. The hauling capacity
is determined by size and sailing speed and the handling capacity by size
and handling speed. The diseconomies of the latter are caused by the fact
that, the larger the size of containerships, the longer the time to be spend
in port. The time spent in port depends on the handling speed, which can
be expressed as:
H = ε0 Sε1 (5)
where handling speed H concerns the number of containers loaded and

unloaded per day. The handling speed depends on factors such as crane
productivity, the number of cranes working the ship simultaneously and
the distribution of containers over the holds. As the speed is assessed over
all ports called at on a roundtrip, some averaging takes place.
Jansson and Shneerson (1987) argued that the handling speed depends
on the number of cranes used which should be proportional with the length
212 S. Veldman
of the ship. Results of regression analysis on WSE (2008) data show that
this would lead to a value of 0.328 for the elasticity of cargo handling speed
to ship size. Cullinane and Khanna (1999) elaborated on the subject, did a
survey under operators resulting in a list of the number of cranes for varying
sizes of ships and apply this for their calculations. Their list corresponds
with an elasticity value of 0.47,a which implies that the cargo handling
increases a bit more than with ship size than as according to Jansson and
Shneerson (1987). For our calculations we take a conservative approach and
apply the elasticity of 0.328.
Ships also need time to prepare for loading and unloading and for
arriving and departure resulting in a fixed component of time per call.
For the assessment of the time ships spend in port the following
assumptions are made:
• Roundtrip distance is 2 × 12,000 nautical miles with an average speed of

25 knots resulting in 40 days spent at sea.
• Per roundtrip 8 ports are called at with a fixed time per port of 6 hours
resulting in 2 days.
• The number of laden and empty containers aboard is 90% of the carrying
capacity resulting in 2 × 90% × 20,000 = 36,000 TEU for the biggest ship.
• The number of containers loaded and unloaded per roundtrip is based on
a TEU/box ratio of 1.6 resulting in 36,000 × 2/1.6 = 45,000 boxes.
• The time needed for loading and unloading of containers for all ports
depends on the product of the number of cranes working simultaneously
and crane productivity. The average number of cranes working is put at
4 cranes for a 6,000 TEU ship and is assumed to increase proportionally
with the ship’s length leading to 5.9 cranes for the 20,000 TEU ship.
Average crane productivity over all ports is put at 30 boxes per hour.
For a 20,000 TEU containership requiring to have 45,000 moves per
roundtrip, the time spend loading and unloading is 45,000/(5.9 × 30 ×
24) = 10.6 days.
Per ship size class the roundtrip time is given in Table 6 and increases
from 46.7 days for the 6,000 TEU ship to 52.5 days for the 20,000 TEU
ship. The total shipping costs of a roundtrip increase from USD 3.4 million
to 7.6 million and per TEU carried the costs decrease from USD 314 to
USD 207.
a The values presented in Fig. 4 of their article were measured from the graph and the
elasticity value was measured with regression analysis.

June 14, 2011
11:46
9in x 6in
On the Ongoing Increase of Containership Size
Table 6 Cost and operational data per roundtrip.
No. of Total Containers

Time at No. containers Handling Fixed Total Roundtrip costs in shipped Cost
Advances in Maritime Logistics & Supply . . .

Size in sea in LOA in cranes handled time in time in port time time in USD in TEU per
TEU 1,000 days metes working 1,000 days days in days days 1,000 1,000 TEU
6 40 291 4.0 13.5 4.7 2 6.7 46.7 3,448 10.8 314

8 40 320 4.4 18 5.7 2 7.7 47.7 4,131 14.4 282
10 40 344 4.7 22.5 6.6 2 8.6 48.6 4,771 18 260
12 40 365 5.0 27 7.5 2 9.5 49.5 5,381 21.6 244
14 40 384 5.3 31.5 8.3 2 10.3 50.3 5,969 25.2 231
16 40 401 5.5 36 9.1 2 11.1 51.1 6,540 28.8 222
18 40 417 5.7 40.5 9.8 2 11.8 51.8 7,099 32.4 214
20 40 432 5.9 45 10.5 2 12.5 52.5 7,647 36 207
b1185-ch10
213
214 S. Veldman
Conclusions on Economies of Ship Size on Long

Distance Trade
The statistical analysis shows that ship size elasticities of Post Panamax
ships with respect to daily costs are 0.733 for capital related costs, 0.0 for
labour and 0.417 for fuel. These elasticities have to be combined with the
operational performance expressed in the elasticity of time spend at sea
(0.0) and the elasticity of time spend in port (0.33). The combined effect
results in total roundtrip costs as a function of ship size.
In order to compare the resulting economies of ship size with the
outcome of other studies, we use the elasticities of shipping costs as a
function of ship size defined as (δc/c)/(δs/s), where c represents costs and s
size. The total cost function is not multiplicative, so that the elasticity value
varies with s and, in fact, decreases as the size of ships increases, which
reflects the increasing importance of time spend in port. The elasticity
values therefore apply to certain ship size intervals, for which it is assumed
that the function is a multiplicative relationship according to:
TC = ζ0 Sζ1 (6a)
TC/S = ζ0 Sζ1 /S = ζ0 Sζ1−1 (6b)
where TC concerns the total costs of a roundtrip and TC/S the cost per
TEU carrying capacity and, given a fixed load degree for all ships of the
range, per TEU carried.
The value of the costs elasticity is 0.65 for the whole range of ships and is
close to the economies of ships size with respect to capital costs, but higher
than those for energy and crew costs. The number of containers carried is
proportional to the size of ships, so that the costs per TEU carried have on
average an elasticity with respect to ship size of ζ1 − 1 = 0.65 − 1 = −0.35.
This means that a 1% increase in ship size will lead to a −0.35% decrease
in shipping costs per TEU. Over the whole range the value of the elasticity
varies from −0.38 for ships of 6,000 TEU at the lower end to −0.30 for
ships of 20,000 TEU at the upper end of the range. Note that if the cargo
handling efficiency increases in the course of time, the diseconomies related
time spend in port becomes less important, which leads to greater economies
of ship size.
From Fig. 2 can be seen that the shipping costs of the time spend at
sea (hauling costs) decrease as a function of ship size and the cost of the
time spend in port (handling costs) increase as a function of ship size. The
diseconomies of ship size of the handling capacity practically appear not
350.0
300.0
250.0
USD/TEU
200.0
Cost at sea
150.0 Cost in port
100.0 Total costs
50.0
0.0
00
00
00
0
00
.0
.0
.0
6.
10
14
18
TEU
Fig. 2 Cost at sea, in port and total costs as a function of ship size.
to play a role for containerships ranging from 6,000 to 20,000 TEU with
a roundtrip distance of 2 × 12,000 nautical miles. For the range of ship
sizes studied the share of port time increases from 14% to 24% of the total
roundtrip time, so that the diseconomies of the hauling capacity have little
negative impact on economies of scale.
Jansson and Shneerson (1987) calculated shipping costs as a function of
ships size for container ships ranging from 240 to 3200 TEU based on their
assumptions on ship size economies as discussed above. At the time they
wrote their book port productivity was substantially lower than now, so
that on a roundtrip of 2 × 10,400 nautical miles the time spent in port was
rather long. For the range of ships they studied the economies of scale as
experienced with capital, fuel and labour costs after a certain size of ships
were more than compensated by diseconomies experienced with the time
spend in port. It appears that economies of ship size exist for the range of
ships from 240 to 1800 TEU and correspond with an elasticity of ship size
of −0.34 for ships at the lower end. From 1800 TEU on costs are practically
constant turning in a slight increase per TEU.
Economies of scale as derived from the shipping costs as calculated by
Cullinane and Khanna (1999) for a long distance route of 11,000 nautical
miles for the range of ships from 1000 to 8000 TEU vary from −0.42 to
−0.20 with an average of −0.25 for the whole range.
Wijnolst et al. (1999) studied comparable shipping costs for the route
Singapore — Rotterdam for existing container ships and two new designs
named Suez-Max and Malacca-Max. The related elasticities of ship sizeb
b Based on a multiplicative relationship between two consecutive cost observations.

216 S. Veldman
appear to be −0.25 for the range from 10,000 to 15,000 TEU and −0.35 for
the range from 15,000 to 18,000 TEU.
4. The Ongoing Increase of Containership Size

4.1. Development of ship size and trade
A certain degree of urgency is needed to stimulate technological improve-
ments intended to realise economies of ship size. Expected high increases in
demand of containerships stimulate ship owners to order, ship engine and
crane builders to innovate and port authorities to accommodate for bigger
ships. All this together makes it possible to realise economies of ship size.
Since their introduction in the late sixties the size of containerships is
increasing gradually. Figure 3 shows that the size of the biggest ship in
service increased from 4,600 TEU in 1988 to 12,500 TEU in 2006. If the
ships on order as in January 2008 are included, the size of the biggest ships
will increase further to 14,200 from 2011 onward. The increase in size of
the biggest ship in operation shows some steep increases by trendsetting
ordering by operators such as Maersk. The figure shows that the increase
in size has a more gradual pattern, if the criterion for “biggest ship” is
measured as the average size of the 2% of the biggest ships of the total fleet
of fully cellular containerships.c
From 1988 to 2007 the size of ships, measured as the average of the
biggest 2%, increased annually with 3.8%. The increase was 3% during first
half and 4.8% during the second half of the period. An explanation of the
lower rate in the first half most likely has to do with the impact of the
Maximum 2% Biggest ships
15000
10000
TEU
5000
0
88
90
92
94
96
98
00
02
04
06
08
10
12
19
19
19
19
19
19
20
20
20
20
20
20
20
Fig. 3 Size of biggest ship and average size of 2% biggest ships per year.
Source: Derived from WSE (2008) Data.
c Note that the due to cancellation of orders the actual average size from 2008 on will be
lower.
Panama Canal on ship size, when ordering Post Panamax ships gradually
got momentum. Accelerations in the increase in ship size are from 1996 to
1998, from 2004 to 2006 and, if the order book as at January 2008 is taken,
also from 2010 to 2012.
The North Europe — Far East trade route has the longest distance
between the world’s major industrial areas and thereby is the one where
the advantage of the biggest ships is best utilised. The 2% biggest ships in
2007 concern 87 units, which correspond with about 11 strings (defined a
liner service providing one roundtrip per week) on the North Europe — Far
East trade route assuming that 8 ships are needed to produce a string. For
the assessment of the average TEU of the top 2% ships we took the fully
cellular container fleet as per 1/1/2008 and on order as basis. This means
that numbers of shipped scrapped between the year concerned and 2008
are disregarded. This means that the real number of ships of the fleet in
that year is greater and thereby also the stated number of ships in the top
2% range. Less ships means that the calculated average will be somewhat
higher than as in reality. It may be expected that this effect is little given
the fast increase in the fleet of ships.
Figures compiled from Drewry (1992), Drewry (2004) and Drewry
(2007) show that the volume of containers carried on the North-Europe —
Far East trade route increased from 2.9 million TEU in 1990 to 13.5 million
TEU in 2007, which is in annual average increase of 9.6%. Over the same
period the increase in size of ships was 4.3%.
An analysis of information of shipping services as published by Drewry
(1992) shows that in 1992 the equivalent of 15 weekly services were offered.
Maximum 2% Total Far East Trade
500
400
300
Index
200
100
0
90
92
94
96
98
00
02
04
06
19
19
19
19
19
20
20
20
20
Fig. 4 Index volume North Europe — Far East trade route and average size 2% biggest
containerships per year.
218 S. Veldman
Total Far East Trade Eastbound Westbound
16000
14000
12000
1000 TEU
10000
8000
6000
4000
2000
0
90
92
94
96
98
00
02
04
06
19
19
19
19
19
20
20
20
20
Fig. 5 North Europe — Far East Trade route volumes eastbound, westbound and total
per year.
Source: compiled from Drewry (1992), Drewry (2004) and Drewry (2007).
Similar information by Drewry (2007) shows that there were 34 services

offered in 2007. The corresponding annual growth rate is 5.6% annually.
The combined annual growth factor from those roundtrips and ship size
comes at (1.056 × 1.043) = 1.10.
This annual increase of 10% per year corresponds rather well with the
increase in total trade volume of the North Europe — Far East trade route,
which is 9.6% for the total volume and 10.8% for the trade of the heaviest
link, i.e., the westbound trade.
4.2. Factors limiting economies of ship size

The economies of ship size as measured in the preceding Section relate to
the size of container ships up to about 14,000 TEU and are extrapolated.
As in the past this extrapolation will meet limitations with respect to
the possibility to continue with a single engine, with increasing cavitation
problems of propeller and rudder, with trading draft in ports, with the
dimensions of the New Panama Canal and, in general, with sufficient
demand to fill the ships. These problems already existed in the past and
have gradually been solved for ships up to 14,000 TEU.
4.2.1. Structural design, engine and cavitation problems

In a technical paper on the state of the art of big containerships Tozer
(2006) states that there are no insurmountable problems with respect to
structural design. With respect to problems associated with a single engine

and cavitation problems with propeller and rudder Tozer (2006) mentions
that propeller design will provide an upper bound on ship size or speed,
rather than the availability of main engines to provide the necessary power.
From the perspective of ship design it is at this stage not possible to assess
a certain upper limit from where on a double set of engines and propellers
are needed.
4.2.2. Trading draft

Draft in seaports is often put forward as a crucial barrier for economies
of ship size. Tozer (2006) states that the trading draft of container ships
generally is less than the maximum draft, as for the important trades the
maximum weight as dictated by the market is up to about 90% of the
maximum. An analysis of the relation between maximum draft and ship
size for Post Panamax ships show that the increase in draft as a function
of size is less than as for ships up to Panamax. The largest Post Panamax
ships have a maximum draft of 16 meters for the biggest ships of Maersk.
This effect is also reflected in the value of the elasticity of maximum draft
with respect to ship size in TEU, which has a low value of 0.168 for Post
Panamax ships against 0.272 for ships up to Panamax. The effect suggests
that for the bigger ships a kind of shallow draft design applies intended to
better deal with problems of draft.
The “shallow draft” design effect as measured for Post Panamax ships
from 6,000 TEU to about 14,000 TEU can be extrapolated to the size class
of the Malacca-Max ships of 18,000 TEU. Assuming that the shallow draft
design will continue to exist to the same degree, we will continue to apply
a draft versus size elasticity of 1/3. This would result in a maximum draft
of about 16 × (18,000/13,000)∧ (1/3) = 17.8 meters instead of 21 meters as
designed for the Malacca-Max containership.
It should also be noted that the maximum draft of 16 meters will
not be reached often, given the difference between maximum draft and
trading draft on the major east-west trade routes. An example as to how
far the trading draft may differ from the actual draft in a systematic
manner concerns the allocation of the MSC Beatrice of 13,800 TEU with
a maximum draft of 14.5 meters on a service calling at Antwerp with a
considerably lower maximum depth.d
d Containerisation International Monthly, May 2009, p. 29.

220 S. Veldman
Fig. 6 Draft versus ship size.

Source: WSE 2008 data.
Table 7 Existing and new Panama Canal restrictions for ships passing.
Existing Panama New Panama

Vessel dimensions (in meters) Canal Canal % increase
Length over all (LOA) 295 426.7 45%

Beam 32.2 55.9 74%
Draft 12.0 18.3 53%
Source: Wikipedia (2009).
4.2.3. The New Panama Canal

In the past the design of containerships was strongly influenced by the
limitations of the Panama Canal with dimensions as given in Table 7. Will
this happen again and how will this influence the design of containerships?
With a maximum of 32.2 meters the vessel’s beam posed the greatest
restriction rather than its LOA maximum of 295 meters. With a beam
restriction of 55.9 meters the New Panama Canal (NPC) allows for the
passage of ships with a 74% greater beam. At present the maximum size
of a Panamax containership given the Panamax beam restriction of 32.2

meters is about 5,000 TEU.
With a beam of 56.6 meters the biggest ships of Maersk are just below
the maximum beam of 55.9 meters. The biggest ships on order by companies
other than Maersk range in size from 13,300 to 14,200 TEU and all have a
beam of less than 50 meters and a LOA of 366 meters. These observations
suggest that the maximum TEU of ships with dimensions optimised with
respect to the NPC are considerably greater than 15,000 TEU, although it
is not known how much bigger.
Of the main world trade routes the deployment of ships on the TransAt-
lantic and TransPacific is affected by the dimensions of the Panama Canal.
These restrictions lead to use of land-bridges extending the TransAtlantic
to the west of North America and the Transpacific to the east part. This
will happen again with the New Panama Canal and what will be the size
of the NewPanamax design? This will affect the increase in the size of
containerships.
4.3. A balance between user and producer costs

In the previous Section the conclusion was made, that economies of ship
size as they exist for capital, fuel and crewing costs for ship up to Panamax,
continue for Post Panamax ships. The combined effect of economies of ship
size of the hauling capacity and diseconomies of ship size of the handling
capacity results in a elasticity of shipping cost per container shipped on
the Europe — Far East trade of −0.35, i.e., a small increase in ship size of
1%, leads to a reduction of shipping costs of −0.35% and being a bit less
for the ships of about 20,000 TEU. It was further concluded that technical
limitations with respect to increases in ship size beyond the size of the
presently largest ships of about 14,000 TEU most probably are still elastic,
as they were in the past. It is, however, unsure how far that goes.
The optimum size of ships is not only determined by shipping costs
as incurred by the ship operator, the producer of shipping services. The
costs incurred by the users of these services, the shippers and receivers
of cargo, also may play a role. Jansson and Shneerson (1987) state, that
the optimum size of ships on a trade is determined by the sum of user
and producer costs. Producer costs are the costs needed for the production
of shipping services as discussed before, while user costs concern all other
logistic costs incurred by the user of a shipping service such as stock carrying
costs. The employment of bigger ships results in less roundtrips to be made,
222 S. Veldman
given a fixed volume of trade. This leads to a longer time span between
two consecutive sailings resulting in more storage costs incurred by users.
This aspect is one of the factors limiting the size of ships on many routes:
employment of bigger ships leads to lower producer costs, but the users may
not like the resulting lower frequency of service. These effects are referred
to as economies of ship size of producer costs and diseconomies of ship size
related to numbers, i.e., related to the number of roundtrips.
Veldman (1993) shows on the basis of cross-section data for 1987 that
the average size of ships employed on the shipping routes connecting the
world maritime regions can be explained by trade level and sailing distance.
The size of containerships employed on secondary container routes is
considerably smaller than the maximum size of ships and the cargo volume
offered on these route is not sufficient to load big ships with a sufficiently
high frequency of service. In other words, the users of the shipping services
on these routes demand a frequency of service which provides a sufficiently
high load factor for the ships employed.
As mentioned above Jansson and Shneerson (1987) worked out an
example of a liner shipping route demonstrating the trade-off between
economies of ship size at sea and diseconomies of ship size in port for vessels
ranging from 240–3200 TEU. With the low cargo handling efficiency of that
time the optimum ship size was about 2400 TEU. By adding stock carrying
and interest costs the minimum of the total of user and producer costs is
reached for much smaller size classes. Veldman (1993) gave an example of
the sum of user and producer costs for the Europe — Far East trade route,
demonstrating how the total cost curve is U-shaped for the interval of ships
ranging from 1000 to 4500 TEU and varies as a function of the value of
trade.
The issue of stock carrying costs plays a role in cases where the number
of roundtrips offered is below what the market is asking. What is the
market? An operator investing in new larger tonnage in order to reduce
its costs to become more competitive has to pay attention to the frequency
of service he is offering for the port pairs connected by his service. This
means that most likely he will not deploy, given his market share, much
bigger ships so that he has to offer fewer services. Instead, he will instead
offer at least the same number of services/roundtrips, with bigger ships.
Then, the increase in ship size will reflect his opinion about the increase of
the total market and his market share therein. This means that in a fast
growing market he will take bigger steps than in a slow growing one. The
big new-building bonanza of the last years is proof of this.
4.4. Factors affecting development of user costs

4.4.1. Numbers of ports being connected
The shipping services offered on a trade route are a mixture of hub-and-
spoke operations and multi-porting operations. On the Europe — Far East
trade route ships may call at about 5 ports at the European end and at
6 ports at the Asian end. As at the Asian end they may call at some ports
twice, inbound and outbound, the actual number of ports called at in Asia
may rather be 5. The itinerary patterns of course differ per operator, some
of who rely more on transhipment than others.
OSC (2006a) and OSC (2006b) present historic port throughput figures,
which can be aggregated by coastal areas. From these figures it appears that
the number of ports for the coastal areas served by the Europe — Far East
trade route, there is an increase in the number of ports. A broad analysis
suggests that over the period 1990–2006 the increase was about 50% per
coastal area. These ports include mainline, feeder line and regional ports
and it is likely that most of them are linked to the worldwide container
network. Roughly, the number of ports connected increase with a factor
1.5 × 1.5 = 2.25, which corresponds with an annual rate of 5 percent. This
increase in the number of port-pairs to be connected puts the increase of
the number of roundtrips of 5.6%, as mentioned in Section 4.1 in a different
light. It suggests that the increase in roundtrips is needed to compensate for
the increase in port-pairs to be served in order to maintain the frequency
of service between ports.
Another way to deal with the increasing number of port pairs is
transhipment. On a typical service the number of port pairs to be connected
comes at 5 × 5 = 25 on a total of, say, 50 × 50 = 2500 different port pairs to
be connected on the West Europe — Far East trade route. If an operator
would like to cover all these port pairs he would need to offer 2500/25 = 100
different services. This is much and it may be clear that some degree of
transhipment is inevitable.
4.4.2. Development of transhipment

