16 Purchaser-Provider Systems

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

This chapter considers the scope for applying the purchaser-provider (henceforth “PP”) principle as a performance budgeting technique. It builds upon the previous chapter on formula funding systems, of which PP is one specific type. The chapter commences with a discussion of the properties of PP systems. It then identifies certain criteria which determine to which types of services the PP mechanism may be successfully applied. Following this is a discussion of the experiment with government-wide budgeting on a PP basis in Australia, which commenced in the late 1990s. The fourth section considers the ways in which agencies might be motivated to focus on their financial result under a PP arrangement. Conclusions then follow.

PP systems aim to simulate market transactions between arm’s-length purchasers and providers (suppliers) for the supply of products at pre-specified prices. They seek to create, by these means, a particularly tight link between results and funding in order to put strong pressure on the provider agency to maximize its technical efficiency. The PP mechanism was introduced in Chapter 1 by way of a short-hand definition based on two properties:

  • government pays the public sector service delivery (“provider”) agency a per unit “price” for its outputs, with the agency making a profit or a loss reflecting any difference between what it is paid and the cost of producing those outputs

  • the agency only receives payment for the outputs it actually delivers to the community (the principle of “payment-for-results”).

This can be regarded as a good working definition, upon which most of the analysis in this chapter will be based. It should, however, be emphasized that the definition should not be construed as suggesting that the unit price need be constant irrespective of the quantity of output “purchased.”1 It requires, rather, that funding be by formula. The definition is also a little too narrow in suggesting that PP arrangements may only be used to “purchase” outputs, and not outcomes, and also in implicitly focusing only on output quantity. It is true that, for reasons which are discussed further below, it tends to be difficult to apply the PP mechanism to the funding of outcomes or output quality. There are, nevertheless, some cases where this is possible. It is convenient in the analysis below to focus on the use of PP arrangements for funding output quantity, and to then consider separately (and more briefly) the scope for paying for outcomes or output quality.

The PP mechanism is capable of substantially boosting technical efficiency when applied to at least some public services (such as the hospitals via the DRG-based case payments system outlined in the last chapter2). The success of the PP principle in certain sectors has generated considerable international interest in extending its application more broadly within the public sector.

The broader application of the PP principle would be a particularly useful way of realizing the performance budgeting goal of linking funding and results in order to boost public sector performance. It is therefore highly desirable to identify the types of public service in relation to which this tool may be used to positive effect. With this in mind, the main objective of this chapter is to identify key characteristics of public services which determine whether or not they are suitable candidates for the application of the PP funding mechanism. A subsidiary objective is to clarify further the nature and modus operandi of PP systems.

The essence of the PP principle is the introduction of the distinction between the price received by the provider for its output and the cost incurred in producing that output, so that the agency will make a loss if its costs of production exceed the price paid. To achieve this, the price must be set at a level which will put pressure on the agency concerned to improve its technical efficiency, the degree of pressure imposed being a matter of policy choice.3

The PP principle therefore requires that the financial result become an important management and operational objective for the provider agency.4 A prerequisite of this is that the agency is able to measure its financial bottom line. To do this, it must adopt business accounting in the place of traditional government accounting, which treats government agencies as cost centers with no financial bottom line. Business accounting requires, first, accrual accounting, because accrual accounting is designed to produce an accounting measure of the costs of producing outputs. Second, it requires that the government’s “payments” to the agency for outputs be recognized in agency accounts as revenue, so that the operating statement measures the difference between the payments the provider receives for delivery of outputs and its cost of production. Such an accounting treatment contrasts with usual government accounting practice, in which funding from the budget is not accounted for as agency revenue because it is more in the nature of a spending allowance than market revenue.5

The question arises as to how, and indeed whether, the introduction of such a “quasi-market” arrangement changes the behavior of the provider agency and motivates it to improve technical efficiency. This issue is discussed in the final section of this chapter. In the meantime, it will be assumed for analytic purposes that the provider agency does have an incentive to worry about its bottom line. More precisely, the assumption is made that the introduction of a PP arrangement transforms the agency into a loss avoider, although not necessarily into a profit maximizer.

When does the purchaser-provider mechanism work?

Purchaser-provider arrangements are appropriate for some types of services, but not for others. This can be clarified by considering a number of concrete examples.

