USING RISK ANALYSIS TO DETERMINE CONSTRUCTION

PROJECT CONTINGENCIES
By Stephen Mak1 and David Picken2

ABSTRACT: A contingency allowance is an amount of money used to provide for uncertainties associated with
a construction project. Traditionally, it is a percentage addition on top of the base estimate. Estimating using
risk analysis (ERA) is a methodology that can be used to substantiate the contingency by identifying uncertainties
and estimating their financial implications. A study of the effect of ERA was carried out to compare the variability
and consistency of the contingency estimates between non-ERA and ERA projects. This paper presents results
of a survey that compares a total of 287 non-ERA and 45 ERA projects. The results show a highly significant
difference in variation and consistency between these groups. It indicates successful use of the ERA method for
public works projects to reduce unnecessary and exaggerated allowance for risk. However, the contingency
allowance for ERA projects was still considered high. Improvement and refinement of the ERA method as well
as recommendations on capital budgeting policy are suggested.

INTRODUCTION single lump sum with no attempt made to identify, describe,

and value various categories and possible areas of uncertainty
Risk has long been recognized in the construction industry. and risk. Often the contingency amount allowed is merely a
Contractors are required to accept a certain level of risk due percentage of the overall project cost. In many cases it
to unforeseen costs that they incur during construction. Risk amounts to an educated guess at best. If there is some form
is also an issue for clients. The term ‘‘clients’’ refers to those of tender documentation provided to bidders, the contingency
persons or organizations investing in the construction of built will usually be transferred to the provisional sums section in
facilities. This risk manifests itself in unforeseen expenditure these documents. In an attempt to deal with the determination
that was not envisaged at the planning stage. It is the risk to of contingencies in a more analytical way the Hong Kong
clients that this paper deals with. When cost consultants are Government implemented a technique called Estimating using
preparing feasibility and subsequent planning stage estimates, Risk Analysis (ERA) in 1993. ERA produces similar base plus
risk can be reflected by the inclusion of a contingency sum. contingency cost estimates at pretender stage. For building
The contingency sum, usually expressed as a percentage projects that usually use the government’s fixed quantities con-
markup on the base estimate, is used in an attmept to allow tract, the magnitude of the final account variations (comprising
for the unexpected. Construction and development is fraught additions and omissions) can be compared with the contingen-
with difficulty, and the basic notion of risk analysis is that it cies included in estimates. This comparison can be used to
is useful to at least make an attempt to identify these risky assess the accuracy of the allowance made for the contingen-
items and attach some financial value to them. These amounts cies at the planning stages. This paper compares the contin-
can then be added to a project budget as items of possible gency estimates and final account variations of public works
expenditure. The intention is that the project budget becomes projects by analyzing data sets of pre-1993 (non-ERA) and
a more realistic representation of the client’s likely outlay. post-1993 (ERA) projects.
At different stages of project development, there are differ- With the traditional approach of guessing at contingencies
ent types of risk. At the feasibility and inception stages, for in mind, it can then be perceived that project proposals (what
example, the client might not have decided exactly on the floor it is intended to build; for example, the size and the level of
area that is required, or an amount of additional floor area over quality) will be affected by the contingency amount. If one
that in a basic scheme may be in abeyance. Such matters will takes a wider view than just one project, for example, in the
represent an uncertainty from the estimator’s point of view. case of public expenditure, the task becomes one of distrib-
The depth of piling for foundations is another typical example uting the available funds among various competing projects.
of uncertainty. Some of the uncertainties will be eliminated or The situation will arise that some proposed facilities are not
clarified as the planning of the project develops toward de- included in a construction program due to the limited funds.
tailed design stage when, for example, the client has decided Having chosen those projects that will be included, the bud-
the floor area required. Some uncertainties will be carried for- geted capital expenditure becomes committed and ‘‘locked
ward to tender stage. The use of risk premium money is re- up’’ for the duration of the program. Sometimes project teams
garded as standard practice in construction (Raftery 1994). The
can inflate the contingency allowances in an attempt to avoid
practice of presenting project cost estimates as a deterministic
the need to seek additional funds if budgets become overspent.
figure comprising a base estimate and the addition of a single
However, this inflation can become exaggerated. In such a case
contingency amount (usually as a percentage addition) has
if, ultimately, there are no heavy calls on the contingency fund
been adopted in the construction industry for a long time for
beyond what might reasonably have been expected for the
budgeting purposes. Usual practice is for this amount to be a
project, budgets can be seriously underspent. The magnitude
1
of the underspending can be so large that it is possible to
Asst. Prof., Dept. of Build. and Real Estate, Hong Kong Polytechnic
Univ., Kowloon, Hong Kong.
identify facilities that were previously foregone but could have
2
Sr. Lect., School of Arch. and Build., Deakin Univ., Geelong 3217, been included in a construction program in the first place. Ex-
Australia. plaining situations such as this can be just as embarrassing for