Hub-and-spoke (HS) operations are a means to increase the coverage of
main lines serving two coastlines thereby also serving the port pairs connect-
ing the coastlines that are not linked directly by multi-porting itineraries.
To make HS operations competitive with multi-porting operations, given a
certain market size, it is required that the mainline part of the costs is as low
224 S. Veldman
as possible by deployment of the largest possible ships. Cost comparisons

with a model by Jansson and Sheerson (1987) show that, given a fixed
market, a HS system is only under exceptional circumstances more cost
effective than a multi-porting system. This concerns situations where the
size of the mainline ship deployed in the HS system is big in comparison
to the ship of the multi-porting system, where cargo handling costs are low
and the geographical setting favourable.
Under the heading “Direct call versus transhipment debate” Baird
(2005) mentions the general opinion under maritime economists of the view
that extra feeder and handling costs make transhipment more expensive
than multi-porting. This supports the findings of Jansson and Shneerson
(1987).
Not all transhipment has to do with hub-and-spoke operations. Inter-
liner transhipment, linking one mainline with the other, does not and
generally takes place in certain specialised ports favourably located at cross
points of east-west and north-south trade routes, such as Algeciras, Salalah
and Singapore. Inter-liner transhipment as far as it takes place within a
trade route however is a means to increase the coverage of port pairs on
that trade route in a similar way as HS operations.
HS operations are an extra reason, i.e., additional to the effort to
just beat each other with bigger ships, to stimulate the introduction of
bigger ships. The introduction of bigger ships itself seems to stimulate HS
operations as the number of ports able to accommodate such ships is less
than for the ships they replace. Introduction of HS operations and big ships
seems to mutually strengthen each other.
How is transhipment developing: is it increasing or decreasing? No
information is available at the level of container flows, but there is at
the level of port throughput. A compilation of data from Drewry (2004),
(2007 and 2008) shows that the share of transhipment in total container
throughput at global level is increasing gradually from 11% in 1980 to
18% in 1990, to 25% in 2000 and further to 27% in 2007. Looking at the
development since 2001 it appears that growth has stopped.
Per coastal region the development is different. For North Europe
the transhipment share is increasing from 14% in 1980 to 18% in 1990
further increasing to 24% in 2002 and fluctuating below 24% since then.
The geography of South Europe is more favourable for transhipment, while
containerisation started later: the share of transhipment increased from 15%
in 1980 to 25% in 1990 and further to 35% in 2000 and is increasing still
further to 43% in 2007.
Table 8 Development of the share of transhipment for coastal regions.
Region 1980 1990 2000 2001 2002 2003 2004 2005 2006 2007
N. Europe 14% 18% 20% 23% 24% 24% 21% 22% 23% 24%
S. Europe 15% 25% 35% 35% 37% 36% 37% 42% 43% 43%
SE Asia 32% 40% 48% 46% 46% 47% 48% 50% 50% 51%
Far East 12% 18% 23% 26% 27% 26% 23% 22% 21% 21%
Total world 11% 18% 25% 27% 27% 27% 26% 27% 27% 27%
Sources: Compiled from Drewry (2004), Drewry (2007) and Drewry (2008).
Note: In case figures in different publications differ, those of the newest one are used.
The region Far East shows an increase from 12% in 1980 to 18% in
1990 and further to 23% in 2000 reaching a peak of 27% in 2003. Since
then it is decreasing gradually to 21% in 2007. The geography of Southeast
Asia is favourable for transhipment. From a high level of 32% in 1980 it
is gradually increasing to 40% in 1990, 48% in 2000 and further to 51% in
2007.
The long term development of the share of transhipment from 1980 to
2007 suggests that at a global level transhipment has reached its ceiling.
Of the coastline regions dominating the Europe — Fareast trade route the
Far East and West Europe seem to have reached a ceiling, while for the Far
East it has turned into a slight decrease. South Europe and Southeast Asia
still experience some growth.
It can be concluded that the increase in trade on the North Europe —
Far East trade route results in an increase in ship size and an increase
in number of roundtrips. The increase in the number of port pairs to be
connected seems to be met by an increase in roundtrips in combination with
an increase in HS operations. Since the first years of the century the increase
of transhipment and thereby of HS operations, seems to have reached a
ceiling.
5. Conclusions
The results of the statistical analysis based on a great sample of con-
tainerships shows that economies of ship size forz Post-Panamax ships,
as expressed in elasticities of costs as a function of ship size, differ only
slightly from those of ships up to Panamax size. Economies of ship size are
reflected in the resulting shipping costs on the North Europe — Far East
shipping route.
226 S. Veldman
The values of the estimated elasticities are similar to those of earlier

research concerning smaller ships and lower cargo handling efficiency in
ports such as by Ryder and Chappel (1979) for ships ranging from 80 to
580 TEU, by Jansson and Shneerson (1987) for ships ranging from 240 to
3200 TEU and by Cullinane and Khanna (1999) for ships ranging from 1000
to 8000 TEU. The results suggest that scale effects are equal or slightly less
for capital related costs and slightly higher for fuel costs.
Economies of ships size act as the motor of demand of bigger ships.
In order not to lead to too high user costs the increase in size has to be
in balance with the increase in trade volumes, in the number of port pairs
to be served, the number of roundtrips and in the share of transhipment
operations.
The technical limitations to the size ships are gradually being pushed
further. The analysis shows that there are good arguments to extrapolate
to the scale effects as measured for ships up to 14,000 TEU further to ship
sizes up to 20,000 TEU. The observations suggest that the ongoing increase
in ship size will continue.
The development shows that over the period 1990 to 2007 the increase
in trade on the North Europe — Far East trade route in west bound
direction, the motor of the increase in ship size, increased with 10.8% per
year, while the increase in the size of the top 2% containerships increased
with from 4,526 TEU to 9,220 TEU, corresponding with an annual growth
rate of 4.3%. Over the same period the number of weekly services offered
on the trade route increased with 5.6% annually, matching fairly well the
difference of the increase in ship size and the increase in trade.
The number of ports located along the coastlines at the end of the trade
route based on OSC (2006) is increasing and rough estimates suggest with a
rate of about 5% annually. To cope with this increased density of port pairs
on the trade route more roundtrips need to be made in combination with
an increase in transhipment. The 5.6% increase in the number of roundtrips
suggests that this is the major factor, while the increase in transhipment,
which seems to have stopped short after 2000, appears to have become a
less important factor.
With the motor of economies of scale continuing to exist for ships in
excess of 14,000 TEU there are good reasons to expect that the average size
of the top 2% ships will increase further at a rate somewhat lower than as
for the increase in demand of trade as it did in the past. Part of this effect
has already taken place as reflected in the number of big ships in the order
book as per 1/1/2008.
The observed stop of the growth of transhipment may have a slight

downward impact on the ratio of the growth rate of ships size versus
demand, i.e., 4.3% versus 10.8%. This effect, however, may be compensated
by a stop in the increase in seaports per coastal area, which has a downward
impact.
So far the long run development. Since the end of 2008 the increase in
worldwide container trade and in the North Europe — Far East trade has
come to a standstill and turned into a decrease, as a result of the credit
crunch in combination with overoptimistic ordering. It will take time for
the world economy to recover and to reach the 2008 level again, for world
trade to recover and reach the 2008 level again and when the increase of
trade will have started again to reach levels where the existing over-tonnage
will have been worked away. If that moment has arrived ordering will start
again and also ordering of container ships in excess of the biggest ships
existing now and now on order.
References
1. Baird, A. (2005). “Optimising the container transhipment hub location
in northern Europe”, Journal of Transport Geography, Article in Press,
www.sciencedirect.com.
2. Cullinane, K. and Khanna, M. (1999). Economies of scale in large container
ships, Journal of Transport Economics and Policy, 33, pp. 185–208.
3. Cullinane, K. and Khanna, M. (2000). Economies of scale in large contain-
erships: optimal size and geographical implications, Journal of Transport
Geography, 8, pp. 181–195.
4. Drewry Shipping Consultants (1992). Strategy and Profitability in Global
Container Shipping.
5. Drewry Shipping Consultants (2004). Container Container Market Review
and Forecast — 2004/05.
6. Drewry Shipping Consultants (2006). Ship Operating Costs, Annual Review
7. Drewry Shipping Consultants (2007). Container Container Market Review
8. Drewry Shipping Consultants (2008). Container Forecaster Annual.
9. Jansson, J.O. and Shneerson, D. (1987). Liner Shipping Economics,
Section 5.5, Chapman and Hall Ltd, London.
10. Ocean Shipping Consultants (OSC) (2006a). The European & Mediterranean
Containerport Markets to 2015.
11. Ocean Shipping Consultants (OSC) (2006b). East Asian Container Port
Markets to 2020.
12. Ryder, S. and Chappell, D. (1979). Optimal Speed and Ship Size for the Liner
Trades, Marime Transport Centre, Liverpool.
228 S. Veldman
13. Tozer, D. (2006). Design challenges of large container ships, Lloyd’s Register,
Paper presented at ICHCA.
14. Veldman, S. (1993). The optimum size of ships and the impact of user
costs, an application to container shipping, Chapter 9 in Current Issues
in Maritime Economics, Edited by K.M. Gwilliam, Kluwer Academic
Publishers.
15. WES (2008). World Shipping Encyclopaedia of Lloyd’s/Fairplay.
16. Wijnolst, N., Scholtens, M. and Waals, F. (1999). Malacca-Max, The
Ultimate Container Carriers, Delft University Press.
CHAPTER 11
A LINEARIZED APPROACH
FOR LINER SHIP FLEET
PLANNING WITH DEMAND
UNCERTAINTY
Qiang Meng∗ , Tingsong Wang and Shahin Gelareh
Department of Civil Engineering, National University of Singapore
P.O. Box 117576, Singapore
∗ [email protected]
In this paper, we focus on the liner Ship fleet planning (LSFP) with cargo
demand uncertainty. The LSFP aims to determine which type of ships and
how many of them are needed, and how to deploy and operate these ships.
We first propose a mixed integer nonlinear programming model for the LSFP
by taking into account cargo shipment demand uncertainty. We find that the
fuel consumption cost of a ship can be approximated by a linear function with
respect to its cruising speed, and we then proceed to build a mixed integer linear
programming model that can approximate the originally proposed nonlinear
programming model. The mixed integer programming model can be thus
effectively solved by some optimization solvers. Finally, numerical examples
are carried out to assess the linearized approach.
1. Introduction
1.1. Background
Due to the increased global trade activities, the maritime transportation
industry has been growing steadily during the past decade. In particular,
highly-containerized liner trade is the fastest growing sector. Liner shipping
occupies the most major place in the global trading transportation.
UNCTAD1 reported that the share of the top 20 liner operators held about
70 percent of the total container capacity deployed in 2008, and it pointed
out that with the continuous advancement of ship building technology and
229
230 Q. Meng, T. S. Wang and G. Shahin
incensement of global container traffic, the liner shipping dominant trend

would continue to strengthen.
Since the ships are assets with expensive operational cost, how to
effectively manage the ship fleet in order to maximize its profit or minimize
its cost is an issue highly concerned by the liner operators. This issue is
termed as a Liner Ship Fleet Planning (LSFP) problem in this paper. There
are two tasks involved in LSFP problem: one is the planning of fleet size
and mix; and the other is the deployment of the fleet. The fleet size refers
to the total number of ships needed to ship cargoes for shippers, while the
mix refers to the types of ships in the fleet. For example, Oriental Oversea
Container Line (OOCL), a liner shipping company in Hong Kong, owns
a fleet with the size of 39 ships constituted of 9 different types in 2009.2
Fleet deployment is the ship-to-route allocation. We will deal with these two
tasks simultaneously in this paper and aim at developing a mathematical
programming model in order to assist a liner operator to manage their
ship fleet effectively, thereby operating the fleet with a cost as less as
possible. The core of this model is a stochastic programming model where
the demand of containers delivered between any two ports is considered to
be random variablea. It is firstly formulated as a mixed-integer nonlinear
programming (MINLP) model with consideration of ships’ cruising speed
as decision variables. We find that the fuel consumption cost for a ship can
be approximated by a linear function with respect to its cruising speed,
and we then proceed to build a mixed integer linear programming (MILP)
model to approximate nonlinear programming model. Finally, the solver
CPLEX is employed to solve the MILP model.
1.2. Literature review

There are a number of studies focusing on ship fleet planning problem, most
of which are surveyed in the following four major review articles or chapter:
Ronen,3,4 Perakis5 and Christiansen et al.6
To the best knowledge of authors, this paper is first to incorporate
uncertain demand into optimization model for LSFP. Thus, the literature
directly related to this paper is focused on those dealing with deterministic
demand, which falls into two categories: modeling for fleet’s size and mix
problem and for fleet deployment problem.
The decision-making of liner ship fleet size and mix is one of the
important tasks in LSFP for both researchers and shipping service oper-
ators. Some related researches have been studied: Dantzig and Fulkerson7
A Linearized Approach for Liner Ship Fleet Planning 231
addressed the fleet’s size planning problem for a tanker fleet, Laderman
et al.8 discussed the same problem for a tramp or an industrial shipping
company and Everett et al.9 proposed a linear programming model for large
bulkers and tankers fleet. A simple approach was proposed by Benford10
to select the optimal mix of available bulkers and cruising speed operated
between two specific ports. Afterwards, Perakis11 solved the problem of
Benford10 by using Lagrangian multipliers approach.
However, those researches studied are either for tramp ships or indus-
trial ships, not for liner ships. Liner ships are different with the first two
shipping modes. It provides a fixed liner service, at regular intervals, between
named ports and offers transport to any shippers. In liner shipping, time
is very important due to the fact that liner ships have to comply with
the schedules even the operation is at low utilization levels (Lawrence;12
Stopford.13 ) Lane et al.14 developed a three phased approach to determine
the fleet size and mix for a liner trade route with known demands for
shipping services between any origin-and-destination pairs. A three-phase
approach was developed by Fagerholt15 to find the optimal fleet design.
Recently, Hsu and Hsieh16 formulated a two-objective model to determine
the optimal liner routing, ship size and sailing frequency of container carriers
by using the Pareto principle. Tabu search algorithm is applied by Brandao17
to seek the optimal fleet size and mix planning for vehicle routing problem.
Some literatures studied the fleet deployment problem, which is another
important task in LSFP: Perakis and Papadakis18,19 proposed a mathe-
matical model to determine the optimal fleet deployment and sensitivity
analysis was performed to study the effects of small or large changes in
one or more cost components on the total cost. Later, the multiorigin,
multidestination fleet deployment problem was studied by Papadakis and
Perakis.20 In this paper, the speeds of vessels were categorized into two
types: full load and ballast speeds. A projected Lagrangian method was
applied to solve the problem. A realistic model for optimal deployment with
detailed description of operating costs of liner ships was proposed by Perakis
and Jaramillo,21 and their proposed model is solved by LINDO solver in
Jaramillo and Perakis.22 Powell and Perakis23 revisited the problem studied
by Perakis and Jaramillo21 and formulated it as an integer programming
model. Millar and Gunn24 formulated two mixed-integer programming
models to dispatch a fishing trawler fleet and used a heuristic method
to solve them. Vukadinović and Teodorović25 discussed the dispatching
problem of barges by using the fuzzy approach. A linear programming
model and a mixed integer programming model by Cho and Perakis26
investigated liner fleet deployment and expressed it as matrix forms by

introducing flow-route incidence matrix. A mathematical program solver
OSL was employed to solve their model. Mourão et al.27 presented an
application of integer programming model to support the decision-making
process of assigning ships with hub and spoke constraints.
Those researches reviewed in the above sections are all for short-term
planning horizon, for the long-term planning, Nicholson and Pullen28 did
the first study on fleet management of sale and replacement and used a
two-stage method to tackle it. Xie et al.29 applied a dynamic programming
model to seek the best liner fleet deployment policy for long-term planning
horizon, and for short-term planning horizon (each one year), it was
formulated as a linear integer programming model. The problem was solved
by a heuristic algorithm.
1.3. Randomness
There is one major source of randomness in the LSFP problem, namely the
cargo shipment demand. Before describing it, we first briefly introduce the
procedures of a typical shipping flow as follows: a shipper books space from
a liner shipping company through a shipping agent at port of shipment by
filling in a shipping application (S/A), if S/A is accepted, the shipper will
receive a shipping order (S/O) to pick up empty containers from a depot
and load the containers. Then, the carrier will offer a mates receipt (M/R)
to the shipper. The shipper bears it to exchange bill of lading (B/L) and
posts it to the consignee. The shipping agent at port of discharge informs
the consignee to deliver the goods when it arrives. After the payment of
all fees, the consignee uses B/L to exchange the delivery order (D/O) and
takes delivery of goods.
As described above, a liner ship serves a large number of shippers; also,
a shipper can choose one carrier from many competitors. In the procedure
of decision making, the schedule and itinerary of the liner service will be
firstly taken into account. Besides those, the freight is also another issue to
be considered. There are some other factors affect the decision making such
as the liner service level, the security of the area along the itinerary and
so on. More importantly, the shippers are allowed to cancel the shipping
contract signed with the carrier. Therefore, the actual amount of cargo
shipment demand is highly uncertain. In practice, it is fluctuant. The
slump of global seaborne trade caused by US subprime mortgage crisis is a
convincing example. This view has been confirmed by the statistical data
from Neptune Orient Lines (NOL), which is the global leader in container
shipping. For the period November 15-December 26, Marcus30 reported
that NOL carried 218,100 FEUs globally, equivalent to 24% fewer boxes
than in the same period in 2007.
1.4. Contributions
The contributions of this paper are threefold: firstly, the assumption that
the cargo shipment demand is deterministic in previous researches is not
reasonable. This motivates us to seek a more reasonable assumption for our
problem. In this paper, we assume that the cargo shipment demand between
two ports on each liner trade route is a random variable following normal
distribution with a given mean value and standard deviation. But this will
create another issue: the liner shipping carrier has to satisfy the container
transportation requirement on each liner trade route. Since the demand
is assumed to be a random variable, how to formulate this constraint? If
the fleet size and mix are deterministic, then its transportation capacity
is deterministic; however, the demand is uncertain, this creates a risk
that the fleet could not satisfy the container delivery requirement. In
order to deal with this problem, we introduce the risk level concept
and apply the chance constraints. The risk level refers to the possibility
that the fleet’s transportation capacity is insufficient. Then, we use the
chance constraints to explore the possibility that the fleet could deliver the
containers. Furthermore, we assume all unsatisfied containers are regarded
lost. Dealing with the demand uncertainty is the main contribution of this
paper. Moreover, we will study the sensitivity analysis of the effect of risk
level on the fleet deployment plan.
Secondly, all of the previous researches either neglected or predeter-
mined the cruising speeds of vessels in their proposed models. However,
experiment data shows that the fuel consumption, the biggest part occupied
in the total shipping cost of a vessel, is significantly affected by its cruising
speed. It increases exponentially with the speed increasing one knot or
more.13 Thus, the cruising speed has an important impact on the shipping
cost of a vessel. Within the consideration of its effect on the optimal fleet
planning, it is regarded as a decision variable in this paper to seek the
optimal cruising speed.
Thirdly, most researches considered the planning of fleet’s size, mix
and deployment separately; however, Agarwal and Ergun31 addressed that
they inter-affect on each other. In this paper, we extended and improved
the models for LSFP built in previous researches by integrating the fleet
size, mix and deployment in the decision-making procedure of the optimal
fleet planning. In the procedure of decision-making, the carrier can use its
own or chartered ships from other carriers to carry containers, even can
charter-out some ones to other carriers.
This paper is organized as follows: Section 2 describes the liner ship
fleet planning problem and introduces the notations and assumptions in
this paper. Section 3 develops a mixed-integer nonlinear programming
model with chance constraints to deal with the cargo shipment demand
uncertainty. Section 4 explores the linearity relationship between the fuel
consumption and the cruising speed and converts the MINLP model into a
MILP model. Section 5 randomly generates numerical examples to illustrate
our model and studies the impact of risk level on fleet deployment. Section 6
summarizes and discusses our work.
2. Problem Description, Assumptions and Notations

Consider a liner container shipping company operating a set of routes with
given liner service frequencies for a short-term planning horizon to pick
up and delivery cargoes from shippers. Given a cargo shipment demand
pattern within a short-term horizon, the liner container shipping company
has to deal with the short-term LSFP problem: to decide which types of own
ships and how many of each type are used/chartered out (if any), or which
types of chartered ships and how many of each type are needed (namely
decide the fleet size and mix); furthermore, how to assign those ships in
the fleet to the liner trade routes (namely decide the fleet deployment)
so that the planned fleet provides a regular liner service on each route to
pick up and delivery cargoes from one port to another. The following will
explain the terms appeared in this paper: itinerary, charter strategies and
chance constraints.
2.1. Itinerary
In most liner services, ship schedules and itineraries are fixed and services
are offered on a regular basis. A liner service itinerary is defined as a
sequence of ports visited by ships. The direction from origin to destination is
called outbound direction, contrarily, is called inbound direction. A ‘voyage’
is defined as one round-trip shown in Fig. 1.
The outbound direction and inbound direction may be asymmetric. As
the Fig. 1 shows, Ningbo Port is not visited in the outbound direction, while
Fig. 1 An example of liner ships’ itinerary.
Fig. 2 The number code of liner ships’ itinerary.
Tianjin Port is not visited in the inbound direction. The sequence of ports
in this itinerary is: Pusan–Dalian–Tianjin–Shekou–Ningbo–Dalian–Pusan.
In order to simplify the expression of ports sequence, we code the ports by
numbers. Due to there are totally five ports in the itinerary, thus, number
1, 2, 3, 4 and 5 are used to present the ports. The smallest and the largest
number present the origin and the destination port respectively. By using
the number code, the sequence of ports is expressed as: 1 — 2 — 3 — 5 —
4 — 2 — 1 (see Fig. 2).
It is noted that in our work, we assume all routes have specific origin
and destination port. This case may not be common in practice because
some liner shipping company does not follow this convention. Sometimes,
the origin and destination port are variable. However, in order to simplify
our problem, we insist to make such assumption and moreover, all ships are
empty before they start their journey.
2.2. Charter strategies

There are three main chartering ways: voyage charter, time charter, and
bareboat charter.
A voyage charter, which is the most popular way in chartering market,

provides transport for a specific cargo from one port to the other for a
fixed price per ton. The ship owner has to complete the terms in charter-
party and pay all fixed and variable costs. The charterer pays for the freight
according to the charter-party. If completed over the due date of lay time
(the period of time that a ship is available for loading and discharging of
cargoes at a port) allowed in the charter-party, the owner will submit a
claim of demurrage to the charterer. Conversely, if completed ahead, the
charterer will submit a claim of dispatch to the owner.
A time charter gives the charterer operational control of the ship
carrying his cargo, while leaving ownership and management of the vessel
in the hands of the ship owner. The length of the charter is often a period
of months or years (period charter). During this period, the ship owner
continues to bear the all fixed costs except the capital repayment, tax and
depreciation, but the charterer directs the commercial operations of the
vessel and pays all variable costs.
Finally, if a charterer wishes to have full operational control of the ship,
but does not wish to own it, a bareboat charter is arranged. The charterer
manages the vessel and pays all costs except the capital repayment, tax
and depreciation. In other words, the owner does not bear any cost except
collecting the rent from the charterer. Due to the owner does not feel
relieved to let the charterer operate his vessel and the management of
a vessel for the charterer is not easy, bareboat charter is less popular
compared with the above two ways though it is simplest. However, in order
to simplify the problem, we assume a carrier can only adopt the bareboat
charter in this paper if necessary.
Due to buying ships is a much huger capital investment compared with
chartered ships, it is excluded in the short-term LSFP. According to the
latest news released by China Shipping (Group) Company, it will book
8 Post-Suezmax ships with the shipping capacity of 12,600 TEUs from
Samsung Engineering Company. The cost of a ship is about $170 million,
and the total trade will exceed $1.3 billion.32 However, the annual rent
fee of a ship is only $22 million.33 Thus, for the short-term fleet planning,
chartering ships is preferred than buying ships.
2.3. Chance constraints