The first is veterinary services for emergency response to animal disease outbreaks. Such a service requires the maintenance on a continuing basis of an animal control disease center, with appropriate specialist veterinary expertise capable of determining the appropriate strategy to fight an outbreak. There needs to be an associated veterinary diagnostic laboratory armed with the specialist staff and equipment to enable it to identify a wide range of unusual and unexpected diseases. Contingency and support plans need to be prepared and periodically updated, and simulation exercises and training undertaken. Research capacity capable of diagnosing hitherto unknown diseases should, at least in wealthier countries, also be maintained (Geering et al., 1999).

Although it is possible to make some peripheral day-to-day use of certain of these facilities (for example, the diagnostic laboratory), the key characteristic of such a service is that it is not maintained in order to produce outputs on a day-to-day basis. It may, in fact, remain ready to respond to emergencies for years at a time without being called into action. A PP arrangement would therefore fail to provide such a service with the funding it requires.

This is an example of a contingent capacity service. As defined in Chapter 4, these 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. As pointed out in that chapter, there are many such services maintained by government, including the armed forces and a range of emergency services. For contingent capacity services, budget funding is not about paying for outputs, but rather about paying for the maintenance of the capacity to deliver essential outputs when and if they are needed.

The second example is investment facilitation services, that is, the service which many governments provide to potential large-scale foreign investors in order to encourage investment and boost growth. Such a service typically covers a range of assistance such as:

  • advice on and facilitation of development approvals (environmental, urban planning, health and safety, and so on)

  • identification, and assistance in the assurance, of necessary infrastructure and utility services

  • advice on industrial relations issues

  • (in many countries) the provision of financial incentives.

The extent of assistance provided to any given potential investor will tend to be highly variable, depending on factors such as the size of the potential investment, the complexity of the regulatory issues involved, and the political importance of the specific project. The cost per potential investor assisted (that is, per unit of output) could therefore be expected to be quite variable and unpredictable, not only within each financial year, but over time.

It would clearly make no sense to finance a service such as this on a purchaser-provider basis, paying the same “price” to the government agency concerned for each project for which it provided assistance. If such an approach were adopted, an obvious response by the agency would be to reduce significantly the degree of assistance offered to the most complex and important potential projects, so as to avoid making unsustainable losses.

Investment facilitation services are an example of an output type which is highly heterogeneous (non-standardized). As discussed in Chapter 3, a service is heterogeneous when there is a deliberate variation in the amount and/or types of activities received by each client/case of the “same” service, particularly in response to differences in client or case characteristics. The consequence of heterogeneity is that the unit cost of delivering the service will vary from one client/case to another, for reasons which reflect variations in case complexity and have nothing whatsoever to do with inefficiency (see Chapter 4).

This is not an isolated example. Heterogeneity is very common among services delivered by government (and, indeed, increasingly common in the economy as a whole). Consider just one other example—child protection services, which refers to the protection of children who are abused or at risk of abuse. The degree and type of assistance provided to each child can and should vary enormously, depending on a range of factors such as the degree of investigation required to ascertain whether abuse is occurring, the severity of the abuse, whether it is feasible to leave the child with the parent, whether psychological treatment is necessary, and the child’s geographical location. Again, paying a uniform amount per child assisted might well lead to a marked reduction in services to the children most in need of assistance.

The tendency of a PP system to result in the underservicing of more complex clients/cases is a familiar phenomenon which economists have labeled “skimping” (Allen and Gertler, 1991; Ellis, 1998). It is not, however, the only type of perverse behavior which can result from the inappropriate use of PP financing of a highly heterogeneous service. The other great danger is “cream-skimming,” which refers to the deliberate avoidance by providers of complex, and therefore high-cost, clients/cases.6

The heterogeneity problem is, of course, in no sense specific to the public sector—consider, for example, most types of legal services. It is for precisely this reason that many private sector services are not supplied by way of contracts with pre-specified output prices, but by a range of other forms of contracting many of which involve a substantial degree of risk sharing (see, for example, Williamson, 1985). The emergence of the “service society” has seen a major shift away from the dominance of standardized mass manufacturing towards a focus on heterogeneous services and goods produced in a working environment “more artisanal than industrial” (Cohen, 2006, p. 14).