Note. Discussion open until September 1, 2000. To extend the closing project teams as seeking additional funds because of budgets
date one month, a written request must be filed with the ASCE Manager being inadequate to meet unforeseen costs.
of Journals. The manuscript for this paper was submitted for review and It is these matters that prompted the Hong Kong Govern-
possible publication on August 17, 1998. This paper is part of the Journal
of Construction Engineering and Management, Vol. 126, No. 2, March/ ment to seek alternative methods of dealing with risk and un-
April, 2000. 䉷ASCE, ISSN 0733-9634/00/0002-0130–0136/$8.00 ⫹ certainty in its capital cost estimating. The following sections
$.50 per page. Paper No. 19034. describe the background to this search for a new methodology.
130 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000
The issues as perceived by the related government departments • The percentage figure is, most likely, arbitrarily arrived at
are outlined, and the methodology eventually adopted is de- and not appropriate for the specific project.
scribed. In addition, this paper describes an investigation of • There is a tendency to double count risk because some
the Hong Kong Government’s experience in using its new ap- estimators are inclined to include contingencies in their
proach. best estimate.
• A percentage addition still results in a single-figure pre-
CAPITAL COST BUDGETING diction of estimated cost, implying a degree of certainty
The Hong Kong Government’s Works Branch gives the that is simply not justified.
background to the issue of risk, and how it should be dealt • The percentage added indicates the potential for detri-
with, in a circular to the various departments responsible for mental or downside risk; it does not indicate any potential
the development and construction of buildings and infrastruc- for cost reduction and may therefore hide poor manage-
ture (Works Branch 1992). The view of financial risk is that ment of the execution of the project.
it emanates from uncertainty about the implications on cost of • Because the percentage allows for all risk in terms of a
the design team not having complete information. There can cost contingency, it tends to direct attention away from
be uncertainty both about the scope of the feature in question time, performance, and quality risks.
and, flowing from that, its value. • It does not encourage creativity in estimating practice, al-
Risks can arise from planning decisions, where the outcome lowing it to become routine and mundane, which can
cannot be adequately costed. The majority of risks, however, propagate oversights.
arise from matters yet to be decided (e.g., incomplete brief or
no site investigation). Risk assessment of the cost of uncertain There are a number of methods in risk analysis practice that
features will be present at all stages of a project. The number are used to deal with uncertainty. For budgeting purposes,
of risks will normally decrease as a project progresses through however, rather than indicate a range of figures, there is a need
the various planning stages of the Public Works Program. Ob- to present an estimate as a fixed amount so that the client can
viously, other risks can arise until planning is complete and arrange for the financing of the project. The estimate is also
during the construction phase. important in reaching a decision on whether to proceed with
Estimates are prepared at various stages of the planning pro- a project, given that the client has only a fixed budget.
cess. The specific stages are identified as Feasibility, Category There is a tendency for estimators to include an inflated
C, Category B, and Category A. As a project becomes more buffer in the contingency estimate. Raftery (1994) has identi-
definite (likely to go forward) its category designation is fied personal bias and differences in personal risk attitude.
changed—Category A being more certain than Category C. Kahnaman and Tversky (1972) referred to this as ‘‘conserva-
As part of the confirmation process, estimates are prepared at tism.’’ The term conservatism originated from studies of hu-
each stage. man information processing in that individuals tended to revise
The Works Branch identified that there is a need to employ their opinions in the light of new evidence to a lesser degree
more scientific approaches for dealing with this category of than would be expected of an optimal information processing
risk in construction projects. These approaches would seek to system. This tendency for humans to be less extreme than
identify the areas of uncertainty and risk on a project. That is, optimal in their judgments has been termed conservatism in
there would still be a contingency allowance, but the nature probabilistic information processing.
of the contingencies would be described, and the amounts al- Moreover, due to the effect of negative sanctions (i.e., im-
lowed against them would be calculated. Briefly, the descrip- posing a penalty for an underestimate, where tender bids are
tion would indicate what the risk is and the likelihood of its above the pretender estimate but no reward/penalty for an
occurrence. The amount allowed would be an estimate, with overestimate), an over-exaggerated contingency is not uncom-
backup calculations, of the likely expenditure associated with mon in many project estimates. For public works projects, this
the risk should it occur. leads to misallocation of resources as more than sufficient
As the project is developed through the various stages, the funds are locked up in projects. Misallocation of resources will
process of risk analysis and identification seeks to reduce the severely disturb the important fiscal functions of public sector
level of uncertainty. The Hong Kong Government approach spending in (1) provision of social goods; (2) distribution of
concentrates on the major risks. income and wealth; and (3) maintenance of a stable economy
As risks are resolved the dollar value is added to the base in the aspects of employment and price levels (Musgrave and
estimate, which is, in effect, the value of the risk free elements Musgrave 1984).
of a project.

