4 Cost Information

Marc Robinson
Published Date:
October 2007
  • ShareShare
Show Summary Details
Marc Robinson1

This chapter focuses on the cost information requirements of performance budgeting, and more generally on the relationship between results and costs. Its objectives are to identify the type of cost information which is of the greatest potential relevance to performance budgeting, to note and discuss some of the key challenges which arise in seeking to obtain that information, and to identify certain significant limits on the potential utility of cost information for performance budgeting purposes.

Because different performance budgeting models seek to link results and funding in different ways, and with somewhat varying objectives, the cost information requirements of those models vary. For expenditure prioritization purposes, it is program costing which is usually seen as of primary importance. In other words, with expenditure prioritization decisions viewed as a matter of deciding how to allocate money between programs, the starting point is to measure program costs so as to permit a comparison—be it informal or more systematic—between program costs and benefits. The approach is based upon the widely-held assumption that at the level of the government-wide budget, it is in general only practical to set priorities in terms of the broad categories of expenditure which programs represent. From this perspective, the appropriate place for more detailed expenditure prioritization is internally within agencies, and it follows that for this purpose agencies will require cost information for more disaggregated program components (for example, sub-programs).

What might be called cost function information is, by contrast, viewed by many as an important tool for building a tighter relationship between funding and performance through formula funding, or budget-related performance targets. The most common version of this idea, as mentioned in Chapter 1, focuses upon unit cost information. As noted there, the idea, in its simplest form, is to use the identity average cost • quantity = total cost as the basis for predicting the results an agency should be expected to deliver with any given level of funding or, equivalently, the funding required to deliver any targeted level of results. This approach has usually been applied to outputs, although it is sometimes applied to outcomes. The notion of linking funding and results through unit costs is, as discussed below, somewhat simplistic, and it is better therefore to express the underlying idea in a more general form: namely as the proposition that cost functions—that is, the functional relationships between cost and any given result (output or outcome)—can be used to set targets or funding formulas.

The role of comparative cost information tends to be stressed by performance budgeting models which aim to boost efficiency and cost-effectiveness by factoring savings targets into performance targets or funding formulas. The principal theme is that measures of the costs at which others—that is, organizations delivering similar services—deliver outputs or outcomes can provide an indication of the magnitude of potential efficiency gain. Such information might, in principle, be obtained by market prices for similar products, by cost benchmarking or by market testing. Time series data on an agency’s own past costs of delivering results may also potentially serve as a source of comparative cost information.

As with all performance information, cost information is unavoidably costly to obtain, in terms of the expense not only of setting up costing systems, but also of the ongoing collection and verification of data. One of the factors contributing to these information costs is the fact that those at the “coal face” who have easier actual or potential access to cost information can be reluctant to provide this information to central decision-makers who may use it to their disadvantage. But even setting aside such strategic and “asymmetric information” considerations, cost information is not usually cheap or easy to obtain.

Because the cost information requirements of each performance budgeting model differ, so also does the expense of obtaining the cost information which each requires. Unit cost information (more generally, cost function information) is, for example, generally significantly more costly than program costing. It follows that the cost of cost information is an important consideration which should be taken into account in deciding what types of performance budgeting model to adopt, and then in deciding whether to adopt these models across the board or on a more selective basis for certain areas of public expenditure.

It is also necessary to strike a balance between accuracy and information costs in the design of the costing system to support performance budgeting. The more accurate information on the costs of delivering results becomes, the better the linkage between results and funding it is likely to permit. However, the marginal cost of more accurate cost information can be high. And even if a great deal of money is spent on sophisticated cost accounting, cost information will generally remain far from perfect. One of the implications of these information cost considerations is that particular caution should be taken in deciding how much money to invest in costing systems in countries with very limited financial resources and shortages of appropriately trained accountants. (This in turn has implications for the choice of performance budgeting model in such countries.) However, as discussed below, it should not be thought that effective performance budgeting requires perfect cost information. Much can be achieved with approximate costing, whether it is for expenditure prioritization purposes, or for more sophisticated forms of performance budgeting such as formula funding and target-setting.

The recognition that cost information does not come free should not lead one to take a negative view towards the role of cost accounting in the public sector. The potential value of results cost information as a tool for improving the efficiency and effectiveness of public expenditure is quite considerable. Traditional government budgeting and financial management systems have made far too little use of such information.

Program costing and expenditure prioritization for allocative efficiency

Programs should, in principle, group together expenditures on services which share common public policy objectives—common intended outcomes—in such a manner that decisions about how much to spend upon competing programs capture the choices which are most central to allocative efficiency. This means choices such as how much to spend on health treatment versus health prevention, or upon primary education versus tertiary education. Budget funding is, however, spent on inputs (or, to be more precise, inputs and transfers), so the costing of programs requires that expenditure on inputs be allocated in accordance with the objectives for which those inputs are used. Expenditure prioritization requires, moreover, not only ex-post information about program costs, but also a capacity to control expenditure by program. It is therefore necessary that the program classification of expenditure be an integral part of the budgetary classification system and of the underlying chart of accounts. The allocation and control of expenditure by objectives is, however, not part of traditional government budgeting and accounting systems.2 For program-based expenditure prioritization, it is therefore necessary to build robust costing and cost control systems based upon the objectives of expenditure.