It should not, however, be assumed that any degree of heterogeneity rules out the use of the PP mechanism. A PP arrangement may well work if, notwithstanding complexity-related variations in cost, the average cost per unit of output is reasonably stable—in other words, if the additional cost of the more complex cases is offset by the lower cost of less complex cases. If this is the case, the provider agency could finance the losses on more complex cases by the profits made on less complex cases. If the provider were a loss avoider (rather than a profit maximizer) it would then be under no pressure to engage in skimping or cream-skimming. Such a “swings and roundabouts” approach to higher-cost patients on the part of hospitals is an important reason for the success of the DRG hospital payments systems, notwithstanding the degree of complexity-related variability which exists even within the DRG output categories upon which the system is based. The DRG example highlights another important point—that average cost is, other things being equal, more likely to be stable if the quantities of output for the service concerned are large,7 as is the case for many hospital services.8

It follows from this that heterogeneity makes services unsuitable for the use of the PP mechanism when it is not possible to define output groups for which average cost is reasonable, stable, and predictable. This would not be true of investment facilitation services, nor probably for child protection services. A little reflection would suggest that a great many government-provided services are like this.

The other important criteria governing the scope for applying the PP mechanism is information costs. As highlighted in Chapter 4, more complex performance budgeting mechanisms such as PP are considerably more demanding of performance information. To implement a PP system, one has to develop reasonably reliable unit cost measures. Price-setting also requires relevant cost information (such as comparative cost information obtained through cost benchmarking) which can inform the purchaser’s decision about just how much price pressure to impose. Such information does not come cheaply, and will only be worth obtaining if the potential savings from increased technical efficiency for the services concerned are substantial. This will tend to rule out the use of PP arrangements except in the case of services which absorb quite a large amount of government expenditure.

In summary, the PP mechanism is most suitable for relatively standardized services produced in large volumes.

Against this background, we can return to the question of PP arrangements based on outcomes and output quality. As mentioned at the outset of the chapter, PP arrangements are normally applied to outputs rather than outcomes. There are some enthusiasts who have called for the widespread application across government of outcome-based PP contracts, on the familiar grounds that it is outcomes and not outputs which are the true objective of public policy (Osborne and Gaebler, 1993, p. 139). However, this is in general impracticable because of the two key problems identified in Chapter 3:

  • the uncertainty which often characterizes the relationship between funding and outcomes, due to the impact of “external factors”: if the provider agency has only limited control over the outcome, an arrangement in which its core funding is outcome-based may expose it to unsustainably large random volatility in revenue

  • the difficulty of measuring many outcomes, leading often to reliance on quite imperfect proxies.

Consider, for example, the practical implications of attempting to fund an investment facilitation service by the number of projects it succeeded in attracting, or of funding a hospital emergency service by the number of lives it succeeded in saving.

The impracticability of funding on the basis of outcomes is greater the higher-level the outcomes concerned, because both the attribution and measurement problems tend to be more severe. Conversely, the limited opportunities which exist to apply the PP principle to outcomes arise in respect to certain proximate outcomes which have a higher degree of controllability.9

By contrast, the incorporation of an explicit quality component in the payment mechanism in PP systems is an idea which is attracting growing interest in sectors such as health (Berwick et al., 2003; Salber and Bradley, 2002). The practical challenge of this is, of course, that of quality measurement. Notwithstanding the complexities of quality measurement, this is an area where, as Smith indicated in Chapter 15, progress may be expected in future.

Accrual output budgeting in Australia: a government-wide PP experiment

As outlined in Chapter 1, there were experiments in New Zealand10 (starting in the mid-1990s) and Australia (late 1990s) which sought to apply the purchaser-provider principle to the budget system as a whole. These experiments yielded important lessons about the limitations of the PP model. What follows focuses particularly on the Australian experience. This section is based partly on extensive interviews conducted by the author with officials of both the national and certain State governments.11

In Australia both the national government and all but one12 of its eight State and territorial governments attempted to implement what came to be known generically as “accrual output budgeting” (AOB). AOB was intended to transform the budget into a mechanism whereby the government would “purchase” outputs from its own agencies “at specified levels of quantity, quality and price” (VDTF, 1997, p. 11). As the then head of the Australian Finance Department put it, “the Government will know what [outputs] it is being offered and at what prices… [and] can then actively decide what it wants to buy and how much it wants to pay” (Boxall, 1998).13 Budget funding was reconceptualized as a “price” paid to agencies for delivery of outputs, as in a simple market transaction between arm’s-length purchasers and providers. It was intended that this system would place government-owned agencies and potential private suppliers of publicly financed services on an equal competitive footing, thus facilitating contestability and creating a “new competitive world” (Boxall, 1998).