CONTINGENCIES ERA
Traditionally, cost estimates are point estimates. That is, sin- To alleviate these usually overexaggerated contingency es-
gle value estimates based on the most likely values of the cost timates, the Hong Kong Government has introduced the tech-
elements. These point estimates may or may not accurately nique of ERA in all public works projects. The Property Ser-
indicate the possible value of the estimate, and they certainly vices Agency (a government agency that manages government
do not indicate the possible range of values an estimate may real estate and property) in the United Kingdom was among
assume (Toakley 1995). When estimating, the most common the first to adopt an ERA style approach that was known as
method of allowing for uncertainty is to add a percentage fig- Multiple Estimating using Risk Analysis (HM Treasury 1993).
ure to the most likely estimate of the final cost of the known ERA is used to estimate the contingency of a project by
works. The amount added is usually called a contingency identifying and costing risk events associated with a project.
(Thompson and Perry 1992). The starting point for the ERA process is a base estimate,
Contingencies are often allowed in cost estimates. The ob- which is an estimate of the known scope and is risk free. The
jective of contingency allocation is to ensure that the estimated contingencies as determined by the ERA process are added to
project cost is realistic and sufficient to contain any cost in- the base estimate. The first step in the ERA process is to iden-
curred by risks and uncertainties. However, Thompson and tify risks by the project team. These risk items are then cate-
Perry (1992) pointed out several weaknesses of using a con- gorized as either (1) fixed; or (2) variable. For each risk event,
tingency amount: an average risk allowance and a maximum risk allowance are
JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 131
TABLE 1. Relationship between Risk Allowance and Risk Cat- Variable risk events are those events that will occur, but the
egory in ERA extent to which they will occur is uncertain. The cost incurred
Average risk Maximum risk will therefore by uncertain and variable. An example is the
Type of risk allowance allowance depth of piles required to be driven. The maximum risk allow-
(1) (2) (3) ance is estimated by the project team members based on past
Fixed risk Probability ⫻ maximum cost Maximum cost experience or records. This means the most expensive type of
Variable risk Estimated separately Estimated separately piling being required at the maximum length. The Hong Kong
Assumption 50% chance of being exceeded 10% chance of being Government’s ERA method requires an assumption to be made
exceeded that there is only a 10% chance that the actual cost incurred
will exceed this allowance. The average risk allowance is es-
timated with an assumption of a 50% chance of being ex-
calculated. The relationship between risk category and risk al-
ceeded. There can be a mathematical relationship between the
lowance is shown in Table 1.
Fixed risk events are those that either happen in total or not average and maximum risk allowance, but it is also legitimate
at all. If the event happens, the maximum cost will be incurred; for these two allowances to be estimated separately.
if not, then no cost will be incurred. An example is the need The rationale for using a 50% chance of being exceeded in
for an additional access road. At the early stages of a project, the average risk allowance is that it is unusual for all identified
the client might be uncertain as to whether an additional access risks (i.e., the worst case) to occur. The swings and rounda-
road will be required, rendering this as a fixed risk item. The bouts effect of the totality of the risk events identified should
scope of the road, should it be required, can be known and be able to cover the most likely costs incurred.
used to determine the maximum risk allowance. The uncer- Having identified all risk events and calculated their average
tainty is whether the road will be needed or not. The maximum and maximum risk allowances, the summation of the average
risk allowance is the cost of constructing this access. The av- risk allowance of all events will become the contingency
erage risk allowance is the probability of the client requiring of the project concerned. Fig. 1 shows a typical ERA work-
it multiplied by the maximum risk allowance. sheet.

