Professional Documents
Culture Documents
Life-Cycle Cost Analysis Dot - 41999 - DS1
Life-Cycle Cost Analysis Dot - 41999 - DS1
FHWA-SA-98-079
To reinforce this publication, the FHWA Office of Engineering, Pavement Division, in cooperation with
the Office of Technology Applications, offers LCCA technical support through FHWA Demonstration
Project No.115, Probabilistic LCCA in Pavement Design (DP-115). DP-115 is a free 2-day
workshop that demonstrates good/best practices in performing life-cycle cost analyses for pavement
design. This workshop is available, upon request, to State highway agencies.
This publication and DP-115 support the FHWA’s response to the Transportation Equity Act for the
21st Century legislative mandate to develop recommended procedures for conducting LCCA on
National Highway System projects.
The United States Government does not endorse products or manufacturers. Trade or manufacturers’
names appear herein only because they are considered essential to the objective of this document.
Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient’’s Catalog Number
FHWA-SA-98-079
4. Title and Subtitle 5. Report Date
16. Abstract
This Interim Technical Bulletin recommends procedures for conducting Life-Cycle Cost Analysis
(LCCA) of pavements, provides detailed procedures to determine work zone user costs, and introduces a
probabilistic approach to account for the uncertainty associated with LCCA inputs. The Bulletin begins
with a discussion of the broad fundamental principles involved in an LCCA. It discusses input parameters
and presents simple examples of traditional LCCA in a pavement design setting. It discusses the variability
and inherent uncertainty associated with input parameters, and provides recommendation on acceptable
ranges for the value of time as well as discount rates. It explores the use of sensitivity analysis in
traditional LCCA approaches. User costs are a combination of delay, vehicle operating costs, and crash
costs. Each of these cost components is explored and procedures are presented to determine their value.
Given the power and sophistication of today’s computers and software, simulation techniques such as
Monte Carlo are recommended for incorporating variability associated with LCCA inputs into final results.
iii
ACKNOWLEDGMENTS
The authors appreciate the following individuals who reviewed this Bulletin:
The authors are grateful for the special support received from the Pennsylvania Department of
Transportation for use of its user cost procedures as a framework for developing the User Cost
chapter of this Bulletin.
iv
CONTENTS
v
Step 5. Quantify Traffic Affected by Each Cost Component ........................................................54
Step 6. Compute Reduced Speed Delay ........................................................................................57
Step 7. Select and Assign VOC Rates ...........................................................................................64
Step 8. Select and Assign Delay Cost Rates ................................................................................65
Step 9. Assign Traffic to Vehicle Classes .....................................................................................65
Step 10. Compute User Cost Components by Vehicle Class ......................................................... 66
Step 11. Sum Total Work Zone User Costs ...................................................................................69
Step 12. Address Circuity and Delay Costs ..................................................................................70
CIRCUITY ...................................................................................................................................................71
CRASH COSTS ............................................................................................................................................72
General ..........................................................................................................................................72
Overall Crash Rates ......................................................................................................................73
Work Zone Crash Rates ................................................................................................................74
Example Crash Cost Calculations .................................................................................................78
vi
FIGURES
Figure 1. 1 Historical trends on 10-year Treasury notes. ............................................................................. 6
vii
TABLES
Table 2.1 PennDOT’s design strategy for new, reconstructed, and unbonded overlay. ............................. 11
Table 2.2 Added time and vehicle running cost /1,000 stops and idling costs (1970 $). ............................. 17
Table 2.3 Added time and vehicle running cost /1,000 stops and idling costs (Aug 96 $). .........................18
Table 2.4 Speed change computations. .......................................................................................................18
Table 2.5 Updated NCHRP 133 values of time ($/Veh-Hr) (Aug 96 $). ......................................................... 19
Table 2.6 Updated MicroBENCOST default values of time ($/Veh-Hr) (Aug 96 $). .....................................20
Table 2.7 Composite earlier research value of time ($/Veh-Hr) (Aug 96 $). ..................................................20
Table 2.8 Travel time ranges as a percent of national wage rate (1995 $/Person-Hr). .................................. 21
Table 2.9 OST-recommended hourly wage rates (1995 $/Person-Hr). .......................................................... 21
Table 2.10 Ranges for hourly values of travel time (1995 $/Person-Hr). ......................................................21
Table 2.11 Value of one vehicle hour of travel time (1995 $). ....................................................................... 22
Table 2.12 Composite listing of travel time values. .....................................................................................23
Table 2.13 Recommended values of time ($/Veh-Hr) (Aug 96 $). ................................................................. 23
Table 2.14 MicroBENCOST default crash cost rates ($1,000, 1990 $). ......................................................... 23
Table 2.15 MicroBENCOST default crash cost rates ($1,000, Aug 96 $). .................................................... 24
Table 2.16 Present value discount factors: single future payment. .............................................................26
Table 2.17 NPV calculation using 4 percent discount rate factors. .............................................................27
Table 2.18 Sensitivity analysis – Alternative # 1. ........................................................................................ 28
Table 2.19 Sensitivity analysis – Alternative # 2. ........................................................................................ 28
Table 2.20 Comparison of alternative NPVs ($1,000) to discount rate. ........................................................29
Table 2.21 Sensitivity to user cost and discount rate. ................................................................................. 30
Table 3.1 Default hourly distributions from MicroBENCOST (all functional classes). ................................38
Table 3.2 PennDOT AADT distribution (hourly percentages). ...................................................................39
Table 3.3 Work zone directional hourly demand (all vehicle classes). ......................................................... 44
Table 3.4 Truck equivalency factors. .......................................................................................................... 46
Table 3.5 Maximum mixed vehicle traffic capacities for trucks in the traffic stream
(4-lane facilities). .........................................................................................................................47
Table 3.6 Maximum mixed vehicle traffic capacities for trucks in the traffic stream
(6 or more lanes). .........................................................................................................................47
Table 3.7 Observed saturation flow rates per hour of green time. ............................................................... 48
Table 3.8 Measured average work zone capacities. .....................................................................................49
Table 3.9 Work zone analysis matrix. ........................................................................................................... 51
Table 3.10 Expanded work zone matrix. .......................................................................................................54
Table 3.11 Summary traffic affected by each cost component. ...................................................................57
Table 3.12 Work zone reduced speed delay. ................................................................................................58
Table 3.13 Queue speed. ............................................................................................................................. 58
Table 3.14 Average queue length calculations. ...........................................................................................62
Table 3.15 Average queue length – alternative approach. ..........................................................................63
viii
Table 3.16 Average queue delay time. ......................................................................................................... 64
Table 3.17 Added time and vehicle running cost/1,000 stops and idling costs (Aug 96 $). ........................64
Table 3.18 Speed change computations. ..................................................................................................... 65
Table 3.19 Recommended values of travel time ($/Veh-Hr) (Aug 96). .......................................................... 65
Table 3.20 Affected traffic by vehicle class and user cost component. ......................................................66
Table 3.21 User cost component # 1 - speed change VOC (55-40-55 mi/h). .................................................66
Table 3.22 User cost component # 2 - speed change delay cost (55-40-55 mi/h). ........................................ 67
Table 3.23 User cost component # 3 - work zone reduced speed delay cost. .............................................. 67
Table 3.24 User cost component # 4 - stopping VOC (55-0-55 mi/h). .......................................................... 67
Table 3.25 User cost component # 5 - stopping delay cost (55-0-55 mi/h). .................................................67
Table 3.26 User cost component # 6 - idling VOC. ......................................................................................68
Table 3.27 User cost component # 7 - queue reduced speed delay cost. .................................................... 68
Table 3.28 Master summary - total (60 day) work zone user cost (Aug 96 $). ............................................. 69
Table 3.29 Master summary - work zone user cost distribution (%). ........................................................... 69
Table 3.30 Average weekday delay – North Ridge earthquake. ..................................................................70
Table 3.31 1995 people injured in motor vehicle crashes by functional class. ............................................. 73
Table 3.32 1995 vehicles miles of travel (millions). ......................................................................................73
Table 3.33 Crash injury rates (people injured per 100 M VMT). .................................................................. 74
Table 3.34 1996 work zone motor vehicles crash fatalities as a percent of all fatalities. ...............................75
Table 3.35 Crash rates on SLC. ....................................................................................................................76
Table 3.36 Crash rates on TLTWO. .............................................................................................................77
Table 3.37 Average overall crash rates. .......................................................................................................78
Table 3.38 Average fatal and nonfatal injury crash rates. ............................................................................ 78
Table 3.39 Crash cost calculation matrix. ..................................................................................................... 79
ix
ABBREVIATIONS
ADT ............................................................................................................................. Average Daily Traffic
AADT ............................................................................................................. Average Annual Daily Traffic
BAMS ..................................................................................................... Bid Analysis Management System
B/C .................................................................................................................................... Benefit/Cost Ratio
CFR ................................................................................................................... Code of Federal Regulations
CPI ............................................................................................................................... Consumer Price Index
DP-115 .................................................................................................................. Demonstration Project 115
DOT ................................................................................................................ Department of Transportation
EUAC ......................................................................................................... Equivalent Uniform Annual Cost
FARS ........................................................................................................ Fatal Accident Reporting Systems
FHWA ........................................................................................................ Federal Highway Administration
GAO ..................................................................................................................... General Accounting Office
HCM .................................................................................................................... Highway Capacity Manual
HERS ............................................................................................ Highway Economic Requirements System
HMAC ................................................................................................................. Hot-Mix Asphalt Concrete
HOT ............................................................................................................................. High Occupancy Toll
HOV ........................................................................................................................ High Occupancy Vehicle
IRI ................................................................................................................. International Roughness Index
IRR ............................................................................................................................. Internal Rate of Return
ISTEA .............................................................................. Intermodal Surface Transportation Efficiency Act
LCC ........................................................................................................................................ Life-Cycle Cost
LCCA ..................................................................................................................... Life-Cycle Cost Analysis
LTPP ......................................................................................................... Long-Term Pavement Performance
mi/h .......................................................................................................................................... miles per hour
MPO ...................................................................................................... Metropolitan Planning Organization
NCHRP ............................................................................ National Cooperative Highway Research Program
NHS ....................................................................................................................... National Highway System
NPV .................................................................................................................................... Net Present Value
NPW ................................................................................................................................. Net Present Worth
OIG ...................................................................................................................... Office of Inspector General
OMB ....................................................................................................... Office of Management and Budget
OST ................................................................................................Office of the Secretary of Transportation
PennDOT ................................................................................. Pennsylvania Department of Transportation
PCCP .................................................................................................... Portland Cement Concrete Pavement
pcplph ........................................................................................................ passenger cars per lane per hour
PV .............................................................................................................................................. Present Value
PSR .................................................................................................................. Present Serviceability Rating
RSL ............................................................................................................................ Remaining Service Life
SHA .......................................................................................................................... State Highway Agency
SLC ................................................................................................................................ Single-Lane Closure
SOV ..................................................................................................................... Single Occupancy Vehicles
TEA-21 ................................................................................ Transportation Equity Act for the 21st Century
TLTWO ......................................................................................................... Two-Lane Two-Way Operation
VMT ........................................................................................................................ Vehicle Mile(s) of Travel
vph ..................................................................................................................................... Vehicles Per Hour
V/C ......................................................................................................................... Volume-to-Capacity Ratio
VOC ............................................................................................................................ Vehicle Operating Cost
WZ ............................................................................................................................................... Work Zone
x
EXECUTIVE SUMMARY
This Interim Technical Bulletin provides technical guidance and recommendations on good
practice in conducting Life-Cycle Cost Analysis (LCCA) in pavement design. It also introduces
Risk Analysis, a probabilistic approach to describe and account for the uncertainty inherent in
the process. It deals specifically with the technical aspects of the long-term economic efficiency
implications of alternative pavement designs. The Bulletin is directed at State highway agency
(SHA) personnel with responsibility for conducting and/or reviewing pavement design LCCAs.
Purpose of LCCA
LCCA is an analysis technique that builds on the well-founded principles of economic analysis
to evaluate the over-all-long-term economic efficiency between competing alternative investment
options. It does not address equity issues. It incorporates initial and discounted future agency,
user, and other relevant costs over the life of alternative investments. It attempts to identify the
best value (the lowest long-term cost that satisfies the performance objective being sought) for
investment expenditures.
LCCA Requirements
The National Highway System (NHS) Designation Act of 1995 specifically required States to
conduct life-cycle cost analysis on NHS projects costing $25 million or more. Implementing
guidance was provided in Federal Highway Administration (FHWA) Executive Director
Anthony Kane’s April 19, 1996, Memorandum to FHWA Regional administrators.
The implementing guidance did not recommend specific LCCA procedures, but rather it
specified the use of good practice.
The FHWA position on LCCA is further defined in its Final Policy Statement on LCCA
published in the September 18, 1996, Federal Register. FHWA Policy on LCCA is that it is a
decision support tool, and the results of LCCA are not decisions in and of themselves. The
logical analytical evaluation framework that life-cycle cost analyses fosters is as important as the
LCCA results themselves. As a result, although LCCA was only officially mandated in a very
limited number of situations, FHWA has always encouraged the use of LCCA in analyzing all
major investment decisions where such analyses are likely to increase the efficiency and
effectiveness of investment decisions whether or not they meet specific LCCA-mandated
requirements.
The 1998 Transportation Equity Act for the 21st Century (TEA-21) has sinced removed the
requirement for SHA’s to conduct LCCA on high-cost NHS useable project segments.
However, the congressional interest in LCCA is continued in the new requirement that the
Secretary of Transportation develop recommended LCCA procedures for NHS projects.
Bulletin Format
The Interim Technical Bulletin discusses the broad fundamental principles involved in LCCA
and it presents widely accepted procedures used in setting up and conducting LCC analysis.
xi
It also discusses input parameters, the variability and inherent uncertainty associated with them,
and provides recommendations on acceptable ranges for a variety of parameters. It presents
examples of traditional LCCA in a pavement design setting. It then provides a detailed, rational
highway capacity-based approach for determining work zone user delay, vehicle operating, and
crash costs associated with alternative pavement design strategies. It explores the use of
sensitivity analysis in traditional LCCA approaches and introduces a probabilistic-based risk
analysis approach to account for the variability of inputs. Several microcomputer software
programs are available for conducting deterministic LCCA on routine pavement rehabilitation
projects. There are also powerful microcomputer-based risk analysis software programs
currently on the market that work well in conjunction with standard computer spreadsheet
applications. The appendix to this Interim Bulletin includes a discussion of supporting
computer software and additional LCCA resource documents.
LCCA Procedures
Life Cycle Cost (LCC) analysis should be conducted as early in the project development cycle
as possible. For pavement design, the appropriate time for conducting the LCCA is during the
project design stage. The LCCA level of detail should be consistent with the level of
investment. Typical LCCA models based on primary pavement management strategies can be
used to reduce unnecessarily repetitive analyses.
LCCA need only consider differential cost among alternatives. Costs common to all alternatives
cancel out, are generally so noted in the text, and are not included in LCCA calculations.
Inclusion of all potential LCCA factors in every analysis is counterproductive; however, all
LCCA factors and assumptions should be addressed, even if only limited to an explanation of
the rationale for not including eliminated factors in detail. Sunk costs, which are irrelevant to the
decision at hand, should not be included.
xii
mixed (i.e., all costs must be in real dollars or all costs must be in nominal dollars). Further, the
discount rate selected must be consistent with the dollar type used (i.e., use real cost and real
discount rates or nominal cost and nominal discount rates).
The discount rates employed in LCCA should reflect historical trends over long periods of time.
Although long-term trends for real discount rates hover around 4 percent, 3 to 5 percent is an
acceptable range and is consistent with values historically reported in Appendix A of OMB
Circular A-94.
Performance periods for individual pavement designs and rehabilitation strategies have a
significant impact on analysis results. Longer performance periods for individual pavement
designs require fewer rehabilitation projects and associated agency and work zones user costs.
While most analyses include traditional agency costs, some do not fully account for the SHA
engineering and construction management overhead, especially on future rehabilitations. This can
be a serious oversight on short-lived rehabilitations as SHAs design processes lengthen in an era
of downsizing.
Routine, reactive type annual maintenance costs have only a marginal effect on NPV. They are
hard to obtain, generally very small in comparison to initial construction and rehabilitation costs,
and differentials between competing pavement strategies are usually very small, particularly
when discounted over 30- to 40-year analysis periods.
Salvage value should be based on the remaining life of an alternative at the end of the analysis
period as a prorated share of the last rehabilitation cost.
User Costs
User costs are the delay, vehicle operating, and crash costs incurred by the users of a facility
and should be included in the LCCA. Vehicle delay and crash costs are unlikely to vary among
alternative pavement designs between periods of construction, maintenance, and rehabilitation
operations. Although vehicle operating costs are likely to vary during periods of normal
operations for different pavement design strategies, there is little research on quantifying such
Vehicle Operating Cost (VOC) differentials under the pavement condition levels prevailing in the
U.S.A. The Technical Bulletin therefore focuses strictly on work zone user cost differences
between alternatives.
User costs are heavily influenced by current and future roadway operating characteristics. They
are directly related to the current and future traffic demand, facility capacity, and the timing,
duration, and frequency of work zone-induced capacity restrictions, as well as any circuitous
mileage caused by detours. Directional hourly traffic demand forecasts for the analysis year in
question are essential for determining work zone user costs.
As long as work zone capacity exceeds vehicle demand on the facility, user costs are normally
manageable and represent more of an inconvenience than a serious cost to the traveling public.
When vehicle demand on the facility exceeds work zone capacity, the facility operates under
forced-flow conditions and user costs can be immense. Queuing costs can account for more
than 95 percent of work zone user costs with the lion’s share of the cost being the delay time of
crawling through long, slow-moving queues.
xiii
Different vehicle classes have different operating characteristics and associated operating costs,
and as a result, user costs should be analyzed for at least three broad vehicle classes: Passenger
Vehicles, Single-Unit Trucks, and Combination Trucks.
User delay cost rates are probably the most contentious of all user cost inputs. While there are
several different sources for the dollar value of time delay, the recommended mean values and
ranges for the value of time (Aug 96 $) shown in the table below appear reasonable. It is
important to note that commercial vehicles support higher values of travel time delay rates and
that passenger vehicles, particularly pickup trucks, represent both commercial and
noncommercial use.
Work zone crash cost differentials between alternatives are very difficult to determine because of the
lack of hard statistically significant data on work zone crash rates and the difficulty in determining vehicle
work zone exposure. However, default dollar value ranges associated with fatal and nonfatal injury
highway crashes are included.
Risk Analysis
LCCA, as a minimum, should include a sensitivity analysis to address the variability within major
analyses input assumptions and estimates. Traditionally, sensitivity analysis has evaluated different
discount rates or assigned value of time, normally evaluating a best and worst case scenario. The
ultimate extension of sensitivity analysis is a probabilistic approach, which allows all significant inputs to
vary simultaneously.
The Interim Technical Bulletin advocates the use of a probabilistic approach to LCCA that
incorporates analysis of the variation within the input assumptions, projections, and estimates. The
prevailing term used in private industry for a probabilistic approach is Risk Analysis. Risk analysis is a
technique that exposes areas of uncertainty, typically hidden in the traditional deterministic approach to
LCCA, and it allows the decision maker to weigh the probability of the outcome actually occurring. The
risk analysis approach combines probability descriptions of uncertain variables and a computer
simulation technique, generally know as Monte Carlo Simulation, to characterize uncertainty. Monte
Carlo simulations randomly draw samples from the individual inputs consistent with their defined
distributions to calculate thousands, even tens of thousands, of what if outcomes. With enough
samples, the program can define an overall composite NPV probability distribution for each alternative
— one that shows the entire range of possible outcomes and the likelihood that any particular outcome
will actually occur. Given the power and sophistication of today’s computers and software, the FHWA
strongly endorses the use of techniques, such as Monte Carlo simulation, for incorporating variability
associated with LCCA inputs into final results.
xiv
CHAPTER 1. INTRODUCTION
The purpose of this Interim Technical Bulletin is to provide technical guidance and
recommend good practice in conducting Life-Cycle Cost Analysis (LCCA) in pavement
design and to introduce Risk Analysis, a probabilistic approach, which describes the
uncertainty inherent in the process. The primary audience for this Bulletin is State highway
agency (SHA) personnel responsible for conducting and/or reviewing LCCA of highway
pavements. This includes State pavement design engineers and pavement management
engineers, as well as district or area supervisors responsible for selecting pavement type and
rehabilitation strategies.
SCOPE
The Interim Technical Bulletin recommends specific procedures for conducting LCCA in
pavement design and discusses the relative importance of LCCA factors on analysis results. In
the interest of technical purity, the discussion includes all relative LCCA factors, even though
not all elements influence the final LCCA results to the same degree. The Bulletin first
addresses the broad fundamental principles involved in LCCA; this is followed by presentation
of the widely accepted procedures used to set up and conduct LCC analysis. It discusses input
parameters and presents examples of traditional LCCA in a pavement design setting. It
discusses the variability and inherent uncertainty associated with input parameters and
recommends acceptable ranges for a variety of parameters. It explores the use of sensitivity
and introduces a risk analysis approach to account for the variability of inputs. Finally, there is a
discussion of supporting computer software. The appendix lists additional LCCA resource
documents specific to pavement design.
While the issue of equity is a highly significant consideration in any public investment decision, it
is not part of the economic efficiency issue. This Interim Technical Bulletin deals specifically
with the technical aspects of the long-term economic efficiency implications of alternative
pavement designs.
LCCA results are a useful decision support tool, but they are not decisions in and of
themselves. Frequently, the analytical evaluation that such analysis fosters is as important as the
LCCA results. As a result, SHAs are encouraged to conduct LCCA in support of all major
investment decisions.
APPLICATION
Fundamental principles of economic analysis have broad application. In general, this Interim
Technical Bulletin presents generic concepts that may be applied to areas other than
pavements. For example, LCCA may be applied to establish funding levels, allocate resources
among program areas, and prioritize project selection.
1
LEVEL OF DETAIL
The relative influence of individual Life-Cycle Cost (LCC) factors on analysis results may vary
from major to minor to insignificant. The analyst should ensure that the level of detail
incorporated in an LCCA is consistent with the level of investment decision under consideration.