The next chapter in this part examines the full range of issues which arise in establishing and using a program classification of expenditure. Here the focus is exclusively on one issue—the costing of programs. One of the more important challenges in costing programs is the attribution of indirect costs. In the context of costing programs, indirect costs refer to the costs of inputs or activities which contribute to more than one program. What constitutes an indirect cost depends upon how programs are defined. Suppose that all programs are defined in terms of outputs and outcomes—that is, as groups of services to external parties3 with common intended outcomes. If this is the case, then all ministry-wide support services and infrastructure (such as human resource management, internal financial management, and the IT network)—together with other “overhead” items such as the salaries of the chief executive and other top ministry-wide managers and the costs of ministry-wide (non-program-specific) policy development—will be indirect costs which will need to be allocated between programs.

Accountants use the term “allocation basis” to refer to the formula or principle used to allocate a specific indirect cost between two or more “cost objects” (in this case, programs). Ideally, the allocation basis used for each indirect cost should accurately reflect that indirect cost’s contribution to each program. But in the case of some indirect costs this may not be easy to determine. In practice, the cost and effort of accurate cost attribution means that approximate, or sometimes even quite arbitrary, allocation bases are often used, particularly for those indirect costs which are most difficult or costly to allocate accurately. It is, for example, not uncommon for many support service costs to be allocated between programs on the basis of the proportion of the total ministry staffing accounted for by each program, without any attempt to actually measure the usage of those support services by each program. In the extreme, measures of the “full” cost of programs derived by such arbitrary cost allocations may be less useful than measures of direct program costs alone.

There are many who see the solution to the indirect cost allocation problem as lying in the general application to the general government sector of activity-based costing (ABC), which aims to develop a much more accurate two-stage costing process. ABC is discussed further later in this chapter. The key point to be made at this stage is, however, that the level of costing sophistication called for by ABC is expensive and demanding of skilled human resources—which helps to explain why, even in the richest countries, ABC tends to be used selectively rather than across the board. Program costing obtained using less accurate, but also much less costly, costing methodologies may be quite adequate to meet the needs of government expenditure planning.

The indirect cost allocation challenge has other important implications—in particular, for choices about how programs are defined. It is particularly relevant to the question of whether or not to have “corporate services” programs which group together ministry-wide support services, infrastructure, and other overheads. The fundamental objective of program classification—the facilitation of expenditure prioritization choices which are relevant to allocative efficiency—would seem to suggest that the answer should be “no,” and that programs ought to be defined in terms of outputs and outcomes, and not in terms of support services and overheads. Support services and overheads are, after all, not incurred for their own sake, but simply in order to facilitate the production of services which are of direct benefit to the community. This implies that the decision on what types of support services are needed and how much money to spend on them should be derived from choices about outputs and outcomes, rather than viewed as a program prioritization choice in its own right.

Reflecting this view, the program classification of expenditure in countries such as Australia4 and the UK does not include corporate service or other overhead costs programs. In these countries, the program structure of, say, an education ministry is comprised exclusively of programs such as primary education, tertiary education, and technical education, with all corporate services and other support services costs attributed to those programs rather than themselves forming a program. There are, however, many other countries (for example, South Africa and Canada) which do have “corporate services” programs—as had, formerly, some of the countries which now have only output- and outcome-based programs.

It is possible to view the decision to make use of corporate services programs as, at least in some cases, a reasonable pragmatic response to the indirect cost allocation challenge. In the early stages of the adoption of program budgeting, any allocation of indirect costs is likely to be particularly arbitrary. Under these circumstances, there are essentially two options:

  • to move immediately to programs based exclusively on outputs and outcomes, accepting that the costing of these programs may be highly inaccurate

  • to begin by assigning many support service and other overhead costs to corporate services programs, and then progressively move over time to attribute many of these costs to output- and outcome-based programs. In countries with very limited financial and skilled accounting resources, it may make sense to view this process as a gradual one.

It should not, moreover, be assumed that the ultimate objective need necessarily be the full allocation of all indirect costs. Some indirect program costs (for example, employment costs of top agency management) are joint costs, meaning that they are costs which would not change even if one of those programs were closed down. It is a conventional principle of management accounting that joint costs are not decision-relevant and should therefore not be allocated to products. Specifically, in the context of program costing, to allocate joint costs to a program is to give a misleading impression of the costs which are at stake in decisions about program expenditure priorities. This suggests that there is nothing wrong in principle with maintaining, within a program structure which is generally output- and outcome-based, a small unallocated “joint cost” program. This issue will be further discussed at the conclusion of this chapter, as will the related issue of the treatment in program costing of so-called sunk costs, which is in turn linked to the choice between accrual and “cash” accounting.

