Professional Documents
Culture Documents
Risk Matrices Implied Accuracy and False Assumptions PDF
Risk Matrices Implied Accuracy and False Assumptions PDF
Abstract
Risk matrices are used during hazard identification and risk assessment processes and provide a construct for people
needing to display the two variable relationship between likelihood and consequence that are considered to be the
elements of risk. The purpose of a matrix is to reduce the continuum of risk into ranges or bands such as high,
medium or low. These bands are often allocated colours such as red for the highest risks to green for the lowest.
Sometimes each band in a matrix is allocated a numerical value or range. The multiplication of likelihood and
consequence implies a quantitative basis although it may not be widely understood. The multiplication operator
produces lines of equal risk that a matrix cannot model accurately and thereby introduces risk reversal errors.
Weaknesses in matrices are further compounded by subjectivity and bias introduced by users and the value of such
tools is brought into doubt. A shift of emphasis from the risk assessment stage to the risk control stage of a hazard
management process may lead to better and more timely decision making and better use of resources.
Introduction
Risk matrices are very commonly used during hazard
identification and risk assessment processes (Cook
2008). They are used to: articulate the level of risk
associated with an identified hazard; to rank risks
and thereby propose actions; to justify a proposal
or action; and to re-assess risk to demonstrate the
effectiveness of a control (residual risk) (Cook 2008;
Cox 2008; Smith, Siefert and Drain 2008). Risk
1VIOSH Australia, University matrices provide a construct for people needing
of Ballarat, University Drive, to display the two variable relationship between
Mt Helen, Victoria, Australia likelihood and consequence that are considered to be
the elements of risk (Standards Australia 2004).
Correspondence; Alexander
A Risk Matrix is a tool used to allocate a level of
Pickering, c/o VIOSH Australia,
risk to a hazard from a pre-defined set. An example is
University of Ballarat, PO Box
663, Ballarat, Victoria 3350. shown in Figure 1. Two dimensional matrices are most
common but not exclusive (Hewett, Quinn, Whitehead
Key words and Flynn 2004) and are lauded as simple, effective
Matrix, risk, likelihood, approaches to risk management (Cox 2008). They
consequence, estimation, bias are used in many countries (Papadakis and Chalkidou
2008) and promoted through international standards
(Standards Australia 2004; Cook 2008).
Consequence
volume 2 issue 1 october 2010 Journal of Health & Safety Research & Practice 9
Risk Matrices: implied accuracy and false assumptions
10 Journal of Health & Safety Research & Practice volume 2 issue 1 october 2010
Risk Matrices: implied accuracy and false assumptions
or levels. The number of risk categories If the numerical value of both likelihood
within a matrix is determined by the and consequence are known, then the
organisational requirement for specific quantitative measure of risk is also known
actions with respect to the risk category based on R = L x C. In this case, a Risk
(Smith et al., 2008, p. 2). For example, Matrix is not required to rank hazards as
within a matrix having three categories this will be self evident.
of risk, the organisation may dictate Consequence values in quantitative
that work must cease when a hazard matrices are often represented by
is categorised as high-risk but proceed ranges because they are dependent on
when categorised as low-risk. Some conditional factors. This lack of point
predetermined actions may be required value is considered to be a weakness
if the risk is categorised as moderate. (Smith, Siefert et al. 2008). Establishing
Within a 5 x 5 matrix having five risk levels this point value through accuracy in the
(for example, low, moderate, high, very estimation of likelihood and consequence
high and extreme) a range of additional is impractical in most cases. Despite it
actions may be included. Risk matrices representing objectivity, the expense in
with too few categories may suffer range time and resources for investigation,
compression, where risks with significant testing and analysis exceeds the capability
variation in likelihood and or consequence of the organisation and the time frame of
might become grouped into the same the project (Smith, Siefert et al. 2008).
category (Cox, 2009, p. 101) (Hubbard,
2009, p.130).
The parameters applied to the x and Matrix Use and interpretation
y axes also vary and some matrices The cell at the intersection of a row and
illustrate risk increasing from left to right a column that respectively represent
and bottom to top. Others represent the the chosen likelihood and consequence
reverse with increasing risk towards the values signifies a discrete risk category
left or top down (Alp, 2004, p. 36). or score and therefore the boundaries
Some matrices are purely qualitative between the cells imply that each cell is
and use words to express likelihood categorical rather than a position on a
and consequence (Bender, 2004, p. 2) risk continuum. However, if R = L x C
(Standards Australia 2004). Qualitative (Donoghue 2001; Standards Australia
analysis is used when quantitative data is 2004; Cox 2008; Smith, Siefert et al.