As aforementioned, the cargo (container) shipment demand between any
two ports on each liner service route is assumed following a normal
distribution. This assumption may lead to another problem: since the

demands are uncertain, one can hardly find any decision which would
definitely exclude later constraint violation caused by unexpected random
effects, in other words, once the decisions in LSFP problem are determined,
the fleet of ships may be unable to fully meet the pickups and deliveries
requirement for its customers, even though the expected demands along the
route do not exceed the fleet capacity. Once such case happens, it implies
losing money for this liner container shipping company. Since it is hardly
unavoidable, the liner container shipping company hopes that it happens
at a low possibility as possible.
In order to reduce the possibility of the occurrence that the liner
container shipping company cannot satisfy the customers’ demand, such
constraint violation can often be balanced afterwards by some compensating
decisions which are considered as a penalization for constraint violation.
However, the compensation cannot be modeled by cost in this paper because
the cargo shipment demand is not realized. In such circumstances, we would
rather insist on decisions guaranteeing feasibility ’as much as possible’. This
loose term refers once more to the fact that constraint violation can almost
never be avoided because of unexpected extreme events. On the other hand,
when knowing or approximating the distribution of the random parameter,
it makes sense to call decisions feasible whenever they are feasible with
high probability, i.e., only a low percentage of realizations of the random
parameter leads to constraint violation under this fixed decision. Therefore,
we formulate the constraint that the liner container shipping company
should satisfy the customers’ demand as a probabilistic form in this paper,
which is called chance constraint. The probability of the constraint violation
is called a risk level in this study. It indicates that if the liner container
shipping company makes a decision which satisfies the chance constraint,
the event that the customers’ demand cannot be met will occur at most
with this probability. For those unmet cargoes, we regarded they are
lost.
Here, we specially explain the term “frequency of liner service”, which
is the most important feature of liner shipping. It is defined as the time
interval between two successive liner services.21 For example, the shipping
company Oriental Oversea Container Line (OOCL) provides a weekly liner
service on the route Asian-Pacific Express.2 In other word, the liner service
frequency on that route is 7 days.
We also make another three assumptions: first, the cargo demand from
ports to ports on a route are regarded to be independent random variables
following normal distributions with given mean and standard deviation;

second, the waiting time from chartering a ship to being serviced is ignored;
thirdly, the number of available ships is finite; fourthly, we assume the ships
are empty when they starts their journey.
2.4. Notations
Notations used in this paper are listed as follows:
Sets
K Set of ship types’ number code
Mr Set of ports’ number code on route r
R Set of liner trade routes’ number code
Z+ Set of positive integer
R+ Set of positive real number
Deterministic Parameters
ckr : Shipping cost of a ship of type k on route r ($/voyage)
cin
k : Cost of chartering-in a ship of type k in the planning horizon
($/ship)
cout
k : Profit of chartering-out ship of type k in the planning horizon
($/ship)
dr : Length of route r (mile)
dij
r : Distance from port i to port j on route r (mile)
ek : Lay-up cost for a ship of type k ($/day)
Mrd : Number code of the destination port on route r
Nkmax : Maximum amount of the ships owned by the carrier of type k
N CIkmax : Maximum amount of the ships of type k chartered-in from
other carriers
L
Skr : Lower bound of the cruising speed of a ship of type k on route
r (mile/hr)
U
Skr : Upper bound of the cruising speed of a ship of type k on route
r (mile/hr)
T: Length of the short-term planning horizon (day)
Tk : Shipping season for a ship of type k in the planning horizon
(day)
Vk : Capacity of a ship of type k (TEUs)
αr : Risk level of the capacity of fleet is insufficient on route r
Random Data
ξrij : Cargo demand from port i to port j on route r in the
planning horizon (TEU)
µij
r : Mean value of ξrij (TEUs)
σrij : Standard deviation of ξrij (TEUs)
ηrij : Flow of containers on the segment from port i to port j on
route r in the planning horizon (TEUs)
Decision variables
fr : Frequency of liner service on route r (day/service)
nkr : Number of owned ships of type k sailing on route r
nin
kr : Number of chartered-in ships of type k sailing on route r
nout
k : Number of chartered-out ships of type k to other carriers
skr : Cruising speed of ships of type k on route r
xkr : Number of voyages of ships of type k on route r
ykr : Lay-up days of ships of type k on route r
3. A Mixed Integer Nonlinear Programming Model with

Chance Constraints
In this section, a MINLP model is developed to minimize the cost of
shipping the containers in the short-term planning horizon.
The expenditure is consists of three components: shipping cost, lay-up
cost and chartered rent. They are calculated as follows:
Expenditure in shipping costs:

Cs = ckr (skr )xkr
r∈R k∈K
Expenditure in lay-up costs:

Cl = ek ykr
r∈R k∈K
Expenditure in chartering-in ships:

C ci = nin in
kr ck
r∈R k∈K
Thus, the total cost equals to:

C= (ckr (skr )xkr + ek ykr + nin in
kr ck ) (1)
r∈R k∈K
The ckr (skr ) in the above Eq. (1) denotes the shipping cost per voyage for a
ship of type k on route r. It includes the port charges, canal fees, fuel cost,
maintenance cost, insurance cost, administrative cost and crew cost, etc.21
As explained in Stopford,13 the cruising speed affects the fuel cost, thus,
the shipping cost per voyage for a ship should be a function with respect
to cruising speed. Based on the cost function formulated by Perakis and
Jaramillo,21 the shipping cost per voyage can be expressed as the following
equation:
ckr (skr ) = λ̄kr s2kr + λ̃kr /skr + λ̂kr (2)
where λ̄kr , λ̃kr and λ̂kr are parameters.
Remark 1 There are some other different cost functions: Cullinane and
Khanna,34 Drewry,35 Shintani et al.36 and Takano and Arai.37 These
researches investigated the shipping cost by various ships’ size. The regres-
sion analysis was applied based on the cost data. Thus, they formulated the
shipping cost as a function with respective to the ship size. It is noted that
the parameters in the regression functions of cost developed in those studies
are dependent on the cost data space. Those regression models are valid
only for a specific data space. In other words, those regression functions of
shipping cost are not available in this paper; if employed, the parameters
need to be calibrated.
Since we suppose liner operators can charter out some ships to other
carriers to earn the profit, our objective function is formulated as follows:

kr ck ) −
(ckr (skr )xkr + ek ykr + nin in
[P] min nout
k ck
out
(3)
r∈R k∈K k∈K
Subject to:

Pr xkr Vk ≥ ηrij ≥ 1 − αr , i = 1, . . . , Mrd − 1;
k∈K
j = i + 1, . . . , Mrd ; ∀r ∈ R (4)

Pr xkr Vk ≥ ηrij ≥ 1 − αr , i = 2, . . . , Mrd ;
k∈K
j = 1, . . . , i − 1; ∀r ∈ R (5)
dr xkr ykr
× + ≤ T, ∀r ∈ R, k ∈ K (6)
24skr nkr + nin
kr n in
kr + nkr
ykr
≥ T − Tk , ∀r ∈ R, k ∈ K (7)
nkr + nin
kr

nkr ≤ Nkmax , ∀k ∈ K (8)
r∈R

max
kr ≤ N CIk
nin , ∀k ∈ K (9)
r∈R

nout
k = Nkmax − nkr , ∀k ∈ K (10)
r∈R
L
Skr ≤ skr ≤ Skr
U
, ∀k ∈ K, r ∈ R (11)
xkr , ykr , nkr , nin out

kr , nk ∈ Z+ ∪ {0}; skr ∈ R+ ∪ {0}, ∀k ∈ K, r ∈ R
(12)
Constraints (4) and (5) are chance constraints to investigate the possibilities
that the fleet’s capacity is sufficient on route r in the outbound trip and
inbound trip respectively. When a ship sails from port i to port j, the
containers on the ship includes those will be unloaded at port j from
previously visited ports and those loaded at port i to the foregoing ports,
it is referred to as a term flow of a segment from port i to port j in this
paper. The containers loaded in the ship on any voyage segment can not
exceed the capacity of the ship. Segment flow on each route is calculated
as follows:
(a) For outbound trip
d

i
Mr
ηrij = ξrpq , i = 1, . . . , Mrd − 1; j = i + 1, . . . , Mrd ; ∀r ∈ R
p=1 q=j
(13)
(b) For inbound trip

d

Mr j
ηrij = ξrpq , i = 2, . . . , Mrd ; j = 1, . . . , i − 1; ∀r ∈ R (14)
p=i q=1
Remark 2 Due to the cargo demand from port i to port j on a route is

assumed to follow inter-independent normal distribution, according to the
probability theory, the flow of containers on the segment from port i to
port j on a route also follows a normal distribution.
(a) For outbound trip
 
Mrd Mrd
i i
ηrij ∼ N  µpq
r , (σrpq )2  (15)
p=1 q=j p=1 q=j
(b) For inbound trip

 d

Mrd j

Mr j

ηrij ∼ N  µpq
r , (σrpq )2  (16)
p=i q=1 p=i q=1
Thus, the above constraints (4) and (5) can be respectively rewritten as
follows:

i Mrd

i Mrd
−1
xkr Vk ≥ Φ (1 − αr )
pq 2
(σr ) + µpq
r ,
k∈K p=1 q=j p=1 q=j
i = 1, . . . , Mrd − 1; j = i + 1, . . . , Mrd ; ∀r ∈ R (17)

d
Mrd j
r pq 2
M j
−1
xkr Vk ≥ Φ (1 − αr )
(σr ) + µpq
r ,
k∈K p=i q=1 p=i q=1
i = 2, . . . , Mrd ; j = 1, . . . , i − 1; ∀r ∈ R (18)
where Φ−1 (•) is the inverse function with respective to probability •.

Constraint (6) ensures the time of a ship sailing on sea and its lay-up
time should not exceed the planning horizon. Rewrite it as follows:
dr xkr
+ ykr ≤ T (nkr + nin
kr ), ∀r ∈ R, k∈K (19)
24skr
where the length of route r, dr , can be calculated by the following
formulation:

dr = dij
r , ∀r ∈ R (20)
i,j∈Mr
The least lay-up days for a ship of type k on route r are through
constraint (7). We can rewrite is as follows:
ykr ≥ (T − Tk )(nkr + nin

kr ), ∀r ∈ R, k ∈ K (21)
Constraints (8) and (9) ensure the number ships of own and chartered-
in should not exceed the maximum available ships. Constraint (10) provides
the calculation formulation of the number of chartering-out ships.
Constraint (11) provides the feasible bounds for cruising speed of a ship of
type k on route r. Constraint (12) requires that all variables are nonnegative.
The frequency is the most important property in liner shipping, it could
be obtained by the following formula:
 
T
fr =  , ∀r ∈ R (22)
xkr
k∈K
where [•] denotes the largest integer less than •.
4. A Linearized Approach
According to one technique report 38, for ships in the range of up to
1,500 TEU, the speed is between 9 and 25 knots per hour, with the majority
of ships sailing at some 15–19 knots. The most popular speed for the
1,500–2500 TEU ships is 18–21 knots. In the 2,500–4,000 TEU range, 90%
of the ships have a speed of 20–24 knots. 71% of the 4,000–6,000 TEU ships
have a speed of 23–25 knots. Finally, 80% of the ships that are larger than
6,000 TEU have speed of 24–26 knot. For the new generation of container
ships we let the speed to be between 25–26 knots per hour. The design
fuel consumption is assumed to be the upper bound of each interval. Table 1
shows the design fuel consumption of ships with different capacities from
1000 TEUs to 10000 TEUs.
Table 1 Design fuel consumption of ships with different capacities.
Capacity (TEU) F ∗ (tons/day) Capacity (TEU) F ∗ (tons/day)
1000 40 4000 145

1250 50 5000 165
1500 60 6000 185
1750 70 7000 205
2000 80 8000 225
2250 90 9000 245
2500 100 10000 265
We use the data from Table 1 and plot the fuel consumption graph
against the speed in the given interval. We can observe that the function
can be nicely approximated by a linear function in the given interval. For
example, for the ships with the capacities of 1,000–1,500 TEU and given
F ∗ in Table 1, the fuel consumption against the speed is depicted in Fig. 3.
Also, the absolute and relative errors of the approximated value with the
real value against the speed are illustrated in the below Fig. 4 and Fig. 5.
18
17
Fuel consumption (tons/day)
16
15
14
13
12
11
10
9
8
15 15.5 16 16.5 17 17.5 18 18.5 19
Speed (k/hr)
Fig. 3 Fuel consumption behaviors for the ships with capacities between 1,000 and
1,500 TEU.
0.4
0.35
Absolute error (tons/day)
0.3
0.25
0.2
0.15
0.1
0.05
0
15 15.5 16 16.5 17 17.5 18 18.5 19
Speed (k/hr)
Fig. 4 Absolute error of approximated and real fuel consumption.

These two figures indicate that the maximum absolute error is less than
0.4 tons per day and the relative error are less than 0.04% compared
with the real fuel consumption. For all range, they are also displayed in
Fig. 6(a)∼(c). From Fig. 6(b), it indicates that the maximum error between
the approximation and the real value of fuel consumption is less than 2 tons
per day with the speed ranges from 20 knots to 24 knots per day. And
0.04
0.035
0.03
Relative error (%)
0.025
0.02
0.015
0.01
0.005
0
15 15.5 16 16.5 17 17.5 18 18.5 19
Speed (k/hr)
Fig. 5 Relative error of approximated and real fuel consumption.
300
250
Fuel consumption (tons/day)
200
150
100
50
0
14 16 18 20 22 24 26
Speed (k/hr)
Fig. 6(a) Fuel consumption vs. speed.

2
1.8
1.6
Absolute error (tons/day)
1.4
1.2
1
0.8
0.6
0.4
0.2
0
14 16 18 20 22 24 26
Speed (k/hr)
Fig. 6(b) Absolute error of approximated and real fuel consumption.
0.04
0.035
0.03
Relative error (%)
0.025
0.02
0.015
0.01
0.005
0
14 16 18 20 22 24 26
Speed (k/hr)
Fig. 6(c) Relative error of approximated and real fuel consumption.
the maximum relative error shown in Fig. 6(c) is less than 0.04%. This
implies that the approximation of the fuel consumption is acceptable and
the performance is very well.
Therefore, the fuel consumption with respect to speed can be approxi-
mated by a linear function as follows:
Fkr = ak skr + bk (23)

Table 2 The slop and intercept of the approximated linear function.
Capacity Max absolute error Max relative

(TEU) ak bk (tons/day) error (%)
1, 000 2.2271 −25.0653 0.3563 0.0393

1, 250 2.7838 −31.3317 0.4453 0.0393
1, 500 3.3406 −37.5980 0.5344 0.0393
1, 750 8.6357 −111.9076 0.6732 0.0148
2, 000 9.8693 −127.8944 0.7694 0.0148
2, 250 11.1030 −143.8812 0.8656 0.0148
2, 500 12.3367 −159.8680 0.9618 0.0148
4, 000 15.2610 −222.9008 1.8791 0.0216
5, 000 18.2567 −291.7981 0.5108 0.0039
6, 000 20.4696 −327.1676 0.5728 0.0039
7, 000 21.8792 −364.2990 0.5875 0.0036
8, 000 24.0138 −399.8403 0.6449 0.0036
9, 000 26.1483 −435.3817 0.7022 0.0036
10, 000 28.2829 −470.9231 0.7595 0.0036
where ak and bk are the slop and intercept and skr is the speed of a ship
of type k on route r. They are listed in the below Table 2.
Now the shipping cost in Eq. (2) can be rewritten as follows:
ckr (skr ) = λ̃kr /skr + λ̂kr (24)

xkr
Let zkr = skr (∀k, r), and substitute it into constraint (11), then we
have:
xkr xkr
U
≤ zkr ≤ L , ∀k, r (25)
Skr Skr
Finally, the new MILP model with chance constraints is proposed as
follows:

min = (λ̃kr zkr + λ̂kr xkr + ek ykr + nin
kr ck ) −
in
nout
k ck
out
(26)
r∈R k∈K k∈K
s.t. constraints (4) to (12).
5. Numerical Example
In this section, a one-year planning for a fleet consisting of 5 types of ships
operated on 8 liner trade routes by OOCL is studied: Central China Express
(CCX), China Pakistan Express (CPX), Gulf Indian Subcontinent Strait
Service (GIS), India US Express (IDX), North & Central China East Coast
Express (NCE), New Zealand Express (NZX), South China East Coast
Table 3 Ports calling sequences in the liner trade routes.
Routes code Port calling sequence and number code
1 (CCX) Los Angeles/Oakland/Pusan/Dalian/Xingang/Qingdao/Ningbo/

Shanghai/Pusan/Los Angles (1-2-3-4-5-6-8-7-3-1)
2 (CPX) Shanghai/Ningbo/Shekou/Singapore/Karachi/Mundra/Penang/
PortKelang/Singapore/Hong Kong/Shanghai (1-2-4-5-9-8-7-6-5-3-1)
3 (GIS) Singapore/Port Kelang/Nhava Sheva/Karachi/Jebel Ali/Bandar
Abbas/Jebel Ali/Mundra/Cochin/Singapore (1-2-4-6-7-8-7-5-3-1)
4 (IDX) Colombo/Tuticorin/Cochin/Nhava heva/Mundra/Suez/Barchlona/
NewYork/Norfolk/Charleston/Barcelona/Suez/Colombo
(1-2-3-4-5-6-7-8-9-10-7-6-1)
5 (NCE) New York/Norfolk/Savannah/Panama/Pusan/Dalian/Xingang/
Qingdao/Ningbo/Shanghai/Panama/New York
(1-2-3-4-5-6-7-8-9-10-4-1)
6 (NZX) Singapore/Port Kelang/Brisbane/Auckland/Napier/Lyttelton/
Wellington/Brisbane/Singapore (1-2-3-4-5-7-6-3-1)
7 (SCE) New York/Norfolk/Savannah/Panama/Kaohsiung/Shekou/Hong
Kong/Panama/New York (1-2-3-4-5-6-7-4-1)
8 (UKX) Southampton/Hull/Grangemouth/Southampton (1-2-3-1)
Source: OOCL website
Table 4 Types of ships with different capacities in the

example.
Type Capacity (TEU) Nkmax N CIkmax
1 (ICE class) 2,808 1 10

2 (F class) 3,218 1 10
3 (P class) 4,500 9 25
4 (S class) 5,714 2 25
5 (SX class) 8,063 12 10
Note: the data in the first three columns is from OOCL

website, the fourth column is hypothetical.
Express (SCE), UK Express (UKX); ICE class, F class, P class, S class and
SX class of ships. The calling ports of each route and the associated number
codes are listed in Table 3 and the types of ships are shown in Table 4. The
data of costs are hypothetical due to the real data is not available.
Our discussion is twofold: first, 8 cases are tested to show the advantage
of the model developed in this paper within consideration of the fleet
size and mix, the cruising speed of ships and the liner service frequency
simultaneously over those without. The comparison are shown in the
Table 5; second, two cases are tested to discuss the effect of risk level on
the fleet plan, shown in Fig. 7 and 8.
Table 5 Comparison of the 8 cases.
Fleet size Optimal Difference Proportion

Case Frequency and mix Speed solution with case 1 with case 1
1 × × × 415,867,758.4 0 100%
√
2 × × 471,188,247.5 55,320,489.1 113.3%
√
3 × × 461,005,592.1 45,137,833.7 110.8%
√ √
4 × 482,959,681.9 67,091,923.5 116.1%
√
5 × × 489,881,584.6 74,013,826.2 117.8%
√ √
6 × 479, 379,980 63,512,221.6 115.3%
√ √
7 × 489,887,214 74,019,455.6 117.8%
√ √ √
8 — — —
√
Note: decision variables are denoted by ×, otherwise, .
From the Table 5, we can find that the Case 1 is the cheapest, in which
the frequency, fleet and speed are taken into account. Compared Case 1
and Case 2, speed is decision variable in Case 1 while it is not in Case 2,
the optimal solution to Case 2 is the feasible solution to Case 1, thus, the
objective value of Case 2 is larger than that of Case 1. Similarly, if the
fleet size and mix are also predetermined and fixed (Case 3 and Case 4),
Case 4 has a larger optimal solution compared with Case 3. For Case 5 and
Case 6, we can get the same conclusion if we set the same value of frequency
in the two cases. If they are different, it is possible that Case 6 is more
economical than Case 5. The similar conclusion can be obtained for Case 7
and Case 8 by the same deduction. It is noticed that there is no feasible
solution to Case 8, in which the frequency, fleet size and mix and speed
are all predetermined and fixed. This implies that whatever deployment
of the fleet, it can not satisfy the demand requirement. If we decrease the
demand, it may have feasible solution, depending on the cargo demand.
In conclusion, any cases without taking the frequency, fleet size and mix
and speed into account simultaneously are not the best fleet planning. The
solution obtained by CPLEX of Case 1 is shown in Table 6.
In order to observe the role of α, the risk level, plays on the fleet
plan, two examples with different transportation capacity of current fleet
are generated and the corresponding results are shown in the following
Fig. 7 and 8. These Figs. roughly indicate that with the decrease of α, the
number of chartered-out ships has the decreasing trend, while the number of
chartered-in ships has the increased trend. Compared the Figs., we found
that Fig. 7(a∼c) are more fluctuant than Fig. 8(a∼c), this implies that
α effects on the first example more than that on the second one. The
transportation capacity of the current fleet in the first example is tighter
June 14, 2011
250
11:46
Table 6 Solution to case 1.
9in x 6in
Ship types: 1 (ICE), 2 (F), 3 (P), 4 (S), 5 (SX)
Number of ships of each
type on each route
No. of ships charter-out
Q. Meng, T. S. Wang and G. Shahin

Routes code Fr No. of voyages Lay-up days Speeds Own Charter-in of each type
Advances in Maritime Logistics & Supply . . .