FIG. 1. Example of ERA Worksheet at Sketch Design Stage

132 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000

 FIG. 2. Example of ERA Worksheet at Pretender Stage

ADVANTAGES OF ERA conservative and excessive percentage add-on contingency and

lead to a better allocation of resources.
The ERA process is usually carried out several times during
the pretender period for any one project. Fig. 2 shows the ERA EMPIRICAL DATA
calculations for another stage of the same project. As the
project develops, some events that were originally identified A summary of completed projects was received from the
as uncertain will be clarified and will be either deleted from government in December 1997 that detailed the contract sum,
the list of risk events or included in the base estimate as a original contingency, amount of additions, amount of
certainty. For example, the requirement for the secondary ac- omissions, final account amount, and start date of 332 building
cess road may have been confirmed. The remaining uncertain- projects, as shown in Table 2. Forty-five of these building
ties will form the final contingency allowance. Fig. 3 shows projects used the ERA process for determining contingencies
how the total average risk allowance evolves at successive and 287 were done by the traditional method. Because all of
ERA exercises as more becomes known about the uncertain the building projects used the government’s standard forms of
items identified. Fig. 4 shows the relative proportion of risk contract, it is possible to compare the contingency allowance
allowance (contingency) and base estimate at different stages against variations. The original contract sums ranged from
of the project. It should be noted that the use of ERA does HK$0.31M to HK$1,331.0M. Only 15 projects had a contract
not necessarily reduce the total cost of a project. It makes it sum larger than HK$100M.
less uncertain. A distinct advantage of ERA lies in its ability A new variable, DEVI, defined as the ratio between the
to retain the traditional method of presenting a project cost amount of contingency and the amount of variation at final
estimate in the form of a base estimate plus a contingency. It account, was used to reflect the accuracy of the original con-
imposes a discipline from the outset to systematically identify, tingency estimate, having included all the additions and
estimate cost, and consider the likely significance of any risks omissions in the final account. In most building contracts, a
associated with a project. It also aids financial control in hav- fixed contract sum is used whereby a contractor is paid ac-
ing risk and uncertainty costs identified before action is taken cording to a sum shown in the contract documents. Only if
to determine precise requirements. Furthermore, the itemized there are variation orders will the contractor be paid accord-
and substantiated contingency forms the basis on which to ingly. Therefore, the summary of variations should, by and
evaluate the impact of risk and uncertainties upon completion large, reflect how accurately the estimator has calculated the
of the project, thereby providing useful data for use in invest- uncertainties at the pretender stage. If the value of DEVI is 1,
ment appraisal. In short, it is a mechanism for accountability then there is a perfect match between contingency and varia-
for public money. Rigorously done, it will reduce the usually tion, and the contingency is capable of covering uncertainties.
JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 133
 FIG. 3. Total Risk Allowance versus Stages of Project Development