There comes a point of diminishing returns as more and more cost factors are incorporated in an
LCCA. For example, slight differences in future costs have a marginal effect on discounted
present value. Including such factors as this unnecessarily complicates the analysis without
providing tangible improvement in analysis results. Including all factors in every analysis is
frequently not productive. The difficulty in capturing some costs makes omitting them the more
prudent choice — particularly when the effect on the LCCA results is marginal at best.
In conducting an LCCA, analysts should evaluate all factors for inclusion and explain the
rationale for eliminating factors. Such explanations make analysis results more supportable when
they are scrutinized by critics who are not pleased with the analysis outcome. This Interim
Technical Bulletin does not provide guidance on determining the appropriate extent of LCCA
on specific projects.
The current FHWA position on pavement-related LCCA has its roots in the Intermodal Surface
Transportation Efficiency Act (ISTEA) of 1991, which specifically required consideration of
“the use of life-cycle costs in the design and engineering of bridges, tunnels, or pavement” in
both Metropolitan and Statewide Transportation Planning. Additional direction came in January
1994 with Executive Order No.12893, “Principles for Federal Infrastructure Investments,”
which requires systematic analysis of benefits and costs when making infrastructure investment
decisions. It also requires that the costs be measured and discounted over the full life cycle of
each project. Further, an Office of Inspector General/Government Accounting Office (OIG/
GAO) 1994 Highway Infrastructure report on cost comparison of asphalt versus concrete
pavements reviewed in the Federal Highway Administration (FHWA) Region 4 States made
specific recommendations on the FHWA’s need to provide additional technical guidance
on LCCA.(1)
In addition, the National Highway System (NHS) Designation Act of 1995 specifically requires
that the Secretary of Transportation establish a program requiring States to conduct life-cycle
costs analysis on NHS projects where the cost of a usable project segment equals or exceeds
$25 million. The FHWA’s Executive Director, Anthony Kane, distributed implementing
guidance on NHS LCCA requirements to FHWA field offices in an April 19, 1996,
memorandum. The implementing guidance focused on the use of good practice rather than
prescribe specific LCCA procedures.
The NHS Designation Act of 1995 also required the SHAs to perform Value Engineering
Analysis on the same high-cost NHS projects. The Value Engineering provisions were
implemented in 23 Code of Federal Regulations (CFR) Part 627 published in the Federal
Register in February 1997, and requirements took effect March 17, 1997.(2)
2
Finally, the 1998 Transportation Equity Act for the 21st Century (TEA-21) removed the LCCA
requirements established in the NHS Act and directed the Secretary of Transportation to
develop recommended procedures for conducting LCCA on NHS projects. Such
recommended procedures are to be developed in consultation with AASHTO and in concert
with the principles defined in Executive Order 12893.
GENERAL DEFINITIONS
Some of the more general definitions used in this Technical Bulletin are listed below. Other
definitions are provided in the sections where they are addressed.
Life-Cycle Cost Analysis (LCCA), was legislatively defined in Section 303, Quality
Improvement, of the National Highway System NHS Designation Act of 1995. The definition as
modified by TEA-21, is “. . . a process for evaluating the total economic worth of a usable
project segment by analyzing initial costs and discounted future cost, such as maintenance, user,
reconstruction, rehabilitation, restoring, and resurfacing costs, over the life of the project
segment.” A usable project segment is defined as a portion of a highway that, when completed,
could be opened to traffic independent of some larger overall project.
In simpler terms, LCCA is an analysis technique that supports more informed and, it is hoped,
better investment decisions. It builds on some well-founded principles of economic analysis that
have been used to evaluate highway and other public works investments for years, but LCCA
has a slightly stronger focus on the longer term. It incorporates discounted long-term agency,
user, and other relevant costs over the life of a highway or bridge to identify the best value for
investment expenditures (i.e., the lowest long-term cost that satisfies the performance objective
sought). LCCA can be applied to a wide variety of investment-related decision levels to evaluate
the economic worth of various designs, projects, alternatives, or system investment strategies to
get the best return on the dollar.
User Costs are costs incurred by highway users traveling on the facility and the excess costs
incurred by those who cannot use the facility because of either agency or self-imposed detour
requirements. User costs typically are an aggregation of three separate components: Vehicle
Operating Costs (VOC), Crash Costs, and User Delay Costs. Chapter 3 discusses each of
these cost components in detail.
Deterministic Approach to LCCA applies procedures and techniques without regard for the
variability of the inputs. The primary disadvantage of this traditional approach is that it does not
account for the variability associated with the LCCA input parameters.
Risk Analysis Approach characterizes uncertainty. This Interim Technical Bulletin advocates
this approach because it combines probability descriptions of analysis inputs with computer
simulations to generate the entire range of outcomes as well as the likelihood of occurrence.
3
Economic Indicators
Several economic indicators are available to the analyst. The most common include Benefit/
Cost (B/C) Ratios, Internal Rate of Return (IRR), Net Present Value (NPV), and Equivalent
Uniform Annual Costs (EUAC). Many of these indicators are thoroughly discussed in the 1992
Office of Management and Budget Circular A-94.(3)
Internal Rate of Return, primarily used in private industry, represents the discount rate
necessary to make discounted cost and benefits equal. While the IRR does not generally
provide an acceptable decision criterion, it does provide useful information, particularly when
budgets are constrained or there is uncertainty about the appropriate discount rate.
Net Present Value, sometimes called Net Present Worth (NPW), is the discounted monetary
value of expected net benefits (i.e., benefits minus costs). NPV is computed by assigning
monetary values to benefits and costs, discounting future benefits (PVbenefits) and costs (PVcosts)
using an appropriate discount rate, and subtracting the sum total of discounted costs from the
sum total of discounted benefits.
Discounting benefits and costs transforms gains and losses occurring in different time periods to
a common unit of measurement. Programs with positive NPV value increase social resources
and are generally preferred. Programs with negative NPV should generally be avoided. There is
fairly strong agreement in the literature that NPV is the economic efficiency indicator of choice.
The basic formula for computing NPV is:
Because the benefits of keeping the roadway above some preestablished terminal service ability
level are the same for all design alternatives, the benefits component drops out and the formula
reduces to:
Nn 1
NPV = Initial Cost + ∑ Rehab Cost k nk
k =1 (1 + i )
where: i = discount rate
n = year of expenditure
The section on Compute Net Present Value (page 25) discusses NPV computations in more detail.
Equivalent Uniform Annual Costs represents the NPV of all discounted cost and benefits
of an alternative as if they were to occur uniformly throughout the analysis period. EUAC is a
particularly useful indicator when budgets are established on an annual basis. The preferred
4
method of determining EUAC is first to determine the NPV, and then use the following formula
to convert it to EUAC:
1(1 + i ) n
EUAC = NPV
(1 + i ) − 1
n
Additional terms are defined as necessary as they occur in the body of the text.
COST ESTIMATES
Estimates of future costs and benefits can be made using constant or nominal dollars.
Constant dollars, often called real dollars, reflect dollars with the same or constant purchasing
power over time. In such cases, the cost of performing an activity would not change as a
function of the future year in which it would be accomplished. For example, if hot-mix asphalt
concrete (HMAC) costs $20/ton today, then $20/ton should be used for future year HMAC
cost estimates. Nominal dollars, on the other hand, reflect dollars that fluctuate in purchasing
power as a function of time. They are normally used to fold in future general price rises resulting
from anticipated inflation. When using nominal dollars, the estimated cost of an activity would
change as a function of the future year in which it is accomplished. In this case, if HMAC costs
$20/ton today, and inflation were estimated at 5 percent, HMAC cost estimates for 1 year from
today would be $21/ton.
While LCCA can be conducted using either constant or nominal dollars, there are two
cautions. First, in any given LCCA, constant and nominal dollars cannot be mixed in the same
analysis (i.e., all costs must be in either constant dollars or all costs must be in nominal
dollars). Second, the discount rate (discussed below) selected must be consistent with the dollar
type used (i.e., use constant dollars and discount rates or nominal dollars and discount rates).
Good practice suggests conducting LCCA using constant dollars and real discount rates. This
combination eliminates the need to estimate and include an inflation premium for both cost and
discount rates.
DISCOUNT RATES
Similar to costs, LCCA can use either real or nominal discount rates. Real discount rates reflect
the true time value of money with no inflation premium and should be used in conjunction with
noninflated dollar cost estimates of future investments. Nominal discount rates include an
inflation component and should only be used in conjunction with inflated future dollar cost
estimates of future investments. The same caveats, as noted above, apply to mixing real dollar
cost and nominal discount rates and vice versa. The OMB Circular A-94, and the annual
updates of appendix A to the Circular, further discuss the real versus nominal dollar and
discount rates issue.
5
Values to Use
Discount rates can significantly influence the analysis result. LCCA should use a reasonable
discount rate that reflects historical trends over long periods of time. Data on the historical
trends over very long periods indicate that the real time value of money is approximately
4 percent.
In the public sector, because investment resources come from Jane and John Q. Public in the
form of taxes or user fees, the discount rate used needs to be consistent with the opportunity
cost of the public at large. The supersafe U.S. Government Treasury Bill is one conservative
indicator of the opportunity cost of money for the public at large. Figure 1.1 reflects the
historical trend of yields on 10-year Treasury notes. The upper curve reflects the nominal rate
of return while the lower curve represents the inflation adjusted real rate of return. For the
period March 1991 through August 1996, the real rate of return ranges somewhere between 3-
to 5-percent and the average close to 4 percent.
8.06%
Yield on a 10-year Treasury note
6.94%
Amount Lost
to Inflation
3.74%
3.21%
Actual yield to investors,
after inflation
The Department of the Treasury made its first offering of Inflation Protected Securities to the
general public in spring 1997. These securities offer a real rate of return and a provision that
adjusts the principal to protect against inflation. The offering was very well received by the
public (there was more demand for the securities than the Treasury Department wanted to sell)
at a yield of just over 3.5 percent.
In 1995 and 1996, the FHWA Office of Engineering, Pavement Division, conducted a national
pavement design review and found that the discount rates currently employed by SHAs to
conduct LCCA in pavement design showed a distribution of values clustering in the 3- to 5-
percent range.
6
Finally, table 1.1 shows recent trends in real discount rates for various analysis periods
published over the last several years in annual updates to OMB Circular A-94. Considering the
above, good practice suggests using a real discount rate, one that does not reflect an inflation
premium, of 3 to 5 percent in conjunction with real/constant dollar cost estimates.
Analysis Period
Year
3 5 7 10 30
92 2.7 3.1 3.3 3.6 3.8
93 3.1 3.6 4.0 4.3 4.5
94 2.1 2.3 2.5 2.7 2.8
95 4.2 4.5 4.6 4.8 4.9
96 2.7 2.7 2.8 2.8 3.0
97 3.2 3.3 3.4 3.5 3.6
98 3.4 3.5 3.5 3.6 3.8
Average 3.1 3.3 3.4 3.6 3.8
Standard Deviation 0.7 0.7 0.7 0.8 0.7
STRUCTURED APPROACH
Analysts should work from formalized, objective LCCA procedures incorporated within the
overall pavement design process. Such procedures should be comprehensive enough to capture
and evaluate the differences between competing pavement design alternatives and subsequent
rehabilitation strategies. The design process should clearly identify when and at what level to
perform the LCCA, as well as the scope and level of detail of such analysis. LCCA procedures
should clearly identify the components and factors that are included in addition to supporting
rationale for selected input values. LCCA input assumptions should be reasonable and conform
to accepted practice and convention. LCCA should recognize the uncertainty associated with
LCCA inputs and the implication of the uncertainty on LCCA results. As a minimum, LCCA
should include a sensitivity analysis of LCCA results to variation in major LCCA inputs. SHAs
are encouraged to incorporate a quantitative risk analysis approach to treat input uncertainty
(see chapter 4).
7
8
CHAPTER 2. LCCA PROCEDURES
This chapter identifies the procedural steps involved in conducting a life-cycle cost analysis
(LCCA). They include:
While the steps are generally sequential, the sequence can be altered to meet specific LCCA
needs. The following sections discuss each step.
The primary purpose of an LCCA is to quantify the long-term implication of initial pavement
design decisions on the future cost of maintenance and rehabilitation activities necessary to
maintain some preestablished minimum acceptable level of service for some specified time.
A Pavement Design Strategy is the combination of initial pavement design and necessary
supporting maintenance and rehabilitation activities. Analysis Period is the time horizon over
which future cost are evaluated. The first step in conducting an LCCA of alternative pavement
designs is to identify the alternative pavement design strategies for the analysis period under
consideration.
Analysis Period
LCCA analysis period should be sufficiently long to reflect long-term cost differences associated
with reasonable design strategies. The analysis period should generally always be longer than the
pavement design period, except in the case of extremely long-lived pavements. As a rule of
thumb, the analysis period should be long enough to incorporate at least one rehabilitation
activity. The FHWA’s September 1996 Final LCCA Policy statement recommends an analysis
period of at least 35 years for all pavement projects, including new or total reconstruction
projects as well as rehabilitation, restoration, and resurfacing projects.(4)
At times, a shorter analysis periods may be appropriate, particularly when pavement design
alternatives are developed to buy time (say 10 years) until total reconstruction. It may be
appropriate to deviate from the recommended minimum 35-year analysis period when slightly
shorter periods could simplify salvage value computations. For example, if all alternative
strategies would reach terminal serviceability at year 32, then a 32-year analysis would be quite
appropriate.
9
Regardless of the analysis period selected, the analysis period used should be the same for all
alternatives. Figure 2.1 shows a typical analysis period for a pavement design alternative.
Rehabilitation
Pavement
Condition
Terminal Serviceability
Pavement
Analysis Period
Life
Typically, each design alternative will have an expected initial design life, periodic maintenance
treatments, and possibly a series of rehabilitation activities. It is important to identify the scope,
timing, and cost of these activities. Depending on the initial pavement design, SHAs employ a
variety of rehabilitation strategies to keep the highway facilities in functional condition.(5,6) For
example, table 2.1 shows the Pennsylvania Department of Transportation’s (PennDOT’s)
typical supporting maintenance and rehabilitation strategy for new, reconstructed, and unbonded
portland cement concrete pavements included in its LCCA procedures. PennDOT’s LCCA
procedures also contain typical supporting strategies for new and reconstructed asphalt
concrete pavements. Note that user cost requirements are also identified.
Performance life for the initial pavement design and subsequent rehabilitation activities has a
major impact on LCCA results. It directly affects the frequency of agency intervention on the
highway facility, which in turn affects agency cost as well as user costs during periods of
construction and maintenance activities. SHAs can determine specific performance information
for various pavement strategies through analysis of pavement management data and historical
experience. Operational pavement management systems can provide the data and analysis
techniques to evaluate pavement condition and performance and traffic volumes to identify cost-
effective strategies for short- and long-term capital projects and maintenance programs. Some
SHAs develop performance lives based on the collective experience of their senior engineers.
10
Table 2.1. PennDOT’s design strategy for new, reconstructed, and unbonded overlay.
Year Treatment
5 Clean and seal 25% of longitudinal joints.
Clean and seal 5% of transverse joints. 0% for neoprene seals.
Seal coat shoulders if Type 1 paved shoulders.
10 Same as year 5.
15 Clean and seal 25% of longitudinal joints.
Clean and seal 10% of transverse joints, 5% for neoprene seals.
Seal coat shoulders, if Type 1 paved shoulders.
20 Concrete patch 5% of pavement area
Spall repair 1% of transverse joints (5 sf/joint).
Slab stabilization: minimum 25% of transverse joint.
Diamond grind 100% of pavement area.
Clean and seal all longitudinal joints, including shoulders.
Clean and seal all transverse joints, 7% for neoprene seals.
Seal coat shoulders, if Type I paved shoulders.
Maintenance and protection of traffic.
User delay.
25 Clean and seal 25% of longitudinal joints.
Clean and seal 10% of transverse joints, 10% for neoprene seals
Seal coat shoulders, if Type I paved shoulders.
30 Concrete patch 2% of pavement area.
Clean and seal all joints with fiber asphalt membrane.
60-#/sy leveling course.
3.5-in ID-2 or 4-in ID-3/ID-2 overlay.
Saw and seal joints.
Type 7 paved shoulders.
Adjust all guide rail and drainage structures.
Maintenance and protection of traffic.
User delay.
35 Seal coat shoulders.
Note: The CPR strategy slated for year 20 can be moved to year 15 at the District’s discretion. However,
when doing this, the overlay at year 30 must be moved to year 25, and another overlay added at year 33.
Current FHWA efforts to analyze pavement performance data collected as part of the Strategic
Highway Research Program (SHRP) Long-Term Pavement Performance Program (LTPP)
should provide an additional valuable resource to SHAs. To support that effort, the FHWA is
also coordinating the development and wide distribution of the DataPave software program to
make LTPP performance data directly available to the SHAs. Specific pavement performance
information is also available in various pavement performance reports developed by SHAs such
as Minnesota and Illinois, just to mention a few.
Work zone requirements for initial construction, maintenance, and rehabilitation directly affect
highway user costs and should be estimated along with pavement strategy development. The
11
frequency, duration, severity, and year of work zone requirement are critical factors in
developing user costs for the alternatives being considered.
Construction quantities and costs are directly related to the initial design and subsequent
rehabilitation strategy. The first step in estimating agency costs is to determine construction
quantities/unit prices. Unit prices can be determined from SHA historical data on previously bid
jobs of comparable scale. Other data sources include the Bid Analysis Management System
(BAMS), if used by the SHA.
LCCA comparisons are always made between mutually exclusive competing alternatives.
LCCA need only consider differential costs between alternatives. Costs common to all
alternatives cancel out, these cost factors are generally noted and excluded from LCCA
calculations.
Agency costs include all costs incurred directly by the agency over the life of the project. They
typically include initial preliminary engineering, contract administration, construction supervision
and construction costs, as well as future routine and preventive maintenance, resurfacing and
rehabilitation cost, and the associated administrative cost. Routine reactive-type maintenance
cost data are normally not available except on a very general, areawide cost per lane mile.
Fortunately, routine reactive-type maintenance costs generally are not very high, primarily
because of the relatively high performance levels maintained on major highway facilities. Further,
SHAs that do report routine reactive-type maintenance costs note little difference between most
alternative pavement strategies. When discounted to the present, small reactive maintenance
cost differences have negligible effect on NPV and can generally be ignored.
Agency costs also include maintenance of traffic cost and can include operating cost such as
pump station energy costs, tunnel lighting, and ventilation. At times, the salvage value, the
remaining value of the investment at the end of the analysis period, is included as a negative cost.
Salvage Value represents value of an investment alternative at the end of the analysis period.
The two fundamental components associated with salvage value are residual value and
serviceable life.
Residual Value refers to the net value from recycling the pavement. The differential residual
value between pavement design strategies is generally not very large, and, when discounted over
35 years, tends to have little effect on LCCA results.
Serviceable Life represents the more significant salvage value component and is the remaining
life in a pavement alternative at the end of the analysis period. It is primarily used to account for
differences in remaining pavement life between alternative pavement design strategies at the end
of the analysis period. For example, over a 35-year analysis, Alternative A reaches terminal
serviceability at year 35, while Alternative B requires a 10-year design rehabilitation at year 30.
In this case, the serviceable life of Alternative A at year 35 would be 0, as it has reached its
terminal serviceability. Conversely, Alternative B receives a 10-year design rehabilitation at year
12
30 and will have 5 years of serviceable life at year 35, the year the analysis terminates. The
value of the serviceable life of Alternative B at year 35 could be calculated as a percent of
design life remaining at the end of the analysis period (5 of 10 years or 50 percent) multiplied by
the cost of Alternative B’s rehabilitation at year 30.
Sunk Costs represent a special category of costs that are irrelevant to the decision at hand.
Analysts should be careful not to include them in LCCA. An example may serve best in
understanding the concept.
The $10 deposit is a sunk cost and is irrelevant to the decision. The decision comes down to
paying Store A the $90 balance for the camera, or paying Store B $80 for an identical camera.
Not all cases of sunk cost are this clear and, again, analysts need to take care to guard against
including them in LCCA. An example more specific to pavement design might involve the
reluctance of a designer to select an alternative with a much lower life-cycle cost because it
would mean wasting the money previously spent on developing final plans for a clearly inferior
alternative.
In the simplest sense, user costs are costs incurred by the highway user over the life of the
project. In LCCA, highway user costs of concern are the differential costs incurred by the
motoring public between competing alternative highway improvements and associated
maintenance and rehabilitation strategies over the analysis period. In the pavement design arena,
the user costs of interest are further limited to the differences in user costs resulting from
differences in long-term pavement design decisions and the supporting maintenance and
rehabilitation implications. User costs are an aggregation of three separate cost components:
vehicle operating costs (VOC), user delay costs, and crash costs.
In the LCCA of pavement design alternatives, there are user costs associated with both normal
operations and work zone operations. The normal operations category reflects highway user
costs associated with using a facility during periods free of construction, maintenance, and/or
rehabilitation (i.e., work zone) activities that restrict the capacity of the facility. User costs in this
category are a function of the differential pavement performance (roughness) of the alternatives.
The work zone operations category, however, reflects highway user costs associated with
using a facility during periods of construction, maintenance, and/or rehabilitation activities that
generally restrict the capacity of the facility and disrupt normal traffic flow.
13
During normal operating conditions, as a general rule, there should be little difference between
crash costs and delay costs resulting from pavement design decisions. Further, as long as the
pavement performance levels remain relatively high, and performance curves of the alternative
designs are similar, there should be little if any difference between vehicle operating costs.
If, however, pavement performance curves and levels differ substantially, significant vehicle
operating cost differentials can develop. Figure 2.2 depicts an exaggerated example of
alternative pavement design strategies.
