This manual is designed to accompany the courses on performance budgeting delivered at the regional Centers for Learning on Evaluation and Results (CLEAR). It presents summaries of the essential content for each of the core topics covered in those courses. This manual can therefore be used on a “stand alone” basis. However, it may be used to best advantage in conjunction with the support of the key readings listed at the end of each section.
CLEAR Training Materials
Table of Acronyms ........................................................................................................................................5
1. Budgeting and the Budget Process........................................................................................................... 8
2. Overview of Performance Budgeting...................................................................................................... 12
Expenditure Prioritization....................................................................................................................... 12
Models of Performance Budgeting......................................................................................................... 14
3. Performance Information Fundamentals ............................................................................................... 20
Performance Concepts: the Results Chain.............................................................................................. 20
Outcomes vs. External Factors................................................................................................................ 23
Outputs vs. Outcomes and Activities...................................................................................................... 24
Performance Indicators .......................................................................................................................... 24
4. Performance Measures and the Budget................................................................................................. 28
Ensuring that Indicators are Useful ........................................................................................................ 30
Linking Funding to Outcomes ................................................................................................................. 31
Linking Funding to Outputs..................................................................................................................... 32
5. Performance Measurement Systems...................................................................................................... 35
Ensuring that Performance Information is Cost‐Effective ...................................................................... 39
6. Evaluation and Performance Budgeting ................................................................................................. 42
Evaluating Program Effectiveness........................................................................................................... 43
Efficiency Review ....................................................................................................................................45
Making Evaluation Relevant to the Budget ............................................................................................ 45
Evaluation and the Budget: The Chilean Example .................................................................................. 46
Internal versus External Evaluations....................................................................................................... 48
7. Performance Auditing ............................................................................................................................. 50
8. Program Budgeting Fundamentals ......................................................................................................... 52
Performance Pressure and Program Budgeting ..................................................................................... 54
Program Performance Information ........................................................................................................ 55
Program Titles and Objectives................................................................................................................ 56
Program Statements............................................................................................................................... 58
The Program Hierarchy ........................................................................................................................... 58
Number and size of programs................................................................................................................. 59
Developing Programs.............................................................................................................................. 60
Integrating Strategic Plans with the Program Budget ............................................................................ 61
9. Program Appropriation and Expenditure Control .................................................................................. 64
10. Program Accounting and Costing.......................................................................................................... 68
Budget Classification and Chart of Accounts.......................................................................................... 68
Financial Management Information Systems......................................................................................... 69
Program Cost Allocation ......................................................................................................................... 72
Administration Programs........................................................................................................................ 75
Salary Costs and Programs...................................................................................................................... 78
Cost Allocation and the Program Hierarchy ........................................................................................... 78
Program Cost Estimation in Budget Preparation.................................................................................... 79
11. Special Topics in Program Budgeting.................................................................................................... 82
Programs and Organizational Structure ................................................................................................. 82
Ministry Boundaries and Internal Support Services ............................................................................... 85
Broader Conflicts between Program and Organizational Structure ....................................................... 87
Program Structures in Developing Countries ......................................................................................... 88
Functions and Programs ......................................................................................................................... 89
Programs and Functions: a Case Study ................................................................................................... 92
Programs and Levels of Government...................................................................................................... 93
12. Formula Funding and Purchaser Provider Systems .............................................................................. 95
Limits to Formula Funding and Purchaser Provider................................................................................ 96
13. Targets and Performance Budgeting .................................................................................................... 98
Gaming and Perverse Effects................................................................................................................ 100
14. Public Financial Management Reform Foundations........................................................................... 102
15. Expenditure Prioritization and Performance Budgeting..................................................................... 105
Expenditure Prioritization Processes More Generally .......................................................................... 109
Planning and Prioritization.................................................................................................................... 111
Expenditure Prioritization at the Sector and Ministry Levels............................................................... 111
16. Accrual Accounting and Performance Budgeting ............................................................................... 114
What are Accrual Accounting and Accrual Budgeting? ........................................................................ 114
The Complexity of Accrual Accounting and Budgeting......................................................................... 119
Accrual Accounting and Program Budgeting ........................................................................................ 120
Accrual Accounting, Purchaser‐Provider and Formula Funding Systems.............................................121
17. Parliament and Performance Budgeting............................................................................................. 124
Parliament and Performance Accountability........................................................................................ 124
Parliament and Expenditure Priorities.................................................................................................. 126
18. Medium‐Term Budgeting and Performance Budgeting...................................................................... 127
Nature and Benefits of MT Budgeting .................................................................................................. 127
MT Ceilings: Fixed vs. Indicative?.......................................................................................................... 129
MT Budgeting and Performance Budgeting.......................................................................................... 131
The Medium‐Term Expenditure Framework (MTEF)............................................................................ 132
Developing Countries and the MTEF .................................................................................................... 134
19. Implementing Government‐Wide Performance Budgeting ............................................................... 137
20. The Managing for Results Context...................................................................................................... 141
21. Concluding Comments........................................................................................................................ 143
Table of Acronyms
AA Accrual accounting
ABC Activity based costing
AOB Accrual output budgeting
BC Budget classification
COA Chart of accounts
COFOG Classification of the Functions of Government
DRG Diagnostic related group
FF Formula funding
FMIS Financial management information system
GBE Government business enterprise
IFMIS Integrated financial management information system
IT Information technology
LOLF Loi organique relative aux lois de finances
MFR Managing for results
MoF Ministry of finance
MT Medium term
MTEF Medium‐term expenditure framework
MTFF Medium‐term fiscal framework
OECD Organization for Economic Cooperation and Development
PB Performance budgeting
PFM Public financial management
PM Performance measurement
PSA Public Service Agreement
SAI Supreme audit institution
ZBB Zero‐base budgeting
This manual is designed to accompany the courses on performance budgeting delivered at the
regional Centers for Learning on Evaluation and Results (CLEAR). It presents summaries of the
essential content for each of the core topics covered in those courses. This manual can
therefore be used on a “stand alone” basis. However, it may be used to best advantage in
conjunction with the support of the key readings listed at the end of each section.
1. Budgeting and the Budget Process
Government budgeting is the allocation and use of resources, and associated decisions about
how the resources used will be acquired, by that part of the public sector which is financed
primarily by compulsory charges such as taxes1
. Equivalently, government budgeting –
“budgeting” for short in what follows — can be said to be about the allocation of resources for
the provision of services and transfers on a non‐market basis, where “non‐market” means
either free or greatly below cost.
The government budget is therefore a financial plan which covers those public sector agencies
– such as ministries and the social security system – which are primarily tax‐financed. It does
not typically cover public corporations, which normally manage their finances on a more
Budgeting is a process involving four stages:
Aggregate fiscal policy formulation: this refers to the determination of the government’s
overarching objectives for the budget deficit, debt and other relevant fiscal aggregates3
should then be translated into decisions about the desired levels of aggregate revenue and
Budget preparation and enactment: in this stage of the budget process the government decides
how much funding it will provide to which agencies and for which purposes. This is given formal
expression in the budget law and budget regulations, which are enacted by the legislature and
the highest executive organs of government (i.e. the president or cabinet of ministers).
Budget execution: this refers to the carrying out of the expenditure plan developed in the
budget – including the entering of contracts and expenditure of funds.
1 We use the term “tax” as shorthand to include other compulsory revenue sources such as mandatory levies or
fees, and fines. Budget financing via borrowing may be considered to be financing by deferred taxes. 2 Public corporations are financed primarily by the prices which they charge the consumers of their services –
prices which are voluntary in that consumers are not legally obliged to purchase the services concerned. The
budget normally only relates to public corporations insofar as they either receive taxpayer subsidies or their profits
are a supplementary source of budget funding. 3 Depending on the policy framework of the country concerned, the other fiscal aggregates in terms of which fiscal
policy is framed might include, e.g., the level of “net worth” (government assets minus liabilities) or the
percentage of revenue/GDP.
Accounting, Auditing and Reporting: this covers the preparation of accounting records of
government spending and revenue, their auditing by both internal and external auditors, and
the provision of reports on budget execution to government agencies, ministers, parliament
and the public.
Budgeting has three key objectives
Sound macro‐fiscal outcomes: this is primarily the product of good aggregate fiscal policy.
Aggregate fiscal policy needs to ensure “fiscal sustainability” – which means essentially that
deficits and debt do not get out of control – because this is important for economic
confidence and stability. If budgeting does not keep debt to sustainable limits, financial
markets can potentially lose confidence in the security of lending to government. The result
may be a sudden increase in the interest rates which government is forced to pay and, in
the extreme, an unwillingness to lend to government. In addition to fiscal sustainability,
aggregate fiscal policy can play an important “stabilization policy” role – in particular,
through providing support to the economy through additional spending or lower taxes
during a recession.
Appropriate prioritization of expenditure: this refers to the allocation of funds to the sectors
and programs which are most effective in meeting social needs. It means, for example, that
if a country is facing a major new health challenge, additional funds will be allocated in the
budget for programs designed to tackle that challenge. The corollary of this is that funds are
moved away from sectors and programs where spending is low priority.
Service effectiveness and efficiency: this means, firstly, that government services are
delivered efficiently and, secondly, that they are designed and managed so as to maximize
Performance budgeting, as we will see, focuses directly on improving the ability of budgeting in
to deliver on the second and third of these objectives.4 “Effectiveness” and “efficiency” are
therefore key words here. Effectiveness refers to the extent to which a service delivers the
benefits which it is supposed to deliver to society. For example, an HIV/AIDS prevention
campaign is effective if it succeeds in reducing the rate of new HIV/AIDS infections. As will be
explained later, effectiveness is about the extent to which programs achieve their intended
outcomes. Efficiency, on the other hand, refers to delivery of services at the lowest possible
4 However, in doing so, it can also indirectly improve macro‐fiscal outcomes.
cost, without sacrificing quality. Expressed differently, efficiency is about minimizing waste in
the production of government services.
Budget preparation should start with, and be carried out within the context of, good aggregate
fiscal policy. In other words, decisions about how much funding to provide agencies should be
constrained by clear policy about the level of aggregate expenditure which is, given
government revenues, consistent with government’s deficit and debt objectives. Key aspects of
budget preparation and enactment include:
The determination of expenditure priorities,
Estimates of spending requirements – particularly associated with ongoing commitments
such as the government wage bill, capital projects underway, and expenditure obligations
on “mandatory” items such as pensions and other social security payments,
The formulation of current and capital budgets, which require different budget preparation
methods but which should at the same time be linked.5
The formulation of budget funding decisions into legal authorizations (“control totals”)
which can be “appropriated” by the parliament or promulgated by the executive
The way in which budget preparation is organized differ considerably between different
countries. Some key aspects of budget preparation which have a particularly close bearing on
performance budgeting – such as expenditure prioritization and medium‐term budgeting
methods – are discussed in this manual.