1 (CPX) 3 74, –,–, –,47 45,–,–,–,60 15,–,–,–,15 –,–,–,–,– 3,–,–, –,2 –,–,1,–,10
2 (GIS) 5 51,21,–,–,– 30,15,–,–,– 18.21,15,–,–,– –,1,–,–,– 2,–,–,–,–
3 (IDX) 4 76,–,–,–,15 45,–,–,–,30 21.11,–,–,–,19.18 –,–,–,–,– 3,–,–,–,1
4 (NCE) 9 19,–,–,–,21 15,–,–,–,30 17.86,–,–,–,20.62 1,–,–,–,1 –,–,–,–,–
5 (SCE) 5 –,–,72,–,– –,–,90,–,– –,–,17.19,–,– –,–,6,–,– –,–,–,–,–
6 (SIX) 16 –,2,–,–,20 –,15,–,–,60 –,15,–,–,16 –,–,–,–,1 –,1,–,–,1
7 (SBX1) 5 –,–,61,11,– –,–,75,20,– –,–,17.93,16.4,– –,–,–,1,– –,–,5,–,–
8 (SBX2) 5 10,–,53,9,– 15,–,90,20,– 24.35,–,21.51, 22.23,– –,–,2,1,– 1,–,4,–,–
The number of own ships are used: 14
The number of charter-in ships is: 23
The number of charter-out ships is: 11
The objective value is: 415,867,758.4
Note: The elements of the five-dimensional vectors in the last six columns denote the correspond value to the ship of the five
types; — denotes not available.
b1185-ch11
5
Route1
4.5 Route2
Route3
4 Route4
Route5
3.5 Route6
Route7
3
Nr. Vessels
Route8
2.5
1.5
0.5
0
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 7(a) Deployment of chartered-in ships in Example 1 with tight capacity.
20
Route1
Route2
19
Route3
Route4
18 Route5
Route6
17 Route7
Nr. Vessels
Route8
16
15
14
13
12
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 7(b) Deployment of own ships in Example 1 with tight capacity.

7
Route1
Route2
6 Route3
Route4
Route5
5
Route6
Route7
Nr. Vessels
4 Route8
0
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 7(c) Deployment of chartered-out ships in Example 1 with tight capacity.
1
Route1
0.8 Route2
Route3
0.6 Route4
Route5
0.4 Route6
Route7
0.2
Nr. Vessels
Route8
0
-0.2
-0.4
-0.6
-0.8
-1
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 8(a) Deployment of chartered-in ships in Example 2 with loose capacity.

26
Route1
Route2
Route3
24
Route4
Route5
Route6
22
Route7
Nr. Vessels
Route8
20
18
16
14
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 8(b) Deployment of own ships in Example 2 with loose capacity.
12
Route1
Route2
Route3
10
Route4
Route5
Route6
8
Route7
Nr. Vessels
Route8
0
0.5 0.6 0.7 0.8 0.9 1
1-α
Fig. 8(c) Deployment of chartered-out ships in Example 2 with loose capacity.

than that in the second example, especially for the number of chartered-in
ships. (see Fig. 7(a) and Fig. 8(a)).
6. Summary and Conclusion

In this paper, we studied the liner fleet deployment problem within the
cargo demand uncertainty. We first reviewed the previous related researches
and addressed their limitations. The risk level is introduced to describe the
probability that a liner operator could not satisfy the shippers’ requirement.
Then we built a mixed integer nonlinear programming model with chance
constraints. Taken the fact that the fuel consumption cost for a ship can
be approximated by a linear function with respect to its cruising speed,
we proceed to build a mixed integer linear programming model that can
approximate the mixed integer nonlinear programming model proposed in
this study. Numerical examples were generated to show the application of
this model and the CPLEX solver was employed to solve them. Lastly, the
effect of risk level on the fleet deployment was studied. Generally we found
that when the transportation capacity of the current fleet was tight, the
fleet deployment was more sensitive to the risk level compared with that
when the transportation capacity of the current fleet was loose. Though
the GAMS solver can be applied into our model, it can not solve the model
with large size. How to efficiently solve the problem is still worthwhile to
be studied.
Acknowledgement
This paper was supported partially by the Neptune Orient Line Company
Research Grant No. R-264-000-244-720.
References
1. UNCTAD, Review of Maritime Transportation, Paper presented at the
United Nations Conference on Trade and Development. New York and
Geneva, USA. http://www.unctad.org/en/docs/rmt2008 en.pdf (2008).
2. OOCL, Sailing Schedule. Available from: http://www.oocl.com/eng/
ourservices/eservices/sailingschedule/ (2009)
3. Ronen, D. (1983). Cargo ships routing and scheduling: Survey of models and
problems, Eur. J. Oper. Res., 12(2), pp. 119–126.
4. Ronen, D. (1993). Ship scheduling: The last decade, Eur. J. Oper. Res.,
71(3), pp. 325–333.
5. Perakis, A.N. (2002). Fleet operation optimization and fleet deployment.
Costas Th.
6. Christiansen, M., Fagerholt, K. and Ronen, D. (2004). Ship routing and

scheduling: status and perspectives, Transport. Sci., 38(1), pp. 1–18.
7. Dantzig, G.B. and Fulkerson, D.R. (1954). Minimizing the number of tankers
to meet a fixed schedule, Nav Res. Logist. Q., 1, pp. 217–222.
8. Laderman, J., Gleiberman, L. and Egan, J.F. (1966). Vessel allocation by
linear programming, Nav Res. Logist. Q., pp. 315–320.
9. Everett, J.L., Hax., A.C., Lewinson, V.A. and Nudds, D. (1972). Opti-
mization of a Fleet of Large Tankers and Bulkers: A linear Programming
Approach, Marine Tech., October, pp. 430–438.
10. Benford, H. (1981). A simple approach to fleet deployment, Maritime Pol.
Mng., 8(4), pp. 223–228.
11. Perakis, A.N. (1985). A second look at fleet deployment, Maritime Pol. Mng.,
12(3), pp. 209–214.
12. Lawrence, S.A. (1972). International Sea Transport: The Years Ahead.
Lexington Books, 1st ed. Lexington, MA.
13. Stopford, M. (1997). Maritime Eco. 2nd ed. London: Routledge.
14. Lane, D.E., Heaver, T.D. and Uyeno, D. (1987). Planning and scheduling for
efficiency in liner shipping, Maritime Pol. Mng., 12(3), pp. 109–125.
15. Fagerholt, K. (1999). Optimal fleet design in a ship routing problem,
International Transactions in Oper. Res., 6(5), pp. 453–464.
16. Hsu, C.I. and Hsieh, Y.P. (2007) Routing, ship size and sailing frequency
decision-making for a maritime hub-and-spoke container network, Math
Comput Model, 45, pp. 89–916.
17. Brandao, J. (2009). A deterministic tabu search algorithm for the fleet size
and mix vehicle routing problem, Eur. J. Oper. Res., 195(3), pp. 716–728.
18. Perakis, A.N. and Papadakis, N. (1987). Fleet deployment optimization
models Part 1, Maritime Pol. Mng., 14(2), pp. 127–144.
19. Perakis, A.N. and Papadakis, N. (1987). Fleet deployment optimization
models Part 2, Maritime Pol. Mng., 14(2), pp. 145–155.
20. Papadakis, N. and Perakis, A.N. (1989). A Nonlinear Approach to the
Multiorigin, Multidestination Fleet Deployment Problem, Nav Res. Log., 36,
pp. 515–528.
21. Perakis, A.N. and Jarammillo, D.I. (1991). Fleet deployment optimization
for liner shipping Part 1. Background, problem formulation and solution
approaches, Maritime Pol. Mng., 18(3), pp. 183–200.
22. Jarammillo, D.I. and Perakis, A.N. (1991). Fleet deployment optimization
for liner shipping Part 2. Implementation and results, Maritime Pol. Mng.,
18(3), pp. 235–262.
23. Powell, B.J. and Perakis, A.N. (1997). Fleet deployment optimization for
liner shipping: an integer programming model, Maritime Pol. Mng., 24(2),
pp. 183–192.
24. Millar, H.H. and Gunn, E.A. (1991). Dispatching a fishing trawler fleet
in the Canadian R. Agarwal, O., Ergun, , 2008. Ship Scheduling and
Network Design for Cargo Routing in Liner Shipping, Transport. Sci, 42(2),
pp. 175–196.
25. Vukadinović, K. and Teodorović, D. (1994). A fuzzy approach to the vessel

dispatching problem, Eur. J. Oper. Res., 76(1), pp. 155–164.
26. Cho, S.C. and Perakis, A.N. (1996). Optimal liner fleet routing strategies,
Maritime Pol. Mng., 23(3), pp. 249–259.
27. Mourão, M.C., Pato, M.V. and Paixão, A.C. (2001). Ship assignment with
hub and spoke constraints, Maritime Pol. Mng., 29(2), pp. 135–150.
28. Nicholson, T.A.J. and Pullen, R.D. (1971). Dynamic Programming Applied
to Ship Fleet Management, Oper. Res. Quart, 22(3), pp. 211–220.
29. Xie, X.L., Wang, T.F. and Chen, D.S. (2000). A dynamic model and
algorithm for fleet planning, Maritime Pol. Mng., 27(1), pp. 53–63.
30. Marcus, H. (2009). NOL sees box volumes crash 24% in weeks. Avail-
able from: http://www.lloydslist.com/ll/news/viewArticle.htm?articleId=
20017612522.
31. Agarwal, R. and Ergun, O. (2008). Ship schedule and network design for
cargo routing in liner shipping, Transport. Sci, 42(2), pp. 175–196.
32. COSCO, released magnitude expansion treatment. Available from: http://
www.shippingchina.com/static/brmaininfo/200708/32899.shtml. (2007).
33. EMC, Evergreen Marine Corporation Negotiate Renting 126000 TEU Ships.
Available from: http://news.gcdcs.com/article/sort015/info-10505.html.
(2007)
34. Cullinane, K. and Khanna, M. (1999). Economics of scale in large container
ships, J Transp. Econ Policy, 33(2), pp. 185–208.
35. Drewry (2001). Ship Operating Costs Annual Review and Forecast. Drewry
Shipping Consultants.
36. Shintani, K., Imai, A., Nishimura, E. and Papadimitriou, S. (2007). The con-
tainer shipping network design problem with empty container repositioning,
Transport Res E-LOG, 43(1), pp. 39–59.
37. Takano, K. and Arai, M. (2008). A genetic algorithm for the hub-and-spoke
problem applied to containerized cargo transport, J Maritime Sci Tech, (in
press, 2008).
38. MAN B&W Diesel A/S, Propulsion Trends in Container Vessels.
Available from: http://www.manbw.com/files/news/filesof4672/P9028.pdf.
Copenhagen, Denmark (2004).
CHAPTER 12
SHIP EMISSIONS, COSTS

AND THEIR TRADEOFFS
Harilaos N. Psaraftis and Christos A. Kontovas
Laboratory for Maritime Transport,
National Technical University of Athens,
9, Iroon Polytechneiou Str.,
GR-157-73 Athens, Greece
[email protected] [email protected]
Emissions from commercial shipping are currently the subject of intense

scrutiny. Various analyses of many aspects of the problem have been and are
being carried out and a spectrum of measures to reduce emissions is being
contemplated. However, such measures may have important side-effects as
regards the logistical supply chain, and vice-versa. Industry circles have also
voiced the concern that low-sulphur fuel in SECAs (the so-called ‘sulphur
emissions control areas’) may make maritime transport (and in particular
short-sea shipping) more expensive and induce shippers to use land-based
alternatives. A reverse shift of cargo from sea to land might ultimately increase
the overall level of CO2 emissions along the intermodal chain. This paper takes
a look at various tradeoffs and may impact the cost-effectiveness of the logistical
supply chain and present models that can be used to evaluate these tradeoffs.
One of the key results is that speed reduction will always result in a lower
fuel bill and lower emissions, even if the number of ships is increased to meet
demand throughput. Another result is that cleaner fuel at SECAs may result
in a reverse cargo shift from sea to land that has the potential to produce more
emissions on land than those saved at sea. Various examples are presented.
1. Introduction
Air pollution from ships is currently at the center stage of discussion by
the world shipping community and environmental organizations. The Kyoto
protocol to the United Nations Framework Convention on Climate Change
-UNFCCC (1997) stipulates concrete measures to reduce CO2 emissions in
order to curb the projected growth of greenhouse gases (GHG) worldwide.
257
258 H. N. Psaraftis and C. A. Kontovas
Although some regulation exists for non-GHGs, such as SO2 , NOx and
others, shipping has thus far escaped being included in the Kyoto global
emissions reduction target for CO2 and other GHGs (such as CH4 and
N2 O). Even so, it is clear that the time of GHG non-regulation is rapidly
approaching its end, and measures to curb future CO2 and other GHG
growth are being sought with a high sense of urgency and are very high
on the agenda of the International Maritime Organization (IMO) and of
many individual coastal states. In the forthcoming UNFCCC, which will
take place in Copenhagen in December of 2009, shipping is expected to
be included in the discussions on future GHG reduction. In that sense,
various analyses of many aspects of the problem have been and are being
carried out and a broad spectrum of measures is being contemplated. These
measures can be considered to fall into three general categories: technical,
market-based and operational.
Technical measures include more efficient ship hulls, energy-saving
engines, more efficient propulsion, use of alternative fuels such as fuel
cells, biofuels or others, “cold ironing” in ports (providing electrical supply
to ships from shore sources), devices to trap exhaust emissions (such as
scrubbers), and others, even including the use of sails to reduce power
requirements. Market-based instruments (MBIs) are classified into two
main categories, Emissions Trading Schemes (ETS) and Carbon Levy
schemes (also known as International Fund schemes). Finally, operational
schemes mainly involve speed optimization, optimized routing, improved
fleet planning, and other, logistics-based measures.
Some of these measures, important in their own right as regards
emissions reduction, may have non-trivial side-effects as regards the
logistical supply chain. For instance, measures such as (a) reduction of
speed, (b) change of number of ships in the fleet, (c) possibly others, will
generally entail changes (positive or negative) in overall emissions, but
also in other logistics and cost-effectiveness attributes such as in-transit
inventory and other costs. Also, industry circles have voiced the concern
that the mandated use of lower-sulphur fuel in some regions or globally
may make maritime transport (and in particular short-sea shipping) more
expensive and induce shippers to use land-based alternatives (mainly road).
A reverse shift of cargo from sea to land would go against the drive to shift
traffic from land to sea to reduce congestion, and might ultimately increase
the overall level of CO2 emissions along the intermodal chain. In that
regard, in Europe one can already see a potential conflict between two
policies: (a) the designation of certain areas as “sulphur emissions controlled
Ship Emissions, Costs and Their Tradeoffs 259
areas” (or SECAs), such as the Baltic Sea, the North Sea and the English
Channel, and (b) the stated Transport Policy goal of shifting cargo off the
roads and onto ships and railways.
Typical problems in the maritime logistics area include one or a combi-
nation of problems from the following generic list (which is non-exhaustive):
• Optimal ship speed

• Optimal ship size
• Routing and scheduling
• Fleet deployment
• Fleet size and mix
• Weather routing
• Intermodal network design
• Modal split
• Transshipment
• Queuing at ports
• Terminal management
• Berth allocation
• Supply chain management
The traditional analysis of these problems is in terms of cost- benefit criteria

from the point of view of the logistics, operator, shipper, or other end-
user. Such analysis typically ignores environmental issues. Green maritime
logistics tries to bring the environmental dimension into the problem, and
specifically the dimension of emissions reduction, by trying to analyze the
tradeoffs that are at stake and exploring win-win solutions.
It is also important to realize that two different settings can be
analyzed, the strategic setting and the operational one. The distinction
between the two is important, and one that is not mentioned frequently.
Let us clarify the difference between the two by an example.
A spokesman from Germanischer Lloyd (GL) has been recently quoted
as follows: “We recommend that ship-owners consider installing less
powerful engines in their newbuildings and to operate those container
vessels at slower speeds,” (Lloyds List, 2008a). By ‘slower speeds’ it is
understood that the current regime of 24–26 knots would be reduced to
something like 21–22 knots. But some trades may go as low as 15–18 knots,
according to a 2006 study by Lloyds Register (Lloyds List, 2008b). An
obvious reason for suggesting such speed reduction is twofold: fuel costs
and emissions.
Implementing the aforementioned speed reduction would only make

sense in a strategic setting, by modifying the design of the ship, including
hull shape, by installing smaller engines in future newbuildings, by modify-
ing the propeller design, etc. In such a setting however, one would have to
also investigate not only differences in emissions produced by these modified
lower-speed designs, but also other possible ramifications. These may
include emissions differentials by the shipyards that produce these ships,
as well as any difference in emissions when these ships would be recycled.
This strategic approach to the emissions problem is also known as the
‘life-cycle’ approach. It is an important component in the quest to formulate
possible strategic decisions and policies to curb emissions from shipping in
the long run.
It is not the scope of this paper to examine all of the problems
identified above from an environmental perspective. That will take years to
accomplish. Rather, the limited number of models examined in this paper
primarily focus on operational scenarios and mainly serve to highlight some
of the trade-offs that are at stake in these scenarios, so as to motivate further
work in this area.
The rest of this paper is organized as follows. Section 2 reports on
relevant background. Section 3 describes some basics on emissions. Section 4
describes a simple logistical scenario to investigate the effects of speed
reduction. Section 5 introduces the concept of the cost to avert a tonne
of CO2 and Section 6 examines the issue of port time in the quest to reduce
emissions. Section 7 examines the effect of speed reduction at SECAs and
Section 8 looks into possible side-effects of cleaner fuels on modal split.
Finally Section 9 presents the paper’s conclusions.
2. Background
We start by stating that even though the literature on the broad area
of ship emissions is immense, the literature on the specific topic (link
between emissions and maritime logistics) is scant. There are a number
of papers that consider the economic impact of speed reduction especially
for container vessels. Andersson (2008) considered the case of a container
line where the speed for each ship reduced from 26 knots to 23 knots and
one more ship was added to maintain the same throughput. Total costs
per container were reduced by nearly 28 per cent. Eefsen (2008) considered
the economic impact of speed reduction of containerships and included the
inventory cost. Cerup-Simonsen (2008) developed a simplified cost model
to demonstrate how an existing ship could reduce its fuel consumption

by a speed reduction in low and high markets to maximize profits.
Corbett et al. (2009) applied fundamental equations relating speed, energy
consumption, and the total cost to evaluate the impact of speed reduction.
The paper also explored the relationship between fuel price and the optimal
speed.
The situation is similar at the policy level: many activities, but little or
nothing relating to the interface between emissions and logistics. Looking
at developments at the IMO (International Maritime Organization) level,
thus far progress as regards air pollution from ships has been mixed
and rather slow. On the positive side, in November 2008 the Marine
Environment Protection Committee (MEPC) of the IMO unanimously
adopted amendments to the MARPOL Annex VI regulations. The main
changes will see a progressive reduction in sulphur oxide (SOx) emissions
from ships, with the global sulphur cap reduced initially to 3.50%, effective
1 January 2012; then progressively to 0.50%, effective 1 January 2020 (IMO,
2008a).
Furthermore, the report of Phase 1 of the update the 2000 IMO GHG
Study (IMO, 2000) was presented, which was conducted by an international
consortium led by Marintek, Norway (Buhaug, et al., 2008). According
to this study, total CO2 emissions from shipping (both domestic and
international) are estimated to range from 854 to 1,224 million tons (2007),
with a ‘consensus estimate’ set at 1,019 million tons, or 3.3% of global CO2
emissions. By comparison, electricity and heat production accounts for 35%
of global CO2 emissions, manufacturing industries and construction 18.2%,
and transport (all modes) 21.7%. Among transport modes, road accounts
for 51% of all CO2 emissions, shipping (including fishing) for 25%, aviation
for 20%, and rail for 4%. However, in terms of energy use and emissions
per tonne-km, shipping ranks as the most environment-friendly transport
mode, as can be seen in the following table.
Among ship types, according to the results of Phase 1, the three top
fuel consuming categories of ships (and thus, those that produce most of the
CO2 emissions) are (i) container vessels of 3,000–5,000 TEUs, (ii) container
vessels of 5,000–8,000 TEUs and (iii) RoPax Ferries with cruising speed
of less than 25 knots. The common denominator of these three categories,
which results in a high level of CO2 emissions, is their high speed, at least
as compared to other ship types.
These findings are in line with those of Psaraftis and Kontovas
(2008, 2009a). According to their analysis, containerships are the top CO2
Table 1 Energy efficiency and emissions to the atmosphere (by mode).
PS-Type S-Type
container container
vessel vessel
(11,000 (6,600 Rail — Rail — Heavy Boeing
Energy TEU) TEU) electric diesel truck 747-400
use
kWh/tkm 0.014 0.018 0.043 0.067 0.18 2.00
PS-Type S-Type
container container
vessel vessel
Emissions (11,000 (6,600 Rail — Rail — Heavy Boeing
(g/tkm) TEU) TEU) electric diesel truck 747-400
Carbon 7,48 8.36 18 17 50 552

dioxide
(CO2 )
Sulphur 0.19 0.21 0.44 0.35 0.31 5.69
oxides
(SOX )
Nitrogen 0.12 0.162 0.10 0.00005 0.00006 0.17
oxides
(NO)
Particulate 0.008 0.009 n/a 0.008 0.005 n/a
mat-
ters
(PM)
Re. emissions for rail; the complete value chain for el-production is considered.
Source: Network for Transport and the Environment (Sweden).
emissions producer in the world fleet (2007, Lloyds-Fairplay database). Just

the top tier category of container vessels (those of 4,400 TEU and above)
are seen to produce CO2 emissions comparable on an absolute scale to
that produced by the entire crude oil tanker fleet (in fact, the emissions of
that top tier alone are slightly higher than those of all crude oil tankers
combined- see Fig. 1 above).
At the latest meeting of IMO’s Marine Environment Protection
Committee in London last July (MEPC 59) there continued to be a clear
split between industrialized member states, such as Japan, Denmark and
other Northern European countries, and a group of developing countries
including China, India and Brazil, on how to proceed. The latter countries
spoke in favor of the principle of “Common but differentiated responsibility”
(CBDR) under the UNFCCC. In their view, any mandatory regime aiming
Fig. 1 CO2 emissions, world fleet (Psaraftis and Kontovas, 2009a).