FIG. 4. Proportion of Total Risk Allowance versus Stages of Project Development

TABLE 2. Summary of Raw Data If the value is higher than 1, then there is surplus in the con-
tingency fund and vice versa. A negative DEVI value means
Statistic Non-ERA ERA that the amount of omissions is larger than additions, that is,
(1) (2) (3) a net reduction in final contract sum.
Number of projects 287 45
Contract sum (minimum) 0.31M 0.99M
Contract sum (maximum) 1,331.01M 208.48M ANALYSIS
Contingency (minimum) 0.15M 0.08M
Contingency (maximum) 110.00M 38.00M It was hypothesized that the ERA process should produce
Contingency/contract sum (minimum) 0.67% 4.02% better project estimates by making more realistic allowances
Contingency/contract sum (maximum) 137.40% 18.23% for contingencies. Better project estimates means the mean and
Final account variation (minimum) 6.00K 41K standard deviation of DEVI should be consistently smaller for
Final account variation (maximum) 85.98M 27.39M projects with ERA than those without ERA.
DEVI (minimum) 0.10 0.30
DEVI (maximum) 45.00 6.64
Table 3 shows that the variance for ERA and non-ERA
projects are 2.9387 and 28.0390, respectively. The variance
134 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000
for non-ERA projects is surprisingly high. This is contrary to The F-statistic supports the hypothesis that the contingency
the law of large numbers due to its sample size compared with allowance was more variable for non-ERA projects and more
ERA projects. This can be attributable to the fact that a fixed consistent for ERA projects. The t-statistic supports the hy-
or blind percentage was added to the base estimate to arrive pothesis that the contingency allowance was much smaller for
at the contingency amount for non-ERA projects. ERA projects.
An analysis of the F-statistic, which is the ratio between the Having said that, it was considered that, on average, the
two variances, shows that the F value is 9.5412, which is much contingency allowance for ERA projects was still very high.
higher than the critical value of F (1.7998) at the 1% signifi- This can be revealed from the mean value of 2.1490 of the
cance level (p < 0.00001). Thus, the hypothesis that the vari- variable DEVI. That is, the contingency allowance was over-
ances of the two populations are equal is rejected. It can there- estimated by 115%. This certainly will also lead to misallo-
fore be concluded that there is a significant difference between cation of resources. There is an urgent need to further refine
the variances of the two populations, which indicates that the the techniques of ERA. Possible solutions may include (1) the
variability of contingency allowance for non-ERA projects was provision of historical project records for making knowledge-
much higher and that the contingency allowance for ERA able estimates; (2) a review of the policy of capital budgeting;
projects was more consistent. and (3) a study of the effects of positive and negative sanctions
Table 3 also shows the mean values of DEVI of the two on estimators. These will be discussed in greater detail in the
groups of projects. It can be seen that the mean for ERA Recommendations and Conclusions sections.
projects is 2.1490 whereas that of non-ERA projects is 3.1498,
indicating that an average of 115% more funds has been set USERS’ FEEDBACK
aside for uncertainties in ERA projects and an average of