Alternative A Alternative B
Pavement Condition
Terminal Serviceability
0 5 10 15 20
Pavement Life (Years)
In figure 2.2, Alternative A represents a traditional longer term strategy with rehabilitation
implemented on a 15-year cycle. Alternative B consists of a minimal treatment on a 5-year
cycle. This figure graphically depicts differences in performance levels for different rehabilitation
strategies. Intuitively, differences in pavement performance can produce differences in vehicle
operating costs. Slight differences in VOC rates caused by differences in pavement performance
characteristics (primarily roughness), when multiplied by several years Vehicle Mile(s) Traveled
(VMT), could result in huge VOC differentials over the life of the design strategy. This is
particularly true for pavement preservation strategies that exhibit poor performance over most
of the analysis period as shown by Alternative B in figure 2.2.
(1) Accurately estimate the pavement performance differences over time (at least yearly).
(2) Quantify the difference in VOC rates for slight differences in pavement performance at
relatively high performance levels.
Most research on VOC rates, as a function of pavement performance, has been conducted
by the World Bank. Figure 2.3 shows the effect of road roughness, as measured by the
international roughness index (IRI), on road user costs in New Zealand.(7) The figure also
shows that additional operating costs (as compared to a smooth road baseline) begin to accrue
14
around an IRI equal to 170 in/mi. According to Darter’s and Al-Omai’s work, an IRI level of
170 is approximately equal to a PRS rating of 2.5. On higher order systems in the United
States, such as the NHS, SHAs typically consider pavements with a PSR of 2.5 to have
(8)
reached their terminal serviceability index and schedule some form of rehabilitation.
The effect of pavement condition on user operating cost at low roughness levels, if any, is not
well documented. There is, however, current research, NCHRP 1-33, Methodology to
Improve Pavement Investment Decision, under way in this area. NCHRP 1-33 has been
120
Base cost for a smooth road
Operating costs (cents/mi)
60
40
20
0
0 100 200 300 400 500 600 700 800 900
initiated to obtain objective information relating costs associating with truck operating expenses
(health claims, cargo damage, vehicle depreciation, maintenance and repair) with road
roughness. This study is scheduled to be complete by 1999.
Additionally, the Cornell University School of Civil Engineering is performing preliminary work
on establishing differential vehicle operating costs associated with pavement condition (i.e., IRI)
for the New York State Department of Transportation.(9)
Even if user operating cost differentials are established between smooth and very smooth roads,
the analyst must still overcome the difficulty in estimating projected year-by-year performance
differences between alternative pavement design strategies.
Considering the prevailing pavement performance ranges encountered in the United States on
higher type facilities, and the lack of precision in projecting year-by-year pavement performance
differentials, this Interim Technical Bulletin does not address computing user vehicle operating
cost differentials during normal operating conditions at this time.
15
On the other hand, during periods of initial construction and future maintenance and
rehabilitation activities (i.e., work zone operation), vehicle operating cost, user delay, and crash
costs can be significantly different between alternative pavement design strategies. As a result,
this Technical Bulletin focuses primarily on work zone user costs.
User cost rates, as used here, refer to the dollar values assigned to each user cost component.
User costs are calculated by multiplying the quantity of the various additional user cost
components (VOC, delay, and crash) incurred by the unit cost for those cost components.
Additional VOCs are determined by multiplying the quantity of additional VOC components
(i.e., additional speed changes, stops, miles, hours of idling) incurred by the dollar value
assigned to each VOC component. By the same token, user delay costs are determined by
multiplying the additional hours of travel time resulting from WZ-caused traffic delay (or
additional miles of travel caused by detours) by the dollar value of an hour of delay for each
vehicle classification. Finally, the additional crash costs are determined by multiplying the number
of additional crashes (by type) by the appropriate dollar value assigned to each crash type.
Chapter 3 presents detailed procedures to calculate work zone user cost quantities of alternate
pavement design strategies, while the unit costs associated with each cost component are
discussed below.
VOC Rates
Table 5 of NCHRP Report 133, Procedures for Estimating Highway User Costs, Air
Pollution, and Noise Effects, can be used to determine VOC rates for stopping/speed
changes and idling, as well as associated delay times for stopping/speed changes. This work is
based on the earlier work by Winfrey, Economic Analysis for Highways.(10,11) A compressed
version of NCHRP 133 table 5 is reproduced as table 2.2.
Table 2.2 shows additional hours of delay and additional VOC associated with stopping 1,000
vehicles from a particular speed and returning them to that speed. Different factors are provided
for passenger cars and single-unit and combination trucks. In addition, the table includes a
vehicle operating cost associated with idling while stopped. The cost factors reflect 1970 prices
based on a $3 per hour value of time for passenger vehicles and $5 per hour for all trucks.
To make these factors applicable to current analysis, the values shown have been escalated to
reflect more current year dollars. The escalation factor for VOC is determined by using the
transportation component of the Consumer Price Index (CPI) for the base year (1970) and the
more current year (August 1996). The transportation component of the CPI was 37.5 in 1970,
16
Table 2.2. Added time and vehicle running cost/1,000 stops and idling costs (1970 $).
and 142.8 in August 1996. The VOC escalation factor used to escalate 1970 prices to August
1996 prices is:
Mileage Rates
In addition to these incremental VOCs associated with changes from normal operating
condition, there is also a fundamental overall baseline VOC mileage rate associated with normal
17
Table 2.3. Added time and vehicle running cost/1,000 stops and idling costs (Aug 96 $).
operating conditions. Typically this is expressed as an overall cents-per-mile rate. These rates
would typically apply to any additional miles that must be driven because of detours.
Some readily apparent values are the marginal cost rates used by the Federal Government.
Federal travel regulations authorize the payment of $0.31 per mile for using privately owned
passenger vehicles for official government travel. The flat mileage rate allowed by the IRS for
business use of a privately owned passenger vehicle is also $0.31 per mile (tax year 1996).
18
Delay Cost Rates (Value of Time)
Of all of the user costs rates, the cost rate assigned to user delay (i.e., the value of time) is by
far the most controversial. As a result, the cost rates for user delay are discussed from several
perspectives. The sources used in this Interim Technical Bulletin include updated values from
earlier NCHRP research, recent guidance provided by the Office of the Secretary of
Transportation (OST), and recently updated values used by the FHWA in its Highway
Economic Requirements System (HERS) Model.
A base case value can be generated from the earlier NCHRP 133 report, which used 1970
dollar values of $3 per hour passenger vehicles and $5 per hour for all trucks. Once again,
these values must be escalated to reflect more current/base year values.
In this case, the escalation factor for the dollar value of time is determined by using changes to
the All Items Component of the CPI for the base year (1970) and the more current year
(August 1996). The All Items Component of the CPI was 38.8 in 1970 and 152.4 in August
1996. The value of time escalation factor to escalate 1970 prices to August 1996 prices is:
Table 2.5. Updated NCHRP 133 values of time ($/Veh-Hr) (Aug 96 $).
19
Table 2.6 shows more recent work (1993) from the computerized MicroBENCOST program
developed under NCHRP Research Project 7-12, Microcomputer Evaluation of Highway
User Benefits, which uses 1990 base year default values. The table includes another escalation
factor to bring these 1990 base year costs to a new August 1996 base year. Once again, the
escalation factor for the value of time is calculated by dividing the August 1996 overall CPI
(152.4) by the overall CPI for 1990 (130.7).
Table 2.6. Updated MicroBENCOST default values of time ($/Veh-Hr) (Aug 96 $).
Table 2.7 brings together the updated values from both the earlier NCHRP 133 and MicroBENCOST
reports and shows the average values per vehicle hour for the three vehicle classifications in August
1996 dollars. Prior to using these values the analysts should escalate their values to reflect current costs
using a similar approach.
Table 2.7. Composite earlier research value of time ($/Veh-Hr) (Aug 96 $).
Pass Trucks
Updated Source
Cars Single-Unit Combination
NCHRP 133 $11.78 $19.64 $19.64
MicroBENCOST 11.37 17.44 24.98
Average Value $11.58 $18.54 $22.31
OST Approach
Recent guidance provided by the U.S. DOT OST to the various DOT modal administrations
is another source that can be used to determine the value of time. OST recommends using a
percentage of the national wage rate for the value of time. OST-recommended procedures
apply different percentages of the national wage rate as a function of vehicle classification and
20
trip type and purpose. Table 2.8 provides U.S. DOT ranges of the percentage of the national
wage rate that should be applied to various combinations of trip type and purpose. The rates
shown are per person hour. As can be seen in table 2.8, business and truck travel are valued
more highly than personal travel, and intercity personal travel is valued more highly than local
personal travel.
Table 2.8. Travel time ranges as a percent of national wage rate (1995 $/Person-Hr).
Table 2.9 provides information on national hourly wage rates used by OST in 1995 dollars.
Based on the information on appropriate percentages provided in table 2.8 and the
recommended national hourly wage rate provided in table 2.9, the OST developed the
recommended ranges for the value of travel time shown in table 2.10. The values associated
with Mixed are the ranges to be used when the distribution between auto business and personal
trips is not known. Hourly values shown for trucks are $16.50 for both local and intercity trip
types.
Table 2.10 lists the ranges of values in dollars per person hour. As most user delay analyses
compute vehicle hours of delay, these values must be adjusted for vehicle occupancy rates.
Table 2.10. Ranges for hourly values of travel time (1995 $/Person-Hr)
21
According to the 1990 Personal Transportation Survey, typical auto vehicle occupancy rates
for personal travel are 1.7 in urban areas and 2.0 in rural areas.(12) The typical vehicle
occupancy rates for trucks and auto business use is much closer to 1.0.
The FHWA uses the HERS Model in conducting national-level analysis of highway
performance, needs, and economic evaluation of proposed highway improvements. Part of the
economic analysis includes determining the value of travel time delay. Table 2.11 shows the
default dollar values of travel time used in the HERS model.
Table 2.11. Value of one vehicle hour of travel time (1995 $).
Autos Trucks
Travel
Category Single-Unit Combinations
Small Medium
4 Tire 6 Tire 3-4 Axle 4 Axle 5 Axle
Business $27.99 $28.29 $20.20 $24.96 $27.02 $31.02 $31.58
Personal 12.78 $12.78 $12.78 NA NA NA NA
% AADT
90% 90% 69% 0% 0% 0% 0%
Personal Use
Weighted
$14.30 $14.33 $15.08 $24.96 $27.02 $31.02 $31.58
Average
FHWA Highway Economic Requirements System Technical Report v3-1 9/97 (Exhibit 8-7).
The values of travel time shown on lines 1 and 2 of table 2.11 are per vehicle hour based on
typical vehicle occupancy rates. The values associated with % AADT Personal Use on line 3
represent the percent of travel that is personal. The Weighted Average values shown on line 4
are for mixed flow of business and personal travel and are used when traffic flow distribution by
travel category is not known. The values shown for trucks are significantly higher than either
early research study or the OST guidance. Part of the higher cost (approximately 30 percent) is
attributable to value associated with the vehicle cargo and the vehicle itself.
Toll Roads
Finally, another source that can be used to estimate the dollar values of time includes tolls on toll
roads in relation to time saved. One of the more interesting revelations of users’ willingness to
pay comes from the treatment of some High Occupancy Vehicle (HOV) facilities in California
and Texas. Both States are experimenting with High Occupancy Toll (HOT) facilities, which
allow Single Occupancy Vehicles (SOV) to use an HOV facility for a fee. These studies deal
primarily with local auto trips and to some extent may reflect market prices. While the values
inferred from such studies tend to indicate lower values of time (approximately $6 per person
hour), they may in reality reflect inefficient pricing policy rather than lower values.
22
Recommended Values of Travel Time (Dollars per Vehicle Hour)
Table 2.12 below is a composite table that brings together the several sources of the value
of time previously discussed.
Based on consideration of these potential sources, table 2.13 reflects the ranges of the value
of travel time per vehicle recommended for use in typical analyses where distribution data
on trip purpose and type are not known.
Trucks
Passenger Cars
Single-Unit Combinations
$10 to 13 $17 to 20 $21 to 24
The MicroBENCOST software package, developed for the NCHRP Research Project 7-
12, includes default crash cost rates. Table 2.14 shows the default crash cost rates by crash
type for both rural and urban settings in 1990 dollars.
Table 2.14. MicroBENCOST default crash cost rates ($1,000, 1990 $).
23
The MicroBENCOST 1990 dollar default values shown in table 2.14 are escalated to August
1996 dollars in table 2.15 using the 1.116 escalation factor developed earlier.
Table 2.15. MicroBENCOST default crash cost rates ($1,000, Aug 96 $).
The MicroBENCOST default cost values for a fatality range from $1,091,000 to $1,182,000,
which is quite a bit lower than the $2,700,000 cost per fatality averted, which was
recommended by the U.S. DOT in its March, 14, 1995 memorandum from the Assistant
Secretary for Transportation Policy to DOT Modal Administrators. This $2,700,000 value is an
update of the value originally recommended in basic guidance distributed on January 8, 1993.
In addition to the traditional direct work zone user costs, there are indirect user costs such as
the impact of user delay on delivery fleet size, the costs associated with a rolling inventory, and
other, indirect effects of delay to manufacturing plants that now dependent on just-in-time
delivery. While such factors will become more important over time, they are beyond the scope
of these guidelines. It is interesting to note that cost related to delivery fleet size and the costs
associated with a rolling inventory are included in the dollar value of delay time rates used in the
FHWA HERS Model.
Expenditure stream diagrams are graphical representations of expenditures over time. They are
generally developed for each pavement design strategy to help visualize the extent and timing of
expenditures. Figure 2.4. shows a typical expenditure stream diagram.
Initial
Construction
Rehabilitations
Cost ($)
Analysis Period
Time Salvage Value
Figure 2.4. Typical expenditure stream diagram for a pavement design alternative.
24
Normally, costs are depicted as upward arrows at the appropriate time they occur during the
analysis period, and benefits are represented as negative cost or downward arrows.
In LCC analysis of pavement design alternatives, the basic benefits of providing and maintaining
some preestablished pavement condition level on any given roadway are outside the scope of
the analysis. The benefits of providing the specified level of pavement condition are considered
to be the same for all pavement design strategies. As a result, the only concerns are the
differential costs among alternatives. The only negative cost (i.e., the only downward arrow)
would be the cost associated with any salvage value.
Under these conditions, the LCCA objective becomes finding the alternative pavement design
strategy that meets the performance requirements at the lowest life-cycle cost.
In its broadest sense, LCCA is a form of economic analysis used to evaluate the long-term
economic efficiency between alternative investment options. Economic analysis focuses on the
relationship between costs, timings of costs, and discount rates employed. Once all costs and
their timing have been developed, future costs must be discounted to the base year and added
to the initial cost to determine the NPV for the LCCA alternative. As noted earlier, NPV is the
economic indicator of choice, and the basic NPV formula for discounting discrete future
amounts at various points in time back to some base year is:
Nn 1
NPV = Initial Cost + ∑ Rehab Cost k nk
k =1 (1 + i )
where: i = discount rate and
n = year of expenditure
1
∑ (1 + i) nk component of the above formula is referred to as the Present Value (PV)
The
factor for a single future amount. PV factors for various combinations of discount rates and
future years are available in Discount factor tables (more commonly referred to as interest rate
tables). PV for a particular future amount is determined by multiplying the future amount by the
appropriate PV factor. Table 2.16 shows discount factors for a single future payment at 3, 4,
and 5 percent discount rates for up to 40 years in the future.
25
Table 2.16. Present value discount factors: single future payment.
Discount Factor Discount Factor
Year 3% 4% 5% Year 3% 4% 5%
1 0.9709 0.9615 0.9524 21 0.5375 0.4388 0.3589
2 0.9426 0.9246 0.9070 22 0.5219 0.4220 0.3418
3 0.9151 0.8890 0.8638 23 0.5067 0.4057 0.3256
4 0.8885 0.8548 0.8227 24 0.4919 0.3901 0.3101
5 0.8626 0.8219 0.7835 25 0.4776 0.3751 0.2953
6 0.8375 0.7903 0.7462 26 0.4637 0.3607 0.2812
7 0.8131 0.7599 0.7107 27 0.4502 0.3468 0.2678
8 0.7894 0.7307 0.6768 28 0.4371 0.3335 0.2551
9 0.7664 0.7026 0.6446 29 0.4243 0.3207 0.2429
10 0.7441 0.6756 0.6139 30 0.4120 0.3083 0.2314
11 0.7224 0.6496 0.5847 31 0.4000 0.2965 0.2204
12 0.7014 0.6246 0.5568 32 0.3883 0.2851 0.2099
13 0.6810 0.6006 0.5303 33 0.3770 0.2741 0.1999
14 0.6611 0.5775 0.5051 34 0.3660 0.2636 0.1904
15 0.6419 0.5553 0.4810 35 0.3554 0.2534 0.1813
16 0.6232 0.5339 0.4581 36 0.3450 0.2437 0.1727
17 0.6050 0.5134 0.4363 37 0.3350 0.2343 0.1644
18 0.5874 0.4936 0.4155 38 0.3252 0.2253 0.1566
19 0.5703 0.4746 0.3957 39 0.3158 0.2166 0.1491
20 0.5537 0.4564 0.3769 40 0.3066 0.2083 0.1420
NPV Computations
Example NPV computations are provided for the following hypothetical problem. The example
is based on a 35-year analysis period.
The initial pavement design will cost $1.1 million and have an associated work zone user cost of
$300,000 at year 0. Additional rehabilitation cost of $325,000 will be incurred in years 15 and
30. Associated work zone user costs in years 15 and 30 will be $269,000 and $361,000,
respectively. The salvage value at year 35, based on a prorated cost of the year-30
rehabilitation design and remaining life, will be $216,000 (10/15 of $325,000). Figure 2.5
shows the expenditure stream diagram for the example problem.
$300
Costs ($ 1,000)
User Cost
$269 $361
User Cost User Cost
$1,100
Initial Cost
$325
$325
Rehab. #1
Rehab. #2
0 15 30 35
Figure 2.5. Expenditure stream diagram for agency and user costs.
26
Note that estimated user costs drop in year 15 and go back up in year 30. This is consistent
with a longer duration initial work zone followed by short duration rehabilitation work zones
impacted by continually increasing traffic volumes over time. Table 2.17 shows the results of
PV computations using 4 percent PV factors for single future amounts for the example
expenditure stream diagram. The bottom line of table 2.17 shows the NPV of the aggregated
individual PVs.
ANALYZE RESULTS
Many times a sensitivity analysis will focus on best case/worst case scenarios in an attempt to
bracket outcomes. Most LCC sensitivity analysis, as a minimum, evaluate the influence of the
discount rate used on LCCA results.
27
Sensitivity analyses may be carried out using common spreadsheet-based applications such as
Microsoft Excel, Lotus, or Quattro Pro. Tables 2.18 and 2.19 present the results of a spreadsheet
analysis of the sensitivity of NPV of two example pavement design strategies to discount rate
ranges from 2 to 6 percent for a 35-year analysis period. Total NPV at discount rates ranging
from 2 to 6 percent are shown at the bottom of columns (e) through (i).
NPV
Activity Year Cost 2.0% 3.0% 4.0% 5.0% 6.0%
(a) (b) (c) (e) (f) (g) (h) (i)
Construction 0 975 975 975 975 975 975
User Cost 0 200 200 200 200 200 200
Rehab #1 10 200 164 149 135 123 112
User Cost # 1 10 269 220 200 182 165 150
Rehab # 2 20 200 135 111 91 75 62
User Cost # 2 20 361 243 200 165 136 113
Rehab # 3 30 200 110 82 62 46 35
User Cost # 3 30 485 268 200 150 1,122 85
Salvage 35 -100 -50 -36 -25 -18 -13
Total NPV 2,266 2,081 1,934 1,815 1,718
NPV
Activity Year Cost 2.0% 3.0% 4.0% 5.0% 6.0%
(a) (b) (c) (e) (f) (g) (h) (i)
Construction 0 1,100 1,100 1,100 1,100 1,100 1,100
User Cost 0 300 300 300 300 300 300
Rehab #1 15 325 241 209 180 156 136
User Cost # 1 15 269 200 173 149 129 112
Rehab # 2 30 325 179 134 100 75 57
User Cost # 2 30 361 199 149 111 84 63
Salvage 35 -217 -108 -77 -55 -39 -28
Total NPV 2,112 1,987 1,886 1,805 1,739
Alternative #1 has a lower initial agency cost, and, because of a shorter construction period, a
lower user cost than Alternative #2. However, Alternative #1 requires three identical 10-year
design rehabilitations compared to two identical 15-year design rehabilitations for Alternative #2.
Out-year user costs for Alternative #1 increase as a result of increased traffic levels over time;
out-year user costs for Alternative #2 first decrease due to a shorter work zone period and then
increase as a result of increased traffic levels over time.
28
Both alternatives have a remaining service life at year 35; Alternative #1 has 5 years and
Alternative #2 has 10 years. The salvage value, as a prorated share of the last rehabilitation, is
50 percent (5 years remaining on a 10-year design) of its last rehabilitation cost for Alternative
#1. For Alternative #1, this translates into 50 percent of the $200,000 year-30 rehabilitation
cost. The salvage value of Alternative #2, on the other hand, is 66.6 percent (10 years
remaining on a 15-year design) of its last rehabilitation cost. This translates into 66.6 percent of
the $325,000 year-30 rehabilitation cost.
Table 2.20 shows a direct comparison of the NPV of both alternatives at several discount rates.
Inspection of table 2.20 reveals that the NPV of both alternatives decreases as the discount
rate increases. This results from the reduced present value of future costs at higher discount
rates. Because the amount and timing of future costs differ between alternatives, the effect of
discount rate on NPV is different for each alternative. In this example, Alternative #1 is more
expensive than Alternative #2 at discount rates of 5 percent and lower, while Alternative #2 is
more expensive than Alternative #1 at discount rates of 6 percent or more. Figure 2.6 shows
these results graphically.
2,100 Alt-1
2,000
1,900 Alt-2
1,800
1,700
1,600
2.0% 3.0% 4.0% 5.0% 6.0%
Discount Rate
29
Table 2.21 separates agency and user cost differences for the same range of discount rates.
Inspection of the table 2.21 reveals that Alternative #2 has a higher agency cost than Alternative
#1 at all discount rates considered. Further, Alternative #2 has lower user cost than Alternative
#1 at all discount rates considered.