Under AOB, the parliamentary budget appropriations mechanism was transformed to make the principal line of funding a “payment for outputs” appropriation based on the prices to be paid for outputs.14 As in New Zealand,15 there was an explicit intention to introduce into the general government sector the business distinction between the cost of production and the price at which products were sold, so that if output “costs exceed funding, agencies incur an operating deficit” (QT, 1998, p. 12). Accordingly, not only was accrual accounting introduced, but the payments for outputs received by ministries from the government were recognized as revenue in their accounts, giving government ministries for the first time a bottom-line profit/loss measure.

As in any purchaser-provider system, it was intended that the price paid for outputs would be set so as to impose pressure on agencies to improve their technical efficiency. The explicitly stated intention in all jurisdictions16 was that, after an initial period where for practical reasons the “price” would be based upon cost, there would be a rapid transition to a position where output prices would be set on the basis of what might be called the efficient cost of production, which would be measured by cost benchmarking or by market price references. As the Australian finance department put it, “departmental output appropriations will progressively be based on market (or benchmark) prices, rather than accrual based input costs” (DOFA, 1999, p. 27). This would mean that if an agency’s actual cost of production exceeded the “efficient” cost, it would run an operating deficit the magnitude of which would be indicative of the extent of its technical inefficiency. This was intended, in the parlance widely used in New Zealand, to transform “budget pressure” into “purchase pressure” (New Zealand Treasury, 1996); that is, to engineer a situation where agencies would be pressured to respond to reduced funding by improving technical efficiency rather than by cutting services.

The Australian finance ministry and a number of its State counterparts explicitly stated, at the time of the introduction of AOB, their intention of implementing the principle of payment-for-results. To be more precise, what they intended to do was to apply this principle in an accounting sense so that “recognition of Departmental output appropriations” would “reflect agencies [sic] delivery of their outputs” (DOFA, 1999, p. 43).17 What this meant was that agency accounts would record as revenue earned, not the actual budget appropriation, but that part of the budget appropriation which the agency was deemed to have actually earned by delivering outputs.18 Simplifying a little, if an agency were provided with $1 million in the budget in the expectation that it would deliver 1,000 units of output, but due to inefficiency spent its full appropriation and only delivered 600 units, it would be recorded as having only earned $600,000 and as therefore having made a loss of $400,000. The formal mechanism for such a system of payment-for-results was actually established in the State of Victoria. Each department in that State was for some years required to present quarterly “invoices” stating the outputs it had delivered. The idea was that the finance ministry would scrutinize the invoice and determine what payment was due and could be recognized as income in the relevant department’s accounts (VDTF, 1998, pp. 302-3).

A number of key difficulties arose during the implementation stage of the AOB reforms. The first was that the demanding performance information requirements of the system were not met. AOB required the clear specification of the outputs to be delivered, and good measurement of the quantity, quality, and cost of what was actually delivered. In practice, however, there was a “lack of rigour in the definition and measurement of outputs” (Carlin and Guthrie, 2001, p. 97; 2003, p. 154). Given the large number of types of service delivered by governments, this was hardly surprising. Even in a country which had moved some years before to develop performance measurement systems, the performance measurement requirements of the new AOB system were a formidable challenge. Output costing systems also tended to be quite inadequately developed, echoing earlier experience in New Zealand, where a 1995 review determined that cost allocation to specific products was so “inappropriate” that the resultant information “fell well short of playing a useful role for management decision-making” (Coopers & Lybrand, 1995).

A second key difficulty arose from a certain imprecision about exactly what it meant when AOB proponents said that budget funding would be based on “prices.” The logic of AOB required the specification of per-unit prices for each output (or, more generally, a funding formula). Without a formula to indicate how much revenue was payable for any given quantity of output delivered, there would be no way that the finance ministry could implement the principle of payment-for-results. It would be unable to determine, in the event of the failure of an agency to deliver the full quantity of output which was expected of it, what portion of its “payment-for-outputs” appropriation it would be permitted to recognize as revenue in its operating statement.

