1. Budget Deviations in Capital Projects
Budget deviations in capital projects are a fact in far too many cases. We pick some examples to provide a general idea of the problem:
- More than 75% of capital projects have budget overruns, which amount to more than 50% of the initial budget in half the projects.
- At least 90% of major capital projects have budget overruns which, on average, is around 30%.
- Only 5% of light and heavy industrial capital projects fit into “best in class” with regard to cost and schedule predictability.
To dispel the possibility that this may be something of the distant past, the following chart of transportation projects may introduce some soberness into the subject.
Figure 1: Evolution of Estimates Inaccuracy, %
The American Heritage Dictionary presents two interesting meanings for “contingency”:
- An event that may occur but that is not likely or intended; a possibility.
- A possibility that must be prepared for; a future emergency.
Budget deviations are not intended – at least not generally and not openly! But they are possibilities that materialize with worryingly frequency, and for which a project should be prepared.
2. Budget Contingency
There are several reasons why budget deviations in a project modify their condition from contingent to actual. Just to name a few important ones: scope changes, errors in the estimation of quantities or prices, poor soil conditions, low construction performance, labour or political unrest, variation of exchange rates, inflation, harsh weather, modification of laws.
Qualifiers can be added as “unexpected”, “improbable”, “extraordinary” or “inaccurately assessed” to explain or justify the deviation, but from a factual perspective a difference between plan and reality occurs and manifests itself in the budget – not always as an additional cost, bust in most cases!
The project manager has tools to control the possible budget deviations according to their nature, starting from a careful budget preparation.
The study of those tools is beyond the scope of this article, which will focus on one option of last-resort: the “budget contingency”, a monetary allowance in the project budget that is made available to cover certain budget deviations that may appear.
In the following pages, we analyse some relevant features of this budget contingency, starting with its precise purpose.
3. Safety Margins in Engineering
Engineers make frequent use of “safety margins” or “safety factors”, a sort of protective cushions placed between the calculated capacity and the calculated load.
Safety margins are required because of uncertainties, limitations of knowledge, but it is worth noting that these safety margins are not intended to cover all possible uncertainties. For example, they are not introduced to cover gross numerical errors in the calculations, inadequate workmanship in construction, material changes, modifications in use, or lack of normal maintenance. But these safety margins can also take into consideration the particularities of the calculation: f.eg. the same foundation can be designed with safety factors ranging from 1.9 to 3.5 depending on the geotechnical information available to the designer. And, finally: the safety factors are not introduced to make the engineer’s life easier or to produce at a higher cost – they are brought in to minimize risks, which can sometimes be fatal or catastrophic.
We can draw some analogies between the engineer’s safety margin and the budget contingency:
- It is introduced because of uncertainties, but it is not devised to cover all possible risks that may cause budget deviations.
- It can be adapted to the expected project conditions.
- It is not a cushion to reduce the project team efforts to control the project cost.
The first of these conditions refers to the scope of the budget contingency: let’s see.
4. Scope of Budget Contingency
There is a certain degree of confusion with the precise scope and meaning of budget contingency. This seems due to the mix of related but different concepts, divergent practices, and the misunderstanding of some statistical ideas.
Without entering into details, let’s try to set some basic points clear:
- An estimate implies uncertainty. We cannot know the real value in the future, it is not possible to predict with total certitude what will happen, at least at the time of issuing the estimate, and we make an informed guess in the form of an estimate.
The proper way of expressing any estimate is with an indication of its uncertainty, with the expected range of possible values. For practical purposes we often tend to omit this in our daily lives. I can precisely state that I have two daughters, but if we talk about the price of concrete, we can say something like “approx. 90 eur/m3”, or “between 78 and 95 eur/m3”. Even better: “the most frequent value is 86 eur/m3, with a range between 78 and 95 eur/m3 for an estimated confidence of 80%”. Simply saying “90 eur/m3” is losing or hiding information that seems quite valuable – but in some cases, perhaps most common cases, it may be enough: language also means economy, agreed!
- It is important to note that the distribution of the possible project costs is typically not centred, symmetrical, it does not follow a bell curve or normal distribution. It is right skewed, the tail of possible overruns is larger than the tail of underbudget cases. This is a fact, clearly shown in Figure 2.
- In general, the preferred statistical indicators are those with “maximum likelihood”. For our purposes, the preferred estimator is the most likely project cost (the mode).