Discount Rate, %
Cost Component
2.0 3.0 4.0 5.0 6.0
Alternative #1 Agency Cost 1,334 1,281 1,238 1,201 1,171
Alternative #2 Agency Cost 1,413 1,366 1,326 1,292 1,264
Agency Cost Advantage Alt #2 vs #1 -79 -85 -88 -91 -93
The decision to include or exclude user costs can significantly affect the LCCA results. In an
effort to put the agency and user costs in perspective, the bottom row of table 2.21 includes an
incremental B/C comparison of the reduction in user costs as a function of increased agency
costs. The incremental B/C data in table 2.21 is computed by dividing the reduction in user
costs (i.e., benefits) associated with selecting Alternative #2 in lieu of Alternative #1 by the
added agency cost(s) associated with selection of Alternative #2.
Similar sensitivity analyses could be conducted using other input variables such as agency cost,
user costs, pavement performance lives, and hourly dollar value of user delay.
In addition to conducting a sensitivity analysis, the analyst should examine the implications of
contractor work hours on queuing costs as well as the anticipated maximum queue lengths and
delay times. Chapter 3 details queue lengths and discusses associated user costs.
A primary drawback of the sensitivity analysis is that the analysis gives equal weight to any input
value assumptions, regardless of the likelihood of occurring. In other words, the extreme values
(best case and worst case) are given the same likelihood of occurrence as the expected value—
which is not realistic. Chapter 4 presents a powerful analytical technique to overcome this
limitation.
30
REEVALUATE DESIGN STRATEGY
Once the net present values have been computed for each alternative and limited sensitivity
analysis performed, the analyst needs to step back and reevaluate the competing design
strategies. As noted in chapter 1, the overall benefit of conducting a life-cycle cost analysis is
not necessarily the LCCA results themselves, but rather how the designer can use the
information resulting from the analysis to modify the proposed alternatives and develop more
cost-effective strategies.
For example, if user costs dwarf agency costs for all the alternatives, the analysis may indicate
that none of the alternatives analyzed are viable. It could indicate that the designer needs to
evaluate the current design strategies’ impacts on future maintenance of traffic . . . that the
design strategies should reflect the need for additional capacity in the out-years to mitigate the
impact on highway users. The solution to out-year capacity problems could include enhanced
structural design of the shoulders in early-year pavement designs to allow for the use of the
shoulder in subsequent rehabilitation traffic control plans. It also could include enhanced
structural design of the mainline pavement to minimize the frequency of subsequent rehabilitation
efforts and designing in features that will make future rehabilitation proceed more smoothly.
Other options available include revising the maintenance of traffic plans, reducing the
construction period, restricting the contractor’s work hours or imposing lane rental fees,
planning for additional lanes/routes, and even examining programs to temporarily shift traffic to
alternative modes of travel. It is important to note that restricting the contractor’s hours of
operation or the number of work days allowed will more than likely increase agency cost.
LCCA results are just one of many factors that influence the ultimate selection of a pavement
design strategy. The final decision may include a number of additional factors outside the LCCA
process, such as local politics, availability of funding, industry capability to perform the required
construction, and agency experience with a particular pavement type, as well as the accuracy of
the pavement design and rehabilitation models. Chapter 3 of the 1993 AASHTO Pavement
Design Guide further discusses such other factors.(13) When such other factors weigh heavily
in the final pavement design selection, it is imperative to document their influence on the final
decision.
Many assumptions, estimates, and projections feed the LCCA process. The variability
associated with these inputs can have a major influence on the confidence the analyst can place
in LCCA results. It all depends on the accuracy of the inputs used. The accuracy of LCCA
results depends directly on the analyst’s ability to accurately forecast such variables as future
costs, pavement performance, and traffic for more than 30 years into the future. To effectively
deal with the uncertainty associated with such forecasts, a probabilistic risk analysis approach
(as presented in chapter 4) is increasingly essential to quantitatively capture the uncertainty
associated with input parameters in LCCA results.
31
32
CHAPTER 3. WORK ZONE USER COSTS
This chapter presents a rational, step-by-step procedure to allow the analyst, based on capacity
flow analysis, to determine user costs associated with establishing a work zone. Chapter 2
discussed user cost rates. This chapter focuses on calculating the quantity associated with each
of the individual user cost components (VOC, user delay, and crashes) and computing overall
user costs in a spreadsheet environment. The chapter ends with a discussion of automated
computer programs.
While there are microcomputer-based user cost analysis software programs currently available,
it is important that the analyst thoroughly understand the principles involved before attempting to
apply them. By understanding the major factors influencing work zone user costs, the analyst
can take steps to minimize the effect of planned future rehabilitation activities on highway users.
Work zone user costs are the increased VOC, delay, and crash costs to highway users resulting
from construction, maintenance, or rehabilitation work zones. They are a function of the timing,
duration, frequency, scope, and characteristics of the work zone; the volume and operating
characteristics of the traffic affected; and the dollar cost rates assigned to vehicle operating,
delay, and crashes. The following sections address each of these issues.
In the end, user costs are computed by multiplying the quantity of additional VOC, delay, and
number of crashes by the unit cost rates assigned to these components.
Work Zone is defined in the Highway Capacity Manual (HCM) as an area of a highway
where maintenance and construction operations impinge on the number of lanes available to
traffic or affect the operational characteristics of traffic flowing through the area.(14)
Each work zone established over the analysis period affects traffic flow and associated user
costs differently and must be evaluated as a separate event. A separate work zone must be
defined and analyzed whenever characteristics of the work zone or the characteristics of the
affected traffic change.
Pavement design performance differences directly affect the frequency and timing of
maintenance and rehabilitation activities. Pavement rehabilitation and maintenance activities
generally occur at different points in the analysis period with different traffic, and they generally
vary in scope and duration. The time that they occur also affects the influence of the discount
factor used in developing NPV.
33
WORK ZONE CHARACTERISTICS
In order to analyze work zone user costs, work zone characteristics associated with alternative
designs and supporting maintenance and rehabilitation strategies must be defined as part of the
development of alternative pavement design strategies. Alternative pavement preservation
strategies must include how often (the number of times) the facility will be under construction,
maintenance or rehabilitation activities, and the year at which work zones are anticipated.
Strategies should also include estimates of the number of days the work zone will last, the hours
of the day the work zone will be in place, and the anticipated maintenance of traffic strategy.
Work zone characteristics of concern include such factors as work zone length, number and
capacity of lanes open, duration of lane closures, timing (hours of the day, days of the week,
season of the year, etc.) of lane closures, posted speed, and the availability and physical and
traffic characteristics of alternative routes. The strategy for maintaining traffic should include any
anticipated restrictions on contractor’s or maintenance force’s hours of operations or ability to
establish lane closures. Specific details in an LCCA should include:
l Specific hours of each day, as well as the days of the week the work zone will be in
place.
The duration of a work zone (the overall length of time a facility or portion of a facility is out of
service or traffic is restricted) can range from sporadic daily lane closures for maintenance to
several months for bridge deck replacements.
As noted earlier, the differential routine maintenance cost between alternative pavement design
strategies tends to be insignificant when compared to initial construction and rehabilitation costs.
To a great extent, the same is true of user costs resulting from routine reactive-type maintenance
activities. Routine maintenance work zones tend to be relatively infrequent, of short duration,
and outside of peak traffic flow periods. As such, analysts should focus attention on user costs
associated with major work zones.
TRAFFIC CHARACTERISTICS
User costs are directly dependent on the volume and operating characteristics of the traffic on
the facility. Each construction, maintenance, and rehabilitation activity generally involves some
temporary effect on traffic using the facility. The effect can vary from insignificant for minor work
zone restrictions on low-volume facilities to highly significant for major lane closures on high-
volume facilities.
34
The major traffic characteristics of interest for each year a work zone will be established include:
(1) the overall projected Average Annual Daily Traffic (AADT) volumes on both the facility and
possibly alternate routes, (2) the associated 24-hour directional hourly demand distributions,
and (3) the vehicle classification distribution of the projected traffic streams. On high-volume
routes, distinctions between weekday and weekend traffic demand and hourly distributions
become important. Further, seasonal AADT traffic distribution also becomes important when
work zones are proposed on recreational routes during seasonal peak periods.
AADT
Current AADT volumes are normally readily available for the base year and projected
compound traffic growth rates can be obtained from the SHA traffic monitoring section. From
these two pieces of information, calculating future year work zone AADT is relatively simple.
While the calculations for future year AADT are rather straightforward, there are two major
issues associated with the reasonableness of such projections on high-volume urban facilities.
The first is whether the roadway in question can handle large projected volume increases; the
second is whether traffic using a facility will continue to use it when work zones are established
and traffic flow is restricted.
In the first issue, design capacity and ultimate capacity are quite different. Design capacity is
generally set to handle the design year 30 highest hour volume at level of service D or E.
However, table 2.1 of the HCM shows measured maximum 24-hour traffic volumes for some of
the more heavily traveled interstate urban freeways that far exceed the traffic volumes normally
associated with level of service D or E. Reasonableness dictates that AADT projections should
not exceed the maximum observed 24-hour traffic volumes contained in table 2.1 of the HCM.
The next section discusses the second issue, which relates to whether traffic using a facility will
continue to use it when work zones are established and traffic flow is restricted.
Traffic Diversion
Traffic demand is generally determined based on facility operating characteristics during periods
of normal operations. Traffic demand during work zone operations may or may not be the same.
Some portion of the traffic normally wanting to use the facility may divert to other routes when
work zones are established.
Vehicles use a given facility because it offers what the vehicle operators perceive as the least
expensive combination of vehicle operating and time delay costs, consistent with safety
requirements. When faced with restricted flow, or even the anticipation of restricted flow,
vehicle operators who normally use a facility will exercise one of several options, which are
somewhat whimsically categorized below.
Hang Toughers will continue to use the facility as they always have. They are primarily users
with little or no option. They (1) must make the trip; (2) must make it at a specific time; and (3)
35
either don’t know or don’t have alternative routes or modes from which to choose. These users
pay the full price of the work zone and have little effect on other facilities in the corridor. In rural
areas the predominate choice of through traffic will be to tough it out, as these users generally
must make the trip and do not have much information on alternative routes unless formal detours
are established.
Time Shifters have the ability and choose to travel on the facility at a different time — generally
a time well outside of the restricted flow period. These users lessen their impact by sharing the
impact with other vehicles by invading their time slot. These users also have little effect on other
highway facilities in the corridor, but they do affect hourly traffic distribution.
Detourees either seek out and use alternate routes or are forced to negotiate detours
established by the highway agency. These operators also lessen their impact by sharing the
impact with other vehicles by invading their routes. They tend to trade off anticipated time
delay for additional travel distances and associated vehicle operating costs. In urban areas this
could include users who switch modes. Detourees can have significant impact on overall user
costs of alternative routes, especially those operating at or near capacity. The sections on
Circuitry (page 69) and Crash Costs (page 70) address the additional user costs associated
with the additional mileage traveled by detoured and diverted traffic.
Trip Swappers have the luxury of totally abandoning the trip or seeking other destinations when
the cost, in terms of time and money, becomes too great. Historically, this group consists
primarily of shopping and social/recreational trip makers. While their behavior may diminish the
user cost impact of the work zone, they adversely affect businesses along the route in question.
More recent trends in people working out of the home and telecommuting may significantly
affect work trips in the future.
Handling Diversion
In simple cases, where either work zone disruption is tolerable or alternative routes are limited,
estimated AADT during the year of the work zone can be anticipated to continue on the facility
and the work zone analysis can be limited to the existing facility.
In more complex situations where existing traffic would face intolerable work zone disruptions, it
is entirely possible that total travel demand and hourly distribution on the facility may change
when the work zone is established. When demand changes, the scope of work zone user cost
analysis may have to expand beyond the existing facility and include user cost changes on major
alternative routes. When preliminary analysis of travel demand shows that work zone user
delays are unreasonably high, the analyst should seek special help from the traffic engineering
section to generate revised traffic demand forecasts during work zone operations.
An alternative approach is to assume that those extraordinary user costs are truly there, and
time and route shifters are merely diminishing their effect by sharing the costs with users of other
routes and time slots. In short, the misery’s still there — it’s just being shared differently.
36
Vehicle Classification
Highway user costs are a composite of the costs of all affected highway users. Highway users
are not a homogeneous group. They include commercial and noncommercial vehicles ranging
from passenger vehicles through the heaviest trucks. These different vehicle types have different
operating characteristics and associated operating costs. Further, the value of user delay differs
between vehicle classes. As a result, user costs need to be analyzed for each major vehicle class
present in the traffic stream.
There are many truck-vehicle classifications representing various size and weight configurations.
Appendix A of the FHWA Traffic Monitoring Guide, includes 13 different vehicle
(15)
classifications. User cost analysis based on 13 vehicle classes is much too detailed for the
level of sophistication in the analysis procedures presented.
On high traffic volume facilities, user delay costs are likely to represent a very significant
component of overall user costs in a pavement design LCCA, particularly when vehicle demand
exceeds capacity. As passenger vehicles represent the bulk of vehicles in the traffic stream,
analysts may find it beneficial to subdivide passenger vehicles into commercial and
noncommercial categories. Further, although vehicle occupancy rates have consistently fallen
over time, they cannot generally be ignored.
The estimated hourly traffic distribution during work zone operations is essential to compare the
unrestricted demand on the facility with the facility’s ability to carry that traffic through the work
zone. Hourly distribution specific to a particular facility can be developed from agency traffic
data or general distribution data developed for various functional class- and area-type
combinations. On high-volume or tourist destination routes, the difference between weekday
and weekend traffic volumes and hourly distributions needs to be considered.
Table 3.1 represents default directional hourly distributions for Urban and Rural roadways
included in MicroBENCOST. The table also includes hourly directional distribution factors
associated with inbound and outbound directions. The hourly demand is computed by
multiplying the AADT by the hourly percent and directional factors for the direction in question.
37
Table 3. 1. Default hourly distributions from MicroBENCOST (all functional classes).
For example, the inbound directional hourly demand volume between 8 and 9 a.m. on an urban
roadway with a 40,000 AADT would be:
Although MicroBENCOST provides for the use of different directional hourly distribution
values for Interstate, Principal Arterial, Minor Arterial, and Major Collector, the default
38
values listed are the same for all functional classes. Pennsylvania, on the other hand, provides
different hourly distributions for various functional class roadways, but it does not provide a
directional split factor. Table 3.2 contains statewide hourly distribution factors used by
PennDOT in developing user delay cost.
CONCEPTUAL ANALYSIS
Before addressing user cost calculation procedures, it is helpful to conduct a conceptual analysis
of a work zone operation. There are seven possible work zone user cost components that can
occur; three are associated with a Base Case situation where traffic operates under Free-Flow
conditions, and four are associated with a Queue situation where traffic operates under Forced-
Flow conditions. The next section conceptually discusses potential user costs involved under
Free-Flow and Forced-Flow (Level of Service F) conditions.
Free Flow
Work zones restrict traffic flow either by reducing capacity or, as a minimum, by posting lower
speed limits. Figure 3.1 shows free-flow conditions at a work zone. All traffic that flows through
39
the work zone must, at a minimum, slow while traveling through it and then accelerate back to
normal operating speed. This is commonly referred to as a speed change and it results in three
work zone-related User Cost components: speed change delay, speed change VOC, and
reduced speed delay. Figure 3.1 presents the cost components associated with free flow.
Free Flow
Construction Zone
Shoulder
Traverse Work
Speed Change Delay is the additional time necessary to decelerate from the upstream
approach speed to the work zone speed and then to accelerate back to the initial approach
speed after traversing the work zone.
Speed Change VOC is the additional vehicle operating cost associated with decelerating from
the upstream approach speed to the work zone speed and then accelerating back to the
approach speed after leaving the work zone.
Reduced Speed Delay is the additional time necessary to traverse the work zone at the lower
posted speed; it depends on the upstream and work zone speed differential and length of the
work zone.
If traffic demand remains below work zone capacity, added work zone user costs are limited to
the above three components and the analysis is relatively simple. In most cases, delay times
remain relatively low and represent more of a minor irritation and inconvenience than a serious
problem.
40
Forced Flow (Level of Service F)
When hourly traffic demand exceeds work zone capacity, traffic flow breaks down and a queue
of vehicles develops, as figure 3.2 shows. It is important to note that the queue forms not in the
work zone itself, but in the upstream approach to the work zone.
Forced Flow
Upstream Queue Area Work Zone
Shoulder
Work Zone
Speed Change Stopping Queue Reduced Speed Delay
VOC & Delay VOC & Delay Idling & Delay (Traverse Work Zone)
Once a queue develops, all approaching vehicles must not only slow down before proceeding
through the work zone itself, but they also must stop at the upstream end of the queue and
creep through the length of the physical queue under forced-flow conditions. As long as
demand exceeds capacity, the length of the queue grows, exacerbating the problem. When
demand eventually falls below capacity, or when capacity is increased above demand by
removing the work zone restriction, vehicles then leave the queue faster than they arrive and the
length of the queue shrinks and eventually dissipates. When capacity is reduced on high-traffic
facilities, it is common for queues to develop in the morning peak traffic period, dissipate, and
then redevelop in the afternoon peak traffic period. In exceptionally congested areas, queues
may form early in the morning and continue throughout the day and into the evening hours.
Queuing situations impose four work zone-related user costs that only apply to vehicles that
encounter a physical queue.
Stopping Delay is the additional time necessary to come to a complete stop from the upstream
approach speed (instead of just slowing to the work zone speed) and the additional time to
accelerate back to the approach speed after traversing the work zone.
Stopping VOC is the additional vehicle operating cost associated with stopping from the
upstream approach speed and accelerating back up to the approach speed after traversing
work zone.
41
Queue Delay is the additional time necessary to creep through the queue under forced-flow
conditions.
Idling VOC is the additional vehicle operating cost associated with stop-and-go driving in the
queue. The idling cost rate multiplied by the additional time spent in the queue is an
approximation of actual VOC associated with stop-and-go conditions. When a queue exists,
stopping delay and VOC replace the free-flow speed change delay and VOC.
The conceptual analysis presented here is geared primarily to freeway conditions. Conceptual
analysis of other facilities with at-grade intersections would also incur speed change, stopping,
delay, and idling cost but at a much higher frequency, because of intersection-control devices
and turning movements.
COMPUTATIONAL ANALYSIS
Once the individual work zones have been identified, each is evaluated separately. This is the
point at which individual user cost components are quantified and converted to dollar cost
values. This section provides an approach for actually quantifying and costing the individual
work zone user cost components encountered. The 12 overall steps involved are:
The following information describes the work zone used as the example problem throughout the
remainder of this section to illustrate the work zone user cost computational steps described
above.
Facility: 3-lane directional outbound Interstate. Posted speed is 55 mi/h and grades on
the facility are less than 2 percent.
Traffic: Base year (1995) AADT of 122,000 vehicles per day. Vehicle classification
counts indicate a traffic stream mix of 90 percent Passenger
42
Vehicles, 5.4 percent Single-Unit Trucks, and 4.6 percent Combination Trucks.
The 10-year projected traffic growth rates for the various vehicle classifications
are 2.2 percent for passenger vehicles and 5 percent per year for both single-
unit and combination trucks.
Work Zone: 5.25-mile-long single-lane closure from 9 a.m. until 3 p.m. and from 8 p.m. until
5 a.m. the following morning. Work zone posted speed is 40 mi/h. The work
zone will be in place for 60 days in 1999.
The first step is to project future year hourly traffic demand volumes for each vehicle class for
the year the work zones will be in place, from current or base year AADT, using compound
traffic growth factors. The following formula applies:
Future Year AADT = Base Year AADT x Vehicle class % x (1 + growth rate)(Future Yr. – Base Yr.)
The AADT on a facility in 1999 can be determined from a 1995 base year AADT of 122,000
using the above formula by applying the appropriate compound growth rate factor (in this
example 2.2 percent for passenger vehicles and 5 percent for trucks) as follows:
Based on these new numbers, total traffic in 1999 will be 134,615, and because of the
differential traffic growth rates for trucks, the 1999 vehicle mix will be approximately 89 percent
for passenger vehicles, 6 percent for single-unit trucks, and 5 percent for combination trucks.
Directional hourly traffic distribution should be determined from agency traffic data on the
roadway being analyzed or from traffic data on similar facilities. If, however, such data is not
available, the default hourly distributions for various roadway types in urban and rural settings
from MicroBENCOST (reproduced in table 3.1) can be used. Table 3.3 contains the future
year directional hourly demand for the example problem generated using the default
MicroBENCOST urban outbound hourly distribution factors and 134,615 future year AADT.
43
Inspection of table 3.3 reveals that a.m. outbound demand peaks at 4,200 vehicles per hour in
the 7 to 8 a.m. period, while the p.m. outbound demand peaks at 6,885 vehicles per hour in the
5 to 6 p.m. time period.
Table 3.3. Work zone directional hourly demand (all vehicle classes).
In analyzing work zone user costs, there are three capacities that need to be determined: (1) the free-
flow capacity of the facility under normal operating condition, (2) the capacity of the facility when the
work zone is in place, and (3) the capacity of the facility to dissipate traffic from a standing queue.
44
Free-Flow Capacity is the maximum capacity a facility can handle under free-flow conditions.
According to the 1994 HCM (p. 2-10), the maximum capacity for a 2-lane directional freeway
under ideal conditions is 2,200 passenger cars per hour per lane (pcphpl) and 2,300 pcphpl for
a 3- or more lane directional freeway. The 1994 HCM points out the need to reduce the above
ideal condition capacities for such real world factors as restricted lane widths, reduced lateral
clearances, the presence of trucks and recreational vehicles, and the presence of a driver
population unfamiliar with the area (p. 3-11). The real-world free-flow capacity of the facility is
determined by applying the following formula:
SFi = service flow rate for LOS i under prevailing roadway and traffic conditions for N
lanes in one direction in vehicles per hour (vph),
MSFi = Maximum service flow rate for LOS i for N lanes in one direction (vph)
N = number of lanes in one direction of the freeway,
fw = factor to adjust for the effects of restricted lane widths and lateral clearances,
fHV = factor to adjust for the effect of heavy vehicles on the traffic stream, and
fp = factor to adjust for the effect of recreational or unfamiliar driver populations.