Setting performance expectation using cost information

The idea of using cost information to link funding to results is a very important theme in performance budgeting theory and practice. As a simplified illustration of the idea, suppose that an African government decides to scale-up substantially the effort to vaccinate the population against yellow fever. If at present it costs, say, $3 per person for such vaccinations, this could be taken to suggest that an additional $1.5 million should permit 500,000 more people to be vaccinated. Such an expectation could be formalized in an explicit performance target set for the government agency managing the vaccination program. Target-setting of this type can serve as a means of helping to safeguard against the ever-present danger that increased funding can be partially wasted (through, for example, disproportionate wage increases or reduced levels of work effort by staff of the agency concerned).

The same government could in principle go further and use comparative cost information to set performance targets in such a way as to create pressure for improvements in the efficiency of service delivery. Imagine that the vaccination effort is handled by a number of regional units within the country concerned, and that it is established that the most efficient of these deliver vaccinations for, say, $2.70 per unit. This information could be used to set targets for expected cost reductions to be achieved, over an appropriate interval of time, by the less efficient regional units. This is an example of cross-sectional cost benchmarking, which may be defined as the process of comparing the costs of producing services with other similar producers, in order to obtain an indication of the scope for savings from improved efficiency. As discussed elsewhere in this volume, such a use of cross-sectional cost benchmarking is at the heart of the incentives for improved efficiency which can be created by well-designed purchaser-provider funding systems. Under such systems, a service delivery unit which delivers a particular service (output) relatively inefficiently will find itself losing money because the price it receives for the service will be less than its costs.

More generally, as outlined in Chapter 1, cost information can be used to link funding to results in order to improve allocative and/or technical efficiency in the following ways:

  • for medium-term expenditure forward estimates

  • in the formulation of performance targets

  • as the basis for formula funding

  • as the basis for the “purchaser-provider” version of formula funding.

If cost information is to be used in such ways, it must be at the level of specific services (output types), rather that at the level of the broad program. Expenditure on vaccinations might, for example, be part of a “preventative health” program in the health ministry’s budget, together with many other services such as health awareness information provision, eradication of breeding grounds for malaria-carrying mosquitoes, and free provision of items such as syringes and condoms to high-risk populations. Because most programs are like this, and comprise “apples and pears” baskets of different services, the costs of each service within the program basket will need to be considered separately if the aim is to use the cost information for target-setting or formula funding. Even when dealing with a group of quite similar outputs, there are benefits in differentiating as far as possible between products with different costs. For example, in a formula funding system for undergraduate education in public universities, it is usual to differentiate between different categories of course based on cost. High-cost courses such as medicine would, for example, attract a higher level of funding per student-year than a low-cost course such as law or liberal arts. The logic of this is that if all courses attracted the same rate of funding, universities would be given a financial incentive to shift their course offering away from high-cost course offerings to low-cost course offerings.

The need to undertake costing at the individual product level means that the indirect cost allocation task will be considerably larger than in the case of program costing.

There are two important practical challenges which arise in seeking to use cost information to link performance and funding. One concerns the nature of the underlying cost/results relationship. The other pertains to the degree to which comparative cost differentials can reflect factors other than differences in efficiency.

In respect to first of these issues, to talk about using unit costs to link funding and results is to simplify the challenge involved. Unit costs normally mean average costs. To measure the average cost of delivering a service at a given time and use this (perhaps with some downward adjustment to promote improved technical efficiency) for target-setting, or any of the other applications identified above, is in effect to assume that average cost is constant. It is, however, basic economics that average cost is not necessarily constant, but may rise or fall with changes in the volume of production. In the short run, for example, if marginal cost is constant, average cost will fall as production increases due to the spreading of fixed costs5 over the increased output.6 In the long run, moreover, average costs may also change with increased production volumes as a result of changing economies of scale. Under these circumstances, using average costs can lead to the setting of performance targets which are insufficiently demanding (or, equivalently in formula funding systems, prices which are too high). Thus, in the example given above, the $3 average cost per vaccination will include program overhead costs and some other costs which do not need to be increased commensurately with any increase in the volume of vaccinations delivered. Performance targets associated with the scaling-up of the program should therefore be set with the expectation that an extra $1.5 million of funding should result in the vaccination of significantly more that 500,000 people.

In principle, inappropriate target- or price-setting of this sort may be entirely avoided by basing the funding/results link on the true underlying relationship between costs and output implicit in the agency’s production processes—that is, on what economists call the “cost function”7—rather than upon arbitrary assumptions (such as the assumption that average costs are constant). In practice, however, it tends to be costly and difficult (if not impossible) to obtain precise information on the underlying cost function, even if a good deal of money is spent on sophisticated cost accounting.