Budget execution, on the other hand, can be seen as a clearly defined sequence of stages, as
Authorization stage: based on the parliamentary budget appropriations, the finance
ministry communicates to ministries their expenditure entitlements. The nature of these
authorizations differs internationally. In some countries, authorizations are made on a
month‐by‐month basis – that is, each month ministries are told how much money they will
have to spend in the coming month.
5 Building major new capital assets such as schools, hospitals and roads has important implications for future
current expenditure, whether on staff and supplies to run the facilities concerned, or on the maintenance of those
Commitment stage: a commitment is a contract or some other obligation to make a future
Verification stage: ensuring that the service or goods previously contracted for have been
delivered, and meet the specifications of the contract.
Payment authorization stage: in the light of verification, approval for the payment to be
made is given by an official independent of the persons who initiated the commitment.
Payment stage: the payment is actually made.
Accounting stage: each payment is entered into the accounting system and its relevant
characteristics recorded. In a well‐developed system, the accounting system will also record
the commitments stage.
Auditing stage: the accounts are audited – firstly by “internal auditors” (within the spending
ministry concerned) and subsequently by “external auditors” (the auditor‐general or audit
Performance budgeting aims to change budget preparation fundamentally, linking it much
more systematically to the effectiveness and efficiency of expenditure. In doing so, however, it
also brings about equivalently major changes in budget execution.
Potter, B. and J. Diamond (1999), Guidelines for Public Expenditure Management (Washington: IMF),
available at http://www.imf.org/external/pubs/ft/expend/index.htm.
Schiavo‐Campo, S. & D. Tommasi (1999), Managing government expenditure (Manila : Asian
Development Bank), obtainable at http://www.adb.org/documents/manuals/govt_expenditure/.
2. Overview of Performance Budgeting
Performance budgeting aims to improve the effectiveness and efficiency of public expenditure
by linking the funding of public sector organizations to the results they deliver. It uses
systematic performance information (indicators, evaluations, program costings etc) to make this
link. The impact of performance budgeting may be felt in improved prioritization of
expenditure, and in improved service effectiveness and/or efficiency.
PB usually also emphasizes giving government agencies and their managers greater flexibility in
the use of resources than they would typically have under traditional tightly‐controlled public
management systems. A key element of this is greater flexibility in the choice of the mix of
inputs which are to be used to deliver services (e.g. how much labor input vs. externally sourced
inputs, the mix of types of externally‐source supplies and services used). An important
implication of this is the need for more flexibility human resources management, a topic
discussed in a later section.
The increased international interest in performance budgeting has been prompted in part by a
recognition that it is all too easy in Government to lose sight of the fundamental objective of
delivering positive outcomes to the community. Public sector organizations which are financed
by taxes and other compulsory charges lack the market disciplines which compel commercial
enterprises, particularly those operating in highly competitive markets, to be customer‐
oriented. Political accountability through the electoral process is, of course, extremely
important, but is not necessarily sufficient to ensure that public sector organizations are highly
focused upon the results they deliver.
Expenditure prioritization is often quite weak in government. A particular concern is the
tendency for much ongoing program expenditure to escape serious scrutiny, and for budgetary
decisions to be mainly focused on what new spending initiatives to adopt. The term
“incrementalism” is widely used to describe this tendency.6
If prioritization processes are not well developed, governments typically rely, if and when they
need to make spending cuts, on “across‐the‐board” cuts – that is, cuts of the same percentage
6 For more on incrementalism, see Robinson and Brumby (2005).
to all ministries. Across‐the‐board cuts are by definition indiscriminate and inferior to a selective
approach which would target cuts at lower‐priority or less effective programs.
Improved expenditure prioritization can help to achieve more sustainable public finances. If, for
example, spending cuts are needed to achieve “fiscal consolidation” (i.e. to get deficits and debt
under control), the capacity to target cuts selectively at the lowest priority programs will help to
make those cuts more sustainable (i.e. the reduce the pressures to reverse them), thus
increasing the probability that the improvement to the overall state of public finances will be
Performance budgeting should be viewed in the broader context of a set of related “managing‐
for‐results” (MFR) reforms. MFR can be defined as the use of formal performance information
to improve public sector efficiency and effectiveness. Its fundamental starting point is maximum
clarity about the outcomes which government is attempting to achieve, and about the
relationship of outputs, activities and resources used to those desired outcomes. Good strategic
planning and business planning are an essential element of MFR. MFR also tends to emphasize
the ex ante stipulation of performance expectations for agencies, work units and individuals
through the use of performance targets and standards.
A standard element of the “strategic human resources management” component of MFR is the
introduction of stronger performance‐based extrinsic incentives (rewards and sanctions) for
public officials. Typically, this is accompanied by greater flexibility of employment, including
greater capacity to sanction or dismiss poor performers, and greater ease in transferring or
terminate employees in programs which the government is eliminating or cutting back.
Examples of other elements of MFR reforms which have been introduced in many countries
Customer orientation measures: an example of such a measure is “client rights charters”,
which define client service rights (e.g. timeliness of service, what type of service etc).
Market‐type reforms: this refers to a range of measures whereby public service providers
are made to operate in a more business‐like manner, under the pressure of certain types of
market forces or mechanisms which simulate markets. Competitive tendering – where a
service provided by a government agency is thrown open to private competition via tender
– is one example. Purchaser‐provider systems are a performance budgeting model which
falls into the category of market‐type reform.
Models of Performance Budgeting
There is not just one model of performance budgeting, but a range of different models. All link
funding and results, but in different ways.
Some performance budgeting systems – program budgeting and zero‐base budgeting, for
example – are government‐wide systems. On the other hand, some are only intended to apply
to particular sectors of expenditure or categories of organization (e.g. a formula funding system
applying to schools).
Performance budgeting systems differ to some extent in their objectives. Some place aim
principally to improve expenditure prioritization, while others are mainly focused on in
improving service effectiveness and/or efficiency. Reflecting the different forms of linkage
between results and funding which they seek to build, performance budgeting systems also
differ in respect to the type of performance information upon which they primarily rely. This
applies both to the type of performance measures which they use, and also to whether they
make use of evaluation as well as indicators. As discussed earlier, all forms of performance
budgeting have in common the idea of greater managerial freedom, particularly in respect to
the choice of input mix.
The most basic form of performance budgeting is that which uses performance information
systematically in the preparation of the government‐wide budget. A common tool used for this
purpose is program budgeting, in which expenditure is classified in the budget by objectives
(outcomes and outputs), rather than solely by economic categories (such as salaries, supplies
and communication costs) and organizational categories (e.g. ministry and department with the
ministry). The primary objective of program budgeting is improved expenditure prioritization.
However, by increasing the attention paid during budget preparation to spending ministry
performance, this type of performance budgeting also aims to increase the pressure on
ministries to improve efficiency and effectiveness.7
It should be emphasized that, in the terminology adopted here, performance budgeting is a
generic concept of which program budgeting is one form of performance budgeting. This needs
to be emphasized because there are some who attempt to distinguish program budgeting from
Zero‐base budgeting (ZBB) is essentially a variation of program budgeting which, in its original
form, called for the comprehensive review and prioritization of all expenditure on a continuing
basis. To achieve this, all programs were to be decomposed for the purposes of budget
preparation into “decision packages” (also known as “service increments”) which would provide
choice about the extent to which each program might be cut back or, alternatively, given
increased funding. Ideally, these decision packages would cover all major options, even to a 100
percent cut in the program concerned (hence the “zero base”). Priority rankings would be
attached to these decision packages – again, on the basis of performance information – and
these rankings would then be used to ensure that the available level of revenue funded those
decision packages which were of highest priority. In this way, ZBB aimed to go even further
than program budgeting in improving expenditure prioritization.
The fundamental problem with ZBB was the practical impossibility of comprehensively
reviewing all expenditure each year. In other words, the “zero base” examination of each
program was not really practicable. Selective expenditure appraisal is all that, in practice, can be
achieved. In recognition of this, ZBB evolved into what is sometimes call alternative budgeting.
Rather than seeking to examine all options for each program, extending even to a one‐hundred
percent cut, alternative budgeting tried to make the task manageable by confining the process
to the examination each year of a much narrower range of options for cuts or increases to each
program – for example, the options of 15, 10 and 5 percent cuts, and perhaps also of 5 or 10
7 Program budgeting has the potential to contribute significantly to aggregate expenditure discipline, and thereby
to fiscal discipline. By giving government an improved capacity to identify low priority or ineffective programs
which can be cut, it makes it easier to create the fiscal space necessary to respond to new policy priorities without
undue growth in aggregate spending. Moreover, if and when fiscal consolidation is necessary, better prioritization
makes it possible to target spending cuts so as to minimize their social cost and increase their sustainability.
Under both program budgeting and zero‐base budgeting, the link between results and funding
is a rather loose, rather than mechanical, one. The results achieved by programs (or decision
packages with programs) is considered closely in the budget preparation process, but there it is
not the case that poor results automatically means cuts – or, conversely, that good results
automatically means increasing funding.
Since the 1980s, a range of newer forms of performance budgeting have been developed which
aim to build a tighter linkage between funding and results. These newer forms of performance
budgeting all aim to increase the pressure on public agencies to improve the effectiveness
and/or efficiency of the services which they delivery. To achieve these, they rely on three
mechanisms, used either in isolation or in combination.
The first of these mechanisms is linking budgets to performance targets.
Setting performance targets – whether for ministries, work units or individuals – is, as noted
above, a general MFR theme. It is when budget funding is linked in some manner to
performance targets that we are talking about a form of performance budgeting. The best
example of a target‐based form of performance budgeting is the UK Public Service Agreement
(PSA) system as it operated between 1998 and 20048
. Under this system, targets were set on a
triennial basis as part of a multi‐year budgeting framework. In other words, high‐level targets
for outcomes and outputs (see the next section for explanations of these key concepts) were set
as an integral part of the process which, every three years, determined the core funding levels
which each ministry would receive for the coming three year period. Targets were set in the
light of funding, and funding in the light of performance against targets. The major question
that this raises – to be discussed later in section 13 – is the concrete nature of this link between
funding and targets.