to reduce GHG emissions from ships engaged in international trade should

be applicable exclusively to the countries listed in Annex I to the UNFCCC,
therefore their strong wish is not to be included in any mandatory set
of measures.
Due to ‘political’ reasons such as above, progress as regards regulating
CO2 and other GHGs continues to be very slow. In fact, the stated objective
to finalize a mandatory Energy Efficiency Design Index (EEDI) of the
environmental performance of new ships has not been reached yet. The same
is true for the Energy Efficiency Operational Indicator (EEOI), which will
be applicable to all ships. As a result, the IMO will not be in a position to
have reached a clear position on these two indices in time for the United
Nations Framework Conference for Climate Change (UNFCCC) that will
be held in Copenhagen in December of this year, when a new climate
agreement is expected to be reached, after Kyoto in 1997.
Without going into technical details regarding these two indices, one
can state that the first index (EEDI) concerns the design of new ships and
the second (EEOI) concerns the operation of all ships, new and existing.
Both indices are ratios, in which the numerator is a complex function of
all energy consumed by the ship, and the denominator includes a product
of the ship’s deadweight (or payload) and the ship’s operational speed.
The fact that speed is in the denominator means that the slower the
ship goes, the higher both these indices will be, therefore the higher the
ship will be ranked in terms of energy efficiency, both for design and for
operation. No doubt about it, faster ships will score low as regards these
indices.
The implication of this is unknown, other than the fact than in
any ranking based on these indices, fast ships will have an unfavorable
environmental performance vis-à-vis slower ships of the same capacity. In
spite of extensive discussions on this topic, it is still not clear exactly how
these indices will be used in future IMO rulemaking. In fact, these indices
still have not been finalized, as certain issues still demand discussion and
agreement.
Progress as far as other measures to regulate GHG emissions, such as
MBIs has been even slower. Reaction to this concept has been even more
pronounced, and it is not clear which among two main schemes, the
Emissions Trading Scheme (ETS) and the Carbon Levy, will be eventually
adopted. Certainly no agreement will be reached before the Copenhagen
UNFCCC conference, and the latest IMO timetable on this issue goes
into 2012.
What does slow progress on GHGs mean? And what if no agreement

is reached at the IMO any time soon? This will certainly increase the
pressure for regional approaches. In fact the European Commission is
following IMO developments very closely, and has stated very clearly its
intention to act alone if IMO’s procedures take longer than previously
anticipated. As regards GHGs, the anticipated approach of the Commission
is to formulate an ETS, similar to that used in other land-based industries.
The Commission has started the procedure for including air transport into
its ETS scheme, and many think it will eventually do the same for shipping.
Many ship owners’ circles have voiced strong concerns that such a scheme
would be complicated and unworkable.
Currently, European legislation mainly concerns the sulphur content of
marine fuels. The maximum sulphur content for marine fuels according
to EU directive 2005/33/EC is in line with MARPOL Annex VI. The
implementation dates are differently from those agreed by the IMO under
MARPOL Annex VI, but the main point is that currently all vessels sailing
in the designated areas (SECAs) should use marine fuels with a maximum
of 1.5% by mass content of sulphur. What is different from MARPOL is
that the EU Directive sets a limit for all passenger vessels operating on
regular service to or from EU ports to a maximum sulphur content of
1.5% (the same as in SECAs). This limit came into effect on August 11th,
2006 (EU directive 2005/33/EC, Article 4a). Furthermore, according to
Article 4b of the same Directive, from January 1st, 2010 a 0.1% limit comes
into effect for inland waterway vessels and ships at berth in EU ports with
some exemptions.
Perhaps more interesting are developments on the logistics side: the
European Commission states in their Freight Transport Logistics Action
Plan launched in October 2007 that “Logistics policy needs to be pursued
at all levels of governance”, which is also the reason behind this action
plan as one in a series of policy initiatives to improve the efficiency and
sustainability of freight transport in Europe. In the Freight Transport
Logistics Action Plan a number of short — to medium-term actions is
presented that will help Europe address its current and future challenges
and ensure a competitive and sustainable freight transport system in
Europe. Among the actions are the “Green transport corridors for freight”.
The Green Corridors are characterized by a concentration of freight traffic
between major hubs and by relatively long distances of transport. Green
Corridors should in all ways be environmentally friendly, safe and efficient.
This is perhaps one of the few EU policy initiatives that aim to establish
a clear connection between environment and logistics, even though this

activity is still very much at its infancy. It is clear that the maritime
mode will be involved in some of these Green Corridors, particularly
those involving the Trans European Transport Networks (TEN-T’s) and
the Motorways of the Sea, and the question is, what ships, what types,
what sizes, what speeds, how will they be utilized, and how will tradeoffs
will be assessed.
In the United States, the Environmental Protection Agency (EPA) has
established a tier-based timeline for implementing NOx emission standards
to marine diesel engines that became effective in 2007. These standards are
similar to those described in MARPOL Annex VI which has been ratified
by the US in October 2008 although the Convention entered into force in
May 2005. Canada has not yet ratified Annex VI, however Canada and
the United States jointly proposed the designation of an Emissions Control
Area (ECA) for specified portions of the US and Canadian coastal waters
covering a total o 200 nm. At MEPC 59, the proposal was agreed in principle
and will be voted during MEPC 60, scheduled for March 2010. If approved
the ECA would enter into force in 2012.
On a local government basis, the State of California which is the
home of the two busiest ports in the US has created a special agency,
the California Air Resources Board (CARB) which is the primary source
for ship emission regulations in California. On July 2008, CARB adopted
the regulation “Fuel Sulfur and Other Operation Requirements for Ocean-
Going Vessels within California Waters and 24 Nautical Miles of the
California Baseline” that sets specific limits on the sulfur content of fuel
used within 24 nm of the Californian coast.
In addition, the two busiest ports in the US (Long Beach and Los
Angeles) both located in Southern California have introduced a series
of voluntary incentive-based programs. On March 2008, the Board of
Commissioners of the ports of Los Angeles (POLA) and Long Beach
(POLB) authorized the Low-Sulphur Vessel Main Engine Fuel Incentive
program to encourage operators to use cleaner fuels within 40 nm or 20 nm
from Point Fermin. The program will pay the operators that will agree to
use fuels that contain less than 0.2% sulphur the price difference between
that fuel and IFO 380. Furthermore, the two ports offer a 15% discount on
dockage fees to vessels that voluntary comply with the SPBP-OGV1 Vessel
Speed Reduction Program and reduce their speed to 12 knots within 20 nm
of Point Fermin while entering or leaving the ports.
3. Some Basics: Algebra of Emissions and Fuel Cost

Before logistical scenarios are examined, some basics have to be established
first. Two are the main attributes of any logistical scenario that is viewed
from a green perspective: the amount of emissions produced, and the cost.
To calculate CO2 emissions, one has to multiply bunker consumption
by an appropriate emissions factor, FCO2 . The factor of 3.17 has been the
empirical mean value most commonly used in CO2 emissions calculations
based on fuel consumption (see EMEP/CORINAIR (2002) and Endresen
(2007)). According to the IMO GHG study (IMO, 2000), the actual value
of this coefficient may range from 3.159 (low value) to 3.175 (high value).
The update of the IMO 2000 study (Buhaug et al., 2008), uses slightly
lower coefficients, different for Heavy Fuel Oil and for Marine Diesel
Oil. The actual values are 3.082 for Marine Diesel and Marine Gas Oils
(MDO/MGO) and 3.021 for Heavy Fuel Oils (HFO). According to the
report of the Working Group on Greenhouse Gas Emissions from Ships
(IMO, 2008b), the group agreed that the Carbon to CO2 conversion factors
used by the IMO should correspond to the factors used by IPCC (2006
IPCC Guidelines) in order to ensure harmonization of the emissions factor
used by parties under the UNFCCC and the Kyoto Protocol. In this paper
we shall use the original value of 3.17 also used in Psaraftis and Kontovas
(2008, 2009a) except for the example in Section 6 where the value of 3.13
has been used, noting that our emissions results will have to be scaled down
by up to 5% if a lower emissions factor is used. Table 2 summarizes various
emissions factors.
As regards SO2 , this type of emissions depends on the type of fuel used.
One has to multiply total bunker consumption (in tonnes per day) by the
percentage of sulphur present in the fuel (for instance, 4%, 1.5%, 0.5%,
Table 2 Comparison of emission factors kg CO2 /kg fuel. (IMO, 2008b).
IPCC 2006 Guidelines

GHG-WG Revised 1996
Fuel type 1/3/1 Default Lower Upper guidelines
Marine diesel and 3.082 3.19 3.01 3.24 3.212

marine gas oils
(MDO/MGO)
Low sulphur fuel oils 3.075 3.13 3.00 3.29
(LSFO)
High sulphur fuel 3.021
oils (HSFO)
or other) and subsequently by a factor of 0.02 to compute SO2 emissions

(in tonnes per day). The factor of 0.02 is exact, and is derived from the
chemical reaction of sulphur with oxygen.
Finally, NOx emissions depend on engine type. The ratio of NOx
emissions to fuel consumed (tonnes per day to tonnes per day) ranges
from 0.087 for slow speed engines to 0.057 for medium speed engines. Also
directly proportional to the amount of fuel used is fuel cost, one of the
most important components of total cost (although by no means the only
one). Fuel cost can be estimated by multiplying the amount of bunkers used
with the price of fuel. In our analysis we assume that the price of the fuel
used by the ship is known and equal to p, assumed constant during the
year. Even though it is assumed a constant in our analysis, p is very much
market-related, and, as such, may fluctuate widely in time, as historical
experience has shown (see Fig. 2 below). But this assumption causes no loss
of generality, as an average price can be used. Also, as the ship will generally
consume different kinds of fuels during the trip and in port, assuming a
unique fuel price is obviously a simplification. But this causes no loss of
generality either, as the analysis can be readily extended to account for
different fuel types on board.
4. A Simple Logistical Scenario: Factors and Tradeoffs

Given that fuel costs and emissions are directly proportional to one
another (both being directly proportional to fuel used), it would appear
that reducing both would be a straightforward way towards a “win-win”
solution. In an operational setting, one of the obvious tools for such a
simultaneous reduction is speed: sail slower, and you reduce both emissions
and your fuel bill. This may sound simple, but its possible ramifications are
not so simple.
Assuming a given ship, and for speeds that are close to the original
speed, the effect of speed change on fuel consumption is assumed cubic,
that is,
3
F V
=
Fo Vo
where F (F0 ) is the daily fuel consumption at speed V (V0 ).
This assumption comes from basic ship hydrodynamics. It means that
F = kV 3 , where k is a known constant, which is a function of the
loading condition of the ship and of other ship characteristics (e.g., engine,
Average monthly fuel oil prices (from www.bunkerworld.com).
Fig. 2
Fig. 3 Ship route.
horsepower, geometry, age, etc). Of course, an implicit assumption in this

analysis is that the ship’s power plant would still be able to function
efficiently if speed is reduced. Speed reduction usually requires reconfiguring
the engine so that its operation is optimized at the reduced load.
Also note that the cubic law is only an approximation, and one that is
usually valid for small changes in speed. If the speed changes drastically,
for instance from 20 to 10 or even 5 knots, one would expect a different
relationship between V and F.
Our simplest logistical scenario to investigate tradeoffs between ship
CO2 emissions and other attributes of the ship operation assumes a fleet
of N identical ships (N: integer), each of capacity (payload) W. Each ship
loads from a port A (time in port TA, , days), travels to port B with known
speed V1 , discharges at B (time in port TB , days) and goes back to port A
in ballast, with speed V2 . Assume speeds are expressed in km per day. The
distance between A and B is known and equal to L (km). Assume these ships
are chartered on a term charter and the charterer, who is the effective owner
of this fleet for the duration of the charter, incurs a known operational cost
of OC per ship per year. This cost depends on market conditions at the time
the charter is signed and includes the charter to the ship owner(s) and all
other non-fuel related expenses that the charterer must pay, such as canal
tolls, port dues, cargo handling expenses, and so on. Not included in OC
are fuel expenses, which are also paid by the charterer, and which depend
on the actual fuel consumed by the fleet of ships. The latter depends on
how the fleet is used.
Obviously, the above rudimentary scenario (a ship going fully laden
one way and on ballast on the return leg) is not the only one that one
may encounter in world shipping markets. This scenario is encountered
mainly in the charter market and specifically in the tanker trades. Bulk
carriers may also be employed likewise; however they are more likely to
also trade in triangular routes, depending on the cargoes that are available.
Containerships and other ships in the liner market definitely do not use
such employment pattern, being engaged in trades that visit many ports.
Even though these operational scenarios are different from the one examined
above, extending our approach to these other scenarios is straightforward,
and the main thrust of our analysis is valid for these scenarios as well.
Assume that each ship’s operational days per year are D (0 < D < 365),
a known input, and that the total daily fuel consumptions (including both
main engine and auxiliaries) are known and are as follows for each ship:
In port: f (tonnes per day)

At sea: F1 , F2 (tonnes per day) for laden and ballast legs (respectively).
As stated earlier, the effect of speed change on fuel consumption is

assumed cubic for the same ship, that is, Fnew /F = (Vnew /V )3 , or, F1 =
k1 V13 , F2 = k2 V23 , where k1 and k2 are known constants. Also as mentioned
in the previous Section, one tonne of fuel burned in the ship’s engine room
will produce FCO2 tonnes of CO2 , where FCO2 is the emissions factor.
In addition to the standard costs borne by the charterer, our analysis
will also take into account cargo inventory costs. The reason is that any
conceivable speed reduction to save fuel costs and/or reduce emissions will
have as a consequence an increase in inventory costs due to late delivery
of cargo and must be taken into account if the analysis is to be complete
from a logistical standpoint. These cargo inventory costs are assumed equal
to IC per tonne and per day of delay, where IC is a known constant. In
computing these costs, we assume that cargo arrives in port ‘just-in-time’,
that is, just when each ship arrives. In that sense, inventory costs accrue
only when loading, transiting (laden) and discharging. We shall call these
inventory costs ‘in-transit inventory costs’. Generalizing to the case where
inventory costs due to port storage are also considered is straightforward.
If the market price of the cargo at the destination (CIF price) is P
($/tonne), then one day of delay in the delivery of one tonne of this cargo
will inflict a loss of PR/365 to the cargo owner, where R is the cost of capital
of the cargo owner (expressed as an annual interest rate). This loss will be
in terms of lost income due to the delayed sale of the cargo. Therefore, it
is straightforward to see that IC = P R/365.
Based on the above, and on a per ship basis, and after some straight-
forward algebraic manipulations, we can compute the following:
Round trip duration:
d = L/V1 + L/V2 + TAB ,

where
TAB = TA + TB (total port time per round trip)
Number of round trips in a year: n = D/d

Therefore n = D/[L/V1 + L/V2 + TAB ] (note that n may not necessarily
be an integer)
Total roundtrip fuel consumption: TF C = TAB f + L(k1 V12 + k2 V22 )
[As a parenthesis, it can be seen here that although the per day fuel
consumption is a cubic function of speed, the roundtrip fuel consumption
is only a quadratic function of speed, as the slower the ship goes, the more
days it stays at sea.]
Total costs in a year:

L
pnT F C + nI C W TAB + + OC
V

2 2 L
= np[TAB f + L(k1 V1 + k2 V2 )] + nI C W TAB + + OC
V1

p[TAB f + L(k1 V12 + k2 V22 )] + IC W TAB + VL1
=D L L
+ OC
V1 + V2 + TAB.
Fuel consumed per tonne-km: TF C /WL

For a fleet of N ships, total fleet costs in a year:

L
pnN T F C + nN kW TAB + + N OC
V

L
= nN p[TAB f + L(k1 V12 + k2 V22 )] + nN I C W TAB + + N OC
V1

p[TAB f + L(k1 V12 + k2 V22 )] + ICW TAB + VL1
= DN L L
+ N OC
V1 + V2 + TAB.
With this basic scenario complete, we are now ready to investigate the
impact of speed reduction.
To investigate what happens if we reduce speed, we assume that we
reduce the speed of all ships in the fleet by a common amount.mm Let this
common reduction (initial speed — final speed) be equal to ∆V ≥ 0.nn
mm Reducing speeds by different amounts is a straightforward generalization.

nn We implicitly assume that we shall not consider a speed increase, or ∆V
< 0, even
though this may be warranted cost-wise. A speed increase will always increase fuel
To reduce speed and maintain annual throughput constant, we have to

add more ships. If these additional ∆N ships are identical in design to the
original N ones, ∆N can be determined by equating nNW (the quantity
of cargo moved in a year with N ships) with the equivalent expression
for N + ∆N ships. ∆N may not necessarily be an integer, although for
illustration purposes one may want to round it to the next highest integer.
It is easy to check that we can compute ∆N from the following equation:
L L

V1 −∆V + V2 −∆V + TAB
∆N = N L L
−1
V1 + V2 + TAB
Before we proceed, we implicitly assume that these ∆N ships are readily

available and can be immediately incorporated into the original fleet at
a cost equal to OC per ship per year, the same as that paid to charter
the original N ships. However, this may not be the case if there is a lack
of supply of available ships, which may have as a result a lower total
throughput and/or an increase of charter rates to levels above OC . Also,
and as we investigate an operational setting, we do not take into account
long-term effects such as emissions produced by shipyards that would build
these extra ships, emissions produced by the ships carrying the additional
raw materials to be used to build these ships, and other similar life-cycle
quantities.
After some straightforward algebraic manipulations, the difference in
total fleet costs (costs after, minus costs before) is equal to
∆(total f leet costs) = N L∆V

“ ”
IC W D 1 1
−pD(2k1 V1 + 2k2 V2 − (K1 + k2 )∆V ) + V1 (V1 −∆V ) + OC V1 (V1 −∆V ) + V2 (V2 −∆V )
× L L
V1 + V2 + TAB
(1)
Or, in simplified form, if V1 = V2 = V (this may not mean that k1 = k2 ):
∆(total fleet costs)

−pD(2V − ∆V )(k1 + k2 ) + IC W D+2OC
V (V −∆V )
= N L∆V (2)
2 VL + TAB
consumption and emissions, but may actually entail lower other costs, such as inventory
or other, leading in turn to lower total costs.
The difference in fuel costs alone (costs after minus costs before) is
equal to
pD(2k1 V1 + 2k2 V2 − (k1 + k2 )∆V )
∆(total fuel costs) = −N L∆V L L
(3)
V1 + V2 + TAB
Or, in simplified form,

pD(2V − ∆V )(k1 + k2 )
∆(total fuel costs) = −N L∆V (4)
2 VL + TAB
An interesting observation is that fuel cost differentials (and, by exten-
sion, total fleet cost differentials) are independent of port fuel consumption f .
Even though this may seem counter-intuitive, it can be explained by noting
that the new fleet string, even though more numerous than the previous one,
will make an equal number of port calls in a year, therefore fuel burned in
port will be the same.
It is also interesting to note that for ∆V ≥ 0 and for all practical
purposes the differential in fuel costs is always negative or zero, as the term
within the square brackets of [3], or the difference 2V −∆V in [4], is positive
for all realistic values of the speeds and of the speed reduction. This means
that speed reduction cannot result in a higher fuel bill, even though more
ships will be necessary.
The same is true as regards emissions, as these are directly proportional
to the amount of fuel consumed:
∆(total CO2 emissions)

(2k1 V1 + 2k2 V2 − (k1 + k2 )∆V )
= −FCO2 N L∆V D L L
(5)
V1 + V2 + TAB
Or, in simplified form,

(2V − ∆V )(k1 + k2 )
∆(total CO2 emissions) = −FCO2 N L∆V D (6)
2 VL + TAB
Total emissions would thus be always reduced by slowing down, even
though more ships would be used. The higher the speed, and the higher the
speed reduction, the higher this reduction would be.
As a parenthesis we note that mathematically expression [6] achieves
its lowest value (that is, emissions reduction is maximized) if ∆V = V .
This option is of course only of theoretical value, for if this is the case the
fleet would come to a complete standstill and the other cost components
(as well as ∆N ) would go to infinity.
In the general case, whether ∆(total fleet cost) in expressions [1] or [2]
is positive or negative, or reaches a minimum value other than zero, would
depend on the values of all parameters involved, for one can see that in-
transit inventory costs and ship other operational costs count positively
in the cost equation. Both these costs would increase by reducing speed,
and this increase might offset, or even reverse, the corresponding decrease
in fuel costs. High values of either IC or OC (or both) would increase the
chances of this happening, and high values of p would do the opposite, as
will be seen in the examples that follow.
A closer look at expression [2]oo provides some interesting insights.
Expression [2] can be written in the following form:

B
∆(total fleet cost) = ∆V −A(2V − ∆V ) + ≡ G(∆V )
V − ∆V
where A and B are positive constants given by:
(k1 + k2 ) IC W D + 2OC
A = N Lp D B = NL
2 VL + TAB V 2 VL + TAB
As we have assumed that ∆V ≥ 0, function G(∆V ) obtains the value of

0 for ∆V = 0 and goes to infinity when ∆V approaches V . Its behavior for
intermediate values of ∆V depends on the values of all parameters involved.
In fact, we distinguish two cases:
Case 1 : The derivative of G(∆V) at ∆V = 0 is ≥ 0 (see Fig. 4a below).
This is mathematically expressed as V ≤ 2A B
, or as

3 IC W + 2 ODC
V ≤ V0 with V0 ≡ (7)
2p(k1 + k2 )
Speed V0 depends on the parameters shown above and can be consid-

ered as a cost-benefit ‘speed threshold’. If the original speed of the ship V
is at or below that threshold, then any attempt to reduce it to save fuel
(and emissions) would entail a net total cost increase, as G(∆V ) will be
monotonically increasing with ∆V .pp It can be seen that this situation is
more likely to occur if IC and/or OC are high and/or p is low.
oo The analysis for expression

[1] is similar, but more tedious.
pp Again, in this case it may be argued that it is best to increase speed, and reduce
the number of ships, or that ∆V < 0. But this is a case that was excluded from the
beginning.
Fig. 4 Possible forms of G(∆V).
Case 2 : The derivative of G(∆V ) at ∆V = 0 is < 0 (see Fig. 4b).

B
This is mathematically expressed as V > 2A , or as V > V0 with V0
defined as in [7] above.
If the original speed of the ship V is above the V0 threshold, then the
option to reduce speed to save fuel (and emissions) could also reduce total
costs. This situation is more likely to occur if IC and/or OC are low and/or
p is high.
In this case, G(∆V ) achieves a minimum (negative) value for some
‘optimal’ value of ∆V = ∆V ∗ , between 0 and V . In fact, G(∆V ) ≤ 0 for
0 ≤ ∆V ≤ ∆V ∗∗ , and G(∆V ) > 0 for ∆V > ∆V ∗∗ , where ∆V ∗∗ is the
other (nonzero) root of G(∆V ) = 0. We note that ∆V ∗∗ > ∆V ∗ . Both
∆V ∗ and ∆V ∗∗ depend on the values of all other parameters.
If this is the case, speed reduction would indeed be beneficial, and
choosing ∆V = ∆V ∗ would achieve maximum total benefits.
We now present several simple examples to illustrate our approach.
Example 1 — Aframax Tanker Fleet
The first example considers a fleet of N = 10 Aframax double hull tankers,
each with a DWT of 106,000 tonnes, and payload W = 90,000 tonnes,
serving the route from Ras Tanura to Singapore, a distance of L = 3,702 nm
(6,871 km). Other input parameters are as follows:
V1 = V2 = 15 knots = 668.16 km/day.