Fortunately, however, it is not necessary to have perfect information on the cost function in order to use cost information to link funding to results. In many contexts, target-setting, formula funding, or “prices” based on approximate cost information can serve to create very positive incentives for improved efficiency. It can, for example, often make sense to use average variable cost measure instead of average cost for funding formula/target purposes.8 Thus, for example, it might be established that, after having netted out the program overhead costs and the like, the unit cost of delivering yellow fever vaccinations is $2.50. On this basis, one could set a target of an additional 600,000 vaccinations to be delivered with an extra $1.5 million funding. It may, of course, be objected that such an approach merely substitutes one arbitrary assumption about costs for another, because it assumes that average variable cost is constant. The point, however, is that an approximate cost model of this type may serve the policy objective well (and certainly much better than a model which assumes average cost to be constant), while not requiring huge levels of effort and expense on complex cost accounting. This has been, broadly, the approach taken in the output-based funding in the hospital sector, and the evidence suggests that it has worked quite well despite the imperfect costing model (Newhouse, 2002, p. 30; Robinson and Brumby, 2005).

In respect to the use of comparative cost information, the additional problem which must be faced is that cost differentials between different production units may reflect factors other than differences in efficiency. For example, schools or hospitals in rural or remote areas may have intrinsically higher cost structures than their metropolitan counterparts due to a lack of scale economies and other factors such as a need to pay higher salaries to staff to retain their services. It would therefore be inappropriate to use the metropolitan cost levels in setting targets or prices for rural services. Methodologies are therefore needed to make use of comparative cost information in such a way as to remove cost differences unrelated to efficiency differentials. The most obvious way of doing this is to confine cross-sectional cost benchmarking to production units which face similar cost conditions—for example, to comparisons of the relative cost levels of different rural hospitals. The scope for doing this depends on the availability of suitable “benchmarking partners”—organizations sufficiently similar for meaningful cost comparisons to be made. It is not, however, necessary that only service units which face exactly the same cost conditions be used as benchmarking partners for costing purposes. As with cost function information, comparative costing methodologies based on approximate data can serve very well to achieve an effective linkage between funding and results.

Cost function information is generally not used to link outcomes to funding for two principal and related reasons: one is outcome measurement difficulties; the other is the impact of so-called “external factors” discussed in Chapter 3. These exogenous characteristics of the service delivery environment or of clients/cases give rise to an intrinsic uncontrollability and uncertainty about the outcomes which can be delivered with any given level of funding. The degree of this uncontrollability and uncertainty varies, and tends to be greater for high-level outcomes than it is for lower-level outcomes. In general, however, it is inappropriate to think in terms of a tight “functional” relationship between costs and outcomes.

It is for this reason that cost-based formula funding and target-setting tends to be advocated much more for outputs than outcomes. There are quite a few public sector services for which a reasonably stable relationship between costs and outputs9 exists—even if that relationship (the “cost function”) may not be easy to determine. It is, however, also true that there are also many public sector services which are characterized by intrinsic uncertainty in the cost/output relationship.

The most important source of uncertainty in the cost/output relationship is heterogeneity. As discussed in Chapter 3, outputs are heterogeneous when the amount and types of activities required to deliver the “same” service varies due to client or case characteristics. Variability in the activities which comprise units of the “same” output gives rise to cost variability, and therefore creates uncertainty about the level of output which can be produced with any given level of funding. For highly heterogeneous services, the use of cost-based formula funding or target-setting is quite impossible. For criminal investigations, for example, it is quite impossible to say in advance how many investigations a particular level of funding will allow the police to handle. Milder heterogeneity may, however, be manageable—methods have, for example, been found to deal with a certain degree of heterogeneity within the context of the DRG funding system for hospitals.

Heterogeneity is not the only factor which can introduce significant uncertainty into the relationship between costs and outputs. Uncertainty also arises in the case of what may be termed contingent capacity services. Contingent capacity services are services the demand for which is unpredictably variable, and for which there is an essentiality and immediacy to supply—in other words, it is considered important that when a demand for the service arises, the service be supplied promptly.10 Fire services are an example. The funding of fire services typically bears only a loose relationship to the actual level of fire-fighting (the output) they normally carry out. The level of funding is, rather, designed to give the fire service a margin of response capacity to deal with more exceptional and unlikely circumstances. In general, for contingent capacity services, it is inappropriate to expect anything other than the loosest relationship between funding and actual levels of output.

There are quite a few public sector services which have the contingent capacity characteristic to a greater or lesser extent. All emergency services—ambulances, rescue services, and the like—fall into this category. Health systems maintain capacity to fight epidemics and outbreaks of new virulent diseases, and agriculture departments maintain a similar capacity in relation to exotic plant and animal diseases and similar problems. Perhaps the extreme example is the military, the principal “output” of which is the fighting of wars. For most countries, the principal purpose of having a well-funded military is to deter potential attackers and thus avoid having to produce the “output” at all.