215% more for non-ERA projects. It is not surprising that both Beyond the statistical analysis of contingency funds, varia-
methods resulted in excess funding reserved for uncertainties tions, and contract sums described above, it was also consid-
as estimators are usually conservative, as discussed above ered useful to examine the views of the estimators who were
(Kahnaman and Tversky 1972; Raftery 1994). However, the required to implement ERA. A questionnaire was developed
discrepancy of nearly two times (115% versus 215%) the con- for this. In the questionnaire survey, estimators who have used
tingency allowance in non-ERA projects has resulted in severe ERA were asked to rank on a scale of 1 to 5 the importance
misallocation of resources. When contingencies are exagger- of three factors for their using ERA (they ranked the factors
ated in this way, some projects under consideration in a given independently instead of ranking the relative importance of the
phase of a public works program have to be foregone or de- factors). It was revealed that ERA was applied to all types of
ferred due to insufficient funds. government building projects such as government offices, fire
It was hypothesised that the ERA process should lead to a stations, and airport projects. ERA was used an average of four
smaller DEVI value than the non-ERA process. A one-tail t- times throughout the pretender stage of a project. A severity
test was used to determine whether the means of the two index was used to measure the perceived importance of vari-
groups are different, with the assumption of a different vari- ous factors in motivating the respondents to use ERA. In par-
ance (supported by the F-statistic). The resultant t-value is ticular, three areas were identified, namely, ERA being a policy
2.4788 with 204 degrees of freedom (Table 4). This is larger requirement, ERA helping to identify accountability, and
than the critical value for t (2.3488) at the 1% significance ERA’s potential for accuracy in determining contingency al-
level (p = 0.0070). The hypothesis that the means of DEVI for lowances. Table 5 shows the summary of the findings.
the two groups are equal is rejected. It can therefore be con- It can be seen that the respondents, who were government
cluded that the ERA group had a much smaller DEVI value officials, saw policy as the most important factor for using
than the non-ERA group of projects. The low significance ERA. Achieving accuracy in estimating was ranked least im-
level of <1% reveals that the contingency allowance for non- portant among the three attributes. These results were no sur-
ERA projects was highly exaggerated. prise as one reason for introducing ERA was accountability at
management level due to severe underspending of budgets in
TABLE 3. F -Test Result for Variable DEVI the early 1990s.
The respondents also expressed difficulties in implementing
Statistic Non-ERA ERA ERA due to a lack of data and records for them to determine
(1) (2) (3) the probability of occurrence of risk items and the cost asso-
Mean 3.1498 2.1490 ciated with these items. These would make estimating less
Variance 28.0390 2.9387 realistic. They expected more feedback to be made available
Observations 287 45
from past projects.
df 286 44
F 9.5412 —
P (F ⱕ f ) one-tail 0.0000 — RECOMMENDATIONS
F critical one-tail (1%) 1.7998 —
Although there has been a great deal of improvement in the
estimate of contingency allowance having implemented the
TABLE 4. t -Test Result for Variable DEVI