Target‐setting raises a range of issues, which will also be discussed in detail later section 13. One
of these is the problem of selecting the right numerical value for the target. In principle, of
course, the target should be neither too tough nor too easy. However, it is often not easy for the
ministry of finance (MoF) or other central decisions makers, without a detailed knowledge of
the area of service delivery concerned, to set the target at an appropriate level. Another issues
raises by target‐setting is the issue of behavioral distortions, an example of which is the danger
that, in striving to fulfill a target set in terms of one dimension of performance, the agency may
8 See P. Smith, “Performance Budgeting in England” in Robinson, Performance Budgeting.
put less effort into another dimension of performance which is not factored into the
The second mechanism used to creating a tighter link between funding and results is formula
funding, which is the subject of section 12. Formula funding means that the level of funding to a
government agency is determined (in part or whole) as a mathematical function of some
explicit variables. A simple example is a formula to estimate school funding requirements over
the medium term as a function of the number of students and the cost per student (based on
demographic projections). This illustrates the idea of a cost‐based formula, the simplest version
of which bases funding on the quantity of output times the unit (average) cost per output – as,
for example, in a school funding system where funding for each school is determined primarily
by its number of students multiplied by an amount intended to cover the cost of one year’s
education of one student at the relevant stage of schooling (e.g. $4000 per year nine student
per year).. In this case, the formula is being used only as a budget estimation tool – that is, to
estimate how much funding should be provided to a particular service area or to specific service
provider units (e.g. specific schools).
Not all formula funding arrangements can be said to be forms of performance budgeting. It is
only when the formula links funding to results (outcomes and/or outputs), and does so with the
intention of boosting efficiency and/or effectiveness, that formula funding becomes a tool for
performance budgeting. This would be the case, for example, if the formula was being used in
effect to set a performance target for the agency. As a concrete example of this, suppose a
ministry was funded on a formula basis to, say, vaccinate one million persons at $20 per
vaccination in the coming year, and was then held to account for if it spend all the money but
only vaccinated 600,000 people.
The third mechanism used by the newer performance budgeting models is performance funding
incentives to agencies. An example of this is performance bonus funding, which operates by
giving agencies additional payments for measured good performance (or perhaps also funding
reductions as a function of measured poor performance). An example is university funding
systems which operate in many countries in which public universities are given bonus funding,
over and above their core funding, based on performance against measures such as the
percentage of graduates who are in professional employment six months after graduation. As
9 For example, if the target is set in term of the unit cost of service delivery, the agency might sacrifice quality in
order to cuts costs.
this example shows, most performance funding incentives are also forms of formula funding,
because the funding incentive provided is determined by a mathematical formula linked to a
Performance funding incentives – like formula funding – are almost invariably used as
instrument for sectoral performance budgeting systems (i.e. covering some sector of
government, such as schools, or hospitals.
The above outlines three key mechanisms used by newer performance budgeting systems to
create tighter links between funding and results. As indicated above, these mechanisms may be
used in isolation or in combination. An important example of the combined application of whoe
of these mechanisms is the purchaser‐provider (PP) model. PP is a combination of the principles
of incentive payments and formula funding (the second and third of the mechanisms
identified). In a purchaser‐provider system, a public sector agency is funded as if it were an
arms‐length provider of goods or services. Government funds the agency by paying a “price”
per unit for the agency’s outputs (and sometimes, but more rarely, outcomes). If the agency’s
cost of production exceeds the price it is paid, the agency makes a loss. Conversely, it retains
any profit made by producing at a cost below the price. In a PP system, funding is only provided
for results delivered. Finally, the price paid is ideally based on, partly or in whole, on some
measure of what it should cost an efficient producer to deliver the output concerned. The aim
of this system is to motivate strong performance via strong financial performance incentives.
The most successful example of a purchaser‐provider system is the “diagnostic‐related group”
system of hospital funding. Broadly speaking, this is a system in which hospital are funded
primarily for the treatments they deliver to patients. Services are categorized into a number of
different output types (“DRGs”), and a specific price is attached to each DRG category. Such a
system provides a powerful incentive for efficiency. At the same time, it raises concerns about
the danger of cost‐cutting which undermines the quality of treatment or has other perverse
behavioral effects. The experience of the DRG system in this respect is, however, encouraging.
Accrual output budgeting (AOB) represented an (unsuccessful) attempt, in New Zealand and
Australia, to apply the purchaser‐provider principle to the government as whole, as opposed to
a particular sector such as hospitals. The idea of AOB was that ministries would be funded not
10 Note that in this case, the formula is not linked to the costs of the performance delivered (for example, no‐one
would pretend to know the additional cost involved in boosting the graduate employment rate by, say, 5 percent).
by being given budgets, but would rather be paid for the outputs they deliver, as if they were
external contractors to government. It was intended that government would be a tough price‐
setter, seeking to create the type of pricing pressure which occurs in a competitive market.
Because in this model ministries were supposed to act like businesses, it was necessary that
they use business (i.e. accrual) accounting – hence the “accrual” in “accrual output budgeting”
(the link with accrual accounting is discussed in a section 16). It also became necessary to
distinguish government’s “purchaser” role vis‐à‐vis its agencies from its “ownership” role. The
AOB experience is discussed in detail in the section 12.
Robinson, M (2007), “Performance Budgeting Models and Mechanisms” in M. Robinson (ed.)
Performance Budgeting: Linking Funding and Results.
3. Performance Information Fundamentals
“Outcomes” and “outputs” play a central role in all models of performance budgeting, and it is
essential for any discussion of performance budgeting that these and related concepts are
Performance Concepts: the Results Chain
In the results chain framework, outputs are produced using inputs (resources) via activities and
processes, and outputs generate outcomes for the community.
The Results Chain
Outputs are goods or services – the “products” – which a ministry or other government
organization delivers to external parties. This usually means services delivered to or for the
direct benefit of the community, although some outputs are services which ministries deliver to
other ministries. Examples of outputs include: medical treatments; advice received by farmers
from agricultural extension officers; students taught; and police criminal investigations. Most
government outputs are services, and for this reason the terms output and service will be used
interchangeably in this manual.
Outcomes are the intended impacts of outputs – more precisely, the changes brought about by
public programs upon individuals, social structures, or the physical environment. Health
inspections of restaurants are an output, the intended outcome of which is that fewer diners
fall sick. Criminal investigations are a police output, and reduced crime the outcome.
Many government services aim to achieve more than one outcome. For example, school
education aims to increase the level of education of the population. But it also aims, amongst
other things, to improve economic performance. Both a higher level of education and a
stronger economy are outcomes. Because it is by means of the first of these that the second is
achieved, a more educated population is said to be an intermediate outcome, and a stronger
economy a higher‐level outcome. The relation between proximate and high‐level outcomes is
one of logical causality (i.e. the proximate outcomes induce the high‐level outcomes).
The outcomes element is the distinctively public sector component of the results chain. By
contrast, the way in which outputs are produced is conceptualized in the results chain in exactly
the same way as in the private sector: outputs are produced by the use of inputs in production
activities and processes. For example, the treatment which seriously injured person receives in
hospital involves the use of a set of inputs (skilled staff, operating equipment and facilities,
medical supplies, electricity etc) and a set of activities including anesthesia, surgery and
nursing, as well as supporting activities such as supplies and facility management.
Inputs, as this example indicates, refer to all inputs, assets and capabilities which are or may be
drawn on in the production process to deliver the outputs and outcomes desired. Although
“inputs” is the conventional results chain term, and therefore will be used here, the term
“resources” actually captures better the scope of what is referred to. Thus inputs which
contribute to the capability to deliver results include not only equipment and buildings by, for
example, organizational culture and staff morale.11
11 We are, in other words, talking about much more than the “factors of production” of economics 101 (labor, land
and capital), and also to draw attention that the inputs an organization can drawn upon are not only purchased
from outside, but created within the organization.
Physical assets are inputs which are sometimes mistakenly thought of as outputs – as in, for
example, the number of roads or bridges provided by government to the community. But it is
not the roads and bridges which are outputs, but rather the service which citizens get from
these roads and bridges. This means, for example, that passenger miles traveled is an output
measure while kilometers of road is an input measure.12
The term activities may seem self‐explanatory, but confusion between activities and outputs is
very common. Some examples can help avoid this confusion:
In a hospital, anesthesia and cleaning are activities rather than outputs because they are
components of the overall service provided to the patient, rather than the complete
service. The patient can’t recover through anesthesia or cleaning in isolation, and it is only
via the combination of all the necessary activities that the complete service (the output) is
delivered. More generally, an output must be capable of delivering the intended outcome.
Bus driving is an activity, whereas passenger trips are the outputs. Similarly, teaching is an
activity rather than an output, and the output is students taught. In these cases, an activity
produces multiple outputs.13
As mentioned above, outputs are defined as services provided by ministries or other
government agencies to external parties. So if one ministry provides services to another
ministry – for example, training, or office accommodation – these are ministry outputs because
the other ministry is a client external to the ministry providing the service. However, a service
which one part of a ministry provides to another part of the same ministry – for example, when
the education ministry’s human resources group recruits teachers for the primary school group
– is not an output but a support service.14
12 This should be qualified by noting that, if a separate public works ministry is responsible for constructing the
roads and bridges and then delivering them to the transport ministry which manages them, then the roads and
bridges would indeed constitute outputs of the public works ministry – because it is delivering them to an external
client. But if the transport ministry manages the construction itself, the roads and bridges are not outputs. In
either case, the output provided by the transport ministry is the use of the road by travelers. 13 Or, potentially, as illustrated by the case of bus driving, no outputs at all. This would be the case if the bus ran
but no passengers got on board. 14 Some would argue that what constitutes an output, as opposed to a support service, depends on your
organizational perspective, and that from the point of view of the human resources group, recruitment is an
output which they deliver. This is a valid observation. However, in the performance budgeting context, the
perspective normally adopted is that of the ministry or other distinct government organization. The point is to
focus entire ministries on the services they deliver to external clients, rather than on internal activities.
The results chain is a very well‐established conceptual framework for the public sector which
goes under a range of names, including the “program logic model” and “logical model”.