TA = TB = 4 days
F1 = F2 = 65 tonnes/day (meaning that k1 = k2 = 2.1791 · 10−7
D = 350 days
f = 50 tonnes/day
p = $218/tonne (December 2008)
p = $600/tonne (July 2008)
In other words, we examine two variants, one with a low fuel price and one
with a high one (all else being equal).
Then we consider reducing speed by one knot, to 14 knots, or
623.62 km/day. It is straightforward to show that we will need 0.60 more
ships to be able to cover the same annual throughput. Rounding off to one
more ship, we will have (Table 3):
We can see that fuel costs are reduced in both variants, the cost
differential being $512,541 in the low fuel price variant and $1,410,663 in the
high fuel price variant, both on a yearly basis. CO2 averted would amount
to 54,400 tonnes, even though one more ship is employed.
Still, this does not necessarily mean that total fleet costs will be
reduced, as these would also depend on inventory and other operational
costs.
Neglecting inventory costs for this example (these will be examined in
example no. 3), we consider what the other operational costs might be in
each of these variants.
In a market as seriously depressed as in late 2008, ship owners have
been said to be willing to charter their ships for a rate of zero, with the
charterer paying only for fuel. In this case, variant 1 would continue to be
Table 3 Aframax tanker comparison.
10 ships going 11 ships going

Quantity 15 knots 14 knots
Total fuel consumed for fleet, (tonnes per year) 218,952 201,778
CO2 for fleet (tonnes per year) 694,077 639,637
Bunker cost for fleet ($/year)
Fuel price p = 218 $/tonne $7,143,419 $6,630,878
p = 600 $/tonne $19,660,787 $18,250,124
profitable, although the net savings, if expressed per day, would be very
meager ($1,404/day).
For the high-market variant however, the $3,865/day savings of fuel
costs are well below what an Aframax could command when the market
was high. Rates as high as $60,000/day have been observed for this type of
ship (or perhaps even higher), meaning that speed reduction during these
periods would be non-sensical from a cost-benefit viewpoint.
Example 2 — Panamax Containership Fleet
Our second illustrative example investigates the effect of speed reduction
in containerships. As said earlier, containerships are the top CO2 emissions
producer in the world fleet (2007 Lloyds-Fairplay database).
Assuming a hypothetical string of N = 100 (identical) Panamax con-
tainerships, each with a payload of W = 50,000 tonnes, if the base speed
is V = 21 knots (both ways) and fuel consumption at that speed is 115
tonnes/day, then for a fuel price of p = $600/tonne (corresponding to a
period of high fuel prices, before the slump of 2008), the daily fuel bill
would be $69,000 per ship. Running the same type of ship at a reduced
speed V − ∆V = 20 knots (one knot down), the fuel consumption would
drop to 99.34 tonnes/day (cube law vs. 21 knots) and the daily fuel bill
would drop to $59,605 per ship, some $10,000/day lower.
Assume these 100 ships go back and forth a distance of 2,100 miles (each
way) and are 100% full in one direction and completely empty in the other.
This is not necessarily a realistic operational scenario, as containerships
visit many ports and as capacity utilizations are typically lower both ways,
depending on the trade route. The scenario of trade routes from the Far
East to Europe or from the Far East to North America, which are almost
full in one direction and close to empty in the other, is probably close to
the assumed scenario. However, a generalization of this analysis to many
ports and different capacity utilizations in each leg of the trip should be
straightforward. For simplicity, assume D = 365 operating days per year
Table 4 Panamax containership comparison.
100 ships going 105 ships going

Quantity 21 knots (case A) 20 knots (case B)
Total fuel consumed for fleet, 4,197,500 3,807,256

(tonnes per year)
CO2 for fleet (tonnes per year) 13,306,075 12,069,002
Bunker cost for fleet ($/year) 2,518,500,000 2,284,353,741
and zero loading and unloading times. For non-zero port times, the analysis
will be more involved but will lead to similar results.
At a speed of 20 knots, we will need 105 ships to reach the same
throughput per year. Then we will have:
The net reduction of CO2 emissions (per year) is 1,237,073 tonnes, and
the fuel cost reduction (per year) is $234,146,259 for 5 more ships, that is,
$46,829,252 per additional ship. Dividing by 365, this difference is $128,299
per day.
This means that if the sum of additional cargo inventory costs plus
other additional operational costs of these ships (including the time charter)
is less than $128,299 a day, then case B is overall cheaper. One would
initially think that such a threshold would be enough. But it turns out that
this is not necessarily the case if in-transit inventory costs are factored in.
Before we do so, we display Table 5, that illustrates the unit value of
the top 20 containerized imports at the Los Angeles and Long Beach Ports
in 2004 (see CBO(2006)).
Table 5 Unit value of containerized imports (1,000 $ per short ton).
Unit value of the top 20 containerized imports at los angeles and long
beach ports, 2004
Weight Unit Value
Value (thousands (thousands
(billions of of short of dollars
HS# Category of import dollars) tons) per ton)
84 Machinery, boilers, reactors, 38.0 698.6 54.3

parts
85 Electric machinery, sound and 31.7 677.0 46.8
television equipment, parts
87 Vehicles and parts, except 12.1 337.4 35.8
railway or tramway
62 Apparel articles and accessories, 9.9 132.4 74.6
not knit or crochet
95 Toys games, and sports 9.4 377.1 25.0
equipment and parts
94 Furniture, bedding, lamps, etc. 9.3 739.8 12.6
61 Apparel articles and accessories, 9.0 132.1 68.4
knit or crochet
64 Footwear 7.8 181.4 43.0
39 Plastics and articles thereof 5.2 409.0 12.8
73 Articles of iron or steel 4.4 467.0 9.4
42 Leather articles, saddlery, 3.8 117.2 32.1
handbags
90 Optic, photographic, and 3.6 41.8 86.2
medical instruments
Note that one short ton is equal to 0.9072 tonnes.
To compute in-transit inventory costs for the above example, we

hypothetically assume that cargo carried by these vessels consists of high
value, industrial products, similar to those in Table 5, and that its average
value at the destination (CIF price) is $20,000/tonne. We also assume the
cost of capital being 8%. This means that one day of delay of one tonne of
cargo would entail an inventory cost of IC = P R/365 = 20, 000∗0.08/365 =
$4.38. This may not seem like a significant figure, but it is.
Computing the in-transit inventory costs for this case gives a total
annual difference of $200,000,000 ($4,200,000,000–$4,000,000,000) in favor
of case A, which moves cargo faster. This figure is significant, of the same
order of magnitude as the fuel cost differential.
Assuming also a time charter rate of $25,000 per day (typical charter
rate for a Panamax containership in 2007), the total other operational costs
of the reduced speed scenario are $958,125,000 per year for 105 ships,
versus $912,500,000 for 100 ships going full speed. Tallying up we find
a net differential of $11,478,741 per year in favor of case A, meaning that
in-transit inventory and other operational costs offset the positive difference
in fuel costs.
Of course, other scenarios may yield different results, and the reduced
speed scenario may still prevail in terms of overall cost, under different cir-
cumstances. For instance, if the average value of the cargo is $10,000/tonne,
and everything else is the same, then the difference in annual inventory costs
drops to $100,000,000, rendering the reduced speed scenario a profitable
proposition (with a total cost reduction of $88,521,259 per year). Actually,
speed reduction remains profitable if the value of the cargo is no more
than about $18,800/tonne (which can be considered as a break-even
CIF price).
All of the above confirm that the drive to reduce emissions may or may
not be a win-win proposition, with the final outcome depending on the
specific parameters of the particular scenario (see Psaraftis and Kontovas
(2000b) for some additional insights).
We end this section by noting that there are cases where adding more
ships may not be necessary. These are cases in which the ship’s schedule by
design includes an amount of idle time in port. Such cases are typical for
RoPax scheduled operations, where there is idle time built into the ship’s
schedule for various operational reasons. In these cases, any delay due to
speed reduction is absorbed by the available idle time and no additional
ships are necessary. For a discussion of this scenario, see Psaraftis et al.
(2009c).
5. The Cost to Avert One Tonne of CO2

What would it take to avert one tonne of CO2 by speed reduction? Or,
put in a different way, as much as the question “what price safety?” is
common, let us now ask “what price emissions reduction?” We address this
question by noting that in expressions [5] and [6], ∆ (total CO2 emissions)
equals minus total CO2 averted by implementing a speed reduction scheme.
We define as the cost to avert one tonne of CO2 (CATC) the ratio of the
total net cost of the fleet due to CO2 speed reduction divided by the amount
of CO2 averted by speed reduction. Then we will have:
2OC
−pD(2V − ∆V )(k1 + k2 ) + IC W D
V (V −∆V ) + V (V −∆V )
CAT C = N L∆V
FCO2 N DL∆V (2V − ∆V )(k1 + k2 )
After some algebraic manipulations, this can be rewritten as
IC W D + 2OC
p
CAT C = D
− (8)
FCO2 V (V − ∆V )(2V − ∆V )(k1 + k2 ) FCO2
It can be seen that CATC is a positive linear function of both IC and

OC and a negative linear function of the price of fuel p. It can also be seen
that the denominator in the bracket is a cubic function of speed, reflecting
the functional relationship between speed and the quantity of CO2 that is
produced.
In addition, the last term in [8], −p/F CO2 , where p is the price of one
tonne of fuel and FCO2 is the CO2 emissions factor, can be recognized as
the cost of the amount of fuel saved (not spent) that would produce one
tonne of CO2 . This is an opportunity cost that we will have to subtract
from the total cost incurred, as it corresponds to the amount of fuel that
would be saved if one tonne of CO2 is averted.
The CATC criterion can be used whenever alternative options to reduce
emissions are contemplated. In that sense, the alternative that achieves the
lowest CATC is to be preferred.
The case in which CATC is negative corresponds to the case in which
reducing speed is cost-beneficial, that is, to the case the function G(∆V )
of the previous section takes on a negative value.
For the containership example of the previous section, the CATC values
for the various scenarios examined are as follows (Table 6):
This table confirms that CATC can vary widely. It is also interesting
to note that the difference in CATC between the 1st and 2nd scenario is
the same as that between the 3rd and 4th scenario ($80.84/tonne in both
Table 6 Values of CATC as per containership

scenarios outlined earlier.
CATC
Scenario ($/tonne of CO2 averted)
p = $600/tonne
P = $20,000/tonne 9.28
OC = $25,000/day
p = $600/tonne
P = $10,000/tonne −71.56
OC = $25,000/day
p = $250/tonne
P = $20,000/tonne 104.94
OC = $15,000/day
p = $250/tonne
P = $10,000/tonne 24.10
OC = $15,000/day
cases). This is not a coincidence, and can be explained by the structure of

expression [8].
In these examples, the influence of in-transit inventory costs in the
value of CATC can also be seen clearly. This means that perhaps one of the
biggest obstacles that needs to be overcome if emissions are to be reduced,
is the unwillingness of the cargo owners to incur inventory costs for their
cargoes. Optimized routing, logistics, and other operational measures that
would reduce this inventory costs would be important.
As regards what threshold conceivably exists for CATC, that is, under
what (positive) value of CATC a speed reduction scheme would still be
considered desirable, this issue is currently open and it is not an easy one
to address. As much as it is obvious that both the shipping community and
society at large wish to reduce CO2 emissions from shipping, it is far from
clear how much they are willing to pay to do so. This is not a surprise,
given the fact that there is wide disparity of views on what should be done
to curb GHG emissions, and the fact that decisions on the CO2 front are
still pending.
In a conceivable CO2 Emissions Trading Scheme (ETS) for shipping, a
monetary value would be put on a per tonne basis, for instance, $30/tonne of
CO2 averted, and emissions reduction measures would be evaluated against
such a threshold. Such market values for CO2 currently exist for other
industries, but not for shipping, for which it is unclear how, or when such
a scheme would be implemented.
Fig. 5 ECX EU allowances future contract prices.

Source: European Climate Excange.
Figure 5 shows the historic 2009 settlement prices of EU allowances

issued under the EU Emissions Trading Scheme and traded at the European
Climate Exchange (ECX). One EUA equals one tonne of CO2 (right-to-
emit).
The concept of CATC, as defined above, can be generalized to measures
other than speed reduction, and can be a useful concept for the evaluation
of policy or other alternatives.
6. The Port Time Factor

This section focuses on the case where total trip time is kept constant,
even though speed is reduced (see Psaraftis et al. (2009c) for more details).
Given the fact that time at sea increases with slow steaming we must
investigate possible ways to decrease time in port. This is not an easy
task. The most feasible way to reduce time in port is through operational
decisions regarding land-side operations (berth allocation, quay cranes
scheduling and vessel stowage). Optimizing terminal operations has received
increasing interest over the last years. Vis and de Koster (2003) review
the relevant literature and illustrate the main logistics processes in a

container terminal whereas Steenken et al. (2004) provide an overview
of optimization methods terminal operations. The problem of allocating
ships to berths (discrete case) or to quays (continuous case) is dealt among
others in Cordeau et al. (2005) and Wang and Lim (2007). The Quay Crane
Scheduling Problem (QCSP) which refers to the allocation of cranes and
to the scheduling of stevedoring operations can be solved with the use of
dynamic programming as proposed in Lim et al. (2004) or be addressed
with a greedy randomized adaptive search procedure like the one analyzed
in Kim and Park (2004). Lee et al. (2006) address a yard storage allocation
problem to reduce traffic congestion and Lee and Hsu (2007) present model
for container re-marshalling. For a circumstantial review of the operations
research literature of problems related to container terminal management
the reader could refer among others to Vis and de Koster (2003) and
Steenken et al. (2004).
We now present a simple scenario to investigate the impact of speed
reduction on ship CO2 emissions and fuel costs in the case that the total
trip time is kept constant.
Assume a ship that loads from a port A, travels to port B (a total
distance of L nm from A) carrying a payload W with a known speed of
V0 (in knots), where she discharges the cargo and stays at port before
departing again.
The daily fuel consumptions and times that the ship spends at sea and
in port are known and are as follows:
At sea:
Fuel consumption F0 (tonnes per day)
L
Total time at sea T0 = 24·V0
(days)
In port:
Fuel consumption f (tonnes per day)
Total time in port t0 (days)
Thus, the total fuel consumption for this trip is F C 0 = F0 · T0 + f · T0 .
Now suppose that the ship operator wants to investigate the scenario of
speed reduction. The new speed V will be a fraction of the original speed
(V = aV 0 where 0 < a < 1) and, hence, there will be an increase of the
L
time at sea, T = 24V = Tao
The effect of speed change on fuel consumption is assumed cubic for
the same ship (and for speeds that are close to the original speed) as
discussed in Section 4. The fuel consumption in port per day will remain
the same, but we assume that the new time in port (t) will be reduced in
order to keep at least the same total trip time with that before the speed
reduction.
For this trip we can compute the difference in fuel consumption as
follows:
∆(Fuel consumption)
= ∆(consumption at sea) + ∆(consumption at port)
= F · T − F 0 · T 0 + f · t − f · t0
L L
=F· − F0 · + f · (t − t0 )
24 · V 24 · V0
V =aV0 L L
= F· − F0 · + f · (t − t0 )
24 · aV0 24 · V0

L 1
= F · − F0 + f · (t − t0 )
24 · V0 a
“ ”3
F V 3
F0 = V0 L V 1
= F · − F0 + f · (t − t0 )
24 · aV0 V0 a
3
L aV0 1
= F · − F0 + f · (t − t0 )
24 · V0 V0 a

L 3 1
= a F0 · − F0 + f · (t − t0 )
24 · V0 a
Thus, the total fuel consumption for slow steaming is:
L
∆(Fuel consumption) = F0 (a2 − 1) + f · (t − t0 ) (9)
24 · V0
As one may notice, the first addend is negative since, by definition,
parameter ‘a’ lies between 0 and 1 and L, F0 and V0 are always positive.
It is obvious that if time in port remains the same (t − t0 equal to 0) there
will be a need to add a number of additional vessels (possibly fractional) in
order to maintain the same throughput per year. The model in this Section
examines assumes that t < t0 , and in fact that ‘t’ is such that the total
trip time, including time in port, remains the same (T + t = T0 + t0 ).
Furthermore, as discussed in Section 3, to find the equivalent CO2 emis-
sions reduction, one has to multiply the reduction in bunker consumption
by the appropriate emissions factor (FCO2 ) from Table 2.

L
∆(CO2 emissions) = FCO2 · F0 (a2 − 1) + f · (t − t0 ) (10)
24 · V0
Again, the fuel cost reduction can be estimated by assuming a constant
fuel price as in the previous examples.

L 2
∆(fuel costs) = p · F0 (a − 1) + f · (t − t0 ) (11)
24 · V0
We now move forward to a realistic example using the following figures
that are based on operational data provided by Det Norske Veritas (DNV),
see Psaraftis et al. (2009c).
A Panamax container-vessel begins its trip at Port A, and then
consequently visits ports B and C before going back. The time that she
spends at sea and in port and the relating fuel consumptions are as follows:
Avg Sailing F0 f
Depart Arrive Distance speed Total time (tn/ T0 (tn/ t0
port port (miles) (kn) TEU (hrs) day) (days) day) (days)
A B 115 20.18 1892 5.70 91.79 0.24 16.58 1.79

B C 6068 23.41 2593 259.20 136.81 10.80 3.26 5.45
C A 6323 22.85 3294 276.70 139.22 11.53 12.15 3.55
Using Eqs. 9,10,11 and we can calculate the reductions in fuel cost and
emissions for each leg. Note that the emissions factor used in this example
is 3.13.
For reasons of simplicity we omit the detailed calculations and we
present the resulting total reductions for this round trip in Fig. 6.
One can observe some significant savings in fuel consumption, CO2
emissions (in fact, all emissions) and fuel cost. However, “there is no free
lunch” necessarily. Compensating for a reduced speed will entail either
additional ships to maintain the same throughput, or the ability to reduce
port time. If the former can be achieved, overall emissions are shown
to be reduced, but the overall cost (including cargo in-transit inventory
cost) may or may not go down (as per previous section). Emissions can
be reduced even further if port time can be reduced so that there is no
need for additional vessels. But this may be a more difficult proposition.
Fig. 6 Reductions in fuel consumption, CO2 emissions and Fuel Costs.
For instance, in the example illustrated above, when speed is reduced by

5%, time in port has to be reduced by 11% to maintain a constant total
trip time. If this sounds feasible, it is non-trivial nonetheless. For a speed
reduction of 15% the total time in port has to be reduced from 10.8 days
down to 6.81, which is almost a 37% reduction. This is a much more
difficult proposition, possibly entailing drastic port re-engineering and/or
infrastructure improvements.
7. Speed Reduction at SECAs

All considerations of the previous sections of this paper can be also applied
to emissions other than CO2 . For instance, one can compute emissions from
other pollutants and also define CATN (the cost to avert one tonne of NOx),
CATS (the cost to avert one tonne of SO2 ),qq and so on.
qq Not to be confused with CATS (cost to avert one tonne of spilled oil) — a criterion
under discussion in Formal Safety Assessment (in the context of environmental risk
evaluation criteria).
The only difference with the previous analysis is that one would have
to substitute for the CO2 emissions factor the appropriate emissions factor
of the pollutant under consideration. For instance if one considers SO2 ,
one has to multiply total bunker consumption (in tonnes per day) by the
percentage of sulphur present in the fuel (for instance, 4%, 1.5%, 0.5%,
or other) and subsequently by a factor of 0.02 to compute SO2 emissions
(in tonnes per day). Even though the amounts of SO2 produced by ships
are substantially lower than CO2 , for SO2 emissions other considerations
are equally important. SO2 is not a greenhouse gas but as it causes acid
rain (among other effects), its reduction is a matter of high priority. To that
effect, SO2 (and generally SOx ) reduction is also high on the IMO agenda,
and in fact regulatory progress on this front is more advanced than for the
CO2 front, as exemplified in the latest MEPC 58 Annex VI developments
as regards the timetable on SOx emissions caps.
To reduce pollution by SOx , special highly sensitive areas have been
designated by the IMO as ‘Sulphur Control Emissions Areas’, or SECAs,
where specific limits in SOx content are set for a ship’s exhaust gases.
Designated SECAs to date are the Baltic Sea, the North Sea and the English
Channel. The IMO does not specify how the SOx emissions targets should
be reached. Among methods contemplated, sea scrubbers are a measure
that is offered on the technology front. Fuels cleaner in sulphur content is
also a method that is proposed (of which more later).
Among potential operational measures, one question that is relevant
is this: Can speed reduction at SECAs work, as a measure to reduce SOx
emissions? This sounds like an easy question to pose, for which however the
answer may not be so easy.
First of all it should be noted that speed reduction, in and of itself, will
not change the proportion of SO2 in a ship’s exhaust. But it will change
the total amount of SO2 produced, much in the same way as this happens
for CO2 . In that sense, speed reduction to reduce SO2 is worthy of note.
Let us assume a ship that goes from port X to port Y, sailing a total
distance of L. At the beginning or the end of the trip, there is a SECA, of
distance d (< L).
Assume there are two options: The first (option A) is to sail the entire
trip at a constant speed of V . The second (option B) is to reduce speed
to v (< V ) within the SECA, so as to reduce SO2 emissions, but go at a
slightly higher speed of V ∗ (> V ) outside the SECA, so that total transit
time is the same.
Total transit time is kept the same so that we do not need more ships
in the supply chain, and shippers do not lose money on in-transit inventory
costs. If total transit time is not the same, we shall have to go through an
analysis similar to that of previous sections.
Let us now pose the question, with total transit time being the same,
which option burns less fuel, A or B? The one that does so would also cost
less, and would also produce less total emissions, not only in SO2 , but also
CO2 and all other pollutants.
The analysis is straightforward and goes as follows:
Let the transit time in both scenarios be T = L/V (in days). If within
the SECA the speed is v (< V ) for distance d, then
L d L−d
= +
V V V∗
Therefore
L−d
V∗ = (> V )
L
V− dv
[the assumption here is that L/V > d/v, otherwise making up the time lost
in the SECA would be impossible.]
Again, we assume that fuel consumption per day obeys a cube law, that
is, is equal to kV 3 . Since we have to multiply by total days, the law becomes
quadratic, as total fuel consumption is TF C = k V 3 (L/V ) = kLV 2 .
Option A: total fuel consumption TF C (V ) = kLV 2