- In a skewed distribution, the mode is not the average or mean (see Figure 2). The consequence of this skewness is that the mode (a maximum likelihood estimator) will underestimate the project budget with a significant probability: the base budget estimate will be below the real value in most cases.
This is the gap that the budget contingency is intended to cover: the difference in estimation between the mode (most likely estimate) and the median (50% probability of occurrence; Figure 2). With therefore an important consequence: the budget contingency should be expected to be spent during the life of the project.
Figure 2: Probabilistic Distribution of Cost Estimates (example)
- Example generated with a simple Monte Carlo simulation
If the budget estimate thus constructed will be systematically below target, and the budget contingency will be required, then why not increasing the base estimate from start and forget about the contingency?
The answer to this fair question has two parts:
- The final budget is not modified by the calculating procedure. There would be no savings by removing the budget contingency, as the base budget estimate would be increased by the same or similar amount through the line items.
- The budget contingency is not intended to act at the level of each line of the budget estimate. It is an allowance that cannot be systematically added to any line item, but that experience has shown that is required in aggregate. If it were possible to estimate a cost in detail, it would become part of the project scope and would not be considered a contingency.
The preparation of the budget estimate may well include design allowances at the line item level, if experience shows that they are systematically required. These allowances are part of the base budget estimate, and they are not part of the budget contingency.
5. The Need of Budget Contingency
Despite the reasons provided above, there is a tendency to see this budget contingency, or at least a significant part of it, as an artificial buffer or cushion introduced to facilitate the project manager’s life. The budget contingency is also the main or only single line in a capital project budget for which there is no material support behind it, as it is neither goods nor services but just a provision for them.
Experience shows that there are pressures to reduce or eliminate the contingency provision, particularly from finance and business departments – as opposed to engineering or project management.
The eventual pressures aimed at removing or reducing the contingency budget can be transferred into an inflation of the allowances at the line items (“buried contingency”). There are cases where a pre-emptive allocation of contingencies as design allowances is made in expectation of a later reduction of the general budget contingency! This is an incorrect procedure, as it is incorrect to artificially manipulate the budget contingency.
Although the saying “need is the mother of invention” holds true in many cases, the most likely effects of artificially limiting the budget contingency are lower project specifications or reduced scope, plus a miserable life for the project manager, and, at the end, a request for budget increase. In the long term, this practice can arguably create a structural misallocation of funds. Clearly not the best for the owner!
If the finance or business departments don’t trust the project budget estimates, then the argument should go to discuss why, not to prejudge the professionalism of the budget by attacking the contingency allowance.
6. The Management Reserve
But there is more: by the definition given to the budget contingency the budget estimate with the contingency will still be below the final budget in 50% of the cases!
If the project owner wants to be protected against this risk, an additional budget provision should be included.
This is the gap that the management reserve is intended to cover: the difference in estimation between the median and whatever probability the owner wants to secure, according to its risk perception and aversion (see Figure 2).
To the reasons provide above can be added the argument that while it seems sound to allocate a total budget that covers the expected median range (50%), freezing the funds for the 80-90% range can be excessively expensive, especially for companies handling a portfolio of capital projects.
Although we consider that this separation of reserves is prudent, some practitioners do not differentiate between budget contingency (up to 50%) and management reserve (up to 80 or 90%) and use a single budget allowance that covers both.
Where the request for capital investment approvals involves complex and time-consuming procedures, the decision on the management reserve is particularly tricky, because the difference between a 50% and 80% confidence can be substantial and kill the project if the management reserve is not accounted for adequately in the financial evaluations; on the other side, requesting additional funds can be organizationally problematic and not the best way of promoting a manager’s career.
7. Scope Variations
The budget or cost estimate is a prediction made at a given moment in time of what will be the final cost of the project in the future. Further, it is an estimate made for a certain project scope. At the initial stages of project development, the scope is often also defined with coarse lines and uncertainty in the details, but this uncertainty cannot accommodate project variations incorporated later.
It is different to say that contingency provides room for underestimation due to incomplete scope definition at the time of preparing the estimate, that to state that scope variations should be covered by the contingency.
Neither the budget contingency nor the management reserve are intended to cover against changes of scope. In fact, they should also not be used to protect against costs associated with external influences on the project, like social unrest, law changes, natural disasters, currency effects, escalation, etc. unless explicitly accounted for in the budget estimate. These risks should be separately managed with the owner’s contingency, not with the project budget contingency.