not available or when the more onerous 2008) then points of equal risk plotted
quantitative methods are impractical. on a matrix form curved lines of the form
(Standards Australia 2004). y=R/x. Figure 2 shows lines of equal risk
Semi-quantitative and quantitative risk for arbitrary and dimensionless values
matrices incorporate in the likelihood of risk (R) increasing at 0.1 intervals
or consequence arguments, data derived between 0.1 and 0.9, superimposed onto
from injury statistics or epidemiological a 5 x 5 matrix. Figure 2 shows the non-
studies, for example. Use of historic data linearity of points of equal risk, that they
may however be problematic as incident do not align themselves with the cells
rates vary over time and data collection or their boundaries and they bisect the
may be biased (Donoghue 2001; Gadd, cells asymmetrically. Thus the equal risk
Keeley and Balmforth 2004; Hopkins curves divide risk categories and render
2004; Hopkins 2005; Smith, Siefert et the plot of the likelihood and consequence
al. 2008). The number of incidents and estimations ambiguous. Changing the
injuries within organisations is usually too position of grid lines or number of rows/
low to provide a basis for quantification of columns does not eliminate the problem
risk (Health and Safety Executive 2001). (Cox 2008).
volume 2 issue 1 october 2010 Journal of Health & Safety Research & Practice 11
Risk Matrices: implied accuracy and false assumptions
0.8
do 0.6
oh
il
ek
i 0.4
L
0.2
0
0 0.2 0.4 0.6 0.8 1
Consequence
Figure 2 Risk matrix showing lines of Figure 3 Equal distribution of risk categories
equal risk conforming to y=1/x
12 Journal of Health & Safety Research & Practice volume 2 issue 1 october 2010
Risk Matrices: implied accuracy and false assumptions
Consequence
volume 2 issue 1 october 2010 Journal of Health & Safety Research & Practice 13
Risk Matrices: implied accuracy and false assumptions
14 Journal of Health & Safety Research & Practice volume 2 issue 1 october 2010
Risk Matrices: implied accuracy and false assumptions
volume 2 issue 1 october 2010 Journal of Health & Safety Research & Practice 15
Risk Matrices: implied accuracy and false assumptions
Canberra, National Research Centre (2008). Risk matrix input data biases.
for Occupational Health and Safety Systems Engineering 1-17.
Regulation. Working Paper 25.
Standards Australia (2004). HB436:2004
Hopkins, A. (2005). Safety, Culture and Handbook: Risk Management Guidelines:
Risk: The Organisational Causes of Companion to AS/NZS4360:2004.
Disasters. Sydney, CCH Australia Limited. Sydney, Standards Australia.
Hubbard, D. W. (2009). The Failure of Risk Standards Australia (2009). AS/NZS ISO
Management: Why Its Broken and How to 31000:2009 Risk managementPrinciples
Fix It. New Jersey, John Wiley & Sons Inc. and guidelines. Sydney, Stasndards Australia.
Iannacchione, A., F. Varley, et al. (2008). Stoklosa, R. T. (1999). Practical application
The Application of Major Hazard Risk of environmental risk management
Assessment (MHRA) to Eliminate Multiple Gorgon LNG Project case study.
Fatality Occurrences in the U.S. Minerals The APPEA Journal 39(1), 606.
Industry. Spokane, National Institute
Swedish Work Environment Fund
for Occupational Safety and Health.
(ASF) (1988). To Measure or Take
Kennedy, A. (1997). Technical Note: Risk Direct Remedial Action? Stockholm,
Management and Assessment for Rolling Swedish Work Environment Fund.
Stock Safety Cases. Proceedings of the
Tversky, A. and D. Kahneman (1992).
Institution of Mechanical Engineers, Part F,
Advances in Prospect Theory: Cumulative
Journal of Rail & Rapid Transit 211(1), 67-72.
Representation of Uncertainty. Journal of
Martin, P. and R. Pierce (2002). Statistics for Risk and Uncertainty 5(4), 297-323.
starters. Buninyong, Six Sigma Publishing.
Viner, D. (1996). Accident Analysis and Risk
McIlwain, J. C. (2006). A review: a decade Control. Melbourne, Derek Viner Pty Ltd.
of clinical risk management and risk tools.
Viner, D. (2003). What are we talking about?
Clinician in Management 14(4), 189-199.
Language and Paradigms for the Safety
Md-Nor, Z., V. Kecojevic, et al. (2008). Professional. Safety In Australia, 25, 6-9
Risk assessment for loader- and dozer-
related fatal incidents in U.S. mining.
International Journal of Injury Control
& Safety Promotion 15(2), 65-75.
Middleton, M. and A. Franks (2001). Using
Risk Matrices. The Chemical Engineer 34-37.
Moses, K. D. and R. W. Malone,
Development of Risk Assessment
Matrix for NASA Engineering and Safety
Center, NASA Marshall Space Flight.
Papadakis, G. A. and A. A. Chalkidou
(2008). The exposuredamage approach
in the quantification of occupational
risk in workplaces involving dangerous
substances. Occupational Safety and
Risk at ESREL 2006 46(6), 972-991.
Payne, S. L. (1951). The Art of
Asking Questions. New Jersey,
Princeton University Press.
Rowe, W. (1988). An Anatomy of
Risk. Malabar, Robert E. Kreiger.
Smith, E. D., Siefert, W. T., Drain, D.,
16 Journal of Health & Safety Research & Practice volume 2 issue 1 october 2010