Option B: total fuel consumption TF C (V ∗ , v) = k(L − d)V ∗2 + kdv 2
Substituting, we get
k(L − d)3 2
TF C (V ∗ , v) = + kdv
d 2
V − v
L
Define the ratio
TF C (V ∗ , v) (L − d)3 d v 2
R= = 2 +
TF C (V ) L L − d Vv L V
It can be shown mathematically that always R > 1 (assuming again

that L/V > d/v). The proof of this is straightforward.
Let us illustrate this with an example:
Let L = 2000 nautical miles

d = 200 (SECA)
V = 20 knots
v = 18 knots within the SECA
1800
V ∗ = 2000 200 = 20.25 knots outside SECA
20 − 18
Then
2
18003 200 20
R= +
20 2
= 0.9226 + 0.081 = 1.0036
2000 2000 − 200 18 2000 18
Other speed and distance combinations will produce other ratios, but
all will be >1.
The conclusion from this analysis is this: Speed reduction in SECAs
will reduce emissions (of all gases, including SOx ) within the SECA, but
result in more total emissions and more total fuel spent if speed is increased
outside the SECA to make up for lost time. The reduced emissions within
the SECA will be more than offset by higher emissions outside (for all
gases). The total fuel bill will also be higher.
Of course, whether or not society may mind polluting the areas outside
SECAs more in order to make conditions in SECAs more friendly to the
environment is a non-trivial issue that is outside the scope of this paper.
Alternatively, if a lower speed is maintained throughout the ship’s
journey, then obviously total fuel and total emissions will be reduced, but
there will be increased costs in the form of more ships needed to carry the
same cargo in a year and more in-transit inventory cost for the shippers
(as per previous sections).
8. SECAs Continued: Effect on Modal Split

We now make a cursory investigation of the case in which a ship involved in
short sea trades uses low- sulphur fuel at a SECA, to reduce SOx emissions.
This fuel is 4–30% more expensive than high-sulphur fuel (see Fig. 3). Hence
freight rates may go up. Furthermore, according to a document submitted
by INTERFERRY to MEPC 58 (doc. MEPC 58/5/11) the rise in fuel prices
over the past years and the cost increase for low sulphur fuel will either
be passed on to customers or will force some operators out of the market.
Increases in fuel prices are cited in this document as key reason for canceling
certain ferry routes including those from Newcastle (United Kingdom)
to Bergen, or Kristansand (Norway). This may induce shippers to use

land transport alternatives (trucking), which will go against stated policies
toward shifting cargo from land to sea and increase CO2 emissions through
the logistics chain. The European Community Shipowners’ Association
(ECSA) has already warned that new sulphur limits agreed at the IMO
could push more freight onto the roads in Europe (Lloyds List, 2008a).
In this paper we shall only examine a hypothetical and rudimentary
example of this scenario, which goes as follows. A modern Handymax bulk
carrier moves a cargo of W = 45,000 tonnes from Bergen to Oslo, Norway,
a distance of L = 371 nm (689 km). The ship sails with a speed of 14 knots
and consumes 30 tonnes of HFO per day.
The ship completes the trip in 1.1 days, after consuming a total of
33.13 tonnes of fuel. Thus total CO2 emissions are 105.01 tonnes, which
corresponds to 3.39 grams per tonne-km. Total SO2 emissions amount to
2.99 tonnes for high-sulphur (4.5%) heavy fuel oil but only to 0.33 tonnes
for low-sulphur (0.5%) marine diesel oil, which is the maximum allowable
sulphur content effective 1/1/2020. This means that the potential savings
in SO2 emissions by switching to cleaner fuel are 2.66 tonnes (CO2 would
be the same).
We obviously have no way of knowing what the fuel prices will be in
2020, and in particular what the availability of low sulphur fuel might be
and how it could impact these fuel prices at that time. However, let us
assume that prices (in 2008 US dollars) are as they were in July 2008,
when they were high (see Fig. 3). Then the total bunker cost for this
trip is $22,545 when using high-sulphur HFO and $37,354 for low-sulphur
MDO.
Suppose now that a (yet unspecified) portion, or even the whole amount
of this cargo is transported from Bergen to Oslo by road, using a modern
and environment-friendly truck with a trailer (long-haul traffic) whose
engine emits 2.6 kgr of CO2 per liter of low-sulphur diesel fuel (10 ppm
of sulphur).
In this case one truck moves a cargo of 40 tonnes with a speed of
60 km/h and a fuel consumption of 43 liters per 100 km when loaded. Each
one-way trip from Bergen to Oslo, a distance of 490 km by road, takes
8.2 hrs or 0.34 days. Total fuel consumption is 0.2107 tonnes per one-way
trip, which corresponds to 0.548 tonnes of CO2 per one-way trip, or 27.95
grams of CO2 per tonne-km.
We first notice that the comparison is not on a completely equal basis,
as the sea trip distance is some 40% longer than the road one. Even so, let
us calculate the total CO2 produced by the road option.
To move the whole cargo of 45,000 tonnes of one shipload one way
by road, it would take 1,125 truck trips, bringing the total CO2 produced
by this option to 616.3 tonnes, almost 6 times as much as that produced by
the ship, and more than 230 times the amount of SO2 potentially saved by
the cleaner ship fuel. Although comparing the volumes of the two gases
may be like comparing apples with oranges, it is important to have these
figures in mind (SO2 produced by the truck fleet is essentially negligible).rr
Of course, not all of the 45,000 tonnes of cargo may want to shift to
road. The proportion that will do so will depend, among other things, on
things such as:
(a) the unit fuel costs of each of the two options (both for low-sulphur and
for high-sulphur fuel)
(b) how the road option is exercised (e.g., it could be 1,125 trucks doing
one trip each, a fleet of 563 trucks doing two trips each, or any other
combination)
(c) the transit times of each of the two options
(d) the inventory costs of the cargo.
Regarding (a), we note the differential of $0.33/tonne of cargo in the price

of fuel (or $14,809 per shipload, or some 66% more). This will translate
into a cost increase of the sea mode. The calculation of the impact of this
cost increase on the modal split between sea and road (which also depends
on points (b), (c) and (d)) was an issue that was open at the time of the
writing of this paper.
9. Conclusions
This paper has taken a look at some of the problems associated with
green maritime logistics. Speed reduction was the main focus of the paper
and some conditions under which such a scheme would reduce overall cost
were identified for some operational scenarios. In addition, some possible
ramifications of using speed reduction and cleaner fuels at SECAs were
investigated. It was seen that caution should be exercised in proposing
measures that may at first glance look environmentally friendly, but in
reality they may have negative side effects.
rr Thisanalysis does not take into account the additional CO2 emitted by refineries to
produce increased amounts of low-sulphur fuel, or the additional CO2 as a result of the
possible congestion by having a large number of trucks on the highway.
The main conclusions of this paper can be summarized as follows:
• Speed reduction will always result in a lower fuel bill and lower emissions,
even if the number of ships is increased to meet demand throughput.
• Due to in-transit cargo inventory costs and other ship costs, total
fleet operational costs may or may not decrease with speed reduction,
depending on the scenario.
• The cost to avert one tonne of CO2 by speed reduction depends on several
factors, being higher for higher-value cargoes.
• Speed reduction to reduce sulphur emissions at SECAs will result in a
net increase of total emissions (including sulphur) along a ship’s route, if
transit time is to be kept the same.
• Cleaner fuel at SECAs may result in a reverse cargo shift from sea to land
that has the potential to produce more emissions on land than those saved
at sea.
Future research vis-a-vis the models presented here involves extending

these models to more complex logistical scenarios, concerning, among other
things, issues such as ship routing and scheduling, maritime and intermodal
transport network design, queuing at ports, emissions at ports, and all the
others listed at the introductory section. Plus, the broader consideration of
such issues in a strategic setting is also important. In the emerging drive
for green maritime logistics, investigating such problems would become
increasingly important in the future.
Acknowledgments
This is an expanded version of the plenary address by the same title, given
by the first author at the International Symposium on Maritime Logistics
and Supply Chain Systems, held in Singapore in April 2009 (MLOG, 2009).
Parts of the paper draw from papers presented at the International Marine
Design Conference (IMDC, 2009), held in Trondheim, Norway, in May 2009,
and the Conference on FAST Sea Transportation (FAST, 2009), held in
Athens, Greece, in October 2009. The research reported in this paper was
funded in part by a gift from Det Norske Veritas (DNV) to the National
Technical University of Athens. Some of the data used in the paper was
provided by the Hellenic Chamber of Shipping and by DNV.
References
1. Andersson, L. (2008). Economies of scale with ultra large container vessels,
MBA assignment no 3, The Blue MBA, Copenhagen Business School.
2. Buhaug, Ø., Corbett, J.J., Endresen, Ø., Eyring, V., Faber, J., Hanayama, S.,
Lee, D.S., Lee, D., Lindstad, H., Mjelde, A., Pålsson, C., Wanquing, W.,
Winebrake, J.J. and Yoshida, K. (2008). Updated Study on Greenhouse Gas
Emissions from Ships: Phase I Report. International Maritime Organization
(IMO) London, UK, 1 September, (included as Annex in IMO document
MEPC58/INF.6).
3. Cerup-Simonsen, Bo. (2008). Effects of energy cost and environmental
demands on future shipping markets, MBA assignment no 3, The Blue MBA,
Copenhagen Business School.
4. Congressional Budget Office (CBO) (2006). The Economic Costs of Disrup-
tions in Container Shipments, U.S. Congress, Washington, DC.
5. Corbett, J.J. and Köhler, H.W. (2003). Updated emissions from ocean
shipping, Journal of Geophysical Research 108.
6. Corbett, J., Wang, H. and Winebrake, J. (2009). Impacts of Speed Reduc-
tions on Vessel-Based Emissions for International Shipping, Transportation
Research Board Annual Meeting, Washington DC.
7. Cordeau, J., Gaudioso, M., Laporte, G. and Moccia, L. (2007). The service
allocation problem at the Gioia Tauro maritime terminal, European Journal
of Operational Research, 176, 1167–1184.
8. Eefsen, T. (2008). Container shipping: Speed, carbon emissions and supply
chain, MBA assignment no. 2, The Blue MBA, Copenhagen Business
School.
9. Endresen, Ø., Sørgard, E., Sundet, J.K., Dalsøren, S.B., Isaksen, I.S.A.,
Berglen, T.F. and Gravir, G. (2003). Emission from international sea trans-
portation and environmental impact, Journal of Geophysical Research, 108.
10. Endresen, Ø., Sørgård, E., Behrens, H.L., Brett, P.O. and Isaksen, I.S.A.
(2007). A historical reconstruction of ships fuel consumption and emissions,
Journal of Geophysical Research, 112.
11. EMEP/CORINAIR (2002). EMEP Co-operative Programme for Monitoring
and Evaluation of the Long Range Transmission of Air Pollutants in Europe,
The Core Inventory of Air Emissions in Europe (CORINAIR), Atmospheric
Emission Inventory Guidebook, 3rd edition.
12. IMO (2000). Study of Greenhouse Gas Emissions from Ships, Study by
Marintek, Econ Centre for Economic Analysis, Carnegie Mellon University
and DNV.
13. IMO (2008a). INTERIM GUIDELINES FOR VOLUNTARY SHIP CO2
EMISSION INDEXING FOR USE IN TRIALS, MEPC Circ. 471.
14. IMO (2008b). Report of the Drafting Group on amendments to MARPOL
Annex VI and the NOx Technical Code, MEPC 58/WP.9.
15. IMO (2008). Liaison with the Secretariats of UNFCCC and IPCC concerning
the Carbon to CO2 conversion factor, MEPC 58/4/3.
16. IMO (2008). The impact of sulphur limits on ferry operations in Northern
Europe, INTERFERRY, MEPC 58/5/11.
17. Kim, K.H. and Park, Y.M. (2004). A crane scheduling method for
port container terminals, European Journal of Operational Research, 156,
752–768.
18. Lee, L., Chew, E., Tan, K. and Han, Y. (2006). An optimization model
for storage yard management in transshipment hubs, OR Spectrum, 28,
539–561.
19. Lee, Y. and Hsu, N.Y. (2007). An optimization model for the container pre-
marshalling problem, Computers and Operations Research, 34, 3295–3313.
20. Lim, A., Rodrigues, B., Xiao, F. and Zhu, Y. (2004). Crane scheduling with
spatial constraints, Naval Research Logistics, 51, 386–406.
21. Lloyds List. (2008a). IMO Sulphur Limits Deal Could See More Freight Hit
the Road, Lloyds List, 10 April.
22. Lloyds List. (2008b). An Efficient Ship is a Green Ship, says GL, Lloyds List,
30 July.
23. Psaraftis, H.N. and Kontovas, C.A. (2008). Ship Emissions Study. National
Technical University of Athens, report to Hellenic Chamber of Shipping,
http://www.martrans.org/emis/emis.htm, May.
24. Psaraftis, H.N. and Kontovas, C.A. (2009a). CO2 Emissions Statistics for the
World Commercial Fleet, WMU Journal of Maritime Affairs, 8(1), pp. 1–25.
25. Psaraftis, H.N. and Kontovas, C.A. (2009b). Ship Emissions: Logistics
and Other Tradeoffs. 10th Int. Marine Design Conference (IMDC 2009),
Trondheim, Norway, 26–29 May.
26. Psaraftis, H.N., Kontovas, C.A. and Kakalis N. (2009c). Speed Reduction
as an Emissions Reduction Measure for Fast Ships. 10th Int. Conference on
Fast Transportation (FAST 2009), Athens,Greece, 5–8 October.
27. Steenken, D., Voss, S. and Stahlbock, R. (2004). Container terminal
operation and operations research — a classification and literature review,
OR Spectrum, 26, 3–49.
28. Vis, I. and de Koster, R. (2003). Transshipment of containers at a container
terminal: An overview, European Journal of Operational Research, 147,
1–16.
29. Wang, F. and Lim, A. (2007). A stochastic beam search for the berth
allocation problem, Decision Support Systems, 42: 2186–2196.

CHAPTER 13
EXPLORING TANKER MARKET

ELASTICITY WITH RESPECT
TO OIL PRODUCTION USING
FORESIM
P.G. Zacharioudakis
OceanFinance Ltd, Greece
[email protected]
D.V. Lyridis
School of Naval Architecture & Marine Engineering
National Technical University of Athens, Greece
[email protected]
Future market freight levels have always been a critical question in decision
support processes. FORESIM is a simulation technique that models shipping
markets (developed recently). In this paper we present the application of this
technique in order to obtain useful information regarding future values of
the tanker market in numerous states of OPEC oil production levels. This
is the first attempt to express future tanker market freight levels in relation to
current market fundamentals and future values of demand drivers. We follow a
systems analysis seeking for internal and external parameters that affect market
levels. Therefore we apply dynamic features in freight estimation taking into
account all Tanker market characteristics and potential excitations from non
systemic parameters as well as their contribution to freight level formulation
and fluctuation. In this way we are able to measure the behavior of future
market as long as twelve months ahead with very encouraging results. The
output information is therefore useful in all aspects of risk analysis and decision
making in shipping markets.
1. Introduction
A popular definition of “forecast” is that it is a reference to future trends
usually in the form of probability that is realized by processing and
297
298 P. G. Zacharioudakis and D. V. Lyridis
analyzing available data. Then a set of questions slip into mind: In a

volatile market such as the one of shipping freight rates, is it possible to
acquire information regarding its future evolvement? How can we predict
the events that will influence the future state of the market? Future shipping
market risk has always been an attractive thematic issue for many maritime
economists. Especially in the era of risk management attempts in shipping,
measuring market risk is a key point to the success of the whole process.
Selecting between Spot and Period Charter or where to place a vessel,
is a tricky question which, nevertheless, can be successfully approached
by using the appropriate risk management tools. There are two major
characteristics of the shipping market that turned risk management to a
necessity: variability and uncertainty. Risk management can do very little to
reduce market variability, but can be very effective in reducing uncertainty
for those involved in risk-taking decisions. Alternatively Freight Forward
Aggrements (FFA’s) are the latest tools in hedging shipping risk aiming to
be a gowning market parallel to the physical one. So whether we speak about
physical market chartering strategies or “paper” market trading strategies,
future market risk knowledge is essential for efficient decision making.
There are two main approaches regarding the estimation of future
market risk. The first one is based on univariate stochastic models. This
approach applies models like the GARCH, ARIMA, Geometric Brownian
Motion (GBM), Ornstein-Uhlenbeck (O-U) process, Jump-Diffusion (O-U
with Jumps) etc based on the admission that all the necessary information
to estimate future values is located in the precedent historical data of the
time series. This admission is quite defective by a simple consideration of the
tanker and bulk shipping market mechanism. For that reason researchers
developed static econometric models (Zannetos, 1966, Norman, 1979
& 1981, Evans 1994) or dynamic (Eriksen & Norman, 1976, Strandenes,
1986, Beenstock & Vergotis, 1989, Lyridis et al. 2004a & b).
Although this paper presents an application in Tanker vessels and more
specifically in Very Large Crude Oil Carriers (VLCC), the methodology can
also be applied in the bulk market. Both markets operate in a system with
numerous interactions. Shipping market mechanism is full of causality terms
balancing the output in the time field — the freight rates. The formulation
mechanism for the tanker market is not clearly known but the fact is that
it is dependent on the global socioeconomic status. The interaction of the
market and the variables is either direct or indirect according to the way
and the time lag that they interact. For example VLCC rates have a direct
and positive correlation with the orderbook in real time. As the observed
Exploring Tanker Market Elasticity 299
phenomenon of the Shipping cycle describes, following a high freight rate

period new vessels enter the market resulting into an increase in the total
transport capacity and subsequently into a drop in the rates. Therefore, the
two variables have a negative correlation when examined under a specific
time lag. All variables, apart from demand for sea transport, are in some
way correlated to market trends and vice versa. However, demand for sea
transport is determined by other factors and not by the state of the shipping
market. For example, while the level of oil production by OPEC has a strong
influence in the market, there is no feedback from the shipping market to the
level of oil production. But which are the variables that influence demand
in the shipping market? In the case of VLCC carriers the demand is related
mainly to the following:
• The growth of world economy

• Oil shocks
• War — hostile acts near oil production facilities
• Oil reserves
• Oil price
• Climate conditions
• Political decisions — OPEC policy
• New reserves
Many shipping parameters have a dominant role in freight rates future

possible realizations. This leads to the statement that the initial state
of the shipping system as described by the fundamental variables has
a leading effect on how this system may react to excitations such as
a demand increase or decrease, a pipeline closure, an oil shock etc. To
be more specific, a congested shipping market with increased volumes of
laid up vessels is expected to show less sensitivity to demand changes
comparing to a balanced market. It is known high volumes of laid up
vessels is a characteristic of markets with low freight rates. The laid up
vessels will absorb any demand increase by entering to operational state.
To the contrary scenario if the market experiences a demand decrease
for transport services then the already low freight rates levels cannot
fall down from a minimum point relatively to the operating expenses.
Another crucial parameter is the volume of tonnage under construction
(order book). This parameter seems to give an important indication for
the future levels of tonnage supply. Hence it is obvious that a systemic
modeling of shipping market would lead to more bounded possible future
states subject to the constrains of the fundamental explanatory shipping

parameters. Forecasting the need for sea transport is very difficult since
it is related to quantitative and qualitative variables with unforeseen
trends. What can be done is to “feed” the forecasting model with different
scenarios and generate a stochastic ‘description’ of the future. This is why
FORESIM was conceived as a complete simulating procedure. The entire
shipping market parameters such as active fleet or scheduled deliveries
play a predetermined role in future freight rate levels. Additionally crucial
parameters that affect freight rate levels (the OPEC oil production in our
case) and have unpredictable random behaviors are stochastically generated
in the corresponding simulation time period.
This paper is structured as follows: methodology is described in the
next section. FORESIM technique is applied in order to simulate tanker
VLCC market. The third section presents the results regarding future
market estimation. The paper finishes with interesting conclusions about
FORESIM application and market characteristics.
2. Methodology
FORESIM is a simulation technique developed especially for the shipping
system. It is used in order to obtain a solid future view of the maritime
trends. It consists of a stochastic model: this simulates oil production levels
and then using the Monte Carlo technique produces the freight rates by
applying a proper Artificial Neural Network. A main feature of FORESIM
technique and what forms its basic innovative aspect is its ability to simulate
future evolution of an econometric system based on its current state only, in
a way where all crucial parameters can affect future system outputs. Hence
in our case the entire shipping market parameters such as active fleet or
scheduled deliveries play a predetermined role in future freight rate levels.
Additionally crucial parameters that affect freight rate levels (the OPEC
oil production in our case) and have unpredictable random behaviors are
stochastically generated in the corresponding forecast time period.
The first step is the definition of the system and the variables to be
simulated. This is what is called a systemic analysis in order to obtain
absolute knowledge of the examinant system. Next step is to construct a
model capable of simulating the physical market. FORESIM uses the power
of Artificial Neural Models in what is called function approximation seeking
for relations between input vector and desired output. Artificial neural
networks are mathematical models imitating human brain functionality
and are used as an advanced pattern recognition technique with application

in time series forecasting. According to the literature, ANNs are suitable
for analysis of non-stationary nonlinear time series. Focusing in tanker
freight forecasting, in comparison to other methods such as linearly based
autoregressive models, artificial neural networks are proven to be at least
as accurate while, in many cases, yielding impressive results (Lyridis
2004a & b). The possible outer system excitations are entered into the
model with the usage of GARCH-family models. Then to model processes
like oil production, where covariance is not constant in the time domain,
GARCH family models has been fairly successful (Bollerlev, and Engle
1994).
The systemic analysis showed that the independent variables that can
be divided in two major categories. The first has to do with variables related
to demand for transport in and the second with those that are related to
the supply of tonnage. Figure 1 shows a schematic approach of the shipping
system regarding the freight rates generation mechanism.
External Factors: Shipping Sector State:
• State of Global Economy and • Freight Rates Levels

Development
• Supply of Transport Services
• Political Events.
• Laid up Vessels
• Military Actions.
• Slow Steaming Vessels
• Oil Reserves.
• Demolitions
• Climate Conditions
• OBO Vessels
• New Oil Deposits • Vessels Losses
• Commodities Prices
• Vessels Deliveries
• New Orders
• Chartering Trends
Production − Transportation:
• New Building Prices
• Oil • Second Hand Prices

• Iron − Ore • Scrap Prices
Demand for Transport Freight Rates

Services Generation Model
Freight Rates
Level
Fig. 1 Schematic approach of freight rates generation mechanism.