Although its concepts are standardized, and have clear meanings attached to them, there is one
significant divergence of usage worth mentioning: the practice adopted by some of dividing
outcomes, as defined above, into “outcomes” and “impacts”. The distinction is either one of
timelines (impacts being more long‐term), or is the distinction between intermediate and high‐
level outcomes. This terminology is not used here, because in practice there is often no clear
distinction at the margins between outcomes and impacts.
One other minor point of terminology: “intermediate” and “high‐level” outcomes go by a wide
range of other names (immediate outcomes, proximate outcomes, end outcomes, ultimate
outcomes). The idea is, however, the same.
Outcomes vs. External Factors
Outcomes are the changes brought about by government intervention. If the level of malaria
falls for reasons which have nothing to do with government actions – because, for example,
there is a drought which reduces mosquito numbers substantially – this fall is not a government
outcome. Neither is it an outcome if the rate of economic growth increases substantially
because, and only because, the world economy is very buoyant. The fall in malaria, or the boost
in the growth rate, are outcomes only to the extent that they are the result of government
Outcomes therefore need to be distinguished from the consequences of external factors.
External factors are factors beyond the control of government which influence the
characteristics of individuals, social structures or the physical environment which the
government is trying to change. The level of rain is therefore an external factor impact
impacting on the malaria rate. The state of the world economy is an external factor impacting
on the rate of domestic economic growth. In assessing outcomes, the challenge is to distinguish
the impact of external factors from that of the government intervention.
External factors are also sometimes known as contextual factors or confounding factors.
External factor may be part of the external environment in which the program operates, or they
may be characteristics of the client. An example of the former is the impact of the state of the
economy on the success rate of a program for the rehabilitation of injured workers in getting its
clients back into the workforce. An example of the client characteristic type of external factors
can be seen in education, where the educational levels attained by children are determined not
only by the quality of the education they receive at school, but by external factors such as their
intelligence, nutrition and level of parental support. Expressed differently, these types of
external factors impact upon “outcome variables” such as the literacy and numeracy rates (see
the next section for the concept of an outcome variable). This makes it difficult to compare the
performance of two different schools by simply looking at levels of student educational
attainment, because differences in the socio‐economic mix of student populations can make
such comparisons completely misleading.
Outputs vs. Outcomes and Activities
It is useful to clarify further the relationship between outputs and outcomes. To be considered
an output, an output must be capable of achieving its intended outcome, but this does not
necessarily mean that each unit of output will be successful in actually achieving the outcome.
Consider the example of a medical treatment of an accident victim who dies notwithstanding
receiving exactly the best treatment in a timely manner. It would not be reasonable to say
under these circumstances that the patient did not receive a service (output). The same point
applies, say, to education, where despite receiving excellent teaching, certain students may fail
to learn much, while others in the case will do very well.
An associated point is that, to be an output, a service must be complete. Consider the example
of a car, which is only a product (i.e. an output) only when all the parts (wheels, engine etc) are
attached and it can therefore function as intended. Another good example is tetanus injections,
where it is necessary to have a full course (of three injections) in order to acquire immunity.
The tetanus treatment outputs should therefore be measured in terms of numbers of persons
who receive the complete course. Otherwise, one will end up counting activities (e.g.
incomplete course of tetanus treatments) which are incapable of yielding any benefit to the
Performance information refers to information on results achieved, and/or costs of achieving
those results. There are two basic types of performance information: performance measures
and evaluation. In this and the next couple of sections, we focus on measures. After that,
evaluation will be discussed. It should, however, be emphasized at the very outset that it is a
mistake to think that performance budgeting is only about the use of performance measures in
the budget process.
Performance indicators are quantitative measures which provide information on the
effectiveness and efficiency of programs and organizations. There is no difference between a
“performance indicator” and a “performance measure” – both terms are used interchangeably
in this manual. We also need to be careful not to confuse objectives, indicators and targets. An
objective is a statement of what one is trying to achieve – for example “reducing death from
HIV/AIDS”. By contrast, a performance indicator is quantified (e.g. “the percentage of the
population which is HIV/AIDS positive”, or “the number of persons dying annually from
HIV/AIDS”). A target goes one step further and sets a precise aim to be achieved by a specific
date (e.g. “reducing the percentage of HIV/AIDS‐positive persons in the population by at least
one‐third by 2020”).
Performance indicators should be selected according to the extent to which they are:
Minimize perverse effects
An indicator is relevant when the aspect of performance it seeks to measure is important to the
objectives of its users. Relevance depends on who the user is. In selecting indicators to present
to parliament and the public, for example, indicators of internal processes (such as time taken
to fill job vacancies) would not be relevant.
An indicator is representative to the degree to which it succeeds in measuring the dimension of
performance which it seeks to measure. Representative indicators are good proxies for what
they are trying to measure. A measure of average time taken to answer client phone calls is, for
example, a poor proxy for the quality of client service if what happens is that calls are answered
quickly but the client is left on hold for an extended period. An indicator is more representative
the less it varies for reasons which have nothing to do with performance, such as the impact of
uncontrollable external factors and statistical uncertainty. To be representative, an indicator
should also be unambiguous. That is, it should be clear whether a change in the indicator
means that performance has improved or deteriorated.
An indicator is cost‐effective if the benefits of using a specific indicator exceed its cost of
collection, processing and verification. Performance measurement systems are not cost‐free.
Like everything else, they should therefore pass a benefit/cost test.
A comparable indicator is one which is used by similar service providers, or by the same
organization in the past. If, for example, there is an international standard measure for a service
– for example, as with the measures provided in Europe by the common set of social policy
indicators – then using that measure makes it possible to make performance comparisons with
other countries. Similarly, performance comparisons become much easier if all relevant entities
– e.g. all local governments, or all hospitals –a use the same indicators. Comparability over time
is useful because continuing to use the same indicators that have been used in the past makes
it possible to identify trends in before which might be impossible to detect if the performance
measures have changed.
Perverse effects, which are discussed further below, are adverse consequences of performance
measures. Examples are:
In seeking to improve measured literacy and numeracy, schools might reduce their focus on
other important dimensions of education, such as the teaching of foreign languages or
Suppose the timeliness of hospital accident and emergency ward treatment is measured by
the percentage of patients who are treated within 4 hours. This can potentially produce the
perverse effect that any patient who cannot be treated with 4 hours may be kept waiting
considerably longer, with priority being given to patients who have arrived later but who
can still be treated within 4 hours.
Perverse effects are in part the consequence of indicator design: for example, the omission of
some key aspects of performance which, with improved design, could have been captured. In
the case of hospital waiting times, the perverse effect referred to would be avoided by using a
measure of average waiting time instead of the measure of those treated within a particular
(arbitrary) time period.
These five criteria for good performance indicators are ideals, and few good indicators will
score highly on all of them. Sometimes, a high rating on one criterion will come at the price of a
lower rating on another. Outcome measures such as the crime rate, for example, score very
high on the relevance criteria but less well on representativeness because they are greatly
influenced by “external factors” beyond the control of government.
In this section, we have focused on performance concepts and performance information in
general. In the specific context of performance budgeting, however, what we are really
interested in is the performance information – evaluations as well as indicators – which is useful
for, and usable by, budget decision‐makers. An important point to bear in mind here is that
decision‐makers can only absorb and use a limited and carefully‐selected amount of
information. It is therefore essential not only to be highly selective in choosing a handful of key
performance indicators relevant to budget decision‐makers, but also to put these indicators
together with analysis which interprets and explains them, drawing overall conclusions about
performance. This analysis should draw closely on the conclusions of formal evaluations and
other review work.
Robinson, M (2007), “Informing Performance Budgeting” and “Results Information”, in M.
Robinson (ed.) Performance Budgeting: Linking Funding and Results.
HM Treasury et al (2001), Choosing the Right Fabric – a Framework for Performance
Information, obtainable at http://archive.treasury.gov.uk/performance_info/fabric.html.
Royal Statistical Society (2003), Performance Indicators: Good, Bad and Ugly, obtainable at
4. Performance Measures and the Budget
The indicators of most value for performance budgeting will tend to differ significantly from the
indicators used for other purposes. What will generally be of greatest value to budget decision‐
makers in determining appropriate program funding levels will be indicators of the results
achieved by programs – the outcomes that they achieve and the outputs which they deliver to
achieve these outcomes. They will represent only a sub‐set of the indicators used internally for
managerial purposes by ministries, which will include not only results‐oriented indicators, but
also activity and input indicators focused on the internal processes, capacities and resources of
There are three main types of output indicators: indicators of output quantity, indicators of
output quality, and indicators of efficiency.
Output quantity indicators measure the volume of service provided. Examples of output
quantity indicators are:
• Number of vaccinations carried out,
• Number of malaria prone districts sprayed,
• Number students taught at seventh grade,
• Number of planning applications determined.
Output quality indicators provide information on the extent to which the service is of a type
which is likely to achieve its intended outcome. Two examples of more readily‐measurable
quality indicators are timeliness indicators and client satisfaction indicators. Examples of
timeliness indicators are:
• Average waiting time of a hospital patient between arrival and treatment,
• Average time for a planning application to be determined,
• Average response time of the fire brigade to a fire.
Client satisfaction measures can be of various types, from simple measures of the level of
satisfaction felt by the client, to more targeted measures such as:
• Client ratings of the courtesy of the service provider,
• The percentage of clients who consider that they obtained the service which they were
Efficiency indicators measure, ideally, the extent to which the service is delivered at low cost
without sacrificing quality. Unit output cost one important type of efficiency indicator, although
this is not a perfect measure as it does not necessarily hold quality constant. The unit cost is
simply the total cost of delivering the output divided by output quantity. Examples of unit cost
• Cost per vaccination (including delivery),
• Cost per planning application determined,
• Cost per visa application processed.
Labor productivity indicators are another important type of efficiency indicator. Examples of
labor productivity indicators include:
• Pension benefit applications processed per staff member,
• Average staff time taken to administer a practical driving license test.
In respect to outcome measurement, the biggest challenge arises from the distinction,
discussed in the last section, between outcomes and external factors. As noted, outcomes are
the change in characteristic of individuals, social structures or the physical environment brought
about by government intervention, and the challenge is to separate out the impact of
government intervention from that of other factors beyond government control.
In practice, many outcome indicators are in fact measures of what can be called the outcome
variable —that is, the characteristics of individuals, society or the physical environment which
public programs seek to change. Measures of the outcome variable do not distinguish the
impact of the government’s intervention from that of external factors. A good example of this
type of outcome indicator is a crime rate statistic (e.g. the number of robberies per thousand of
population). We know that the crime rate is not only influenced by government interventions
such as policing, but is also greatly affected by social and economic developments beyond
government control (e.g. the level of poverty). Crime rate statistics make no attempt to
eliminate the impact of the latter type of external factor.