Statistic Non-ERA ERA TABLE 5. Summary of Feedback

(1) (2) (3) Policy
Mean 3.1498 2.1490 Rank requirement Accountability Accuracy
Variance 28.0390 2.9387 (1) (2) (3) (4)
Observations 287 45
1 (most important) 15 2 1
Hypothesized mean difference 0 —
2 2 9 7
df 204 —
3 3 5 6
t 2.4788 —
4 1 3 4
P (T ⱕ t) one-tail 0.0070 —
5 (least important) 0 2 3
t critical one-tail (1%) 2.3448 —
P (T ⱕ t) two-tail 0.0140 — Number of respondents 21 21 21
t critical two-tail (1%) 2.6001 — Severity index 86.90 57.14 48.81

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 135

ERA technique, the contingency allowance was still consid- it handles risk analysis is consistent with the traditional prac-
ered to be overly exaggerated (115% more on average). This tice of estimators at pretender stage to produce a deterministic
reinforces previous findings that estimators are conservative, figure in advising clients of likely bid levels. Empirical find-
as reported by Edwards (1968) and Raftery (1994). From the ings from a government department showed that the use of
feedback of estimators who were using ERA, it was found that the ERA approach has improved the overall estimating accu-
they used ERA because they were required to do so and they racy in determining contingency amounts. The samples in this
they viewed ERA as a mechanism of accountability more than study revealed significant improvement in project estimates as
a way of improving the accuracy of estimating. Nonetheless, the mean and variance of ERA projects are consistently
they also expressed the insufficiency of historical data for them smaller than those of non-ERA projects. Further research
to make better estimates of uncertainties. should be carried out to increase the sample size, particularly
Therefore, several measures can be considered to improve that of ERA projects. Clustering of projects into different sizes
the effectiveness of the ERA technique. First, historical data may also be useful to explore the effects of project size (com-
of finished projects have to be made accessible. In view of the plexity) and type (building, civil engineering, HVAC) on the
nature of construction projects with data being stored in dif- uncertainty of estimates. To further improve the accuracy of
ferent formats such as site diaries, calculation sheets, payment contingency estimates, a number of measures are suggested
forms, and even annotated drawings, the use of information that include the use of information technology to make his-
technology is very desirable. To this end, there needs to be
torical data more readily accessible, the implementation of
further studies.
equal sanction/reward for both under- and overestimates, and
Second, a sanction or rewarding system has to be imple-
adjustment of the amount of budget requests by the Treasury.
mented, particularly in view of the relatively passive reception
of the ERA technique. Accurate estimates, for instance, within
⫾50% of contingency, should be rewarded whereas the pen-
alty for an overestimate should be similar to that of an under- APPENDIX. REFERENCES
estimate.
Third, for even better accountability to be achieved in the Edwards, W. (1968). ‘‘Conservatism in human information processing.’’
Formal representation of human judgment, B. Kleinmuntz, ed., New
expenditure of public money, the totality of public works York. Wiley,
spending should be scrutinized instead of on a project-by- HM Treasury. (1993). ‘‘Managing risk and contingency for works
project basis. When it is extremely difficult to ask for budget projects.’’ Guidance Note No. 41, Central Unit on Procurement, Her
top-up in case of an underestimate, there is a tendency to over- Majesty’s Treasury, U.K.
estimate. If the swings and roundabouts principle is applied to Kahneman, D., and Tversky, A. (1972). ‘‘Subjective probability: A judg-
the whole of public works spending, then projects that are ment of representativeness.’’ Cognitive Psychology, 3, 430–454.
underestimated can be compensated by savings from projects Mak, S., and Raftery, J. (1992). ‘‘Risk attitude and systematic bias in
estimating and forecasting.’’ J. Constr. Mgmt. and Economics, 10(4),
that are overestimated. Taking a more macroview, this should
303–320.
lead to the ultimate goal of good allocation of resources. Musgrave, R. A., and Musgrave, P. B. (1984). Public finance in theory
Finally, a long-term record of estimating accuracy should and practice. McGraw-Hill, New York.
be kept so that the Treasury Department can make appropriate Raftery, J. (1994). Risk analysis in project management. E&FN Spon,
adjustments to applications for public works budgets, based on London.
performance of previous years. This will, in one way or an- Thompson, A., and Perry, J. G. (1992). Engineering construction risks:
other, correct the effect of reporting bias as described in Raf- A guide to project risk analysis and risk management. Thomas Telford,
tery (1994). London.
Toakley. (1995). ‘‘Risk management applications—a review.’’ Australian
CONCLUSIONS Inst. of Build. Papers, 6, 77–85.
Works Branch Government Secretariat. (1992). ‘‘Works Branch Technical
ERA is a simple to use method for implementing the use of Circular No. 21/92: Estimating using risk analysis.’’ Works Branch
risk analysis in estimating for construction projects. The way Internal Document, Hong Kong Government, Hong Kong.

136 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000