Activity-based costing and budgeting

Over approximately the past 15 years, there has been considerable interest within the public sector in a number of countries in the use of activity-based costing (ABC) as the basis for costing and budgeting. ABC is an accounting methodology which seeks to develop a “model of organizational resource consumption” (Kaplan and Cooper, 1991, p. 275) to provide accurate allocation bases through which indirect costs can be allocated to products. In a public sector context, ABC can be applied to costing at the program level, or it can be taken further to cost individual services (outputs). It involves a two-stage costing process in which the first stage is the costing of broadly-defined activities,11 and the second stage is the allocation of the costs of these activities to programs and possibly also (depending upon the form of performance budgeting which this costing information is intended to support) to more specific outputs within those programs. Human resource management might, for example, be a cost center, in which case the first stage of the process is the accurate measurement of the overall cost of human resource management, and the second stage is the development of accurate “cost drivers” for attributing human resource management costs between the programs or individual outputs which it supports. To ensure accuracy, the cost drivers used in the second stage would be based upon measures of the quantity, number, and relative resource-intensiveness of each type of service (recruitments, promotional processes, disciplinary processes, separations, and so on) which human resource management provides to each program/output.

Public sector proponents of ABC often argue that it is essential for performance budgeting—that, in the words of one US official, “you can’t have performance and budget integration unless you actually have ABC” (quoted in Peckenpaugh, 2002a, p. 11). There is, however, a dearth of systematic analysis of actual experience with ABC, and there is no consensus in the (largely anecdotal) available literature that it has been a success in the public sector. In the US, for example, it has been said that “ABC has proved exceedingly difficult to implement,” and that many organizations which experimented at the beginning of the current wave of interest have not continued with it (Peckenpaugh, 2002b, p. 42).

Proponents see ABC as relevant to performance budgeting in two ways. The first is as a tool for better expenditure prioritization. The second is as a means of developing a tighter linkage between planned outputs and funding.

In respect to expenditure prioritization, the core argument is that mentioned earlier—that ABC permits much improved program costing and, more generally, that it provides a “systematic way of determining how to apply limited resources to the right activities to produce the right results” (Williams and Melhuish, 1999, p. 36). As noted earlier, it is easy to accept that more accurate program costing can improve the quality of expenditure prioritization choices. ABC is, however, a very high-cost accounting technology. As Brown et al. (1999, p. 18) put it, “installing an ABC system is technically complex, requiring talented personnel and a considerable amount of time and mean dollars.” The question is therefore how far, even in rich countries, the benefits of ABC justify the very considerable cost of the system. The answer to this question is not entirely clear.

For the purposes of linking planned outputs and funding, proponents call for the application of activity-based budgeting (ABB). In ABB, the ABC costing process is reversed, and one goes from planned outputs to the activities required to produce those outputs and then to the cost required to carry out those activities (Cooper and Slagmulder, 2000). What is being suggested is, in essence, that the two-stage process offers a better means for working out the true output cost functions. There is not much evidence on how well this works in practice, but it would seem to make sense for appropriate services. However, it should be emphasized that the use of ABB does not in any way circumvent the problems identified above as limiting the scope for using cost function information in budgeting.12

Reflecting these problems, it has been argued that “in most instances ABC is not appropriate for public sector applications” (Mullins and Zorn, 1999, p. 54). This may be taking the argument a little too far. It seems more plausible to argue that ABC may be best viewed as a tool for selective application to relatively standardized services, the greatest concentration of which will in general be found at the lowest level of government (including local government).

Decision-relevant costs and performance budgeting

As mentioned earlier, standard managerial accounting theory suggests that joint costs should not be attributed to programs because they are costs which will not change as a result of decisions taken about the program, even if the programs concerned were to be closed down. This argument is consistent with the basic economic principle that decisions should be based upon opportunity costs.

Closely analogous to this is the question of how to treat fixed costs—essentially the costs of using capital—in costing of programs or individual services for performance budgeting purposes. It is clear that all variable costs should be taken into account, because these will change not only if a program or output is discontinued, but also if the level of production changes. By contrast, fixed costs do not, by definition, change if the volume of production is changed. They are therefore not relevant to decisions about whether to increase or decrease programs at the margin, or, in principle, to formula funding or cost-based performance targets. Fixed costs can, however, be relevant in part to decisions about whether or not to close programs down. This is because, if a program is discontinued, is it often possible to recover some of the fixed cost by selling the capital assets used by the program. But to the extent that fixed costs cannot be avoided in this manner, it is a mistake to consider them in determining whether to continue to program because there is no point crying over spilt milk.13

Stated in a more general form, the economic principle is that sunk costs (costs which cannot be recovered whatever one does) are never decision-relevant. What is decision-relevant is variable cost and, depending upon the nature of the decision, avoidable cost (that is, variable costs plus the avoidable portion of fixed costs).

Notwithstanding this, many accountants favor “full-cost” accounting for both product costing in the private sector and, in the public sector, for program or output costing. Full-cost accounting (also known as “full absorption” costing) calls for the attribution of all costs, even if they are joint costs or sunk costs, to products. The naïve version of the argument for full cost accounting simply takes it as self-evident that proper decisions cannot be made unless all costs are taken into account. The more sophisticated version argues that “all costs are variable in the long run—and thus subject to managerial action” (Sharp and Christensen, 1991, p. 32), and that taking all costs into account therefore promotes an appropriately longer-term perspective on decision-making.

The question of the treatment of fixed costs is inextricably linked with the accounting basis chosen for the budget. This points to the issue of the relationship between accrual accounting and performance budgeting.