The official documentation at the time of inception of the system made it clear that, reflecting the market analogy upon which AOB was based, many of the system’s proponents did indeed have in mind a model based on unit prices of outputs. Thus, for example, the Victorian finance ministry stated that “[budget] funding will be based on a single quantity and unit cost measure for each output. Major quantity measures should allow a quantity x unit cost = expenditure equation to be derived where possible” (VDTF, 1997, p. 18).

Very little consideration had, however, been given to the practicability of the principle of unit pricing, and it quickly became clear that there were many services for which formularized funding based upon unit prices was difficult to implement. There was great puzzlement as officials sought to apply the principles of AOB to heterogeneous and contingent capacity services such as defense, international representation, and environmental regulation. In the upshot, there was very little success across the country during the implementation phase in developing funding formulas to underpin the “prices” set in budget appropriations. The greatest efforts made in this respect were at the national government level, where the finance ministry attempted to give effect to the New Zealand idea of “purchase agreements” (New Zealand Treasury, 1995, 1996) with individual ministries (see Box 16.1). This had limited success, and led to only a small sub-set of the services delivered by that government being placed on a formula funding footing.

Box 16.1.Setting output prices for immigration services

The experience of the Australian immigration department—an agency which produces more relatively standardized services which are suited to formularized output funding than most other national government agencies—exemplifies the difficulties which arose with formula funding as the basis of PP arrangements. At the time of introduction of AOB, the finance ministry and immigration department entered into a formal “purchasing agreement”a which set unit prices for many of the department’s outputs. There were, for example, unit prices for each of 14 different categories of visa application. Prices were also specified for each of three different categories of “removals” of illegal immigrants—that is, the management of forced repatriations. Even though the purchasing agreement only covered those immigration services which appeared more suited to the purchaser-provider treatment, the problem of heterogeneity still manifested itself. For example, one of the visa output categories used for funding purposes covered a number of complex types of visa application including spouse visa applications—that is, applications for entry status by persons on the grounds of marriage to a citizen. (Immigration’s mandate in respect to such applications is to prevent abuse of spousal entry through contrived marriages contracted solely for immigration purposes.) It became clear, however, that there is such great variability in the extent of investigative activity which spouse visa applications require as to make this particular type of visa application entirely unsuited to funding via a fixed per-case amount. Immigration might, for example, be faced in a given year with a set of spouse visa applications of more than usual average complexity, in which case it would find itself under considerable financial pressure to investigate cases with less than adequate thoroughness.

a An interim purchasing agreement with the immigration ministry was formulated in 1999-2000, and by 2001 this had developed into a three-year agreement covering 2001-04 (DIMA/DOFA, 2001).

Closely related to this was the third problem—the difficulty of implementing the principle of payment-for-results. Victoria was the only jurisdiction which made an across-the-board attempt to do so. Its quarterly invoicing system did not achieve its intended purpose. During the time the invoicing procedures remained in place, there was no instance where an agency was not permitted to recognize its full payment-for-outputs appropriation as revenue. The Victorian finance ministry made a couple of initial attempts to so penalize departments which it saw as underperforming, but these were blocked at the political level: no minister was prepared to accept what would in effect be a public censure of his or her department’s performance. The fundamental technical problem with the system was the inevitable subjectivity of any assessment of departmental delivery performance in a context where there was no unit price for outputs and, in many cases, not even good measures of output quantity.

At the national government level, the initial enthusiasm of the finance ministry to apply the principle of payment-for-results led to robust interaction with the Auditor-General. The latter was concerned that revenue recognition based upon highly subjective assessments of output delivery performance would create wide scope for the manipulation of agency accounts. In this debate, the Auditor-General was in a strong position because of his power to qualify agency accounts, and Finance was obliged to largely give way. The upshot was agreement that budgetary payment-for-outputs appropriations would be recognized as revenue at the time the appropriations legislation became law, unless there was a purchasing agreement or something similar which would clearly state how output delivery performance would be measured (DOFA, 2002, p. 20; ANAO, 2002, pp. 24-5). Because purchasing agreements made so little headway, the second option remained a dead letter.