The perfect outcome measure would be one which excludes the impact of all external factors
and which measures only the change in the outcome variable brought about by government
intervention. Developing such measures is, however, exceedingly difficult. In a few areas,
modern performance measurement techniques have gone quite a long way in developing
outcome measures which remove, to at least some extent, the effects of external factors. A
good example of this is the so‐called value‐added education performance measures, which
attempt to adjust for student characteristics. For example, value‐added school league tables
present measures of the comparative performance of schools which adjust for these differences
in student population. However, although this and other techniques have been developed to
obtain better outcome indicators which are less effected by external factors, the overall scope
for eliminating external factors from outcome measures is quite limited. As a result, the great
majority of outcome measures are affected, to varying degrees, by external factors.
Ensuring that Indicators are Useful
To be useful to central decision‐makers, who invariably have great demands on their limited
time, program performance information needs to be readily digestible. A couple of key program
indicators may, for example, be more useful to top decision‐makers than a comprehensive
compendium of dozens of indicators. The detail is of more interest to program managers than to
A notable innovation in this context has been the development of summary measures which
incorporate a wide range of performance information into one or more overall performance
ratings for the program. A good example of this is the measures of program performance which
was developed under the US Program Assessment Rating Tool (PART) which operated under the
Bush presidency. Under the PART system, the US finance ministry (the Office of Management
and Budget) set about rating all federal government programs over a period of 5 years. Each
program was rated on a scale of 4 ratings, ranging from effective to not effective (there was also
a “results not demonstrated” rating, used where there is insufficient information to form a
judgment). These summary ratings were intended to be much more informative and readily
understood than the large body of more detailed measures and evaluations which underpinned
them. The program ratings, and the reasoning behind them, were all made public (on the
website ExpectMore.gov). Moreover, the PART system was designed from the outset as a tool
for performance budgeting, because the program ratings were used to inform the preparation
of the president’s budget proposal to Congress (and were also intended to influence Congress’
Summary performance measures – of which PART is by no means the only recent international
example – have the enormous advantage of ready comprehensibility. They can also be far more
effective than more detailed indicators in putting real performance pressure on government
agencies. This should not, however, blind us to the very considerable technical challenges
involved in appropriately defining summary measures. Inappropriate selection of constituent
variables, or poor technical design in respect to the aggregation of these variables, can produce
summary measures which are quite misleading.
Linking Funding to Outcomes
In considering the role of performance measures in budgeting and funding decisions, it is
important to explicitly consider certain key challenges which arise in linking outcome and/or
outputs to funding. These have an important bearing on the choice of model of performance
budgeting. We start by considering the outcomes/funding link, and after that look at the
In respect to outcomes, the impact of external factors on measured outcome variables has
major implications for the type of linkage between outcomes and funding which it may be
possible to create in a performance budgeting system. As noted above, most outcome
indicators do not separate – or separate only to a limited degree – the outcomes achieved by
government intervention from the impact of uncontrollable external factors. When the
influence of external factors is extensive, it may be difficult or even impossible to predict the
outcome which any particularly level of funding of a public service will produce. This is a major
obstacle to attempt to create a tight link between funding and outcomes – for example, by
basing a purchaser‐provider system on outcomes. For this reason, performance bonus funding
based on outcomes tends to be used only to provide small additional funding rewards (e.g. 5
percent additional funding), because if there was too strong a link between funding and
unpredictable outcomes, the financial stability of the government agency concerned would be
adversely affected. At the same time, outcome measures are crucial to program budgeting, and
must constitute a key element in expenditure prioritization decisions.
This raises the question of outcomes targets and funding. The UK PSA system relied heavily on
outcome indicators, over many of which the government had limited control. How, and to what
extent, can one link funding to targets which are only partially controllable? Clearly there has to
be flexibility in the funding/target relationship under such circumstances. There are, however,
many who take the view that flexibility is not enough and who dispute the wisdom of seeking to
link budget decisions to outcome targets. In their view, target‐setting in government should be
confined mainly to outputs or other variables over which government has a high degree of
control (e.g. activities).
Linking Funding to Outputs
The main focus in PB system such as formula funding and purchaser‐provider is the creation of
links between the quantity of output (i.e. volume of services provided) and the level of funding.
For many outputs produced by government, there is a much stronger link between funding
provided and outputs delivered (or deliverable) than is the case for outcomes. This is
particularly true for standardized outputs, which are outputs where every client receives pretty
much the same level of service, so that unit cost should the same.15
However, quite a few government services are not standardized. They are, rather,
heterogeneous outputs. This means that the level of service provided to different clients, or in
different cases, is deliberately varied so as to address differences in client conditions or
circumstances. Police criminal investigations are a classic example – the amount of effort put in
per case, even for the same types of case (e.g. murder investigations) varies enormously
depending on the circumstances of the case. Even in school education, which is quite
standardized for the great majority of students, heterogeneity is present when additional
teaching and care activity is devoted to children suffering an intellectual or physical disability.
Substantial heterogeneity undermine the predictability of the relationship between the funding
level and the outputs which the agency can be expected to produce, thus also influencing how
tight a link it is possible to create between funding and outputs. This particularly affects the
scope for applying formula funding and purchaser‐provider systems, as discussed in section 12.
This means that these forms of performance budgeting can only be applies selectively to the
right types of services.
15 At least if we leave aside cost difference arising from factors like geography.
There is one other type of service for which tight links between outputs and funding are
problematic. This is contingent capacity outputs, of which a fire department is a good example.
The fire department maintains capacity to provide at very short notice an output (firefighting)
for which the demand is highly unpredictable. It would be unrealistic to seek to build a very
close link between the number of fires attended by the fire service and the level of funding. Fire
services cannot therefore be funded on a per‐output basis, but must instead be funded in such
a way as to deliver a certain level of capacity to fight fires.
The discussion to this point has focused on output quantity. There is also the question of
potentially linking funding to output quality. It would be highly desirable to be able to include
the output quality as well as quantity dimension in, for example, a purchaser‐provider funding
arrangement. One of the concerns about “perverse effects” raised by purchaser‐provider
systems is that, in funding only for output quantity, one creates incentives for agencies to cut
costs by reducing quality. Including a quality component in funding could, in principle, resolve
this problem. In practice, however, this is not easy, given the limits to our capacity to measure
quality and the consequently highly imperfect nature of most quality measures. In general, the
best hope for linking funding to output quality is through some element of performance bonus
funding based on quality measures (similar to outcome bonuses) – in other words, by adding on
to a system in which the main funding is based on output quantity a small additional element of
Finally, there is the question of linking funding to activities – as suggested those who advocate
“activity based budgeting”. This will be discussed further in the session on formula funding. A
key question which arises in basing funding on activities, however, is that of how far the
activities which are being funding necessarily lead to results (outputs and outcomes) which the
public cares about. Remember that while outputs are services, activities are “merely” work
processes (see section 1). An organization which is internally‐focused rather than externally
focused on results may put excessive emphasis on activities for their own sake. To fund such an
organization partly on the basis of internal support activities – such as meetings held or policy
documents developed – could worsen this problem.
In conclusion, the tightness of the link which can be created between results and funding
depends on the extent to which the underlying connection between funding provided and the
results one can expect is clear‐cut. If the relationship is an uncertain one – due to, for example,
external factors or heterogeneity – then a tight link may be impossible. This is why, for example,
it is hard to use the purchaser‐provider mechanism to fund organizations for outcomes.
Conversely, program budgeting system will work even when there is considerable uncertainty in
the relation between results and funding, precisely because program budgeting links funding to
results only fairly loosely. This is why program budgeting is widely used as the basis for whole‐
of‐government performance budgeting system, whereas mechanisms such as formula funding
purchaser‐provider can only work when applied to certain sectors or types of services (e.g.
hospitals or schools).
The issues discussed in this section also have an important bearing on the role of indicators
versus evaluation: namely, that the imperfections of performance measures mean that one
cannot rely upon measures alone for the performance information to be used for expenditure
Robinson, M (2007), “Results Information” in M. Robinson (ed.) Performance Budgeting: Linking
Funding and Results.
5. Performance Measurement Systems
The performance measurement (PM) system refers to the mechanisms for data collection, the
processing of data into indicators, the validation of those indicators (i.e. ensuring that the
numbers are reliable and are not either manipulated or statistically questionable), and their
presentation to users. A PM system is a subset of the broader monitoring and evaluation
In discussing the performance information requirements of performance budgeting, it needs to
be borne in mind that performance budgeting is not – or should not be – an isolated reform. As
emphasized earlier, it is part of a set of broader reforms, often referred to as “managing for
results”, which are designed to focus public management more on results delivered, and less on
internal processes. These broader reforms include civil service reforms designed to increase the
motivation and incentives of public employees; organizational restructuring to increase the
focus on service delivery and improve coordination (e.g. creation of agencies and reduction of
numbers of ministries); and institutional change to strengthen public accountability for
performance. Action on these and a range of related fronts is necessary if the efficiency and
effectiveness of public expenditure is to be tangibly improved. Improved performance
information is fundamental to each of these elements of MFR reform because all need to be
underpinned by better performance information. For example, a crucial ingredient in making
the civil service more performance‐oriented is the improvement of information on the activities
and outputs of individuals, workgroups and agencies.
This implies that government‐wide performance information strategy should be designed to
meet not only the needs of performance budgeting, but of MFR processes more generally. It is
not appropriate, for example, to attempt to develop a system of performance measurement
aimed exclusively at budgeting applications, and entirely separate sets of measures used for civil
service management, accountability or other purposes. The government‐wide performance
information system should be developed as an integrated whole. For example, if data collection
relies on surveys, it will in general be much better to use one survey to collect a range of
different information from a particular target group than to conduct a number of separate
Efficiency in the operation of the performance measurement system is particularly important
because there are significant financial and human resource costs in the production of pertinent
and reliable performance indicators.
The first stage in the production of performance indicators is data collection. The most readily
available source of raw data for indicators is usually client service records such as hospital
treatment records, school enrolment and attendance data, and data on regulatory approvals
granted. A good PM system will need therefore, firstly, to improve the reliability of client service
records and, secondly, to expand the type of data collected in clients service records so that it
provides the basis for a broader set of indicators.