Accrual accounting and performance budgeting

Does performance budgeting require accrual accounting? Or can it at least be said that accrual accounting supports performance budgeting better than traditional cash accounting?

Applying accrual accounting to performance budgeting means that in linking funding to results, funding is measured using the accrual concept of expenses14 rather than in terms of the expenditure.15 Because this requires the use of the accrual concept of expenses to express the budget authorizations or other funding provided to agencies, it means going beyond the use of accrual accounting for financial reporting purposes to a form of accrual budgeting.

As noted earlier, in making expenditure prioritization decisions, all variable and avoidable costs are relevant. Accrual accounting can facilitate better prioritization between programs by making sure that relevant avoidable costs are not omitted when program costs are measured—and, conversely, that certain irrelevant expenditures are not included. This is because accrual accounting includes in its measure of expenses any costs of production for which payment is deferred to future years. For example, when government employs people, part of the costs of the services which those people deliver to the public in any given year takes the form of entitlements to pension payments in future. These pension costs are clearly a variable cost. An accrual measure of program costs will include these deferred employment costs, whereas a cash accounting measure will not. Failing to take such deferred costs into account may make programs look cheaper than they actually are, and this effect will be greater for those programs which are most public employment intensive. Because such omission can distort program choices, it is desirable that these and other deferred costs be taken into account when making decisions about program priorities.

The use of a cost measure which does not distinguish between costs of production on the basis of whether they are paid this year or in the future is also better for the purposes of performance budgeting systems which seek to develop tighter cost-based links between funding and results. The cash expenditure on delivering a service can, for example, be manipulated by expenditure timing shifts which have nothing to do with changes in the cost of delivering services. Such timing shifts can result from the bringing forward or deferral (“leading and lagging”) of the payment of accounts, from deferred-payment lease arrangements for equipment or other supplies, or from a range of other transactions.16

These clear benefits of accrual accounting for performance budgeting can be realized by placing budgeting on a partial accruals basis. Partial accrual budgeting means giving agencies expense budgets which are intended to cover all expenses of current production which involve payments in either the present or future years. By contrast, under full accrual budgeting, agencies would be given expense budgets from which they would have to cover all expenses—which means not only expenses requiring payments in the present or future years, but also expenses arising from expenditures made in the past. By far and away the most important expense arising from expenditure in the past is depreciation. In essence, therefore, the difference between partial and full accrual budgeting is that only under the latter are agencies required to cover depreciation from their expense budgets.

The question then is whether there are gains to performance budgeting from going beyond partial accrual budgeting to place budgeting on a full accrual basis, including depreciation in agency budgets. This is more debatable. For those who take the view that sunk costs should be disregarded for decision-making purposes, full accrual budgeting is problematic insofar as depreciation represents a sunk cost,17 the inclusion of which will, as discussed, give an exaggerated impression of potential savings which could be realized if the program were closed down. From this perspective, it can be argued that it is better for performance budgeting purposes to budget on a partial accrual basis, and that the appropriate time to consider capital costs for budgetary control purposes is when evaluating potential capital projects in the capital budgeting process. In further support for this position, it might also be argued that most expenditure prioritization decisions are about reducing or increasing the size of programs, rather than eliminating them altogether. This implies that for most budgetary decisions, it is only variable cost—information on which requires only partial accrual accounting—which is relevant. When a program is identified as a potential candidate for complete closure, a special ad hoc study can be undertaken to identify avoidable fixed costs.

The contrary position could be argued on two principal grounds. The first is that, as mentioned above, taking all costs into consideration encourages a long-term rather than short-term approach to decision-making. The other is that while some portion of depreciation is a sunk cost, the rest is avoidable cost. In practice, the degree to which depreciation is an avoidable cost varies considerably between different types of programs and services—depending on the “recoverable value” of the assets concerned. There are certainly some instances where depreciation is predominantly an avoidable cost.18

There is one other argument which is relevant here. This is the proposition that budgeting on a full accruals base is necessary to support a particular form of performance budgeting—the “purchaser-provider” model, which seeks to extend the principle of formula funding to simulate the market process whereby purchasers pay “prices” to buy products from arm’s-length producers. Exponents of the purchaser-provider model argue for the inclusion of depreciation because it is necessary for producers to aim to recover the full costs of production (including depreciation) through sales revenue. The merits of the purchaser-provider model are examined in Chapter 16.

With this one caveat, it can therefore be concluded that budgeting on a partial accrual basis certainly assists performance budgeting, but that the merits for performance budgeting purposes of budgeting on a full accrual basis are less clear-cut. (Note that these comments pertain only to budgeting, and not to financial reporting. It is perfectly possible to have full accrual financial reporting while budgeting on a partial accrual basis.)

The conclusion that accrual budgeting—at least in a partial form—assists performance budgeting should not, however, be interpreted as meaning that accrual accounting is a prerequisite for the introduction of performance budgeting. The distortions in the measurement of variable costs which result from the use of cash accounting—or at least from “modified” cash accounting—are probably often not large. There is, moreover, a long history of performance budgeting—particularly program budgeting and variants thereof—operating in the context of cash-based accounting and budgeting systems.