The importance of investigating these sets of variables is very high since

they determine the freight rates as the result of the equilibrium between
supply and demand. A few from the variables used in modeling the market
are the following:
• Freight rates
• Active fleet
• Demand for transport in the specific market
• Orderbook
• Demolitions
• Laid-up vessels etc.
By using expert judgment and statistical tools to measure correlation

and to avoid colinearity, the input vector is constructed. Table 1 shows the
input vector for the three months ahead simulation of VLCC WS freight
rates (Ras Tanura — Rotterdam) and the corresponding Variance Inflation
Factor (VIF). The input vector consists of 8 variables:
By applying the same process the corresponding input vector for the
twelve months ahead is shown in Table 2.
Table 1 Input vector for three months ahead model.
Independent variable VIF
Oil price 2.858

VLCC supply 4.213
OBO supply 6.260
VLCC demolition prices 1.946
VLCC WS rate 3.727
OPEC production 4.148
OPECDIF 3 (percentage difference after three months) 1.104
ARBITRAGE of oil prices 4.941
Table 2 Input vector for twelve months ahead model.
Independent variable VIF
VLCC supply 3.179

OBO supply 8.144
VLCC demolition prices 1.749
VLCC orderbook 5.001
WS 2.829
OPEC production 4.211
OPECDIF 12 (percentage difference after twelve months) 2.029
It is remarkable that as expected the Orderbook variable is shown to

have a statistical important influence on the dependent variable of freight
rates after twelve months. It also expected that variables like oil price and
Arbitrage can influence the freight rate generation mechanism only in a
short term basis. By constructing appropriate input vectors for up to a
twelve months period ahead FORESIM is capable of simulating the VLCC
market.
Oil production time series data like many time series usually exhibit
a characteristic known as volatility clustering, in which large changes tend
to follow large changes, and small changes tend to follow small changes.
Engle’s test is applied in Oil production time series data to seek for the
presence of ARCH effects. Pre-estimation process includes the Opec time
series transformation using the following equation:
Opec i+1
Opecret i = −1 (1)
Opec i
Under the assumption that the transformed Oil production time series
data is a random sequence of Gaussian disturbances (i.e., no ARCH effects
exist), this test statistic is also asymptotically Chi-Square distributed
(Engle, 1995). The test results reveal that the ARCH effect is present hence
serial dependence of volatility exists. Following this specific preprocessing
procedure by applying Ljunx-Box-Pierce Q-test (Gourieroux, 1997) it is
clear that no serial dependence of mean exists hence there is no need to use
a conditional mean model such as ARIMA.
To feed the technique with possible future oil production volumes after
a fit process an Exponential GARCH model is used. In order to fit a model in
data set, log-likelihood function — LLF-criterion is calculated. In addition,
Akaike and Bayesian information criteria were used to compare alternative
GARCH models based on parsimony and penalize models with additional
parameters. (Box, and Jenkins, 1970). Table 3 shows the results.
The E-GARCH(1,1) (Student t distributed) include a term to capture
the leverage effect, or negative correlation, between examinant variable
Table 3 Input vector for twelve months ahead model.
Selection Criteria
Model LLF AIC BIC
EGARCH11ARMA00T 819.07 −1628.15 −1608.34

Table 4 Input vector for twelve

months ahead model.
Coefficient Value
C 0.00011918
K −0.10279
GARCH(1) 0.98584
ARCH(1) 0.18242
Leverage(1) −0.10488
returns and volatility (Nelson 1991). As estimated in the fit process the

model will have the following coefficients as shown in Table 4.
Hence the form of the E-GARCH that will be used to generate paths
of transformed Opec oil production is the following:
yt = 0.00011918 + εt
Vart−1 (yt ) = Et−1 (ε2t ) = σt2
2 (2)
log σt2 = −0.10279 + 0.98584 log σt−1

|εt−1 | |εt−1 | εt−1
+ 0.18242 −E − 0.10488
σt−1 σt−1 σt−1
Where:

|εt−1 | ν − 2 Γ ν−1
2
E{|zt−1 |} = E = · ,
σt−1 π Γ ν2
Due to the fact that εt is Student’s T distributed.
The estimated ν freedom degrees equal to3.2314 for the specific
distribution
hence by calculating the Gamma function values, the term
|εt−1 |
E σt−1 takes the value of 0.6609.

By substituting the E |εσt−1
t−1
|
term to the log σt2 equation:
log σt2 = −0.172109 + 0.98945 log σt−12

|εt−1 | εt−1
+ 0.19252 − 0.092892 (3)
σt−1 σt−1
A common question is which network to use in each case, as the
researcher is faced with a large number of options. In this paper model the
relation between the dependent and independent variables we use a special
class of MultiLayer Perception networks (hereafter MLP), the modular feed-
forward networks (Fig. 2).
Input Hidden Hidden Output

Layer Layer 1 Layer 2 Layer
Fig. 2 Structure of the general modular artificial neural network.
These networks are trained by a supervised learning momentum

algorithm (Moreira, 1995). The weight update process is the following:
wij (n + 1) = wij (n) + η · δi (n)xj (n) + a(wij (n) − wij (n − 1)) (4)
Where:
η: learning rate
δi (n): current error
xj (n): current input vector
α: momentum rate
Modular ANN don’t have full interconnectivity between neurons and
the layers are divided to modules. Each module cooperates with others in
order to solve part of the whole problem. Due to the partial interconnectiv-
ity a decreased number of weights is necessary and therefore the demand
for training cases is decreased. The specific topology has two hidden layers
with two modules per layer. The number of neuron per module is variable,
subject to optimization. The transfer function is shown in Fig. 3:
tanh(x) = (ex − e−x )/(ex + e−x ) (5)
And the output of the transfer is given by the following equation:
output = tanh(w1 x1 + w2 x2 + · · · + wn xn ) (6)
It has to be mentioned that a modeler should consider basically two
issues in order to obtain the ability of generalization (Jhee, M.J. Shaw).
First, the explanatory model should transfer all necessary information to
Fig. 3 Transfer function of the general modular artificial neural network.
Fig. 4 Goodness transfer function of the general modular artificial neural network.
ANN. It is a matter of experience, deep knowledge and assiduous research

effort to invoke all significant informational variables. The second crucial
issue is to train the ANN using as a training data set a representative sample
of data on which ANN will be used to simulate a forecast. ANN are trained
using a cross validation dataset in order to avoid overtraining issues and luck
of generalization ability. The cross validation data set consist of randomly
selected cases which are kept out of the training process. By this way the
ANN are trained and validated under a wider range of shipping market
situations. Figure 4 shows example of ANN’s fit on the corresponding test
dataset (30 cases) for the nine months ahead case.
The results shows that fit on test data is excellent. This means that
the information provided to the networks — current market fundamental
variables and future demand indicator — is sufficient in order to estimate
Table 5 Goodness of fit results for ANN.

Goodness of fit criterion ANN(+3) ANN(+6) ANN(+9) ANN(+12)
MSE 0.003703 0.002843 0.006776 0.002872

NMSE 0.082971 0.081254 0.087043 0.075537
%Error 13.665811 11.780595 13.676121 14.865287
Fig. 5 Naı̈ve model vs modular artificial neural network performance.
future market values. Table 5 shows the results of goodness of fit process

using Mean Square Error, Normalized MSE and percentage of error criteria:
Comparing the results with the naı̈ve model (future value is equal to
the present one) it is clear that modular ANNs show a sufficient and robust
error performance in the corresponding time spans of three, six, nine and
twelve months. Figure 5 shows the results for the three, six, nine and twelve
months ahead models:
The next step of FORESIM is simulation. The stochastic component
(E-GARCH) feeds ANN with a pre defined number of possible future
demands indicators and the corresponding input vector containing the
fundamental variables describing the current state of the market. Then
the ANN produce an output vector for the corresponding time span.
This vector represents the estimated future freight rate values for every
possible excitation from the demand indicator. The error terms of the ANN
are stochastically estimated using a number of goodness of fit tests: chi-
square (Snedecor and Cochran, 1989), Kolmogorov-Smirnov (Chakravart,
Laha, &Roy, 1967) and Anderson-Darling (Stephens, 1974). By using simple
Monte Carlo simulation every output value of the ANN is recalculated by
adding a possible error term. An example of a FORESIM simulation case
Fig. 6 Histogram of simulation results (case: three months ahead June 2003).
(simulated case: WS rate three months ahead June 2003) can be shown in
the next histogram (Fig. 6).
3. Simulation Results
The superior advantage of FORESIM technique is the ability of simulating
future market states in accordance to the fundamentals parameters and
possible external excitations. According to FORESIM results a crucial
parameter regarding future market levels is the level of OPEC oil pro-
duction. By estimating the output of various cases of future oil production
an investigation regarding the relationship between future market and oil
production is feasible. Table 6 shows the bivariate Pearson Correlation
Coefficient between Demand for VLCC Tankers and the variable of Opec
oil production.
Table 6 Correlation of demand with opec production.
Pearson correlation Opec production Demand for tanker services
Opec production 1 .918

Demand for tanker services .918 1
Table 7 Correlation of market rates with opec

production.
Pearson correlation Opec production VLCC WS
Opec Production 1.000 .641

VLCC WS .641 1.000
Fig. 7 WS vs OPEC oil production (millionbarrels per day) (12 Months Ahead March
2003).
Table 7 shows the bivariate Pearson Correlation Coefficient between

VLCC Tankers freight rates and the variable of Opec oil production.
The results depict the importance of Oil production in future market
freight levels as shown from Correlation Coefficients values. During a
What — If analysis the output of a system is examined after considering
various excitations. Shimojo (1979) has proposed a “J-shaped” diagram
to describe freight rates versus demand for transportation. FORESIM is
capable of simulating the future freight rates levels of a specific shipping
system by applying various future oil production values. By assuming a
specific starting point with well defined supply attributes, Orderbook value,
oil production etc the method is capable of estimating the relation between
future freight rates and future OPEC oil production. For example in March
2003 oil production was 28684 million barrels per day and the freight rate
at Ras Tanura — Rotterdam route was 111 WS. Assuming Ceteris paribus
for all shipping variables Fig. 7 shows the relation between future (after 12
months) freight rates and future (after 12 months) oil production:
• Due to very high values of orderbook at that time (47.8 mil ton DWT),
the large delivery rate would push freight rates to lower levels if demand
for transport services would stay constant. In fact after 12 months
although oil production increased by 2% freight rates decreased from

111 to 98 WS. According to Fig. 7 in order freight market to hold at
111WS the oil production had to increase at 30300 million barrels per
day.
• According to the j-shape curve for oil production below 20000 million
barrels per day the balance is lower freight rates but as oil production
increase the slope of the curve also increase.
• Although ANN have the ability to generalize close to the training values,
for oil production values over the 40000 million barrels per day and lower
than 15000 million barrels per day there is an uncertainty regarding the
results due to the fact that these extreme values were not in the training
set (outliers).
Figure 8 shows simultaneously the estimations for 12 months after

January 2003 (red curve) and March 2003 (blue curve). The shipping
fundamentals are more or less the same hence the expected reaction of
the system is similar.
By applying totally different shipping system fundamentals to the
method the results are totally different. At June 1994 the percentage of
laid up vessels was 34.4% while at March 2003 it was only 6%. The
fundamentals at June 1994 show a non-volatile shipping system to oil
production fluctuation. This is shown at Fig. 9.
Figure 10 shows the reaction of the shipping system at two totally
different time periods. At May 1987 there is a large amount of laid up
vessels and the OPEC oil production is at 18000 million barrels per day
when at June 1994 it was 26000 million barrels per day.
Fig. 8 WS vs OPEC oil production (millionbarrels per day) (12 months ahead March
2003 — squares — and January 2003 — diamonds).
Fig. 9 WS vs OPEC oil production (millionbarrels per day) (12 months ahead March
2003 — squares — and June 1994 — diamonds).
Fig. 10 WS vs OPEC oil production (millionbarrels per day) (12 months ahead May
1987 — crosses — and June 1994 — triangles).
As shown in Fig. 10 the curve of estimated freight rates after 12 months

of May 1987 is shifted to the left compared with the one of June 1994.
That complies with the fact that the available tonnage at May 1987 was
112 million ton. DWT while at June 1994 it was 130 million ton. DWT.
4. Conclusions
In this paper we presented the use of an innovative simulation technique.
Although FORESIM was developed based on special shipping market
characteristics, it has a wide range of applications in econometric systems.
Focusing on shipping needs we concluded that risk is dominant in every
decision. Our research aims to measure risk and provide initially to tanker
owners a decisional framework to manage risk. To achieve this, firstly a
tanker market analysis since 1979 was necessary in order to reveal shipping
market mechanism, establish the most important factors affecting the
market and decide whether a stochastic calculus was needed. The analysis
showed that the crucial non shipping external variable affecting tanker
market is the OPEC oil production level. This variable embodies political,
economical, direct and indirect excitations to the shipping market. There is
no better way to model and quantify excitations such as wars involving oil
producing countries affecting productivity or economical crashes or OPEC
decisions leading in many cases to oil shocks. Oil Production time series
includes all economic and political facts in a global level that may affect a
globalize market such as Tanker shipping market.
Subsequently, and keeping the aforementioned statement in mind, we
tried to simulate the behavior of this variable using an E- GARCH model.
When this was satisfactorily achieved, modular Artificial Neural Networks
was trained to forecast future values of freight rates. Having real historical
data for the route Ras Tanura –Rotterdam we constructed the ANN so as to
predict the Tanker market after, three, six, nine and twelve month periods
and tested it against randomly selected out of training sample data. To be
precise, in fact, separate ANNs for each point in future were constructed,
trained, and tested. The results showed that all ANNs were adequately
capable of simulating future freight rates.
The special characteristics of FORESIM technique are shown in
Table 8.
Table 8 Simulation vs Physical System.
Systems
Simulated system with the

Physical shipping market use of FORESIM
Systems Freight rate generation Use of explanatory ANN to

characteristics mechanism capture causality relations
and interactions
Random excitations from Use of stochastic models to
external non systemic express randomness
parameters
Non stationary system Ability to add/remove
parameters and readjust
weights (adaptive system)
Dynamic system-variability Ability to use technique in real
of shipping states in time time (real time output)
The procedure was developed in order to produce future freight rates

realizations depended to the current state of the market. The procedure is
the first to introduce the concept of generating freight rate realizations
conditional upon the current or the preceding market states and of
embedding explanatory and stochastic modeling. Therefore, it creates tool
for acquiring quality information regarding the trend of the market taking
into consideration unforeseen parameters as well as the present status of
the market.
The main applications of FORESIM are the following:
• Decision support for trading Future Freight Agreements (FFAs) and

various shipping market derivatives;
• Chartering strategy — spot or time charter, duration, etc.;
• Risk management for shipping investments, in combination with cash flow
and monte carlo simulations providing distribution for financial variables;
• Estimation of financing risk such as probability of default etc.
In general it can be concluded that FORESIM represents one very

promising tool in simulating freight rate time series.
References
1. Veenstra, A.W. (1999). Quantitative Analysis of Shipping Markets (Delft
University Press Postbus).
2. Gourieroux, B.V.C. (1997). ARCH Models and Financial Applications,
Springer-Verlag.
3. Beenstock, M. and Vergottis, A. (1989). An econometric model of the world
tanker markets. Journal of Transport Economics and Policy, 23, pp. 263–280.
4. Beenstock, M. and Vergottis, A. (1993). Econometric Modeling of World
Shipping (Chapman & Hall).
5. Berndt, E.R., Hall, B.H., Hall, R.E. and Haussman, J.A. (1974). Estimation
and inference in nonlinear structural models, Annals of Economic and Social
Measurement, 4, pp. 653–665.
6. Bollerlev, T. and Engle, R.F. (1994). Arch Models, Handbook of Economet-
rics, Volume IV, Elsevier Science.
7. Box, G.E. and Jenkins, G.M. (1970). Time Series Analysis: Forecasting and
Control, Holden Day, San Francisco.
8. Drewry Shipping Consultants Ltd.: The Drewry Monthly.
9. Frees, E. (1996). Data Analysis Using Regression Models (Prentice Hall).
10. Enders, W. (1995). Applied Econometric Time Series (John Wiley & Sons).
11. Engle, R.F. (1995). ARCH Selected Readings, Advanced Texts in Economet-
rics (Oxford University Press).
12. Eriksen, I.E. (1977). The Demand for Bulk Services. Norwegian Maritime
Research, 2, Págs. 22–26.
13. Evans, J.J. (1994). An analysis of efficiency of the bulk shipping markets,
Maritime Policy and Management, 21(4), pp. 311–329.
14. Demuth, H. and Beale, M. (1996). Neural Network Toolbox for use with
MATLAB (The Math Works Inc.).
15. Hentschel, L. (1995). All in the Family: Nesting Symmetric and Asymmetric
GARCH Models, Journal of Financial Economics, 39, pp. 71–104.
16. Li, J. and Parsons, M. (1997). Forecasting Tanker Freight Rate Using Neural
Networks (Maritime Policy and Management January Issue).
17. Jang, J.S.R., Sun, C.T. and Mizutani, E. (1997). Neuro-Fuzzy and Soft
Computing (Prentice Hall).
18. Koopmans, T.C. (1939). Tanker Freight Rates and Tankship Building,
Haarlem, The Netherlands.
19. Lloyd’s: Shipping Economist, Clare Longley (publisher)
20. Lyridis D.V., Zacharioudakis P.G., Mitrou, P. and Mylonas, A. (2004a). Fore-
casting tanker market using artificial neural networks, Maritime Economics
and Logistics, 6(2), pp. 93–108.
21. Lyridis, D.V., Zacharioudakis, P.G. and Ousantzopoulos, G. (2004b).
A Multi-Regression approach to forecasting freight rates in the dry bulk
shipping market using neural networks, Proceeding of the Annual Conference
of the International Association of Maritime Economists, IAME 2004, Izmir,
Turkey, Vol. II, pp. 797–811.
22. Azoff, M. (1994). Neural Network Time Series Forecasting of Financial
Markets (Wiley).
23. Beenstock, M. and Vergottis, A. (1993). Econometric Modelling of World
Shipping (Chapman & Hall).
24. Monthly Energy Review of the Energy Information Administration, U.S.
Department of Energy (http://www.eia.doe.gov/).
25. Nelson, D.B. (1991). Conditional Heteroskedasticity in asset returns: A new
approach, Econometrica, 59, 1991.
26. Norman, V.D. and Wergeland, T. (1981). Nortank — A simulation model of
the freight market for large tankers. Centre for Applied Research, Norwegian
School of Economics and Business Administration, Report 4/81. Bergen,
Norway.
27. Norman, Victor D. (1979). Economics of Bulk Shipping. Institute for Ship-
ping Research, Norwegian School of Economics and Business Administration,
Bergen.
28. Pindyck R. and Rubenfeld, D. (1998). Econometric Models and Economic
Forecasts (McGraw-Hill).
29. Trippi, R.R. and Turban, E. (1996). Neural Networks in Finance and
Investing (Irwin).
30. Strandenes, S.P. (1986). NORSHIP — a simulation model for bulk shipping
markets. Center for Applied Research, Norwegian School of Economics and
Business Administration, Bergen, Norway.
31. Svendsen, A.S. (1958). Sea Transport and Shipping Economics, Bremen.
32. Crooks, T. (1992). Care and Feeding of Neural Networks.

33. Shimojo, T. (1979). Economic Analysis of Shipping Freights (Kobe
University).
34. Timbergen, J. (1939). Selected Papers, Klaasen, L.H. et al. Editors, North
Holland Publishing Company, Amsterdam, The Netherlands.
35. Timebergen, J. (1931). Ein Schiffbauzyclus? Weltwirtschaftliches Archiv, 34,
pp. 152–64.
36. Jhee, W.C. and Shaw, M.J. Time Series Prediction Using Minimally
Structured Neural Networks: An Empirical Test.
37. Zannetos, Z. (1966). The Theory of Oil Tankship Rates (MIT press).

Advanced in Maritime Logistics PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced in Maritime Logistics PDF

Uploaded by

Copyright:

Available Formats

Advances in Maritime Logistics

and Supply Chain Systems

British Library Cataloguing-in-Publication Data

ADVANCES IN MARITIME LOGISTICS AND SUPPLY CHAIN SYSTEMS

Typeset by Stallion Press

Steven - Advances in Maritime Logistics.pmd 1 6/14/2011, 2:01 PM

Part 1: Regional Developments and Performance

2. Recent Development of Maritime Logistics 49

3. Scenario Analysis for Hong Kong Port Development

4. Models for Port Competitive Analysis in Asia-Paciﬁc Region 91

5. Is Port Throughput a Port Output? 117

6. A Framework for Modelling and Benchmarking

5. Methods for Evaluating and Benchmarking Maritime

7. A Performance Evaluation Strategy Towards Dealers

Part 2: Ports and liners operations 171

2.1. Yard allocation for import containers . . . . . . . 174

9. Integration of AGVS in Intermodal Rail Operations

10. On the Ongoing Increase of Containership Size 201

3. Shipping Costs of Post Panamax Containerships . . . . . 209

11. A Linearized Approach for Liner Ship Fleet Planning

12. Ship Emissions, Costs and Their Tradeoﬀs 257

8. SECAs Continued: Eﬀect on Modal Split . . . . . . . . . 290

13. Exploring Tanker Market Elasticity with Respect to Oil

This page intentionally left blank

Regional Developments and Performance Analysis

The paper by L. H. Lee, E. P. Chew, L. Zhen, C. C. Gan, and J. Shao

Ports and Liners Operations

originally proposed nonlinear programming model. Numerical examples are

December 2010, Singapore

This page intentionally left blank

history. Particularly, international maritime trade has played a signiﬁcant

Philippines and Thailand) had experienced a disproportionate share of population and

Maritime Trade Evolutions and Port City Developments in Asia 7

2. Evolution of Asia’s Maritime Trade

In the following sub-sections, we explore the evolution of Asia’s mari-

2.1. Ancient maritime trade between India and China,

Maritime Trade Evolutions and Port City Developments in Asia 9

world military expansion of Yuan Dynasty. Particularly, the maritime route

2.2. Maritime networks in colony times, 1500–1950

to control the trade in the Straits from the seventeenth to eighteenth

Maritime Trade Evolutions and Port City Developments in Asia 11

and culture, triggering innovation with local experience and improving

2.3. Asia’s maritime trade under globalization

Maritime Trade Evolutions and Port City Developments in Asia 13

play an important role in economic development, many saw an expansion in

Globalization Maritime Trade

Colonial Economic Trade

Ancient Official and Private Trade

Fig. 2 Evolution of maritime trade in Asia.

e Directcarrier services augment the connectivity of a port by increasing the number of

Fig. 3 Evolution of ports in Asia, modiﬁed from [Lee et al., 2008].

3. Asia Port Developments in the 1990s and Beyond

f Japan has experienced a negative GDP growth of −0.7% in 2008.

Maritime Trade Evolutions and Port City Developments in Asia 15

Fig. 4 Ports in Northeast Asia.

3.1. Northeast Asia

3.1.1. Greater China region

The Hong Kong port is well-regarded as a highly eﬃcient international

Maritime Trade Evolutions and Port City Developments in Asia 17

60,000 tons in 2009. In line with the Taiwanese government’s strategy to

Maritime Trade Evolutions and Port City Developments in Asia 19

Located at the eastern tip of the Korean Peninsula, the gateway on

Maritime Trade Evolutions and Port City Developments in Asia 21