Client records provide some information about outcomes. For example, in schools, examination
and graduations are recorded, and in hospitals deaths are recorded. However, in general, client
service records provide information mainly about the volume of services (quantity of outputs)
delivered to clients and the activities which these outputs comprise. A good PM system
therefore needs to go beyond the constraints of client service records in order to construct
more and better outcome and output quality measures. In this respect, the use of surveys
(including, but not confined to, client satisfaction surveys) is the key element. Survey data can
not only provide good measures of “outcome variables”. It can also provide more timely
measures, by capturing outcome variables which will only much later manifest themselves in
demand for government services. For example, survey‐based data on alcohol and tobacco
consumption rates provides a much earlier warning of the extent of the problem which
preventative health services need to address than do levels of alcohol and tobacco‐related
disease, which arises only after years of abuse. Surveys can pick up measure variables relevant
to outcomes which take time to manifest themselves.
The second stage in the production of indicators is data processing – i.e. the transformation of
raw data into performance indicators. A key question here is the choice of processing
technology. There is a full range of options here, ranging from the most high‐tech through to the
relatively simple. At one end of the spectrum, performance measurement may be built into a
large Integrated Financial Management Information System (see the section on program costing
and accounting for more on IFMISs). The advantage of this – if the system works properly – is
that the performance indicators are directly linked to the relevant financial data (e.g. program
performance indicators are automatically linked to program expenditure data; output quantity
measures to output costs etc).
At the other end of the spectrum, it is perfectly possible to manage a performance
measurement system based on Excel spreadsheets and an Access data base. This is a simpler
and lower‐cost option. Some IT experts will suggest that incorporating performance measures in
the IFMIS should be the preferred approach. However, the more complex and multi‐faceted
IFMISs are, the more chance they have of never working, or not working well. This is particularly
true in developing countries. In many countries, the simpler option will work better.
The third stage in the production of indicators is indicator validation, which refers to the
assurance of the reliability of the indicators. This is important to safeguard against errors and
manipulation, as well as against methodological deficiencies which make the indicators
unreliable. There are two aspects of indicator validation. The first concerns the validation of the
capacity of agency systems to ensure indicator quality – for example, the adequacy of training
and data entry checks.
Performance Measurement Protocols
The UK Royal Statistical Society proposed the idea of a Performance Measurement Protocol in its
excellent 2003 report Performance Indicators: The Good, the Bad and the Ugly, suggesting that: “Before
introducing performance monitoring in any public service, a PM protocol should be written. This is an
orderly record not only of decisions made but also of the reasoning or calculations that led to those
decisions. A PM protocol should cover objectives, design considerations and the definition of PIs,
sampling versus complete enumeration, the information to be collected about context, the likely
perverse behaviors or side‐effects that might be induced as a reaction to the monitoring process, and
also the practicalities of implementation. Procedures for data collection, analysis, presentation of
uncertainty and adjustment for context, together with dissemination rules, should be explicitly defined
and reflect good statistical practice.”
The other aspect of indicator validation is the audit of indicators – that is, the assessment of the
reliability of indicators by persons separate from, and independent of, the units responsible for
their production. Audit may be internal (i.e. carried out by an audit unit within the relevant
ministry), or external. External may mean that the audit is carried out by a government‐wide
body independent of the ministry which produces the indicators, or it may mean – going one
step further – audit by a body independent of executive government (that is, a “supreme audit
There are two broad approaches to the auditing of performance indicators. The first is the
auditing of specific indicators. The other approach is a systems verification, which focuses on
audit review of the performance measurement systems (including quality assurance processes)
within each agency with the objective of forming an opinion as to whether these systems are
such as to be likely to produce reliable performance indicators.
A quite separate issue – not to be confused with the audit of performance measures (which
focuses on assessing their reliability) – is that of external review of the appropriateness of the
indicators chosen by spending ministries. The choice of measures for performance budgeting
purposes cannot be left to spending ministries alone, because the performance information has
to serve the needs of central budget decision‐makers. For example, if when the environment
ministry is asked to come up with key performance indicators for its conservation program, it
suggests the use of a measure of the number of policies which it develops, it should be firmly
told that an activity indicator such as this is not satisfactory, and that it needs to develop
outcome measures (such as independent estimates of the remaining populations of endangered
species). The MoF and/or other central agencies must therefore provide guidance to spending
ministries about the types of indicators they need to develop, and must subsequently reviewing
the appropriateness of indicators suggested by ministries.
The question of whether the supreme audit institution should become involved in questioning
the choice of indicator is more controversial. In some countries this happens (e.g. the US, where
the Government Accountability Office frequently critiques indicators). By contrast, in British‐like
parliamentary systems, it is frequently asserted that to critique the choice of indicators would
be to go too close to the critique of government policy, which is something which supreme
audit institutions are traditionally barred from doing in those systems.
The final stage of the PM system is the presentation of the indicators. This needs to be done in
such a manner as to make the indicators as readily understandable and usable as possible.
Usability first and foremost depends on the selective of a relatively small number of indicators
which are of greatest relevance to the user concerned. There is nothing which reduces the value
of indicators more than presenting a user with hundreds of indicators most of which are of little
interest to the user, but through which he or she is obliged to work in order to find what is of
It is crucial in reporting performance indicators that they be placed into context, rather than
presented simply as numbers in isolation. The means, in particular, the inclusion in a
performance report not only of the indicators themselves, but of a narrative discussion of the
trends which the indicators show and, in association with that, information on – and an
assessment of extent of – the impact of relevant external factors. A “best practice” in external
performance reporting – which not all countries will necessarily be able to implement, given
financial considerations – is to create a multi‐tiered system in which users can start at the
broadest level of higher‐level outcome indicators and then “drill down” to two or more lower
levels of more detailed indicators, at the ministry and program level (many of which will be
more focused on outputs, activities and input/resources).
Need for More Robust PM Systems in Africa
“A problem in African countries, and perhaps in some other Regions, is that although sector ministries
collect a range of performance information, the quality of data is often poor. This is partly because the
burden of data collection falls on overworked officials at the facility level, who must provide the data for
other officials in district offices and the capital but who rarely receive any feedback on how the data are
actually being used, if at all.
“This leads to another chicken‐and‐egg problem: Data are poor partly because they aren’t being used;
and they’re not used partly because their quality is poor. In such countries there is too much data, not
enough information. Thus, another lesson for the institutionalization of a government M&E system is
the need to build reliable ministry data systems—to help provide the raw data on which M&E systems
depend. An audit of data systems and a diagnosis of data capacities can be helpful in this situation. It
would provide the starting point for any necessary rationalization of data collections or improvements in
their quality. It would give the data the credibility necessary to be used.”
Keith McKay (2007), How to Build M&E Systems to Support Better Government, World Bank Independent
Ensuring that Performance Information is CostEffective
The development of performance information systems is not simply a matter of developing the
best and most comprehensive results and cost information possible. Rather, it is about a
benefit/cost judgment. Performance information does not come free. It is costly in both
financial and human capacity terms to design, build and then operate on a continuing basis the
systems concerned. Careful judgments therefore need to be made about how far to go in
respect to choices such as the number of performance measures to be developed, as well as
related questions such as the sophistication of program evaluation methodology and costing
These choices face even the wealthiest countries. But they are particularly pressing for
countries with more limited financial and skilled human resources. Such countries should be
particularly selective and strategic in the development of performance measures. As discussed
later, they should also, in many cases, make use of quite simple program evaluation
methodologies. And, as mentioned above, they should not seek to develop efficiency measures
which require relatively complex managerial accounting (e.g. allocation of indirect costs to
outputs). The temptation of adopting what appear at the time to be cutting‐edge practices —
whether it be accrual accounting and budgeting at present, or purchaser‐provider models ten
years ago — should also be studiously avoided.
It is also a considerable challenge to build the capacity necessary to operate a good
performance information system. For performance budgeting purposes this demands, in
particular, great change in the skill set and competences of the MoF. Rather than being and
exclusively economic/accounting body, the MoF must develop competence in policy analysis
and in the development of performance information to support that policy analysis. Only in this
way can it develop the capacity to advise executive government well about expenditure priority
choices, in order to make effective performance budgeting possible.
Implementation strategy – including in relation to PM systems – is discussed in section 19.
However, it is worth noting here that many developing countries are unrealistic in their
expectations as to how long it should take to develop a comprehensive performance
measurement system. Countries which are particularly advanced in this area – such as the UK
and USA – took decades to get where they are today. It is a serious mistake for developing
countries which are relatively new to performance measurement to set out to develop
thousands of program indicators in very short time periods (e.g. 2‐3 years). It is far better to
start off modestly, focusing on a relatively small number of the most relevant indicators.
Crucially, much attention should be paid to ensuring that collection, processing and validation
systems are adequate in respect to that selective group of indicators. This is much more
valuable than generating a much larger group of indicators which cannot be relied upon.
UK National Audit Office (2000), Good Practice in Performance Reporting in Executive Agencies
and Non‐Departmental Public Bodies, obtainable at
US Government Accounting Standards Board’s (2005) Government Service Efforts and
Accomplishments Performance Reports: a Guide to Understanding, obtainable at
6. Evaluation and Performance Budgeting
Performance budgeting is often represented as being only about the use of performance
indicators in the budget. This is wrong, because it overlooks the crucially important role of
evaluation. It is for this reason that throughout this manual reference is made to the
performance information base of performance budgeting, rather than to performance
Evaluation is the subject of another CLEAR course, and the general nature and methodologies
of evaluation are therefore discussed here only very briefly. Further information can also be
found in the readings listed at the end of this section. In general terms, however, evaluation is
the formal assessment of programs, projects, organizations, or policies using systematic
methodologies, with the intention of forming as objective an assessment as possible of their
efficiency, effectiveness, design or management. One can distinguish between retrospective (ex
post) evaluations, which are evaluations of programs etc which are already operating, and
prospective (ex ante) evaluations, which are appraisals of possible new programs etc before
they are implemented.
A Definition of Evaluation
“The systematic and objective assessment of an ongoing or completed project, program or policy, its
design, implementation and results. The aim is to determine the relevance and fulfillment of objectives,
development efficiency, effectiveness, impact and sustainability. An evaluation should provide
information that is credible and useful, enabling the incorporation of lessons learned into the decision–
making process of both recipients and donors. Evaluation also refers to the process of determining the
worth or significance of an activity, policy or program. An assessment, as systematic and objective as
possible, of a planned, ongoing, or completed development intervention.”