The fourth key problem was the difficulty of giving effect to the stated intention to pay prices based on the efficient cost of production. AOB required finance ministries to determine the efficient cost of production of each of the myriad of outputs produced by budget sector agencies. This would have required a monumental effort even if the necessary analytic techniques had been available. However, the AOB enthusiasm for the use of market price comparisons and benchmarking overlooked the considerable practical limitations upon the use of those techniques. Benchmarking can be a most valuable management tool for services where good benchmarking partners are available. However, there are many areas of government where it is very difficult or even impossible to find other organizations sufficiently similar to permit cost-effective benchmarking. Moreover, even where benchmarking is feasible, it remains a costly and demanding process. As for market price comparators, these tend to be available only for a minority of government outputs (indeed, where they are available, the question arises as to why the government is producing the service concerned rather than buying it from the private sector or leaving it to market provision).

The difficulty of determining the “efficient” price was experienced most intensely in the course of an “output pricing review” process which was initiated at the national government level at the outset of the AOB experiment. The intention of the finance ministry was to conduct an intensive review of every ministry budget over a period of a couple of years in order to determine the appropriate prices which the government should be paying for departmental outputs. This process turned out to be not only exhausting, but also ultimately unsatisfactory, to both Finance and the line agencies concerned. Output pricing reviews which were intended to be accomplished in one year extended to two and sometimes three years. Even then, it turned out to be frustratingly difficult to draw strong conclusions about either the scope for efficiency gains or priority-setting issues. Benchmarking to determine efficient cost proved to be very difficult.19

The consequence of this series of implementation difficulties was that in no jurisdiction did it prove possible to realize the original vision of AOB—that is, of a system in which agencies would experience intense “purchase” pressure to improve efficiency as a result of being forced to earn revenue for outputs actually delivered at prices reflecting efficient operation. This mirrored earlier experience in New Zealand.20 At root, this failure reflected both the inappropriateness of the PP model for many government services, and the formidable informational challenges of implementing such a system.

By 2003, there was a widespread view within the bureaucracy in Australia that AOB had not been a success. There was also, according to senior officials, considerable ministerial dissatisfaction with the system.21 The result was the dismantlement of many of the purchaser-provider elements of the system. At the Commonwealth level, for example, the output pricing review process was a formally dropped as the result of a Budget Estimates and Framework Review recommendation that as of July 1, 2003, these reviews be “rolled into a broader input-based cost analysis model involving a detailed analysis of agency expenditure.” Benchmarking essentially disappeared as an element of the central budget process. The “purchasing agreements” and most other formularized output funding arrangements between Finance and ministries—including that with Immigration—also disappeared. The AOB appropriations framework, with its “payment-for-outputs” appropriations, remained, but essentially as a matter of form.

The PP experiment was a bold attempt to revolutionize public budgeting so as to apply strong pressure for improved efficiency. It was, in the words of one senior official, a “brave experiment…that didn’t quite come off” (quoted in Kelly and Wanna, 2004, p. 106).

Profits, losses, and the motivation to perform

It is via the financial bottom line that PP aims to motivate government agencies to perform better. But why should the agency be concerned about its financial result? Manifestly, it cannot be because its owners’ incentives have changed: if government is both owner and purchaser, the financial consequences for it of any given level of agency inefficiency are the same whether there is a PP contract or not. Clearly, the objective of the PP mechanism must be to improve the motivation of agency management and staff (“agents”) to perform, just as in the case of internal purchaser-provider arrangements within private enterprises. What are therefore crucial are the consequences of the agency’s financial performance and how these impact on agents.

A crucial question here is whether the “revenue earned” by the provider agency for the delivery of outputs actually determines the funding made available to it by government. In sectoral PP arrangements with, say, hospitals and universities, this is the case. By contrast, as outlined above, AOB envisaged that revenue earned would be an accounting entry distinct from the amount of budget funding.