The following discussion describes the procedures to adjust maximum freeway capacity in terms
of pcplph to real-world mixed vehicle conditions when there are trucks in the traffic stream (the
most commonly encountered condition). For adjustments resulting from restricted lane width,
reduced lateral clearance, and other adjustment factors when the work zone is not in place,
refer to chapter 3 (p. 3-12) of the 1994 HCM.
The capacity adjustment resulting from the presence of trucks is based on the fact that trucks
are larger than passenger vehicles and thus physically occupy more roadway space. Further,
trucks, particularly when fully loaded, tend to be less nimble and maneuverable than passenger
vehicles, especially on long, steep up grades.
Truck equivalency factors are used to adjust highway capacity for the presence of trucks in the
traffic stream. Truck equivalency factors are a function of the percent trucks in the traffic stream
and the degree and length of the maximum vertical grade on the facility. Table 3.4 reproduces
the truck equivalency factors for various combinations of percent trucks, grades, and grade
lengths found in table 3-4 of the 1994 HCM.
Inspection of table 3.4 clearly shows that truck equivalency factors increase as grades and
grade length increase. The effect of grade and grade length diminishes as the percent of trucks in
the traffic stream increases. It is also clear from table 3.4 that at grades of less than 2 percent,
the truck equivalency factor is 1.5, regardless of the length of the grade or the percent of trucks
in the traffic stream. Because the example problem used throughout this chapter has by
definition less than 2 percent grades, the truck equivalency factor for the example problem will
be 1.5.
45
Table 3.4. Truck equivalency factors.
Tables 3.5 and 3.6 are maximum freeway capacity (mixed vehicles per lane per hour) look up
tables based on truck equivalency factor and percent trucks in the traffic stream. Table 3.5 is for
2-lane directional freeways and table 3.6 is for 3- or more lane directional freeways. Table 3.6 is
the appropriate table for the example problem. Table 3.6 (6- or more lane facility), with a truck
equivalency factor of 1.5 and future year percent trucks of 11 percent, reveals by extrapolation a
free-flow capacity of 2,180 vehicles per lane per hour, or 6,540 vph for all 3 lanes.
46
Table 3.5. Maximum mixed vehicle traffic capacities for trucks in the traffic stream (4-lane facilities).
Table 3.6. Maximum mixed vehicle traffic capacities for trucks in the traffic stream (6 or more lanes).
47
Queue Dissipation Rates
Capacity during queue dissipation is less than the capacity for free-flow conditions, even
though the lanes are unrestricted (1994 HCM, 2-29). The reduction can easily be as much as
200 vph. According to the 1994 HCM, “. . . various observations of freeway queue depar-
ture rates range from as low as 1,500 pcphpl to as high as 2,000 pcphpl.” This implies that a
separate and distinct temporary dissipation capacity rate exists after a work zone is re-
moved. This rate comes into play when work zones are only in place for certain hours of the
day (i.e, when work zones are removed during peak traffic flow periods).
Removal of restrictions in front of a queue can also be analyzed much like the dissipation of a
queue at a red traffic signal. That is, the first cars move out rather slowly while follow-on cars
take a little less time and finally stabilize at a saturation flow rate further back in the queue.
Table 3.7 shows observed saturation flow rates (capacity) departing from traffic signals as
given in table 2-13, page 2-32, of the 1994 HCM.
As noted earlier, various observations of freeway queue departure rates range from as low as
1,500 pcphpl to as high as 2,000. Using an average of 1,818, with a standard deviation of
Table 3.7. Observed saturation flow rates per hour of green time.
144 from analysis of the traffic signal analogy above, there is a 68 percent probability that the
queue dissipation rate would be somewhere between 1,674 and 1,962. Alternately, there is a
95.5 percent probability that it would be somewhere between 1,530 and 2,106. Further
discussion of queue departure rates can be found on page 6-7 of the 1994 HCM.
Using a 50 percent reliability, the queue dissipation capacity selected for the example problem is
1,818 vehicles per lane. With 3 lanes open, total dissipation capacity becomes 5,454 vph.
48
Work Zone Capacity
Traffic capacity in the work zone can be estimated from research on the capacity associated with
various lane closures on multilane facilities. Table 3.8 reflects observed work zone mixed vehicle
flow capacities at several real-world work zones under several lane closure scenarios.(16)
Table 3.8 indicates, for example, that a 3-lane directional facility with 1 lane closed and 2 lanes
opened to traffic (line 5 of table 3.8) will have a total average work zone capacity of 2,980 vph,
and only 1,490 vehicles per lane per hour. As the 1,490 vehicles per lane per hour represent
the mean capacity, it incorporates a 50 percent reliability factor (i.e., half of the time the
capacity will be greater than 1,490 and half the time less than 1,490). Figure 3.3 shows the
capacity ranges observed for the average work zone capacities in table 3.8. This same data is
plotted as a descending cumulative probability distribution in figure 3.4.
3 Lanes - 1 Open
2 Lanes - 1 Open
5 Lanes - 2 Open
4 Lanes - 2 Open
3 Lanes - 2 Open
Capacity, Vehicles/Hour/Lane
Figure 3.3. Range of observed work zone capacities.
49
Figure 3.4 is used to incorporate a reliability factor in the value selected for the work zone
capacity. Figure 3.4 is used by selecting the desired percent reliability factor from the Y axis,
then intersecting the appropriate work zone situation, and estimating the corresponding capacity.
The x-axis intercept represents the adjusted work zone directional mixed vehicle flow capacity
per lane for the work zone configuration and reliability factor selected.
For the example problem, an 80 percent reliability factor will be used to determine work zone
capacity. By entering the figure at an 80 percent reliability and intersecting the curve for a 3-lane
directional facility with 2 lanes open, the work zone capacity, determined by inspection, is
approximately 1,415 vehicles per lane or 2,830 vph. Using an 80 percent reliability is roughly
equivalent to saying that the work zone capacity will be at least equal to 2,830 vehicles per hour
80 percent of the time. It also means, however, that the capacity of the work zone can be less
than 2,830 for 20 percent of the time.
80
(4,2)
Reliability Factor (%)
60
40
20
Figure 3.4. Cumulative distribution of observed work zone capacities. (Source: HCM, 1994)
50
Step 4. Identify the User Cost Components
With the roadway capacities established, the fourth step is to compare the roadway capacity
with the hourly demand for the facility. One of the most difficult problems in analyzing work zone
user costs is keeping track of all the input values and the resulting computations. One of the
most effective methods of keeping track is to set up the problem in a microcomputer
spreadsheet software program. Most of the following tables, such as table 3.9, were created in
a spreadsheet software program. It provides a convenient way to compare capacity and hourly
demand, and it forms the basis for determining the user cost components that come into play.
AADT 134,615
Num. of
Hour Queue Lanes Operating Cost
Queued
(24-Hr Demand Capacity Rate Open Conditions Factors
Vehicles
Clock)
(a) (b) (c) (d) (e) (f) (g) (h)
0–1 856 2,830 -1,974 0
1–2 614 2,830 -2,216 0 Free Flow Free Flow
2–3 509 2,830 -2,321 0 2 WZ in place Only
3–4 350 2,830 -2,480 0 No Queue Costs (3)
4–5 405 2,830 -2,425 0
5–6 961 6,540 -5,579 0
Free Flow
6–7 2,540 6,540 -4,000 0
3 No WZ No Costs
7–8 4,200 6,540 -2,340 0
No Queue
8–9 3,477 6,540 -3,063 0
9 –10 3,150 2,830 320 320
10 –11 3,417 2,830 587 907 WZ Delay
Forced Flow
11 –12 3,639 2,830 809 1,715 and
2 WZ in place
12 –13 3,769 2,830 939 2,654 Queuing
Queue Exists
13 –14 3,837 2,830 1,007 3,661 (5 costs)
14 –15 4,051 2,830 1,221 4,881
15 –16 4,725 5,454 -729 4,152
16 –17 5,849 5,454 395 4,548 Forced Flow Queuing
17 –18 6,865 5,454 1,411 5,959 3 No WZ Only
18 –19 4,289 5,454 -1,165 4,794 Queue Exists (4 costs)
19 –20 2,730 5,454 -2,724 2,070
20 –21 2,354 2,830 -476 1,594 Forced Flow WZ Delay
21 –22 1,998 2,830 -832 762 2 WZ in place and Queue
22 –23 1,610 2,830 -1,220 0 Queue Exists (5 Costs)
23 –24 1,259 2,830 -1,571 0 No Queue Free Flow
Total 67,453
51
In table 3.9, column (b) shows the directional hourly travel demand determined in table 3.3 on
page 43, while column (c) shows the capacities just determined. The capacities shown include:
(1) Work zone capacity of 2,830 vph when the work zone is in place (midnight to
5 a.m., 9 a.m. to 3 p.m., and once again from 8 p.m. to midnight),
(2) Free-flow capacity of 6,540 vph during free flow operation when the work zone is
not in place (5 to 9 a.m.), and
(3) Queue dissipation capacity of 5,454 vph for the 3 to 8 p.m. time period when the
work zone is removed but a built-up queue exists and starts to dissipate.
Because the major work zone user cost impact is the result of the user delay component of
traversing any queue that may develop, it is important to determine whether or not a queue will
form, and if it forms, how long it will take to dissipate. To answer this question, the directional
hourly demand is compared to the available capacity for each hour of the day.
The Queue Rate in column (d) is the difference between hourly capacity of the facility and the
unrestricted hourly demand (demand minus capacity) during each hour of the day. The queuing
rate is the hourly rate at which vehicles accumulate to, or, if negative, dissipate from any queue
that may exist. Column (e), on the other hand, represents the cumulative number of vehicles
backed up in the queue at the end of each hour.
To assist in the analysis, table 3.9 includes several additional temporary columns:
The most comprehensive approach in analyzing table 3.9 would be to conduct an hour-by-hour
analysis. The analysis can be significantly simplified by grouping and analyzing hours of the day
with similar operating characteristics. The analysis below centers around 5 periods of the day
(one is repeated) that share similar operating characteristics.
Period 1
Inspection of table 3.9 shows the work zone is in place from midnight until 5 a.m. and that
capacity is restricted to 2,830 vph. As traffic demand is lower than capacity, the facility
operates under free-flow conditions. There is no queue and no vehicles have to stop. Under
these conditions the work zone results in three free-flow user costs: the VOC and delay cost of
the speed change associated with slowing down for the work zone, and the delay cost of
traversing the work zone at a reduced speed.
Period 2
From 5 to 9 a.m. the work zone is removed and, with no built-up queue, capacity increases to
the free-flow rate of 6,540 vph and exceeds demand throughout the period. Under these
52
conditions, there’s no queue and no work zone, and therefore there are no work zone- or
queue-related user costs.
Period 3
At 9 a.m., when the work zone is reestablished, capacity falls back to 2,830 vph until 3 p.m.,
when the work zone is removed for the evening rush hours. During this period demand exceeds
capacity and a queue forms. By 10 a.m. the queue grows to 320 vehicles and continues to grow
to 4,881 vehicles by 3 p.m., when the work zone is again removed. During the 9 a.m. to 3 p.m.
period, there are a total of five user cost components. They are the four forced-flow user costs
associated with queuing (stopping VOC and delay costs, idling VOC, and delay cost of
crawling through the queue) as well as the free-flow delay in traversing the work zone. The
speed change delay and VOC cost factors have been replaced by the delay and VOC stopping
cost factors.
Period 4
The work zone is removed at 3 p.m. Because there are 4,881 vehicles already queued on the
roadway, capacity only increases to 5,454 vph (the queue dissipation rate), rather than the
normal free-flow capacity of 6,540 vph.
From 3 to 4 p.m., demand is less than capacity and the queue starts to dissipate, falling to 4,152
vehicles at 4 p.m. Rush hour demand from 4 to 6 p.m. again exceeds capacity. The queue
grows again to its maximum length of 5,959 at 6 p.m. before dissipating over the next 2 hours to
2,070 vehicles at 8 p.m., when the work zone is reestablished. During this entire period from 3
until 8 p.m., there is no work zone and therefore no free-flow user cost. There are, however,
queuing user costs (stopping VOC and delay costs, and idling and delay cost of crawling
through the queue).
Period 5
At 8 p.m., when the work zone is reestablished, there is a standing queue of 2,070 vehicles.
While the capacity falls to 2,830 vph, it exceeds the hourly demand and the queue continues to
dissipate. The queue finally completely dissipates somewhere around 10:30 p.m. During this 8
to 11 p.m. period, conditions are identical to period 3.
Period 1 … Revisited
Finally, by 11 p.m. the queue is completely gone and traffic flow reverts to free-flow operation
through the work zone for the final hour of the 24-hour period. During this last hour, only free-
flow user costs are incurred. The roadway operating characteristics of the 11 p.m. to midnight
period are identical to those of Period 1 discussed earlier.
53
Step 5. Quantify Traffic Affected by Each Cost Component
The next step is to quantify the number of vehicles involved with each cost component. For
simplicity of analysis, hours of the day are aggregated into the periods with similar operating
characteristics. As discussed earlier, the hours are midnight to 5 a.m., 5 to 9 a.m., 9 a.m. to
3 p.m., 3 to 8 p.m., 8 to 11 p.m., and 11 p.m. to midnight.
Table 3.10 is a modification of table 3.9. The three columns that described operating conditions
(f through h) have been replaced with four columns (f) through (i) that provide information on
the number of vehicles that have to (f) traverse the work zone, (g) traverse the queue, (h) stop
for the queue, and (i) those that merely have to slow down. These four columns are used to
identify the number of vehicles involved in the seven user cost components. The table is
subdivided to cluster and subtotal the hours of the day with similar operating conditions
as just discussed.
54
Vehicles That Traverse the Work Zone – Column (f)
The traffic that traverses the work zone in column (f) is generally the traffic demand on the
facility during the hours the work zone is in place. Although this is the case under free-flow
operating conditions, under forced-flow conditions, the maximum number of vehicles that can
traverse the work zone is limited to the capacity of the work zone, which in the example
problem is 2,830 vph.
From midnight to 5 a.m., the number of vehicles traversing the work zone is the demand during
the period or 2,734 vehicles. During the 9 a.m. to 3 p.m. period, the demand exceeds capacity
and the number traversing the work zone is limited to the 2,830 vph capacity of the work zone,
for a total of 16,980 during the period. From 3 to 8 p.m., the work zone is removed and there is
no work zone to traverse.
From 8 p.m. until midnight, the number of vehicles is at first limited to the capacity of the work
zone because, while hourly demand is less than capacity, there is a built-up queue on the
roadway that has to dissipate. Once the queue is dissipated, some time after 10 p.m., the
number of vehicles traversing the work zone area then reverts to the hourly demand. From 8 to
11 p.m., 8,032 vehicles traverse the work zone, and an additional 1,259 vehicles traverse the
work zone, from 11 p.m. to midnight.
A total of 29,005 vehicles traverse the work zone over the 24-hour period, as shown at
the bottom of column (f).
All vehicles that approach the work zone when a physical queue exists must stop and work their
way through the queue before entering the work zone. Traffic that arrives as the queue starts to
develop will have a rather short queue, while traffic arriving when the queue is fully developed
will have a much longer queue to traverse. On the other hand, vehicles arriving as the queue is
dissipating will have a continually shrinking queue length to deal with. Nonetheless, all vehicles
that arrive while a queue is present must traverse it.
It is important to note that because the facility is operating under a forced-flow condition, the
hourly volume of vehicles traversing the queue shown in column (g) is limited to the capacity of
the work zone area. This is because the only way out of the queue is through the work zone
area. During the period from 9 a.m. until 3 p.m., this capacity is 2,830 vph for a total of 16,980
vehicles.
During the period from 3 until 8 p.m., even though the work zone is removed, the capacity only
increases to the queue dissipation rate of 5,454 vph with 3 lanes open. During this 3 to 8 p.m.
period, a total of 27,270 vehicles traverse the queue.
Finally, from 8 to 10 p.m., the hourly volume traversing the queue is once again limited to the
2,830 vph work zone capacity. During the 10 to 11 p.m. period, when the queue finally
55
completely dissipates, the number of vehicles traversing the queue is limited to those vehicles
queued at 10 p.m., plus those vehicles arriving early in the hour before the queue is completely
gone. The 1,767 vehicles shown for 10 to 11 p.m. represent the prorated portion. A total of
7,427 vehicles traverse the queue during this 8 to 11 p.m. period. From 11 p.m. to midnight
there is no queue and therefore no vehicles traverse it.
A total of 51,677 vehicles traverse the queue over the 24-hour period, as shown at the
bottom of column (g).
A total of 51,677 vehicles must stop over the 24-hour period, as shown at the bottom of
column (h).
This is the same number, 51,677 vehicles, that traverse the queue shown at the bottom of
column (g).
Column (i) reveals that, in this example, only a small portion of the daily traffic has to just slow
down to traverse the work zone. The number of vehicles that just have to slow down prior to
traversing the work zone (as opposed to coming to a complete stop) are those vehicles
encountering the work zone under free-flow conditions. In the example presented here, that
situation exists between midnight and 5 a.m. and some time after 10 p.m. through midnight.
During these times, the hourly number of vehicles that must merely slow down equals the
demand during that hour. The number of vehicles that merely slow down during the 10 to
11 p.m. period is less than traffic demand. That is because the first part of the hour is used to
dissipate the last of the queue, and only the last portion of the hour operates under free-flow
conditions. The 605 vehicles shown for 10 to 11 p.m. represent a prorated portion of the hourly
demand.
A total of 4,597 vehicles must slow down over the 24-hour period, as shown at the
bottom of column (i).
56
General Comment
It is important to note that, in this example, a queue develops at midmorning and remains until
sometime after 10 p.m. In other situations with different traffic levels, hourly distributions, and
work zone configurations, it is quite possible for a queue to develop in the morning, completely
dissipate, and then reappear in the afternoon. Because the number of vehicles queued would
differ, the length of the queue and associated user costs would also differ and would need to be
analyzed separately.
Table 3.11 reproduces summary data on the traffic affected for each of the cost components
from table 3.10.
Before computing actual user cost, the analyst must know the number vehicles subjected to
speed changes, the number of vehicles that stop, and the delay time through both the work zone
and the queue. The number of vehicles that undergo speed changes and that stop is directly
related to the affected traffic, which has already been determined. Although the number of
vehicles delayed through the work zone and queue area have been determined, the amount of
delay can only be computed after knowing the work zone and queue area lengths and the
speeds through them.
The delay time through the work zone and through the queue is computed in the same manner.
In each case, the delay is determined by subtracting the time it takes to traverse either the work
zone or queue length when they are present from the time it takes to travel the same distance
when they are not present. Both calculations depend on the length to be traversed and the
appropriate travel speeds when a work zone and/or a queue are present and when they are not.
Work zone and queue delay computations for the example problem follow:
57
Work Zone Reduced Speed Delay
The work zone reduced speed delay is computed using the delay formula just discussed. Table
3.12 shows the results.
Queue reduced speed delay is computed in the same manner, however, in this case the queue
speed and queue length are not known. It is therefore necessary, in this case, to determine the
queue speed and queue length for each of the three analysis time periods where queues exists.
Speed through the queue can be determined by using the Forced-Flow Average Speed versus
Volume to Capacity (V/C) ratio graphs for level of service F contained in the earlier editions of
the Highway Capacity Manual. Using the volume through the queue and the Free-Flow
capacity of the facility, the V/C ratio is calculated for each period and used to find the
corresponding speed. Table 3.13 gives volume and capacity information along with the V/C
calculations for the three different queue periods (9 to 3, 3 to 8, and 8 to 11).
58
Because the only way for traffic to exit the queue is through the work zone, the volume through
the queue section is limited to the capacity of the work zone. The capacity of the work zone
restricts the volume to 2,830 vph during the 9 a.m. to 3 p.m. and the 8 to 11 p.m. periods. The
queue dissipation capacity restricts volume through the queue to 5,454 vph during the 3 to 8
p.m. period.
The capacity of the 3-lane directional facility where the queue forms (upstream of the work
zone) is the same as the capacity in the 3-lane unrestricted upstream section operating in a
free-flow condition just prior to the queue. That free-flow capacity was determined earlier to be
6,540 vph.
Using these values, the V/C ratios are calculated and used in conjunction with figure 3.5 to
determine the respective queue speeds shown at the bottom of table 3.13.
Speed 18 mi/h
Speed 8 mi/h
V/C = .83
V/C = .43
Inspection of figure 3.5 reveals that the speed through the queue for the 9 a.m. to 3 p.m. and the
8 to 11 p.m. time periods will be approximately 8 mi/h based on the computed 0.43 V/C ratio.
The speed for the 3 to 8 p.m. period is approximately 18 mi/h based the computed 0.83 V/C
ratio.
59
Queue Length Calculations
The queue length varies throughout the day with changes in directional hourly demand and
capacity through the work zone section. The number of vehicles in the queue starts out small
when it first forms around 9 a.m., grows to a maximum at about 6 p.m., and trails off to nothing
as it totally dissipates somewhere before 11 p.m. (around 10:40 p.m.). Some vehicles face very
short delays, while others face considerable delays. Figure 3.6 is a plot of the number of queued
vehicles at the end of each hour. It graphically displays the growth and dissipation of the number
of queued vehicles over the course of the day.
7.0
6.0
Vehicles (1,000)
5.0
4.0
3.0
2.0
1.0
0.0
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time
Queue delay computations are generally based on the average queue length over the queue
period. An average queue length can be computed for each hour that a queue exists or an
overall average queue length for the entire 9 a.m. to 11 p.m. period can be developed by
making some simplifying assumptions about queue growth and dissipation rates. Such
assumptions, if valid, can greatly reduce the number of necessary delay time computations.
If the number of queued vehicles grew at a uniform rate to some maximum value and then
dissipated at a uniform rate, the graph in figure 3.6 would be a triangle. If it were a triangle, the
average queue length for the entire 9 a.m. to 11 p.m. period would simply be one-half of the
maximum queue length that occurs during the day. The maximum queue length during the day
occurs when the maximum number of vehicles are queued. In the example problem, this would
be at 6 p.m.