Keith McKay (2007), How to Build M&E Systems to Support Better Government (World Bank Independent
Evaluation can support the budget process by helping either to:
Identify programs or components of programs which can potentially be cut: this means
programs which are not cost‐effective and which cannot readily be made cost‐effective
through policy design or management changes,
Identify savings which can be made by improving the efficiency of service delivery.
Evaluating Program Effectiveness
The evaluation of program effectiveness has a particularly important role to play in those forms
of performance budgeting which focus on the allocation of resources in the government‐wide
budget, of which program budgeting is the most important form. In such systems, the primary
focus is upon making budgeting as performance‐informed as possible. As we have seen, this
means in particular that:
Decisions about expenditure prioritization – where to allocate limited tax resources – are
informed by good information on program effectiveness,
Decisions about funding for specific ministries and agencies – and in particular decisions on
their requests for additional resources – are informed by reliable information on how
effectively the ministry or agency has used funding it has received in past budget.
Evaluation is crucial in this context because performance indicators are frequently insufficient
in isolation to permit judgments on program or agency effectiveness. As we have seen, some
program outcomes cannot be measured, or can be measured only very imperfectly, and many
outcome indicators are heavily contaminated by external factors. Evaluation is very important
as a means of making judgments about the likely impact of external factors on outcome
variable. It can enable us to make some useful judgment about probable effectiveness of
programs even when we lack any outcome measures (e.g. via program logic evaluation – see
below). More generally, whether for budgetary or any other purpose, performance indicators
alone rarely suffice (see box).
Role of Evaluation in a Balanced M&E System
“The most prevalent type of performance assessment practice … is indicator‐based monitoring …
Performance [indicator] monitoring, in and of itself, represents a relatively crude way to inform
decision‐making. In many cases, there is a need for a much more nuanced, in‐depth understanding of
the processes involved in particular programs or policies, which evaluations are much better equipped
to provide. Given the specific strengths and weaknesses of monitoring and evaluation, which are
potentially complementary to each other, the ideal approach is one that relies on an appropriate
balance between the two types of activity.”
Ariel Zaltsman (2006), Experience with Institutionalizing Monitoring and Evaluation Systems In Five Latin American
Countries, World Bank Independent Evaluation Group.
Evaluation of effectiveness also an important role to play in systems of budget‐linked
performance targets liked the British Public Service Agreements system. As noted earlier, in
such systems targets are set as an integral part of the budget process, with the aim of ensuring
that the stringency of the targets is related to the extent of funding provided. After the event,
scrutiny of actual performance against target is very important in this type of system, and it is
here that evaluation can make an important contribution. For example, when the targets are
outcome targets, a crucial part of the assessment of performance against target is
consideration of whether, say, a failure to meet a target is the “fault” of the agency concerned,
or whether it is instead due to the impact of unanticipated external factors. Evaluation can be
very useful for this purpose. More generally, evaluation has a significant role to play in any
performance management system which places considerable emphasis on the setting of
performance targets, irrespective of how closely target‐setting is linked to the budget.
There are two types of evaluation which are relevant to the assessment of program
Outcome evaluations (also often referred to by evaluation specialists as “impact
evaluations”): Outcome evaluations aim to directly measure the effectiveness of programs
by taking outcome indicators and using sophisticated methods to assess the link between
the program intervention and the measures changes in outcome indicators, in the process
explicitly dealing with problems such as external factors. The methods employed include so‐
called “experimental” techniques (which entail the comparison of program beneficiaries
with control or comparison groups at two or more points in time), and the use of regression
analysis to separate the effect of external factors from other causal factors which impact on
Evaluations of program logic: these assess whether the program is designed in such a way
as to make it likely that it will achieve its intended outcome. To evaluate program logic, the
first step is to clarify exactly how the program is supposed to achieve its outcomes.
Expressed in term of the “results chain”, the key questions are: What intermediate
outcomes is the program expected to deliver? How is it that those intermediate outcomes
are expected to generate, or contribute to, the program’s intended higher‐level outcomes?
Once the program logic is clarified, the next step is to ask whether it is reasonable to
assume that the program will achieve its intended outcomes. For example, given what we
know about relevant economic theory, is it reasonable to assume that a specific industry
policy will deliver its intended outcomes?
The use of evaluation to identify savings from improved efficiency of service delivery in relevant
to any form of budgeting. In this context, the form of evaluation we are talking about is
efficiency review. To the extent that efficiency reviews go beyond recommending steps to
improve efficiency to providing quantified estimates of potential savings, they are of potentially
great value to ministries of finance.
Efficiency review has a particular relevance to forms of performance budgeting such as formula
funding and purchaser‐provider systems, in which funding is based on the unit costs at which
the funding authority believe the service delivery agency should be able to produce outputs.
Specifically, the findings of efficiency review can inform that funding authority about what unit
cost to use for funding purposes. For example, suppose that the government has been funding
the health ministry for vaccinations at the ministry’s present unit cost $20 per head, but that an
efficiency review indicates that this current unit cost is greater than it should be and could be
reduced by 10 percent over two years. Under these circumstances, it would be reasonable for
the funding authority to plan to gradually reduce the amount it pays to $18.
Although efficiency reviews are useful for this type of performance budgeting, evaluations of
effectiveness clearly are not. This is because these types of performance budgeting focus are
directed to improving efficiency rather than improving the allocation of resources.
Making Evaluation Relevant to the Budget
Informing the budget process is not by any means the only role of evaluation. Broadly spending,
one can distinguish valuation for budgetary purposes from evaluation for policy/management
improvement purposes. The latter type of evaluation aims to help institutions improve policy
design – that is, help institutions to change the nature of the services they deliver to the
community so as to make them more effective in achieving their intended outcomes – without
focusing on identifying programs which can be cut. Or it can aim to help institutions improve
processes and management so as to make the delivery of services more efficient, but without a
focus on quantifying budgetary savings which can be made as a result of such efficiency
A good performance budgeting system therefore requires the conduct of selected evaluations
specifically intended to inform the budget process – that is, designed to give budget decision‐
makers better information upon which to base budget decisions. Expressed differently, it
cannot be assumed that, simply because a government conducts substantial evaluation work,
that this evaluation work will necessary meet the needs of performance budgeting.
This has implications, firstly, for the selection of topics to be evaluation. If evaluation is going to
make its maximum potential contribution to resource allocation, the programs (or elements of
programs) and topics chosen for evaluation should be those which appear prima facie likely to
yield budgetary savings.
Secondly, budget‐linked evaluation needs to deliver its findings quickly and at the right time to
be taken into account in budgetary decisions. It is important here to bear in mind that
evaluations can be conducted in great depth, if desired, making extensive use of surveys,
interview and other data gathering techniques, in combination with quite sophisticated analytic
techniques. Such in‐depth evaluations may generate more reliable results, but it tends to take
considerable time, which may mean that it is not well geared to serving the needs of budget
Because of the important of timely information, considerable emphasis is being placed these
days on rapid evaluations for budgetary purposes. Rapid evaluations are evaluations carried out
in quite short time frames (e.g. 3‐6 months). Rapid evaluations tend to focus on the evaluation
of program logic rather than outcome evaluation. As indicated above, evaluations of program
logic aim to assess how plausible it is that the program would generate its intended outcomes.
They are in principle inferior to outcome evaluations. However, outcome evaluations tend to
take considerable more time to produce and to cost considerably more. The speed and low‐cost
character of rapid evaluation makes it very attractive in practical times, particularly – but by no
means exclusively – in low income countries.
Evaluation can make its greatest contribution when linked to systematic spending review (SR),
which is discussed in Section 15. SR refers to the systematic scrutiny of ongoing program
expenditure to identify options for cuts. The MoF will in general manage and conduct spending
review, providing advice on potential cuts to the political leadership. In the process, the MoF
can make best use of evaluation by commissioning rapid evaluations focused on programs
which it or the political leaderships feels may be of doubtful effectiveness.
Evaluation and the Budget: The Chilean Example
The best developed example of budget‐linked evaluation linked to spending review is that
which has been progressively developed in Chile from the late 1990s. In Chile, performance
budgeting is part of a broader performance management system, known as the "system of
evaluation and management control". This system aims to improve the effectiveness of policy‐
making and management throughout central government, to create performance incentives for
civil servants, and to make the budget results‐oriented. The degree of use which Chile makes of
systematic evaluation as a basic tool of its performance management system is quite
exceptional by contemporary international standards, at least outside Latin America.
Evaluations are managed by the MoF, and the topics for evaluation are centrally determined.
There are three different types of ex post evaluation in the Chilean system:
Outcome (Impact) Evaluations: for example, an evaluation of whether a labor market
program had achieved its stated objective of bringing the long‐term unemployed
reintegration back into the workforce, carried out mainly via a “longitudinal” analysis of the
track record of participating long‐term employed persons to ascertain the rate at which
they obtained and retained work, and how this compared with other long‐term employed
persons who did not participate in the program.
Evaluations of Government Programs: the main element of which is the evaluation of
program logic (see above).
Institutional Evaluations – going under the potentially misleading name of “comprehensive
expenditure evaluations” (Evaluaciones Comprehensivas de Gasto), these evaluations look
at specific institutions or sectoral groups of institutions. They examine a range of issues
including the consistency of institutional and sectoral objectives, organizational structures,
production and management processes, resource use and service‐delivery performance.
The coverage of ex post evaluation has progressively increased over time, with 33 evaluations in
2009 and 39 in 2010. The programs and organizations to be evaluated are selected by the MoF
in consultation with the Congress. Evaluations are then carried out by external evaluators
(consultants or research institutions) contracted by the MoF, which provide clear terms of
reference and methodological guidelines to the evaluators. All final evaluation reports are
made available to the Congress and public, and their summaries are included in the budget
information papers in the form of “Executive Minutes”. The MoF and the relevant ministry
discuss the recommendations of the evaluations and agree on the actions which should be
taken in response to evaluation recommendations. This then becomes the subject of a formal
agreement, the implementation of which is the monitored in subsequent years by MoF.
In addition to ex post evaluation, Chile has recently extended its evaluation system to ex ante
analysis of new spending proposals. The MoF requires institutions to present all new spending
proposals in a standard format designed, amongst other things, to make absolutely explicit
their intervention logic. "New spending proposals" refers not only to proposed new programs,
but also to significant discretionary expansions of existing programs. Associated with this, MoF
introduced in 2009 a new formal mechanism of ex ante evaluation of new spending proposals.