If revenue earned determines funding, an agency which is making losses will receive a level of government funding insufficient to cover its operating costs. This would place it under strong pressure to reduce its costs, and might ultimately put it at risk of closure. The unpleasant consequences for agents could be expected to motivate loss-avoiding behavior. Such motivation may also be reinforced by the non-self-interested commitment of staff to the objectives of their organization, which would further strengthen their desire to see their organization survive (see Chapter 18). The result would be to transform the agency into a loss avoider. The impact may, however, be asymmetrical, generating little motivation to make profits. This would be consistent with evidence that the DRG payments system works principally by motivating hospitals to seek to avoid losses—the “pressure matters” thesis (see Chapter 18).

Such an approach requires that the provider agency be accorded key operational freedoms normally accorded to public enterprises and private businesses. In order to be able to cut expenses when required, the agencies would require considerable flexibility in areas such as hiring and firing staff and setting salaries. Crucially, it would also require the power to borrow in order to be able to remain afloat in the face of losses, because it could not be expected that any and all losses could be offset immediately by cutting wages and other expenses.22 Such a degree of operational freedom is not usually enjoyed by government ministries—even when accorded the enhanced autonomy advocated by “management-for-results”—and creates certain risks which need to be carefully weighed prior to any decision to adopt this version of the PP mechanism.

A further important consideration in deciding whether to take such an approach is the credibility of the threat (implicit or explicit) of closure of chronically loss-making agencies. This threat will be most credible if there are alternative service providers (for example, other hospitals) or if the service concerned is one which the government could readily cease providing. Conversely, if the supplier agency is the sole supplier of a crucial public service, no threat of closure is likely to be credible.

It was perhaps because such an approach seemed unsuited to the funding of many core government ministries that AOB exponents envisaged the alternative of treating revenue earned as purely an accounting entry.23 The implication of such an approach is, of course, that any loss made as a result of the cost of production exceeding the output price will also be a purely accounting loss, with no direct impact on the resources available to the agency. To make the agency focus on its financial performance, it would be necessary for the government as owner to establish specific sanctions and rewards for management based on the financial result. The efficacy of such an approach would, however, depend critically on the credibility of financial result as a measure of agency performance, which would in turn depend on factors including the appropriateness of the output prices set. If the PP mechanism was applied to the wrong type of service, with the consequence that the price was an arbitrary one, it could not be expected that the financial result would have credibility as a performance measure for either the agency or the government.


The analysis in this chapter suggests that the purchaser-provider form of performance budgeting can serve as a powerful instrument with which to pressure government agencies to improve technical efficiency. It is a mechanism that can be effectively used as the basis for sectoral funding systems for services which are relatively standardized and produced in large volumes. There are, however, many government services which are unsuited to the application of the purchaser-provider principle, including services which are characterized by high levels of heterogeneity or which are “contingent capacity” in nature. Because the purchaser-provider model is only suitable for selective application, it is an inappropriate model upon which to base a government-wide budgeting system.


One related additional point on definition: PP systems do not necessarily require that all funding provided is a function of results delivered. Some PP systems are hybrids in which, in addition to the payment-for-results, agencies receive some element of funding which is not based upon results delivered.

See Robinson and Brumby (2005) for a review of the empirical evidence on the success of this system.

Imposing pressure for improved technical efficiency requires, in a static context, that the price set is below the costs of production prevailing prior to the introduction of the PP system. The greatest pressure would be imposed by setting prices to reflect cost which would prevail if there were 100 percent technical efficiency. In a context where there are multiple providers of the service, one less stringent possibility is to set prices equal to the average costs of all providers so as to put pressure on those with below-average levels of efficiency (an approach mentioned in Chapter 4 in relation to hospital funding). In a dynamic context, pressure would be imposed over time by setting an initial price equal to the costs of production, but then increasing it more slowly than the pre-PP trend rate of cost increase for the service concerned. This was, approximately speaking, the objective when the first version of the DRG system—the Prospective Payment System—was introduced in the United States with the objective of restraining the rate of growth of hospital costs.

This does not, however, necessarily mean that these entities are expected to operate as profit maximizers. What it does certainly require is that there is a strong expectation that they should operate with the aim of avoiding losses.

Accrual accounting and the treatment of budget funding as revenue are quite separate matters. The introduction of accrual accounting in a government context does not in itself require that budget funding be treated as revenue. If there is no purchaser-provider arrangment, there is no logical reason for treating budget funding as revenue, nor any case for producing an agency-level profit/loss figure. To put the matter in business terminology, it is because PP seeks to turn government agencies into profit centers (rather than cost centers) that both elements of business accounting are required.