Because figure 3.6 does not reflect uniform queue growth or dissipation rates, the more detailed
hour-by-hour analysis is more appropriate. For the example problem, the first step is to
determine the average number of vehicles queued in each hour. This is simply the arithmetic
60
average of the number of queued vehicles at the beginning and end of each hour. The computed
hourly average number of queued vehicles is plotted in figure 3.7.
6.0
5.0
Vehicles (1,000)
4.0
3.0
2.0
1. 0
0.0
Time Period
The next step is to determine the hourly queue lengths from the average number of queued
vehicles for each hour.
The average number of queued vehicles in each hour is converted to average hourly queue
lengths by dividing the average number of vehicles in the queue during the hour by the change in
traffic density between the upstream free-flow section and the queue section during that hour.
Traffic density is, by definition, the number of vehicles on a mile of roadway (vehicles/mile). It
can be computed by dividing vehicle flow through the section (vph) by the average speed
through the section (mi/h). Dividing traffic flow (vph) by traffic speed (mi/h) produces units of
vehicles per mile [(vehicles/hour) / (miles/hour) = (vehicles/miles)].
Traffic densities and queue lengths are computed in table 3.14 along with the average queue
lengths. Column (a) is the hour for which the queue length is being calculated. The queue and
upstream volumes (columns b and c) for the hours shown in column (a) are taken from table
3.10. The queue speeds (column d) are taken directly from figure 3.5. The upstream speed
(column e) was given as 55 mi/h in the problem definition.
The density in the queue (column f) is determined by dividing the queue volume in column (b) by
the queue speed from column (d). The density of the upstream section (column g) is determined
by dividing the upstream demand in column (c) by the upstream speed from column (e). The
change in traffic density in column (h) is just the difference in the upstream and queue traffic
densities (column f minus g).
61
The Average Number of Queued Vehicles (column i) is the arithmetic average of the number of
vehicles queued at the beginning and end of each hour. Finally, the Average Queue Length in
miles (column j) is computed by dividing the average number of queued vehicles for the hour
(column i) by the change in traffic density (column h) during that hour.
The queue lengths in table 3.14 are grouped by analysis period (i.e., 9-15, 15-20, and 20-23)
in order to determine the average queue length for the three periods. The average queue length
for each of the three periods is simply the average of the average hourly queue lengths for each
hour in the period.
For the 9 a.m. to 3 p.m. period, the average hourly queue length is 0.54 miles between 9 and 10 a.m.
and continually builds to a maximum hourly average length of 15.25 miles between 2 and 3 p.m. The
overall average queue length for the 9 a.m. to 3 p.m. period is 6.87 miles.
For the 3 to 8 p.m. period, the average hourly queue length starts at 20.81 miles between 3 and 4 p.m.,
builds to 29.48 miles between 5 and 6 p.m., and then drops back to 13.55 miles between 7 and 8 p.m.
This happens despite of the fact that the work zone is not in place during this entire 3 to 8 p.m. period.
The overall average queue length for the entire 5-hour period is 21.97.
Finally, during the 8 to 11 p.m. period, the average hourly queue length starts out at 5.89 miles between
8 and 9 p.m. and drops to 3.71 miles between 9 and 10 p.m. Somewhere between 10 and 11 p.m. the
queue finally dissipates and the average hourly queue length for this period is 1.17 miles. The overall
average queue length for the 8 to 11 p.m. period is 3.59 miles.
62
Queue Length Calculations – Alternate Approach
A simplified approach to calculating queue lengths entails assigning the number of queued
vehicles to the available lanes and multiplying the number of vehicles per lane by an assumed
average vehicle length that includes the space between vehicles. An example problem on page
6-7 of the 1994 HCM assumes an average vehicle length of 40 feet/vehicle. Using the same
assumed value of 40 feet/vehicle and the average number of queued vehicles for each of the
queue periods from table 3.14, the average queue length are calculated in table 3.15 using this
alternate approach.
Average
Average No. Average Queue
Average No. Feet
Queued Vehicle Length
Period No. Queued of Per
(Vehicles/Lane) Length (Miles)
Vehicles Lanes (b)/(c) Mile
(feet) (d)x(e)
(f)
(a) (b) (c) (d) (e) (f) (g)
9:00 - 15:00 1,950 3 650 40 5,280 5.0
15:00 - 20:00 4,586 3 1,529 40 5,280 11.8
20:00 - 23:00 1,130 3 377 40 5,280 2.9
The average queue lengths calculated in table 3.15 under the alternative approach are somewhat
shorter for the 9 a.m. to 3 p.m. and the 8 to 11 p.m. periods (periods when the work zone is in
place). This indicates the assumed vehicle length of 40 feet is about 25 percent too short. The
average queue length calculated for the 3 to 8 p.m. period, when the work zone is not in place,
is significantly shorter (approximately 50 percent) than the average queue length computed using
the roadway density approach. During this period the queue speed is significantly higher and it is
highly unlikely that the assumed 40 feet/vehicle spacing can be maintained.
Once the maximum queue lengths and the speeds through the queue for all three queue analysis
periods have been determined, the average delay for vehicles that transverse each of the queues
can be computed. The average delay is computed by determining the time necessary to
transverse the average queue length at the forced-flow queue speed and subtracting the time it
would normally take if the queue were not present. The speed without the queue present is the
same as the speed in the upstream section (in this case it is given as 55 mi/h).
The average queue delay time for each of the three queue analysis periods is computed in table
3.16. The travel times shown in column (c) are computed by dividing the average queue length
in column (a) by the queue speed in column (b). The travel times for column (d) are computed
by dividing the average queue length in column (a) by the 55 mi/h upstream speed. The average
queue delay times shown in column (e) and (f) are determined by subtracting the travel times
necessary to traverse the queue length (columns c minus d). The daily hours of queue delay will
be computed by multiplying the average queue delay time per vehicle by the number of affected
vehicles in each period.
63
Table 3.16. Average queue delay time.
Table 3.17 is reproduced from chapter 2 and shows additional hours of delay and additional
VOC (in August 1996 dollars) associated with stopping 1000 vehicles from a particular speed
and returning them to that speed for the three vehicle classes. Different factors are provided for
Passenger cars and both Single-Unit and Combination trucks. In addition, the last row of table
3.17 shows the VOC rate associated with idling while stopped.
While this table is designed to determine stopping cost, it can also be used to determine the cost
and time factors associated with slowing from 55 mi/h to 40 mi/h and returning to 55 mi/h. This
is accomplished by subtracting the cost and time factors for stopping associated with each
Table 3.17. Added time and vehicle running cost/1,000 stops and idling costs (Aug 96 $).
Added Time (Hr/1,000 Stops) Added Cost ($/1,000 Stops)
Initial (Excludes Idling Time) (Excludes Idling Time)
Speed Pass. Trucks Pass. Trucks
(mi/h) Cars Single-Unit Combination Car Single-Unit Combination
5 1.02 0.73 1.10 2.70 9.25 33.62
10 1.51 1.47 2.27 8.83 20.72 77.49
15 2.00 2.20 3.48 15.16 33.89 129.97
20 2.49 2.93 4.76 21.74 48.40 190.06
25 2.98 3.67 6.10 28.67 63.97 256.54
30 3.46 4.40 7.56 36.10 80.23 328.21
35 3.94 5.13 9.19 44.06 96.88 403.84
40 4.42 5.87 11.09 52.70 113.97 482.21
45 4.90 6.60 13.39 62.07 130.08 562.14
50 5.37 7.33 16.37 72.31 145.96 642.41
55 5.84 8.07 20.72 83.47 160.89 721.77
60 6.31 8.80 27.94 95.70 178.98 798.99
65 6.78 9.53 NA 109.02 195.84 NA
70 7.25 NA NA 123.61 NA NA
75 7.71 NA NA 139.53 NA NA
80 8.17 NA NA 156.85 NA NA
Idling Cost ($/Veh-Hr) 0.6927 0.7681 0.8248
64
speed from one another. For the example problem, the last line of table 3.18 (columns a through
c, and columns d through e, respectively) shows the delay and speed change cost factors
associated with going from 55 mi/h to 40 mi/h and back to 55 mi/h.
The section on Delay Cost Rates (page 19) discusses user delay cost rates. Table 2.13 is
reproduced here for convenience as table 3.19.
Trucks
Passenger Cars
Single-Unit Combinations
$ 10 to 13 $17 to 20 $ 21 to 24
While any values within the ranges shown in table 3.19 are considered reasonable, for purposes
of the example problem, the Delay rates selected correspond to the mean values determined in
table 2.7 (see page 20). The following values are used in connection with all delay cost
computations in the section on Compute User Cost Component by Vehicle Class:
At this point it is necessary to distribute the directional traffic affected by the various cost
components to the appropriate vehicle classes for each cost component. Table 3.20 lays out the
matrix used to assign the overall traffic associated with each of the user cost components to the
appropriate vehicle classes.
65
Table 3.20. Affected traffic by vehicle class and user cost component.
Continuing with the example problem, column (a) of table 3.20 lists the total daily traffic for
each user cost component. The volumes listed in column (b), (c), and (d) reflect the distribution
of the overall traffic for each of the cost components to the appropriate vehicle classes. The
header for each column lists vehicle class distribution factors used. These percentages were
developed for the example problem in the Project Future Year Traffic Demand section
(page 43). They reflect the projected distributions in 1999 (the year of the work zone will be
established) based on the differential growth rates assigned to passenger vehicles and trucks.
Column (e) is the sum of columns (b), (c), and (d) and is just a mathematical check to ensure
that the traffic assigned to the vehicle classes totals back to the original traffic volume.
Daily user costs by vehicle class for each cost component are computed by multiplying the
affected traffic by the appropriate unit cost rates (either VOC or delay) for the various
components. The individual costs are computed in tables 3.21 through 3.27.
The added VOC rates used in column (b) of table 3.21 are from the bottom line of columns (d)
through (f) in table 3.18, while the rates used in column (b) of table 3.22 are from the bottom
line of columns (a) through (c) of table 3.18.
The Idle cost rates in column (c) of table 3.26 are taken from the bottom line of table 3.17.
Table 3.21. User cost component # 1— speed change VOC (55-40-55 mi/h).
Added VOC Total
Affected Cost ($)
(55-40-55) Costs ($)
Vehicle Class Vehicles per day
$/1,000 Vehicles (60 Days)
(a) (b) (c) (d)
Passenger Cars 4,092 30.77 126 7,554
Single-Unit Truck 276 46.92 13 777
Combination Truck 230 239.56 55 3,304
Total Speed Change VOC 4,597 194 11,634
66
Table 3.22. User cost component # 2 — speed change delay cost (55-40-55 mi/h).
Table 3.23. User cost component # 3 — work zone reduced speed delay cost.
A dded Tim e C osts ($)
A ffected C ost ($) Total C ost
(55-0-55) per day
V ehicle C lass V ehicles (per Veh-H r) (60 days)
(H rs/1,000 V eh)
(a) (b) (c) (d) (e)
Passenger Cars 45,993 5.84 $11.58 3,110 186,621
Single-U nit Truck 3,101 8.07 $18.54 464 27,835
Com bination Truck 2,584 20.72 $22.31 1,194 71,665
T otal Stopping D elay 51,677 4,769 286,121
Table 3.25. User cost component # 5 — stopping delay cost (55-0-55 mi/h).
67
Table 3.26. User cost component # 6 — idling VOC.
Added
Affected Idle VOC Rates Costs ($) Total Cost
Queue Time
Vehicle Class Vehicles ($/1000Veh-Hr) per day (60 days)
Period (Hours)
(a) (b) (c) (d) (e)
9 a.m. to 3 p.m. 15,112 0.7340 7,684 461,038
Passenger 3 p.m. to 8 p.m. 24,270 0.8211 692.70 13,804 828,227
Cars 8 p.m. to 11 p.m. 6,610 0.3837 1,757 105,422
Subtotal 45,992 - - 23,245 1,394,687
9 a.m. to 3 p.m. 1,019 0.7340 574 34,464
Single-Unit 3 p.m. to 8 p.m. 1,636 0.8211 768.10 1,032 61,913
Trucks 8 p.m. to 11 p.m. 446 0.3837 131 7,881
Subtotal 3,101 - - 1,738 104,258
9 a.m. to 3 p.m. 849 0.7340 514 30,840
Combination 3 p.m. to 8 p.m. 1,364 0.8211 824.80 923 55,403
Trucks 8 p.m. to 11 p.m. 371 0.3837 118 7,052
Subtotal 2,584 - - 1,555 93,295
Total Idling VOC 1,592,241
Table 3.27. User cost component # 7 — queue reduced speed delay cost.
Added Delay Cost
Affected Costs ($) Total Cost
Queue Time Rate ($)
Vehicle Class Vehicles (per Veh-Hr)
per day (60 days)
Period (Hours)
(a) (b) (c) (d) (e)
9 a.m. to 3 p.m. 15,112 0.7340 128,454 7,707,264
Passenger Cars 3 p.m. to 8 p.m. 24,270 0.8211 $11.58 230,760 13,845,629
8 p.m. to 11 p.m. 6,610 0.3837 29,373 1,762,360
Subtotal 45,992 - - 388,588 23,315,253
9 a.m. to 3 p.m. 1,019 0.7340 13,865 831,884
Single-Unit 3 p.m. to 8 p.m. 1,636 0.8211 $18.54 24,907 1,494,428
Trucks 8 p.m. to 11 p.m. 446 0.3837 3,170 190,220
Subtotal 3,101 - - 41,942 2,516,533
9 a.m. to 3 p.m. 849 0.7340 13,903 834,202
Combination 3 p.m. to 8 p.m. 1,364 0.8211 $22.31 24,977 1,498,593
Trucks 8 p.m. to 11 p.m. 371 0.3837 3,179 190,750
Subtotal 2,584 - - 42,059 2,523,546
Total Queue Speed Delay 28,355,331
68
Step 11. Sum Total Work Zone User Costs
Table 3.28 shows a master summary of all costs, and table 3.29 shows the percent distributions
of those costs. The first three cost components in tables 3.28 and 3.29 represent the cost
associated with free-flow, while the remaining four cost components represent the forced-flow
queuing costs.
Examination of tables 3.28 and 3.29 immediately reveals that the high user costs are not a
LCCA problem, but are a traffic control problem. Further inspection reveals that more than
90 percent of the user costs result from the queue delay component. An additional 5 percent is
Table 3.28. Master summary — total (60 day) work zone user cost (Aug 96 $).
Table 3.29. Master summary — work zone user cost distribution (%).
69
associated with the queue idling costs and another 2 percent is associated with queue stopping
VOC and delay. Therefore, approximately 97 percent of the user costs can be avoided by not
allowing the queues to develop in the first place. In the example problem, the queuing situation
could be drastically reduced, if not completely avoided, if work zone operations could be
limited to evening work between 7 p.m. to 7 a.m. By limiting the contractor to evening work
hours only, the queue cost in the 9 a.m. to 3 p.m. period would be completely eliminated and
the evening rush hour would not have to deal with the built-up queue from the midday work
zone!
The contractor’s productivity rate would suffer dramatically during the midday use of the facility
because the contractor’s delivery vehicles would have to deal with the same delays as the
general traffic stream. It is therefore not a large penalty on the contractor to be unable to work
during midday.
Other alternatives to lowering the work zone-related user costs include adding capacity prior to
the development of large future traffic demands, accelerating contractor production to reduce
the overall duration the work zone is in place, and limiting the overall frequency of rehabilitation
activities.
While the numbers may appear unreasonably large, they are not out of line when compared to
user costs developed for highway facilities damaged by the North Ridge, California, earthquake.
Table 3.30 shows estimates of the daily user costs associated with having the damaged facilities
out of service.
Note: In the example problem, congestion delay is all work zone related. Prior to establishing
the work zone, the highway facility could handle every hourly demand. In other cases, this may
not be true—the facility may be congested on a regular basis without work zones. In such
cases, routine congestion costs must be subtracted from the work zone-related congestion.
The final step in calculating user costs is to address circuity and crash costs. The following
sections address these costs.
70
CIRCUITY
Circuity is a term used to describe the additional mileage that users travel, either voluntarily or
involuntarily, on a detour to avoid a highway work zone. If traffic is forced to detour, circuity
costs are the:
(1) Full VOC costs ($0.31 per mile) for passenger cars times the excess distance the
detour imposes.
(2) Appropriate $/hour delay rates times the excess detour time.
(3) Difference in crash rates times the exposure rate and cost per crash.
If no formal detour is established and circuity is the result of voluntary self-imposed diversions,
then a consumer surplus approach must be employed.
Problem Statement
A 5-mile rural section of Parkway between interchanges will be closed for 90 days to
reconstruct a bridge taken out by flash flooding. The reconstruction will require traffic to be
detoured to a 10-mile section of rural 2-lane minor arterial highway. Capacity on the minor
arterial is currently underutilized and congestion is not expected to be a major problem. AADT
on the parkway is 10,000 vehicles per day.
71
CRASH COSTS
General
The highway safety community has replaced the term accident with the term crash because the
term accident implies that they are unavoidable. In reality, highway crashes to a large extent are
avoidable. This Interim Technical Bulletin uses the term crash costs to describe what was
previously called accident costs.
Crash costs associated with work zones and work zone-generated circuitous travel are a
function of the generally higher crash rates in work zones and on alternate/detour routes than
on the primary facility in the absence of work zones. Crash rates are based on the number of
crashes as a function of exposure, typically vehicle miles of travel (VMT). Crash rates are
commonly specified as crashes per 100 million vehicle miles of travel (100 M VMT).
Overall crash rates for the various functional classes of roadway are fairly well-established.
Crash rates for work zones, however, are not. While there is a limited amount of work zone
crash history data, the validity of the data used to compute the crash rates is sometimes suspect.
Sometimes, crashes that occur in work zone-generated queues are not classified as work zone
crashes. Even more importantly, most of the time it’s difficult to accurately quantify the work
zone exposure rate (i.e., the length of the work zone and number of hours a day, and the
number of days the work zone and resultant queues are in place). Further, the crash rate, while
significantly higher in work zones than nonwork zones, is sometimes still low enough that there
aren’t any crashes in a given work zone because the exposure period is just too short to allow
for statistically valid results. Finally, the problem is compounded by the fact that work zones
differ significantly in the way they treat maintenance of traffic. For example, some use
permanent barriers, while others use cones or drums; some narrow lanes, while others maintain
lane width and shoulders. Although there appears to be a general rule of thumb indicating that
crash rates in work zones are about three times the normal rate for the facility, there does not
appear to be much statistically significant research data to support this rule of thumb.
With these limitations on the availability and validity of crash rates in mind, crash costs can be
developed by multiplying the unit cost per crash, by the differential crash rates between work
zones and nonwork zones, by the vehicle miles of travel during the duration of the work zone.
The duration includes both the time the work zone is in place and the time that queues are
present. Vehicle miles of travel can be estimated by multiplying the percent of the facility’s
AADT that will be affected by the work zone and resultant queues, by the length of the work
zone and queue, and the number of days the work zone will be in place.
72
Overall Crash Rates
Crash costs on detours and alternate routes used by diverted traffic are generally easier to
calculate because crash rates on the primary facility without a work zone and on alternate routes
are generally better known. However, even when the crash rates are better known, the exposure
rate may still be difficult to determine as it requires some indication or estimate of diversion to
alternate routes when no formal detours are established.
Crash rates by crash type for the various roadway functional classes are computed by using crash
and VMT data contained in the FHWA’s most current annual Highway Statistics report.(17)
Table 3.31 contains 1995 data on crash fatalities and nonfatal injuries listed in the FWHA’s 1995
Highway Statistics.Table 3.32 contains 1995 Highway Statistics data on VMT. The data in
these two tables are used to generate the crash injury rates contained in table 3.33. The overall
crash injury rates shown in table 3.33 are people injured per 100 M VMT and do not specifically
represent work zone injury rates.
Table 3.31. 1995 people injured in motor vehicle crashes by functional class.
Rural Urban
Facility Type
Nonfatal Nonfatal
Fatalities Fatalities
Injuries Injuries
(a) (b) (c) (d) (e)
Interstate 2,340 55,924 2,110 253,823
Other Freeways/Expressways - - 1,290 86,676
Other Principal Arterial 4,498 120,271 5,868 773,446
Minor Arterial 4,300 166,241 3,755 515,586
Major Collector 5,111 187,648 - -
Minor Collector 1,565 60,143 - -
Collectors - - 2,568 218,837
Local 3,910 236,441 4,455 601,383
Total 21,724 826,668 20,046 2,449,751
Source: 1995 Highway Statistics: tables Fl-220 and Fl 221.
73
Table 3.33. Crash injury rates (people injured per 100 M VMT)
Rural Urban
Facility Type
Nonfatal Nonfatal
Fatalities Fatalities
Injury Injuries
(a) (b) (c) (d) (e)
Interstate 1.0 25.0 0.6 74.3
Other Freeways/Expressways - - 0.9 57.2
Other Principal Arterial 2.1 55.8 1.6 208.8
Minor Arterial 2.8 108.6 1.3 175.8
Major Collector 2.7 100.8 - -
Minor Collector 3.1 120.4 - -
Collectors - - 2.0 172.4
Local 3.7 224.8 2.2 292.1
Source: Computed from 1995 Highway Statistics data.
Table 3.33 shows the crash injury rates for rural and urban areas. These were computed by
dividing the number of rural injuries (columns b and c) and urban injuries (columns d and e) in
table 3.31 by the appropriate VMT contained in columns (b) and (c) of table 3.32. The crash
injury rates listed in table 3.33 are the number of fatalities (column b and d) or the number of
nonfatal injuries (columns c and e) per 100 M VMT.
Limited information on work zone crashes is contained in the FHWA’s Construction Cost and
Safety Impacts of Work Zone Traffic Control Strategies.(18) The study evaluated the
differences in Single Lane Closure (SLC) and Two-Lane Two-Way Operation (TLTWO …
cross over) traffic control strategies on rural 4-lane divided highways. The report includes crash
data from 48 construction projects in 11 States. The traffic volumes on the routes were relatively
low, generally within a 10,000 to 30,000 ADT range.