The details of what should be precisely included, or not, in the budget contingency vary depending on the practitioners and reference sources. For instance, some authors indicate that estimating errors should not be included in the contingency (as this would foster malpractice), while others explicitly include within the scope of contingency “errors and omissions in the construction documents”. On this regard we recommend consistency and an explicit declaration of what is included, or excluded, in the budget contingency.
The budget estimate is typically refined along the project execution with more developed estimates: “cost planning is an iterative process, performed in steps of increasing detail as more design becomes available”. However, at a given moment of time decisions have to be taken (“decision gates”) and it is good practice that these decisions are adequately documented. In particular, the project definition should include a reasonably and explicitly detailed baseline with a scope and budget estimate so that later scope variations can be adequately identified.
8. How much Budget Contingency?
Indeed, determining the amount of budget contingency is a matter of professional responsibility and should be defined very carefully.
There are several methods in practice to estimate the adequate budget contingency, but they can be reduced to the following general classification:
8.1 Simulation of Budget Contingency
The simulation can potentially yield the most detailed results, but it requires resources for data gathering, plus the creation and calibration of the computer model.
Until recently this has been available only for large projects. The situation is rapidly changing as new powerful software has appeared in the market which facilitates the mechanical part of the calculation.
The creation of meaningful models and the extraction of sensible results remains the crucial task beyond the running of thousands of simulation cases. Without entering into details, important aspects to be taken into consideration include:
Usually the variables of interest are defined from 3 estimating points for unsymmetrical probability distributions: identifying the points requires information, of course, but the selection of a particular distribution is not solved by the availability of many options in a pop-down menu: the user needs expertise.
In particular, the improbable cases (the long tail) can be largely underestimated by inadequate distributions and narrow risk evaluation.
Very often there is a significant correlation among some of the variables which can also too often be missed by the models.
The simulation tends to be in trouble to say anything about the unexpected, which often is either not included in the analysis or treated as a large lump-sum.
8.2 Percentage Based Estimates
On the other extreme of the methods for calculating the budget contingency, the fixed percentage method, although reportedly the most frequent, is probably a too coarse guide that skips a proper risks evaluation, only suitable for small or rather repetitive projects.
In any case, it can be brought towards the expert judgment method by introducing ranges or using some scope breakdown and different percentages applied on each part, all in order to somehow bring nuance and qualify the project particularities.
The expert judgment method is based on an evaluation of the project risks and how they can be expected to manifest as cost.
The immediately connected matter of interest is how to manage the risks so that the required contingency can be reliably diminished.
A fundamental aspect of (budget) risk is that it is controlled by the time window: the further we look into the future, the bigger the uncertainties. Or, the earlier in the project life, the larger the risks, comparatively.
This almost truism hides the fact that different organizations can handle the same project and its risks very differently: the key variable is actually not time, but information: as project information increases, risk is expected to decrease. And therefore, along with a project’s life, the required budget contingency will (typically) decrease. This “estimating tunnel” is well presented in the following chart derived from AACE. Note the asymmetry of the positive and negative deviations!
Figure 3: Estimated Accuracy Range of the Budget Estimates vs Degree of Project Definition
AACE’s estimate class designations (Class 1 to 5) are a useful but necessarily general guidance in the selection of the appropriate accuracy range of the budget contingency.
Table 1 in the following page complements this chart, while Table 2 provides a comparison across different organizations. This overview is completed with Table 3, where further instances of practice are presented.
Although the degree of project definition is clearly an influential factor in determining the estimated class, it is not the only one, and the process should not become a mechanical assignment.
The differences presented in Table 2 and Table 3 highlight the need to avoid the automatic specification of a contingency range. AACE itself has different reports for specific industrial sectors, where detailed indications on the corresponding estimating class are provided.
In particular, it should be noted that different organizations may have defined different probabilistic targets for their budgets, which directly affects the contingency amount.
As an example, Table 4 presents a simple but seemingly effective procedure for quantifying a number of rather qualitative risk factors (“factor-based approach”).
Table 1: Cost Estimate Classification Matrix for Process Industries (AACE)
Table 2: Comparison of Classification Practices (AACE)
Table 3: Contingency Rates
Table 4: Contingency for Strategic Estimates (Transport and Main Roads, Australia)