It should therefore not be assumed that performance budgeting and accrual accounting need to be introduced simultaneously. Decisions about the introduction and sequencing of accrual accounting need to be taken carefully, bearing in mind a range of considerations apart from any moves to performance budgeting (Diamond, 2002). These include the substantial resource implications of accrual accounting, whether basic public expenditure systems (for example, enforcement of budgetary spending limits and commitments control) are working well, and a number of other potential uses for accrual accounting information (for example, for aggregate fiscal policy). It should also be noted that accrual budgeting arguably creates some financial control risks of its own, although it is not possible to discuss these here.


Developing cost information to underpin performance budgeting is not a matter of simply seeking the fullest range of information on the costs of achieving results, using the most sophisticated and accurate accounting methodologies. Cost information strategy must, rather, be guided by choices about the form of performance budgeting being implemented, because different forms of performance budgeting have different information requirements. It must also be guided by recognition of the high marginal cost of better cost information and the consequent need to strike a balance between these costs and the decision-making benefits of better-quality information.

For performance budgeting which focuses upon improved expenditure prioritization through program choices, the challenge is to develop acceptably accurate measures of the decision-relevant costs of programs. Indirect cost attribution is one of the more important challenges which need to be addressed for this purpose. Careful thought needs to be given to the degree of precision being sought in indirect cost allocation, as well as to the implications of the design of the program structure for the magnitude of the indirect cost allocation task.

If the aim is to use cost information as the basis for linking results and funding more tightly, through formula funding or cost-based performance targets, considerably more cost information will be required. This is particularly true if it also desired to use comparative cost information to set efficiency targets. Given these high information costs, careful consideration will need to be given on a service-by-service basis as to whether the potential benefits are likely to outweigh these costs. It will also need to be recognized that in many cases the relationship between results and costs is not a tight functional one but is, rather, quite loose. This is particularly true for outcomes. However, it is also true for many public sector outputs, particularly those of a highly “heterogeneous” or “contingent capacity” nature.

Accrual accounting is a useful tool for better performance budgeting. However, it is not a prerequisite for performance budgeting, and it is perfectly possible to introduce performance budgeting in a cash (or modified cash) accounting and budgeting context.

Appendix: the cost-results relationship

Assume that for a particular “result” (output or outcome), there is what might be called a “prevailing (short-run) cost function” C = fp(Q) which indicates the quantity of results (Q) which the agency receiving funding is capable of producing with funding C. For this purpose, it is assumed that input prices and the level of fixed costs are constant, as is the case for standard cost functions. However, this cost function assumes a given (prevailing) level of technical efficiency (measured by a hypothetical index measuring the degree of technical inefficiency with which the entity concerned produces the result). (By contrast, of course, conventional cost functions assume there is no technical inefficiency.) If the result concerned is, as is usually the case, an output rather than an outcome, then it is assumed that Q measures the quantity of output of a defined level of quality. If such a cost function exists, then the starting point of a target-setting regime is that with funding of C* the agency could be expected to deliver a level of results of at least fP1(C*) This amounts to demanding that if the agency receives, say, an increase in funding, it should be expected to maintain at least its prevailing level of efficiency, and not to allow some portion of the funding increase to be absorbed in extra “slack.” The results target could be tightened further to put pressure on the agency to increase its level of technical efficiency—for example, by stipulating an expectation that gains in efficiency should produce additional results of a, in which case the performance target would be fP1(C*) + α. Equivalently, formula funding could be based on the prevailing cost function C = fp(Q), with C being the amount of funding provided to deliver results Q.

Basing targets or formula funding on unit costs (average costs) implicitly assumes that C = fp(Q) takes the form b.Q = C, which requires, improbably, that there are no fixed costs. Assume instead that there are fixed costs a, and that the prevailing production function takes the simple linear form a + b.Q = C (in other words, marginal cost is constant). Suppose that in this context, it is decided to base a funding formula or output target on average cost. Then suppose that in year 1, Qx is produced at total cost C1, so that average cost in that time period is C1/Q1, and this average cost measure is used to link funding and outputs for year 2. Suppose that the target output in time period 2 is Q2 Funding provided would then be a.(Q2/Q1) + b.Q2, whereas it “should” be a + b.Q2. The error built into the formula will be a.((Q2 - Q1) /Q1), which will be positive if target output is increasing, and negative if it is falling. If marginal cost is falling, rather than constant—which is what the cost curves employed in basic microeconomics assume within a certain range—then the funding error will be even greater.

If the cost function proper—that is, the conventional cost function assuming no technical inefficiency—is symbolized by C = f(Q), then the prevailing degree of technical inefficiency is measured by the excess cost of production, which is /p(Q) - f(Q)- The practical informational challenge is to derive some approximate measurement of this. There is a large productivity literature that seeks to quantify technical inefficiency using parametric and non-parametric methods (see the survey in Jacobs et al., 2006).


The author would like to thank Peter C. Smith for his valuable comments on an earlier draft of this chapter.