The report concluded that there were no significant differences in overall crash rates between
the two traffic control alternatives for the sites and volumes studied. It further concluded that
there was no significant difference in total crashes rates between before versus during
construction for all projects. However, the report did conclude that the severity (the number of
fatal and nonfatal injury crashes) increased significantly during construction periods for both the
SLC and TLTWO traffic control strategies.
74
The 1996 Fatality Analysis Reporting System (FARS) data supports this finding. Column (a)
of table 3.34 lists ranges of the percent of total motor vehicle crash fatalities that occur in work
zones, while column (b) lists the number of States that fall within a range.(19)
Table 3.34. 1996 work zone motor vehicle crash fatalities as a percent of all fatalities.
Inspection reveals that most SHAs are experiencing a disproportionate percentage of fatalities
(more that 1 percent) occurring in work zones relative to the VMT that takes place in work
zones.
Construction Cost and Safety Impacts of Work Zone Traffic Control Strategies also
includes crash rates on the study routes before and during the time that work zones were
established. The crash rates provided in the report are broken out by type of work zone, type of
construction work, and type of traffic control used. However, because of the limited number of
projects in the study, the statistical validity of such highly disaggregated analysis is questionable.
75
As a result, tables 3.35 and 3.36 present only the overall work zone crash rate data for the two work
zone traffic control strategies contained in the report.
Columns (e) and (f) in tables 3.35 and 3.36 contain a computed crash rate in the form of the annual
number crashes per mile of highway per 10,000 ADT for the time periods before and during which the
work zones are in place. Column (g) shows the difference. The units used in the original report (annual
number crashes per mile of highway per 10,000 ADT) are converted to more commonly used units
(crashes per 100 M VMT) in columns (h) through (j). Inspection of tables 3.35 and 3.36 reveals some
projects had lower crash rates during the period of construction than they had during the period before
76
Table 3.36. Crash rates on TLTWO.
(a) (b) (c) (d) (e) (f) (g) (h) (i) (j)
LA I-20 7.2 23.9 6.607 2.332 -4.285 181.1 61.2 -119.8
UT I-15 8.8 5.2 5.447 2.815 -2.631 149.2 77.1 -72.1
WV I-64 2.1 27.0 5.495 2.931 -2.565 150.5 80.3 -70.2
LA US-190 2.4 16.0 9.505 7.097 -2.408 260.4 194.4 -66.0
MI I-96 12.0 12.8 5.104 3.062 -2.042 139.8 83.9 -55.9
FL SR-95 2.8 4.9 5.173 4.803 -0.369 141.7 131.6 -10.1
MI I-94 10.0 29.0 2.884 2.535 -0.350 79.0 69.5 -9.6
KY WKP 5.5 4.2 2.048 1.756 -0.293 56.1 48.1 -8.0
UT I-84 10.7 3.8 3.265 3.019 -0.246 89.5 82.7 -6.7
AZ I-10 2.2 12.0 2.496 2.304 -0.192 68.4 63.1 -5.3
LA I-20 8.8 13.5 1.703 1.606 -0.097 46.7 44.0 -2.7
OR I-5 13.2 22.6 0.978 1.043 0.065 26.8 28.6 1.8
LA I-20 4.7 27.6 1.590 1.877 0.287 43.6 51.4 7.9
MI I-94 10.0 18.5 3.886 4.185 0.299 106.5 114.7 8.2
LA I-59 6.6 13.0 3.314 3.615 0.301 90.8 99.0 8.2
NC I-40 16.9 25.0 0.980 1.700 0.720 26.8 46.6 19.7
KY I-75 4.0 23.0 1.469 2.498 1.029 40.2 68.4 28.2
OR I-5 7.1 24.2 0.712 1.780 1.068 19.5 48.8 29.3
NC I-40 11.7 17.0 2.050 3.415 1.365 56.2 93.6 37.4
WV I-77 2.6 9.3 2.516 4.193 1.677 68.9 114.9 45.9
NC US-1 1.8 15.0 2.674 4.457 1.783 73.3 122.1 48.8
MI I-69 6.1 15.5 1.397 4.239 2.841 38.3 116.1 77.9
Total 157.1 363.0
Average 3.241 3.057 -0.184 88.8 83.6 -5.1
STD 2.191 1.384 1.697 60.0 38.0 46.8
Source: FHWA-RD-89-210.
the work zone was established. Further, from the bottom lines of the tables, the overall average
crash rate on SLC was slightly (though not significantly) higher during work zone operations,
while the overall average crash rate on TLTWO was slightly (though again not significantly)
lower during work zone operations.
77
Work zone crash rate data is also included in An Evaluation of Lane Closure Strategies for
Interstate Work Zone.(20) Data contained in this report was based on the study of 26 Interstate
4R Projects in Indiana. Tables 3.37 and 3.38 provide data on overall and injury crash rates.
Note that the rates in tables 3.37 and 3.38 are per 10 M VMT.
The following example problem is used to lay out a rational approach to calculating the
additional crash costs associated with a work zone-generated detour.
Problem Statement
A 5-mile rural section of Parkway between interchanges will be closed for 90 days to
reconstruct a bridge taken out by flash flooding. The reconstruction will require traffic to be
detoured to a 10-mile section of rural 2-lane minor arterial highway. Capacity on the minor
78
arterial is currently underutilized and congestion is not expected to be a major problem. AADT
on the parkway is 10,000 vehicles per day.
Solution
Table 3.39 shows the matrix used to calculate the differential crash costs. Column (c) lists the
crash rates for the facility and crash types listed in columns (a) and (b). The rates themselves are
taken from columns (b) and (c) of table 3.33. The crash fatality rate is 2.8 fatalities per 100 M
VMT on rural 2-lane minor arterial highways and 2.1 on rural parkways (other principal arterial
routes). The injury crash rates per 100 M VMT is 108.6 on rural 2-lane minor arterial highways
and 55.8 on rural parkways (other principal arterial routes).
In addition to the higher fatality and nonfatal injury crash rates, detoured and diversionary traffic
on average have a greater crash exposure because of the generally greater travel distances.
Column (g) of table 3.39 lists the exposure for the two alternative routes in units of 100 M
VMT. The values listed were computed by multiplying the ADT (column d) by the number of
days (column e) by the length of each alternative route (column f) and dividing the answer by
100 million.
The actual number of crashes expected (column h) are calculated by multiplying the crash rates
in column (c) by the exposure in column (g).
Finally, the crash cost listed in column (j) are computed by multiplying the expected number of
crashes in column (h) by the crash cost rates in column (i). The crash cost rates used in column
(i) were selected from the updated MicroBENCOST default crash cost rates discussed earlier
in Crash Cost Rates on page 23.
Further information on work zone crash rates is available in the participant’s handbook for
National Highway Institute Course # 38003, Design and Operation of Work Zone Traffic
Control.(21) Pages 2-7 discuss Before and After work zone crash rates. Results from Virginia
and Texas are included.
79
Virginia data for 1991 indicate an increase of 57 percent on multilane highways and 168 percent
on 2-lane urban highways with the variation being a function of traffic, geometry, and
environment. In Texas, 1984 to 1988 data show freeway construction crashes increased by
28.7 percent on the main line and 2.4 percent on frontage roads. It also shows a 37.4 percent
increase at nighttime compared to 24.4 percent during the daytime. The data also suggest that
the average changes in severe mainline crash rates were consistent from project to project, while
other mainline crash categories varied significantly from site to site.
Microcomputer software programs such as MicroBENCOST and QueWZ are available for
conducting LCCA on routine pavement rehabilitation projects. The simplified procedures
outlined in this chapter are not likely to be sufficient when attempting to analyze user cost on
highly complex urban freeways. In such cases, the analyst would do better to use the battery of
urban planning models used by the major metropolitan planning organizations (MPOs). Traffic
assignment models, run on the network level can provide better estimates of traffic diversions
from the route in question and the effect of such diversions of traffic flow on alternate routes.
These models can be run to determine the network-level changes in vehicle miles of travel
(VMT) and hours of traffic delay with and without the work zone in place. When network level
changes in VMT and Delay are known, user-cost calculations become much simpler.
80
CHAPTER 4. RISK ANALYSIS APPROACH
DEFINING RISK
The concept of risk comes from the uncertainty associated with future events — the inability to
know what the future will bring in response to a given action today. Risk can be subjective or
objective. Subjective risk is based on personal perception — intuitively deciding how risky a
situation may be. For example, many people may feel that flying is more risky than driving. This
perception of risk may be related to the consequences of failure, as well as the ability (or
inability) to control the situation. Objective risk is based on theory, experiment, or observation.
Often the facts of the situation may dispute intuitive feelings. For example, in 1996 there were
1,070 aviation fatalities; in the same year there were also 40,676 highway fatalities. Because
individuals’ perceptions of risk vary, decisions incorporating risk management concepts will
depend to a large extent on the decision maker’s tolerance for risk.
Risk analysis is concerned with three basic questions about risk: 1) What can happen? 2) How
likely is it to happen? 3) What are the consequences of its happening? Risk analysis answers
these questions by combining probabilistic descriptions of uncertain input parameters with
computer simulation to characterize the risk associated with future outcomes. It exposes areas of
uncertainty typically hidden in the traditional deterministic approach to LCCA, and it allows the
decision maker to weigh the probability of an outcome actually occurring.
Many analytical models treat input variables as discrete fixed values, as if the values were certain.
In fact, the majority of input variables are uncertain. Economic models used in a typical LCCA
are no exception. In conducting LCCA, it is important to be aware of the inherent uncertainty
surrounding the variables used as inputs into the analysis. Uncertainty results from the
assumptions, estimates, and projections made in conducting the analysis. Table 4.1 summarizes
LCCA input variables and the general basis used to determine their values.
81
Table 4.1. LCCA input variables.
Traditionally, this uncertainty is often ignored in an LCCA. For example, the analyst may make a
series of best guesses of the values for each input variable and compute a single deterministic
result. The problem with this approach is that it often excludes information that could improve
the decision.
In some cases a limited sensitivity analysis may be conducted whereby various combinations of
inputs are selected to qualify their effect on analysis results. However, even with a sensitivity
analysis, this deterministic approach to LCCA often conceals areas of uncertainty that may be
crucial to decision-making process. Often, stakeholders seize upon the uncertainty associated
with LCCA inputs and vigorously debate the validity of the results. Traditional, deterministic-
based LCCA results such as these generate endless debate over which alternative truly has the
lowest life-cycle cost. This process encourages division and unproductive debate.(22)
The need to make strategic long-term investment decisions under short-term budget constraints
is forcing SHAs to consider risk as a criterion for judging a course of action. Risk analysis
exposes areas of uncertainty for the decision maker. Based on this new information, the decision
maker has the opportunity to take mitigating action to decrease exposure to risk. With the
emergence of user-friendly computer software, an SHA can easily integrate quantitative risk
analysis concepts into the decision-making process
GENERAL APPROACH
Figure 4.1 illustrates the basic risk analysis approach in LCCA. It shows the NPV formula
typically used as the economic indicator in an LCCA. As shown in figure 4.1, a risk analysis
approach uses random sampling from probabilistic descriptions of uncertain input variables
82
Combine Variability of Inputs to Generate a
Probability Distribution of Results
(initial cost, future cost, discount rate, and year of rehabilitation) to generate a probabilistic
description of results. By performing the Monte Carlo computer simulation, thousands, even
tens of thousands of samples are randomly drawn from each input distribution to calculate a
separate what-if scenario. With the speed and memory of today’s personal computers, such
iterations can be conducted in a matter of minutes or even seconds. The results generated from
each what-if iteration are captured for later statistical analysis.
Similar to the inputs, risk analysis results are presented in the form of a probability distribution
that describes the range of possible outcomes along with a probability weighting of occurrence.
In other words, by randomly drawing samples from the model’s input distributions, the computer
combines the variability inherent in the inputs and summarizes it in the form of a probability
distribution. With this information, the decision maker knows not only the full range of possible
values, but also the relative probability of any particular outcome actually occurring. This is
exactly the information that the decision maker needs in order to make an educated decision.
By including all possible values for the analysis inputs in relation to their probability of
occurrence, risk analysis elevates the LCCA debate from contesting the validity of results to
deciding best public policy. Armed with this new information, risk analysis provides the decision
maker with the opportunity to take mitigating action to decrease exposure to risk. Moreover,
risk analysis provides those vested with the appropriate authority and accountability, namely
executives and elected officials, the opportunity to make decisions about taking risk. (23)
Although the risk analysis approach is calculation intensive and would be impractical to
complete by hand, the approach can be easily incorporated into deterministic analysis by using
user-friendly computer software programs. Today, software is available that includes advanced
risk analysis techniques such as Monte Carlo simulation. Risk analysis software can be stand-
83
alone programs (Microsoft Visual Basic, C++, etc.) or simple add-ins to spreadsheets such as
Microsoft Excel or Lotus.
Because of the flexibility that spreadsheets offer in solving a wide variety of problems this
Interim Technical Bulletin explores the use of risk analysis in a spreadsheet environment.
The two most often used spreadsheet add-in risk analysis software programs are @RISK and
Crystal Ball.They are very similar in operation and capability. For presentation purpose, this
Technical Bulletin discussion uses @RISK in conjunction with Microsoft Excel spreadsheets.
Risk analysis has been used for a number of years in other industries. As a result, a number of
approaches have evolved to conduct such analysis. The following sections outlines a general
approach that may be used to conduct a risk analysis. The five proposed steps are:
The following sections address each of these steps in detail. The example problem presented
below illustrates how risk analysis may be applied in a pavement design LCCA setting. To keep
the example simple, user costs are omitted.
Alternative A Alternative B
Cost Item
Average Std Dev Average Std Dev
Initial Agency Cost
26.5 0.75 20.0 2.5
($ Millions)
Future Rehabilitation
7.0 0.5 6.0 1.0
Cost ($ Millions)
84
Table 4.3. Estimates of pavement service life.
Alternative A Alternative B
Most Most
Min Likely Max Min Likely Max
Initial Pavement
20 25 30 12 15 18
Design (Years)
Future Rehabilitation
10 13 15 5 7 10
Design (Years)
The first step in conducting a risk analysis is to identify the structure and layout of the problem.
This usually involves reducing the problem to its most basic elements and describing it in the
form of an analytical model. Flow charts sometimes supplement the model and clarify
relationships between variables. In this example problem, the model is defined by the formula for
NPV:
1
NPV = InitialCost + ∑ FutureCosts
(1+ i )n
Figure 4.2 shows the pavement service life curves for each alternative strategy based on the
mean service life values from table 4.3. Based on the mean pavement service life, Alternative A
Alt A
Alt B
Pavement Condition
Terminal Serviceability
RSL
0 5 10 15 20 25 30 35
Pavement Life (Years) 35-Year Analysis Period
85
requires one rehabilitation; whereas, Alternative B requires three rehabilitations over the analysis
period of 35 years. Figure 4.3 shows the expenditure streams for each alternative based on the
mean service life values from table 4.3.
$26.5 Alt A
Alt B
Cost ($ Millions)
$20.0
$7
$6 $6 $6
35
0 5 10 15 20 25 30
$1.5
Pavement Life (Years) 35-Year Analysis Period $1.6
Based on the mean values of estimated service lives, both Alternatives A and B have remaining
service life (RSL) at the end of the 35-year analysis period. The value of the RSL must be
properly accounted for and appropriately discounted. In a deterministic LCCA, treatment of the
RSL of each alternative would be relatively straightforward. In the probabilistic risk analysis
approach, the RSL issue becomes more complex because of the range of possible input values
and the random sampling of model inputs.
The NPV model must take into account the entire range of probable pavement service lives for
both the initial design and future rehabilitation designs. As such, there is a discrete probability
that the last rehabilitations (as defined using mean service life values) shown in figure 4.2 could
take place earlier than the mean service life values would indicate. At the same time there is a
discrete probability that the last rehabilitations shown might well only be required at some point
beyond the end of the analysis period. As a result, probabilistic-based risk analysis models must
account for all the possibilities in determining the number, timing, and RSL of future rehabilitation
requirements. Typically, logic IF statements provided in most spreadsheet programs or
programming languages can be used to facilitate discounting future costs over the entire range of
probable service life values.
In developing the structure and layout of the model, it is crucial to identify dependencies among
uncertain variables to avoid producing incorrect results. Because user costs have been omitted
in the example problem, the number of variables has been significantly reduced. Initial cost,
future costs, discount rate, and year of rehabilitation activity are the uncertain variables in this
model. The next step is to describe their uncertainty.
86
Step 2. Quantify Uncertainty Using Probability
The next step is to develop probability distributions for the uncertain variables identified in the
previous step. A probability distribution allows the analyst to describe the complete range of
values the variable may assume and weights their likelihood of occurrence accordingly.
Figure 4.4 illustrates some of the more common probability distributions. Those shown include
triangle, normal, and uniform distributions in histogram format.
The horizontal axis provides a range of possible values and the vertical axis provides a relative
frequency weighting of the occurrence of any particular value. For the histograms shown, the
probability is equal to the area under the curve and the total shaded area is equal to 1.0.
Figure 4.5 shows a cumulative ascending probability distribution. In this case the cumulative
probability is read directly off the vertical axis. For the example shown, there is a 75 percent
probability that project costs will be less than or equal to $9 million. Sometimes, the cumulative
distribution is shown in the descending format, in which case the probability represents the
probability of exceeding any particular value.
1.00
0.0 4 8 12
Project Cost, ($ Millions)
87
Discrete and Continuous Variables
Probability distributions may be developed using either objective or subjective methods. The
objective method uses hard data (such as bid price list and observed capacity) to formulate the
distribution; the subjective method uses expert opinion.
When existing data is available, statistical analysis packages can be used to automatically fit the
probability distribution to the data. These programs compare the more common distribution
types with available data. Along with recommendations on the distribution types that best
describe the variability of the data, many distribution fitting programs provide statistical
indicators such as Chi-squared, Anderson-Darling (A-D), and Kolmogorov-Smirnoff (K-S)
that describe the goodness of fit. These statistics indicate how closely the probability distribution
fits the data.
Figure 4.6 illustrates the use of group interviews for developing subjective probability
descriptions of uncertain variables. As shown, expert panels are convened to establish the
boundaries and general shape of input distributions. This process of eliciting information from
experts is similar to the well-known Delphi method. Such meetings are structured to elicit all
expert opinion. Background information is provided to participants by using scatter plots, trend
charts, basic statistics, and histograms to summarize available data. Discussion topics include
not only the over all uncertainty of the input variables, but also the possible interrelationships and
Thin Overlay
Frequency
Relative
8 10 12 14
Performance Life, (years)
88
codependencies among the input variables. Follow-up activities may include individual
interviews with meeting participants to ensure all opinions are included. Formal surveys and
questionnaires may also be used in the process to ensure that the resulting distribution covers the
entire range of possible values. To facilitate buy-in on model results, it is important to emphasize
involving all stakeholders in this process.
The six distributions most often used to model the opinion of experts include:
Returning to the example problem, the normal distribution will be used to model the variability
for agency costs using the average and standard deviations provided in table 4.2. When using
hard data to determine the type of distribution to use, it is important to check the reasonableness
of the goodness of fit statistics associated with the selected distribution. When measured data
is not available, a triangular distribution may be used as a rough estimate of the distribution’s
shape. For the example problem, a triangular distribution will be assumed and the data on
minimum, most likely, and maximum service life data in table 4.3 will be used for the distribution
input parameters.
If nothing were known about the shape of a distribution, it might be more appropriate to use the
uniform distribution and thereby give equal weight to probability of all input values. Little is
known about the variability of the discount rate in the example problem other than it ranges from
3 to 5 percent. However, recalling the OMB discount rates shown in table 1.1, a symmetrical
triangular distribution with a minimum, most likely, and maximum of 3, 4, and 5 percent will be
used to estimate the actual distribution. The symmetry of the triangular distribution has the added
advantage of not giving advantage to a particular alternative based on the influence of higher or
89
lower discount rates. Typically, lower discount rates favor higher initial costs and lower future
costs. Higher discount rates typically favor lower initial costs and higher future costs. Table 4.4
summarizes the input distributions that will be used in the analysis.
Probability
Relative
Initial Agency std dev
Normal Alt A – Normal(26.5, 0.75)
Cost
Alt B – Normal(20,2.5)
Probability
Relative
Future Rehab std dev
Normal Alt A – Normal(7, 0.5)
Cost
Alt B – Normal(6,1)
Probability
Relative
Alt A – Triang(20,25,30)
Service Life – Triangular
Alt B – Triang(12,15,18)
Initial Constr. min max
Alt A – Triang(3,4,5)
Discount Rate Triangular
Alt B – Triang(3,4,5)
min max
90
Incorporating Probability Distributions in Spreadsheet Models
Probability distributions can be directly incorporated into spreadsheet models using custom
distribution functions available as part of the @RISK and Crystal Ball software programs. Each
of the programs includes approximately 30 additional spreadsheet functions that are add-ins to
the existing spreadsheet’s standard function set. Each of these additional functions represents a
probability distribution (NORMAL, BETA, TRIANG, etc.).
Normal(20,2.5)
std dev
91
Distribution functions such as this may be used anywhere in the spreadsheet where there is
uncertainty about the value of the variable. @RISK functions may be used the same way that
any of the other normal spreadsheet functions are used. They may be included in mathematical
expressions having cell references or be used as part of IF–THEN expressions, clarifying
logical relationships.
After the structure of the model has been established and distributions developed and
substituted for the uncertain input variables, the next step is to run a simulation of the model to
obtain results. A simulation is essentially a rigorous extension of a sensitivity analysis that uses
different randomly selected sets of values from the input probability distributions to calculate
separate discrete results. The results are arrayed in the form of a distribution covering all
possible outcomes. This process of using random numbers to sample from probability
distributions is known as Monte Carlo sampling.(24) Figure 4.8 depicts the process of sampling
from an input distribution.