The avoidance of high-cost cases is known as “dumping.” Cream-skimming can involve either or both dumping and “creaming,” which is the deliberate selection of low-cost clients/cases (and, in the extreme, the delivery of services to clients/cases who don’t need them).

This is related to what statisticians call the “law of large numbers.”

Another key point highlighted by the DRG example is that the severity of the output heterogeneity problem can sometimes be reduced by defining output types more narrowly so as to reduce the cost variability (to make them more “iso-cost,” in DRG terminology). Take, for example, the treatment of hip fracture patients, where it transpired that there was such variability in case complexity that it was problematic to pay the same amount for the treatment of every patient. The response, as the DRG system evolved, was to split hip fracture patients into a number of groups—such as elderly hip fracture patients and hip fracture patients with secondary medical problems—attracting different prices. This is an example of what Smith refers to in the previous chapter as finer “risk adjustment.” There are, however, limits to how far heterogeneity may be resolved by adopting finer output classification. DRG systems also make use of special payment arrangements for high-cost “outlier” cases to further reduce the heterogeneity problem. This, however, raises the further problem of how to prevent the inappropriate classification of cases as outliers.

Such as the Ethiopian university funding system described in Chapter 24, which pays universities for successful course completions (a proximate outcome measure) rather than simply for courses undertaken (an output measure). What is crucial here is that the proximate outcome upon which payment is based is sufficiently controllable by the provider agency to make the revenue risk manageable.

For the New Zealand experience, see, for example, Schick (1996) and Robinson (2000).

Including, but not confined to, intensive interviews conducted in Canberra in 2003. State government officials were also interviewed in Victoria, Queensland, and Western Australia.

The exception being Australia’s largest State, New South Wales. The very different approach to performance budgeting taken in that State is the subject of Chapter 14.

It was curious given these clear statements of the output-purchase philosophy of the system that the national finance ministry subsequently rejected the “purchaser-provider” label for its new budgeting system (DOFA, 2001), and claimed that it was different from the New Zealand system in that it supposedly put equal emphasis on outcomes as on outputs.

The new appropriation regime—which included “capital injection” appropriations and a distinction between departmental and “administered” expenses—is discussed in detail in Robinson (2002a).

Where Ian Ball, the principal finance ministry architect of the system, described it in the following terms: “it means revenue and cost flows are split. ‘Surplus’ then has a real meaning and can be used as a key ownership performance measure” (Ball, 1992, p. 21).

See Robinson (2002b, p. 19, n6) for references to statements of this position in the official documentation of State governments.

As the Western Australian Treasury put it, that the “recognition of appropriation revenues” were to “be tied to output acquittal performance” (WAT, 2001a, p. 13).

As the West Australian Treasury (2001b, p. 5) put it: “Agencies will only be entitled to recognise appropriation drawdowns as revenue ‘earned,’ when the associated output delivery has been successfully acquitted. Appropriations drawn prior to the acquittal process, will be recognised as revenue in advance (a liability). If an agency’s outputs are not acquitted (in full or in part), an operating loss will be reported for the period.)”

Some agency personnel interviewed suggested that an additional reason for the lack of success of benchmarking in certain areas where intra-government benchmarking was attempted (for example, policy advice) was an uncooperative stance taken by many government departments to sharing cost data.

Where Coopers & Lybrand (1995, p. 40) found in a review commissioned by the finance ministry that there was “little evidence of significant purchase pressure (as opposed to budget pressure).”

The author was told this by a number of senior officials during interviews in 2003. It is also confirmed in Kelly and Wanna (2004).

In a private sector context, if losses are made, it is rarely the case that employees bear the full consequences of any losses, particularly in the short term. A key ownership function is risk taking, and economists analyse the appropriate allocation of risk between owners and their agents by reference to “optimal risk sharing.” It is one of the virtues of market mechanisms that external capital—whether in the form of equity or loans—breaks the short-run nexus between a firm’s earnings and its operating costs.

The New Zealand Treasury (1996, p. 38) at one stage floated the idea that “in the future departments may possibly be required to show their outputs have been produced before the Government pays the ‘price’ and hands over the appropriated output funding.” The idea was not implemented.


    Other Resources Citing This Publication