In a traditional system, the only thing coming close to this is the functional classification of expenditure under the standardized international Classification of the Functions of Government (COFOG). This is, however, an ex-post expenditure classification for information analysis purposes, rather than expenditure control.

Or transfers.

As indicated in Chapter 2, in most of the countries referred to in this paragraph, programs go by some other name—for example, “outcomes” or “output groups” in Australia and “Requests for Resources” in the UK.

And indirect variable costs.

See this chapter’s appendix for a formalization of these and related points.

Economic theory distinguishes, of course, between short-run and long-run cost functions. In principle, if one is using the cost function for formula-funding or short-run target-setting, the short-run production function would be appropriate. By contrast, for long-term expenditure projections, the long-run cost function would be appropriate.

In other words, to base the formula on average variable cost • quantity + fixed costs = total cost

At least if one holds constant a set of relevant variables, as discussed in the appendix.

To be precise, contingent capacity services also have a third characteristic—that the time lags involved in adding to production capacity (arising from factors such as the specialized nature of the human/physical capital used) exceed the desired service response time.

For the purposes of ABC, activities are defined into grouping of work tasks of a similar type. Thus ABC might employ the broad “activity” of human resource management, rather than distinguishing between the multiple specific types of human resource management activity, such as the processing of job applications.

Crucially, ABC/B involves a homogeneity assumption (Lathshaw and Cortese-Danile, 2002, pp. 31-2; Brimson, 1991, pp. 109-10), and is therefore only appropriate for standardized products (Mullins and Zorn, 1999, pp. 41-2). As discussed above, many public sector services are heterogeneous in nature, with the activity content of the service delivered varying between clients/cases. Similarly, ABB cannot deal with costs which are driven by capacity requirements rather than actual production volumes (Cooper and Slagmulder, 2000, pp. 85-6). Contingent capacity services, discussed above, are an extreme example of the role of such costs.

It should go without saying that this only applies after the point at which capital expenditure has been undertaken and the relevant production asset has been constructed. Prior to that point, all costs are avoidable and need to be assessed in project investment appraisal.

Expenses are, approximately speaking, the total costs to the inputs used to produce services in the current financial year irrespective of whether those inputs are paid for in the current year, were paid for in past years (for example, capital) or will be paid for in future years (deferred expenses).

Money paid out this year. Under “modified” cash accounting, funding is measured in terms of a mix of expenditure and commitments (obligations to make future payments).

The scope for such manipulation is reduced significantly by accounting for such items on a commitments basis, although this introduces some other distortions.

Accounting depreciation does not seek to measure the avoidable component of capital costs. If one wished to measure the avoidable cost associated with holding physical assets, one would have to base depreciation upon year-to-year changes in “net recoverable value” (NRV). This is quite different from standard accounting depreciation measures, even if based upon current cost accounting and its variants (for example, “deprival value,” “fair value,” and so on).

It also depends to some extent on which accounting methodology is used to calculate depreciation.


    BrimsonJ.A.1991Activity Accounting: An Activity-Based Costing Approach (New York: John Wiley and Sons).

    BrownR.E.M.J.Myring and C.G.Gard1999Activity-Based Costing in Government: Possibilities and Pitfalls,Public Budgeting and Finance Vol. 19(2) pp. 321.

    CooperR. and R.Slagmulder2000Activity-Based Budgeting—Part 1,Strategic FinanceSeptember pp. 856.

    DiamondJ.2002Performance Budgeting—Is Accrual Accounting Required?IMF Working Paper WP/02/240 (Washington: IMF).

    JacobsR.P.Smith and A.Street2006Measuring Efficiency in Health Care: Analytic Techniques and Health Policy (Cambridge: Cambridge University Press).

    KaplanR. and S.Cooper1991The Design of Cost Management Systems (Englewood Cliffs, NJ: Prentice-Hall).

    LathshawC.A. and A.Cortese-Danile2002Activity-Based Costing: Usage and Pitfalls,Review of BusinessWinter pp. 302.

    MullinsD. and C.Kurt Zorn1999Is Activity-Based Costing up to the Challenge when it Comes to Privatization of Local Government Services?Public Budgeting and Finance Vol. 19(2) pp. 3758.

    NewhouseJ.P.2002Pricing the Priceless: A Health Care Conundrum (Cambridge, Mass.: MIT Press).

    PeckenpaughJ.2002aThe Money Trail,Government Executive Vol. 34(5) p. 11.

    PeckenpaughJ.2002bTeaching the ABCs,Government Executive Vol. 34(4) pp. 416.

    RobinsonM. and J.Brumby2005Does Performance Budgeting Work: An Analytical Review of the Empirical LiteratureIMF Working Paper WP05/210 (Washington: IMF).

    SharpD. and L.F.Christensen1991A New View of Activity-Based Costing,Management Accounting Vol. 73(3) pp. 324.

    WilliamsC. and W.Melhuish1999Is ABCM Destined for Success or Failure in the Federal Government?Public Budgeting and Finance Vol. 19(2) pp. 2236.

    Other Resources Citing This Publication