Random
1.0 Number
Cumulative Probability
0.8
.65
0.6 .57
.48
0.4 .41
0.2
0.0 X .65
Values Sampled from Input Distribution
As shown in figure 4.8, a series of random numbers between 0 and 1 are generated by the
computer along the cumulative probability scale of the input distribution. Values corresponding
to each random number are sampled along the x-scale. For the example shown, when the
computer generates the random number 0.65 in figure 4.8, a corresponding value of x.65 is
sampled. The sampled value is then combined with other distribution samples to compute a
single result. It is important to note that the computer uses a uniform distribution to generate the
random numbers and all values along the cumulative scale of the y-axis have equal probability of
being selected. Therefore, x-axis values corresponding to portions of the distribution curve
where the slope is steeper (more vertical) have a greater likelihood of being sampled compared
to x-axis values that correspond to portions of the curve with flatter slopes.
92
In a simulation, each iteration represents a possible scenario or outcome. The results of each
iteration are captured, compiled, and subjected to rigorous statistical analysis. This process of
sampling from a probability distribution is repeated until the specified number of iterations are
completed or until the simulation process converges (i.e., the point at which additional iterations
do not significantly change the output distribution).
Simulation techniques, such as the Monte Carlo, typically require a large number of iterations to
ensure sufficient opportunity to sample low probability values. This is especially true when highly
skewed distributions are used to describe the input variables. When the number of iterations
performed is low, clustering can occur. In this situation, sampled values are tightly clustered
around high probability outcomes and the low probability outcomes may not be adequately
sampled. Clustering is particularly important when low probability outcomes can have an
inordinately serious effect on results. Such outcomes must be accounted for in the simulation,
but, to do this, they must be sampled. This has led to the development of a sampling technique
known as Latin Hypercube, which forces a more representative sampling with a much lower
number of iterations. (25,26,27)
Latin Hypercube is a stratified sampling technique where the probability scale of the cumulative
distribution curve is divided into an equal number of probability ranges. The number of ranges
used is equal to the number of iterations performed in the simulation.
Figure 4.9 illustrates Latin Hypercube sampling, in this case, showing four iterations, and
therefore four distinct divisions of the cumulative probability scale. As shown in figure 4.9, within
a division (in this case, division number 2), a random number between 0 and 1 is generated and
serves as the basis for selecting the input value. In the case shown, 0.4 is generated and a
1.0
Random
1 Number
Cumulative Probability
0.8
1
0.6
2 .40 Random Number
Between 0 and 1
0
0.4
3
0.2
4
0.0
X.40
Values Sampled from Input Distribution
Figure 4.9. Latin Hypercube sampling showing four iterations.
93
corresponding value is sampled from the distribution. In future iterations, the previously sampled
section is not sampled again. Because the Latin Hypercube sampling technique memorizes the
sections it sampled, it therefore achieves convergence much more quickly when compared to
the Monte Carlo. In addition to being more efficient, this technique ensures low probability
values are sampled and included in the simulation.
Figures 4.10 and 4.11 compare the effect of Monte Carlo and Latin Hypercube sampling when
the number of iterations is limited to 100. As shown, 100 samples were randomly drawn from
the normal distribution function for Initial Agency Cost Alternative B — Normal(20, 2.5).
Figure 4.11 shows that the stratified Latin Hypercube sampling technique achieves a distribution
Normal(20,2.5)
0.14
Relative Probability
.4
.4
.4
.4
.4
.4
.4
.4
.4
16
17
18
19
20
21
22
23
24
25
Normal(20,2.5)
0.14
Latin Hypercube Sampling
Relative Probability
0.12
0.1
0.08
0.06
0.04
0.02
0
12.7 14.0 15.4 16.8 18.2 19.5 20.9 22.3 23.6 25.0
94
that more resembles the familiar bell-shaped normal curve when compared to pure random
Monte Carlo sampling. Because of stratified sampling, it is possible to achieve convergence in
fewer iterations as compared to the Monte Carlo simulation.
A cardinal rule of risk analysis modeling is every iteration of a risk analysis simulation must
be a scenario that can actually occur. Failure to recognize the dependent relationship
between input variables is usually a prime source of error in simulation modeling. Suppose, for
example, that as part of a risk model, probability distributions describe both the random
variability for AADT and pavement service life. Typically, pavement engineers have related
shorter pavement life to higher traffic volumes. Therefore, it would be irrational for a simulation
model to sample the high side of the traffic probability distribution curve, and at the same time,
sample the high side for pavement life. Formal treatment of this relationship between AADT and
pavement service life is usually accounted for in the simulation by using a correlation matrix.
Given that user costs are omitted from the example problem, the number of variables has been
greatly reduced and the analyst does not have to correlate variables. Another source of error is
sampling input distributions outside reasonable bounds. For example, sampling from a normal
distribution for project cost may produce misleading results, particularly when the values
sampled are negative. The use of logical statements as well as establishing bounds for the
distribution may be used to overcome errors such as this.
Figure 4.7, previously shown, is a spreadsheet that shows the results of 10,000 iterations
processed using a Microsoft Excel @RISK model.(28) The simulation run time was 47 seconds
using a Pentium 166 MHz Computer with 80 MB RAM. Before analyzing the simulation results,
it is important to test the assumptions to make certain the model is robust across the entire range
of possible values for the uncertain inputs. The model should be tested as each component is
added to make certain the results are reasonable. This is especially true for large and complex
models. A visual tool to examine a model during a simulation is to construct a plot of the
performance curves and cash flows. @RISK will allow a simulation to be executed one step at
a time. By stepping through the simulation and viewing the changing shape of the output graphs
in real time, it is possible to detect errors that may exist in the model.
Another tool that may be used to test a risk model is to incorporate the use of a random number
generator seed. A seed value is a value that is used to start the sampling process and all
subsequent random numbers will rely on this value. Providing that the model is not changed, a
second run of the simulation using the same seed value will produce exactly the same results.
This can be very useful in testing the effect of varying distributions on risk results. By using a
seed value, the analyst can be certain that any change in the result is caused by changes in the
model and is not a result of randomness of the sampling. After the model is tested and a
simulation is performed, the next step is to analyze and interpret the results.
95
Step 4. Analyze and Interpret Results
If the analysis had been conducted according to the traditional deterministic approach, all that
would be available to base the decision on would be the means of the output distributions:
Alternative A $28.79 million and Alternative B $27.32 million, respectively. Based on the NPV
means, it is readily apparent that Alternative B is less than Alternative A by $1.47 million
(~ 4.8 percent). Some SHAs would probably consider this difference in means insignificant and
may exclude lowest NPV as a consideration for selecting the alternative. Without an analysis of
the variability about the mean, a decision such as this may prove to be a poor choice —
depending on the decision maker’s tolerance for risk.
Interpretation of risk analysis results goes beyond a simple comparison of which alternative on
average costs less by including an analysis of the likelihood that any particular outcome will
occur. There is no presumption that any particular alternative is better. Figure 4.12 shows the
risk profile of the NPV for Alternatives A and B in histogram form, where the probability is the
area under the curve. As Figure 4.12 shows, the entire range of conceivable outcomes is
0.6
0.5
Relative Probability
0.4
0.3
0.2
0.1
0
18 23 28 33 38
Net Present Value ($ Millions)
Alternative A Alternative B
arrayed with the estimated probability of each outcome actually occurring. The main advantage
of the histogram is that it readily shows the variability about the mean. The wider the distribution,
the greater the variability. As shown, the outcome for Alternative B is more uncertain than
Alternative A.
96
In interpreting the risk profile in figure 4.12, it is important to distinguish between upside risk
and downside risk. Downside risk for project cost implies cost overrun — chance of financial
failure. Upside risk for project costs implies cost underrun — opportunity for low cost. As
figure 4.12 shows, Alternative B has greater upside risk compared to Alternative A. However,
as discussed below, it is important to quantify the probability of cost overrun for Alternative B.
Figure 4.13 shows the risk profiles for Alternatives A and B in cumulative form. As shown,
there is a 60 percent probability that project costs for Alternative B will be less than
1
Relative Cumulative Probability
0.8
60%
0.6
0.4
0.2
$28.27
0
18 23 28 33 38
Net Present Value ($ Millions)
Alternative A Alternative B
$28.27 million. This means that for the 10,000 iterations that were processed, 60 percent of the
calculated values for NPV were less than $28.27 million. The variability for the proposed
alternative is inversely proportional to the slope of the cumulative curve. In other words, the
steeper the slope, the less variability. The flatter the slope, the greater the variability. As shown,
the slope for Alternative B is flatter than that for Alternative A, and is therefore more variable.
A risk analysis provides much more information than a simple deterministic solution. As table
4.5 shows, additional information comes in the form of basic statistical measures of simulation
results that reveals the underlying uncertainty associated with each alternative. In interpreting the
risk involved with each alternative, it is important to identify the magnitude of the extremes of the
distributions shown in figure 4.12. As table 4.5 shows, Alternatives A and B have minimums of
$25.4 and $13.3 million and maximums of $33.04 and $40.35 million, respectively. The
standard deviation for Alternative B is much greater than Alternative A. Analysis of the
distribution tails reveals a 10 percent probability that the NPV of Alternative B will be less than
97
Table 4.5. Risk profile statistics for Alternatives A and B.
As part of the risk assessment, a sensitivity analysis can be performed on simulation results to
identify significant input variables that are important in determining the output distributions. The
results of this analysis are usually displayed in the form of a Tornado plot, as shown in figures
4.14 and 4.15. The higher the correlation coefficient, the more significant the input variable is on
determining the results. The variables listed at the top of the graph are more significant than
those at the bottom. The degree of correlation may be calculated using either the rank order
correlation or stepwise least squares regression. Rank order correlation makes no presumption
98
Initial Agency Cost
Discount Rate
Discount Rate
about the relationship between the input and output variables. Least squares regression assumes
a linear relationship between input and output variables. This is important to note because
models that incorporate divisions and power functions often violate this later assumption. Both
figures 4.14 and 4.15 use rank order correlation. Typically, correlation coefficients less than
about 0.6 are not significant.
Figure 4.14 shows Initial Agency Cost has a correlation coefficient of 0.72. This means that
if Initial Agency Cost moves one standard deviation (in either direction), then the NPV for
Alternative A will move 0.72 of a standard deviation in the same direction. If Design Period/
99
Life Initial Construction moves one standard deviation (in either direction), then expectations
are that the NPV for Alternative A will move 0.52 standard deviations in the opposite direction,
because the relationship is reversed as indicated by the negative correlation coefficient.
To further the evaluation, the analyst can explore the low probability area of the outcome
distribution curve for Alternative B, where the NPV is greater than that for Alternative A,
(i.e., greater than the 90th percentile). Figure 4.16 illustrates the need to identify the key input
variables that produce low probability scenarios — the extremes, or tails, of the results
distribution for Alternative B.
1
Relative Cumulative Probability
0.8
0.6
Which Inputs
0.4 “Drive” the Shaded
Tail of Alt-B?
0.2
0
18 23 28 33 38
Net Present Value ($ Millions)
Alternative A Alternative B
A scenario analysis can identify those variables that may cause a project to have significant
cost overrun. The analysis is easily accomplished using Excel/@RISK, which uses the following
procedure to identify significant inputs for a particular scenario.
100
Table 4.6 identifies the significant inputs in the upper 10 percent of the distribution for
Alternative B as Design Period/Life Initial Construction, Agency Cost Initial Construction,
and Agency Cost Rehabilitation. Now that the drivers in the tail are known, the decision
maker may choose to take some mitigating action against these significant inputs to reduce
exposure to upside risk. For example, the decision maker may choose to aggressively control
agency costs.
In order to make a decision based on risk analysis results, it is important for the decision maker
to define the level of risk the organization can tolerate. Decision makers who can tolerate little
risk prefer a small spread in possible results, with most of the probability associated with
desirable results. Conversely, if decision makers are risk-takers, then they will accept a greater
amount of spread, or possible variation in the outcome distribution. Most decision makers can
reach a consensus decision after weighing the probability for upside and downside risk. In our
example, clearly Alternative B appears to be the better alternative since there is far greater
likelihood of cost savings compared to Alternative A. Also, the probability of cost overrun,
compared to Alternative A, appears to be quite low — less than 30 percent.
The reason that this step is called a consensus decision is because all stakeholders were
involved in developing the input probability distributions early in the process. As a result, this
overall process should diffuse any negative reaction special interest groups may have with the
outcome. The result is that the entire risk analysis process facilitates consensus building among
stakeholders so that action, in the best public interest, may be taken.
A risk analysis is of little value if the results cannot be understood. Thus, this section offers
practical advice in presenting risk analysis results to decision makers. The first step in presenting
risk analysis results is to know the audience. Here are some basic questions to ask prior to
presenting the results.
Does the audience need a risk primer? One of the primary reasons people have difficulty
interpreting risk analysis results is they often have little understanding of basic statistics which is a
101
prerequisite to interpreting risk results. Therefore, it may be necessary to provide an
introductory session to train decision makers on fundamental risk analysis concepts such as
Monte Carlo sampling and probability distributions before presenting risk analysis results. The
purpose of such a session is to provide just enough information so the decision maker can
understand and interpret risk analysis results. There is no reason to include a formal treatment of
the statistical theory underlying the concepts.
Does the audience buy in to the risk analysis approach? Although simulation techniques
have been around since the early 1940s, it is only recently that these techniques have been
employed in the spreadsheet world of the analyst. Risk analysis is therefore a relatively new way
of treating uncertainty and involves an entirely new approach to understand and interpret the
results. Risk analysis is an analytical tool to aid the decision maker to select the best alternative.
By exposing previously hidden areas of uncertainty critical information is revealed thereby
providing the decision maker the opportunity to take mitigating action to decrease exposure to
risk. Most importantly, risk analysis provides those vested with the appropriate authority, namely
executives and elected officials, the opportunity to make decisions about risk taking. As a
compliment to LCCA, risk analysis elevates the debate—from the validity of LCCA results—
to taking action that is in the best interest of the general public.
Does the audience buy-in to the analysis? Anyone examining risk analysis results should be
cautious. First and foremost is the model correct? There is nothing worse than making high
stakes decisions based on a model that is wrong. It is important that the structure and logic of
the model be verified by an independent party, particularly, if risk analysis becomes an integral
part of the decision making process.
Does the model follow generally accepted procedures? For example, there are a number
of different procedures available to determine user costs. Are the procedures that are employed
documented and do stakeholders, in general, accept the procedures? This is important to
obtain buy-in on risk results. How were the input distributions determined? Were stakeholders
involved in the process? Does the range for the distributions fall within a generally accepted
range? Finally, the level of detail in a risk analysis report should be commensurate with the
problem at hand.
The presenter of a risk analysis report should be able to answer key questions regarding the risk
analysis results. In the presentation, it is recommended that the developer of the risk model be
present to answer key questions about the structure and logic of the model. In presenting risk
analysis results it is important not to overburden the audience with statistics. A simple 1-page
summary to include summary statistics, histogram and cumulative distributions, and perhaps a
Tornado graph to show the important inputs in the analysis should be sufficient. When conveying
probability graphs to upper management it is highly recommended that either cumulative ascend-
ing or cumulative descending graphs be presented. The presenter should choose a presentation
method and stay with it. Structure and logic charts are also quite useful in presenting the simula-
tion process as well as discussing judgements about the model relationships. Place model
assumptions at the back of the report. These are important, but not so important as to be in the
summary sheet.
102
REFERENCES
1
Highway Infrastructure Report, Office of Inspector General, Government Accounting Office,
1994.
2
Federal Register, Code of Federal Regulations, Part 627, Vol.62, No.31, February 14, 1997,
p. 6866.
3
Office of Management and Budget Circular No.A-94, October 29, 1992.
4
LCC Final Policy Statement, Federal Register, Vol.61, No.182, September 18, 1996,
p. 35404.
5
James W. Mack, Life Cycle Costing: A Discussion Regarding Its Use for Comparing
Alternative Designs, Appendix A.
6
Pavement Rehabilitation Manual, Federal Highway Administration, Washington, DC, 1990.
7
Transfund Project Evaluation Manual, The World Bank, May 1997.
8
Bosher Al-Omai and Michael I. Darter, Relationship Between IRI and PSR, Civil Engineering
studies No. 60, Transportation in Engineering Series, University of Illinois, Champaign-Urbana,
September 1992.
9
Arnim Meyburg et al., Review and Development of Life-Cycle Cost and Network Analysis
Procedures for NYSDOT Highway Pavements, Draft, August 1997.
10
NCHRP Report 133, Procedures for Estimating Highway User Costs, Air Pollution, and
Noise Effects, 1972.
11
Robly Winfrey, Economic Analysis for Highways, International Textbook Company, 1969.
12
Nationwide Personal Transportation Survey, Bureau of Transportation Statistics in
Cooperation with the Federal Highway Administration, 1990.
13
Design for Pavement Structures, American Association of State Highway and
Transportation Officials, Washington, DC, 1993.
14
Special Report 209, Highway Capacity Manual, 3rd ed., Transportation Research Board,
National Research Council, Washington DC, 1994.
15
Traffic Monitoring Guide, FHWA, 3rd ed., February 1995.
16
Highway Capacity Manual, p. 6-11.
103
17
1995 Highway Statistics, Federal Highway Administration, Publication No. FHWA-PL-96-
017, November 1996.
18
Construction Cost and Safety Impacts of Work Zone Traffic Control Strategies, FHWA
Publication No. RD-89-210, December 1989.
19
1996 Fatality Analysis Reporting System, National Highway Traffic Safety Administration,
DOT-HS-82649, December 1997.
20
An Evaluation of Lane Closure Strategies for Interstate Work Zones, Final Report for
the Joint Highway Research Project, FHWA/JHRP-95/1, November 1995.
21
Design and Operation of Work Zone Traffic Control, National Highway Institute
Course #38003, pp 2-7.
22
David Lewis, “The Future of Forecasting—Risk Analysis as a Philosophy of Transportation
Planning, TR News 177, April 1995.
23
Ibid.
24
R.Y. Rubinstein, Simulation and the Monte Carlo Methods: John Wiley and Sons, New
York, NY, 1981.
25
R.L. Inman, J.M. Davenport, and D.K. Zeigler, Latin Hypercube Sampling (A Program
Users Guide), Technical Report SAND79-1473, Sandia Laboratories, Albuquerque, NM,
1980.
26
R.A. Startzman and R.A. Wattenbarger, An Improved Computation Procedure for Risk
Analysis Problems With Unusual Probability Functions: SPE Hydrocarbon Economics and
Evaluation Symposium Proceedings, Dallas, TX, 1985.
27
R.L. Inman, W.J. Conover, and R.J. Beckman, A Comparison of Three Methods for
Selecting Values of Input Variables in the Analysis of Output from a Computer Code,
Technometrics, 1979, p. 211, 239-245.
28
David Vose, Quantitative Risk Analysis: A Guide to Monte Carlo Simulation Modelling,
John Wiley & Sons, New York, July 1997.
104
APPENDIX: RESOURCES
PUBLICATIONS
Federal Highway Administration
Construction Costs and Safety Impacts of Work Zone Traffic Control Strategies, Vol.2:
Information Guide, FHWA, Publication No. FHWA-RD-89-210, December 1989.
FHWA Highway Pavements Course, Chapter 14-1, National Highway Institute Course
No.13114, Publication No. FHWA-HI-90-028, May 1990.
James S. Gillespie, FHWA/VTRC 98-R12, Estimating Road User Costs as a Basis for
Incentive/Disincentive Amounts in Highway Construction Contracts, February 1998.
LCC Interim Policy Statement, Federal Register Vol.59, No.131, July 11, 1994, p. 35404.
Matt Witczak, Ph.D., Probabilistic-Based MCP (Micro-Computer Program) for LCCA for
Asphalt Concrete Pavements, University of Maryland at College Park, May 1994.
Motor Vehicle Accident Costs, Technical Advisory, FHWA, October 31, 1994.
Ratkim Pal and Kumares C. Sinha, An Evaluation of Lane Closure Strategies for Interstate
Work Zones, Joint Highway Research Project, FHWA/INJHRP-95/1, Purdue University/
Indiana Department of Transportation, November 1995.
Searching for Solutions; A Policy Discussion Series, No.16, Exploring the Application of
Benefit/Cost Methodologies to Transportation Infrastructure Decision Making, (May 1995
Conference Proceedings), July 1996.
Project reports are issued at the conclusion of each NCHRP research project.
Report 133 Procedures for Estimating Highway User Costs, Air Pollution, and Noise
Effects, 1972.
Synthesis publications are a series of reports of studies on current practice in the field of
highway transportation. Projects are sponsored by AASHTO.
105
Synthesis Reports 110 and 77 address future maintenance needs and costs.
Synthesis Report 142: Methods of Cost Effectiveness Analysis for Highway Projects,
1988.
Synthesis Project 20-5, Topic 27-12 Road User and Mitigation Costs in Highway
Pavement Projects, Final Draft, November 1997.
Research Projects
NCHRP 1-34 Guidelines for Subsurface Drainage; Final Report, February 1998.
Project 2-18 FY91 Research Strategies for Improving Highway User Cost-Estimating
Methodologies; Hickling, Louis, Brod, April 1994.
AASHTO
A Manual on User Benefit Analysis of Highway and Bus Transit Improvement, 1977.
Anthony Boardman, Aidan Vining, and W.G. Waters II, Cost and Benefits through
Bureaucratic Lens: Example of a Highway Project.
106
COMPUTER SOFTWARE
McTrans and PC Trans are microcomputer software distribution centers for transportation-
related software, including public domain software such as MicroBENCOST and QueWZ,
which were developed with Federal funds. For information, contact:
McTrans PC Trans
phone: 352-392-0378 phone: 785-864-5658
fax: 352-392-3224 fax: 785-864-3199
Web site: www.mctransce.ufl.edu Web site: www.kuhuh.cc.ukans.edu./~pctrans.
email: [email protected] email: [email protected].
Other Software
There are several powerful microcomputer-based risk analysis software programs currently on
the market that work well in conjunction with Lotus and Microsoft® Excel spreadsheet
applications. Authors of this Technical Bulletin have used @Risk and Crystal Ball.
For information on DataPave, contact FHWA Customer Support 423-481-2967. Web site is:
http://DataPave.fhwa.dot.gov, http://www.LTPPDatabase.com/main.htm,
email:[email protected]
107