More recently, it has added to this a technical assistance service to entities under which MoF
provides advice on how to develop and present good quality new spending proposals.
Internal versus External Evaluations
Evaluation for budgetary purposes is only one part of a well‐developed evaluation system. It is,
in particular, appropriate that all ministries and agencies be encouraged to carry out systematic
programs of internal evaluations as a management improvement tool. Many countries have
developed government‐wide evaluation policies to achieve exactly this.16 The focus of such
spending ministry‐led evaluations will tend, in general, to be more operational than that of
evaluations initiated by the Ministry of Finance or other central government agencies to inform
budget decisions. It is inappropriate for the MoF and political leadership to rely on evaluations
led by ministries themselves as the information source for budget decisions. Such evaluations
not only have a more operational focus, but they will not necessarily be focused on the
programs which are the most likely candidates for budget cuts. Moreover, the knowledge that
the results of the internal evaluations will be used by the center to cut programs may lead
ministries to manipulate evaluation conclusions to make sure that they are not too damaging.
In a well‐developed evaluation system, the program of specifically budget‐linked evaluations
should therefore be only the tip of the iceberg of a much broader evaluation effort.
In conclusion, evaluations is critically important as part of the performance information base for
performance budgeting. However, to be useful, evaluation has to be of the right type – which
means both that it is timely, and that it is geared to answer the questions of most relevance to
Guzmán, Marcela (2007), “The Chilean Experience” in M. Robinson (ed) Performance
HM Treasury (1997), Appraisal and Evaluation in Central Government, London: The Stationary
Office, available at http://www.hm‐treasury.gov.uk/d/green_book_complete.pdf.
McKay, Keith (2007), How to Build M&E Systems to Support Better Government, Washington:
World Bank, available at http://www.worldbank.org/oed/ecd/better_government.html.
OECD (1998), Best Practice Guidelines for Evaluation, PUMA Policy Brief No. 5, Paris: OECD,
available at http://www.oecd.org/dataoecd/11/56/1902965.pdf.
16 For a case study of the Mexican system, see Manuel Fernando Castro et al (2009), Mexico’s M&E System: Scaling
Up from the Sectoral to the National Level, ECD Working Paper Series No. 20, Washington, World Bank (obtainable
World Bank (2004), Monitoring and Evaluation: Some Tools, Methods and Approaches,
Washington: The World Bank, available at
7. Performance Auditing
Defining performance auditing is not straightforward, because it is not possible to distinguish
performance auditing methodologically from evaluation. In practice, the term is usually used to
refer to evaluations which are carried out by supreme audit institutions (SAIs). SAI is the generic
term used internationally for bodies such as the UK National Audit Office, the US Government
Accountability Office, and the French Cour des Comptes, which are independent (to varying
degrees) of executive government and designed to hold it to account. However, within the audit
profession, the term “performance auditing” is sometimes also used to refer to evaluation
carried out by internal auditors – that is, by auditors who are part of the executive government
bodies which they audit. It may therefore be that the best definition of performance audit is
that it is performance‐related evaluation carried out by individuals or institutions with a primary
professional background and responsibility for financial audit.
Because SAIs are not part of executive government, performance auditing carried out by SAIs
are by definition not intended to serve (other than incidentally) as instruments to aid executive
budget preparation. Generally speaking, such SAI performance audits are better seen as
instruments to assist the parliament in more effectively carrying out its performance
accountability role. This distinguishes them fundamentally from executive government
evaluations systems, including the type of budget‐linked evaluation which ministries of finance
should be building as part of the performance information framework of performance
budgeting (as discussed in the evaluation section). However, in those countries where
parliament has significant independent power over the allocation of resources in the budget,
SAI performance audits can also have a more direct influence on the budget and be seen as an
integral part of the performance budgeting system. The section on parliament and performance
budgeting discusses these different parliamentary budget roles.
It is useful to distinguish two different types of performance audits, the first of which is the
substantive performance audit. Such audits aim to form explicit judgments on the effectiveness
and/or efficiency of programs, government agencies or functions. The other is the system
performance audit, which focuses instead on assessing the extent to which specific categories of
management system or practice are conducive of effectiveness and efficiency. Substantive
performance audit tends to be more common, and more wide‐ranging, in political systems
characterized by the separation of powers between the parliament and executive, and
consequently be considerable parliamentary independence vis‐à‐vis executive government. By
contrast in parliamentary systems where government ministers tends to control parliament,
there is sometimes a tradition that the SAI should not challenge government policy, which
(depending on how “policy” is interpreted) can have the effect of narrowing significantly the
scope of substantive performance audit.
In a number of countries, a type of executive performance auditing carried out by executive
government bodies which are separate from the ministries which they audit. For example, the
Indonesian Financial and Development Supervisory Board – a body like an SAI but part of
external government (there is a separate SAI which reports to parliament) – carries out
performance audits. And in France, the introduction of the new performance budgeting system
in 2005 was accompanied by the development of a form of executive systems audit carried out
by an Interministerial Committee for Program Audit, and focused on issues crucial to the
evolution of the new systems.
Shand, D. (2007), “Performance Auditing and Performance Budgeting” in Robinson (ed.)
8. Program Budgeting Fundamentals
Program budgeting is the planning, authorization and execution of expenditure in terms of
programs. Programs group together expenditure on specific public policy purposes, such as
environmental conservation or higher education. The classification of expenditure in terms of
programs turns the budget into an instrument for explicit choices about expenditure priorities
such as how much to spend on preventative health vs. treatment health; how much on tertiary
education vs. primary education; and how much on strengthening the army vs. promoting
Improving expenditure prioritization is therefore the primary objective of program budgeting
(although, as discussed below, program budgeting also acts by increasing the pressure on
ministries to perform). Expenditure prioritization refers to the allocation of funds to the sectors
and programs which are most effective in meeting social needs (see Section 1). In the public
sector, decisions about the allocation of resources are to a large degree be made by planning –
a process whereby either the government as a whole, or individual spending ministries, decide
what types of goods and services will be provided to the community, and to whom. This
contrasts with the market economy, where the allocation of resources between competing
products and services consumer is largely decided by customer demand, without the need for
conscious planning. Program budgeting is an instrument for integrating planning and budgeting.
Program budgets are very different from traditional budgets, in which funds are allocated to
each spending ministry not by objective but instead primarily by “line item”. Line‐items are
allocations of funds to types of inputs such as salaries, supplies and capital expenditure. The
problem with a budget based on line items is that it indicates next to nothing about the policy
objectives of the expenditure concerned. A line‐item budget is therefore essentially useless as
an instrument for expenditure prioritization. It was precisely to overcome this weakness that
the concept of program budgeting was originally developed. Program budgeting also differs
from traditional budgeting in that it calls for the substantial reduction of line‐item controls over
how spending ministries use their budgets. This is because program budgeting – and
performance budgeting more generally – call for greater freedom at the ministry level in the
choice of the inputs used to deliver services in return for greater accountability for the results
which ministries deliver to the community. This does not, however, mean that budget
allocations to line items entirely disappear under program budgeting (see Section 9).
To facilitate improved expenditure prioritization, programs (and their constituent sub‐programs
– see below) need to be defined in such a way as to capture the choices about spending
priorities which are made at the government‐wide level (i.e. by presidents, cabinets, ministers
and ministries of finance), and by the spending ministries themselves.
To capture such choices, programs are, first and foremost, categories of expenditure directed at
achieving a common outcome. For example, a nature conservation program covers expenditure
on a range of interventions such as the enforcement of laws banning the hunting of native
species, marketing campaigns designed to raise public awareness of the importance of
protecting the natural environment, actions to prevent the destruction of natural environments
which endangered species depend upon, and the “culling” of feral species (e.g. cats, introduced
fish etc) which may threaten native species. What all these interventions have in common is
that they all aim to preserve native fauna and flora – which is an outcome.
While programs are defined in large measure in terms of the common outcome which they aim
to achieve, in many cases they have other defining characteristics which also relate to the
choices about expenditure priorities which governments and spending ministries wish to make.
Budget decision‐makers are, for example, often also very much concerned with who benefits
from expenditure. For this reason, programs are some cases defined in terms both of their
intended outcome and their target client group or region. For example, education ministries
typically have separate programs for primary and secondary education. These programs have
the same intended outcome – educated and socialized young people – but they target different
client groups (i.e. young people in different age ranges). Similarly, in health treatment services,
the government may wish to make explicit decisions about the regional allocation of health
budget resources – the result of which may be a program structure in which there is a hospital
program which is comprised of regional sub‐programs.
Similarly, programs are sometimes defined in terms of both their intended outcome and the
mode of intervention used for achieving this outcome. For example, the intended outcome of a
“preventative health” program is reduced death and disability from disease and accidents, but a
defining characteristic of this program is that it aims to achieve this outcome by preventative
Programs are linked to the “results chain” not only via the outcomes concept, but also via the
concept of outputs. As explained in Section 3, outputs are goods or services – the “products” –
which a ministry or other government organization delivers to external parties. In stating that a
nature conservation program is comprised of a range of different “interventions” aimed at
protecting native flora and fauna, what we are saying is that the program is a group of different
outputs which share a common intended outcome (as perhaps other common characteristics
such as a common client group or mode of intervention). To take another example, a “crop
industries” program groups together a range of outputs such as extension services, fertilizer
subsidies and marketing support all of which have the objective of boosting the crops industry.
And a preventative health program might include outputs such as sanitation promotion
publicity campaigns, the placement of notices warning people against swimming or washing in
lakes or rivers with waterborne diseases, and the promotion of exercise and fitness in the
community. Listing the most important outputs (types of services) which fall under each
program is an important part of developing a good quality program classification.
In this context, it is important not to make the mistake of confusing outputs with activities, and
of viewing programs as being comprised of constituent activities rather than constituent
outputs. As explained in Section 3, activities are the work processes which are used to produce
outputs. For example, in health treatment, it is the treatments received by patients which are
the outputs, and the activities include nursing, surgery, cleaning and the maintenance of
hospital records. Programs are defined in terms of what matters most to the community – the
services (outputs) received by clients, and the outcomes which those services generate. They
are not defined in terms of activities and inputs, which are more of relevance for internal
management purposes within ministries.
In summary, programs are first and foremost groups of outputs with shared outcomes,
although they may also have other defining characteristics. Only when defined in this way are
programs a useful tool for expenditure prioritization. There are, as discussed below, limited
exceptions to the principle of outcome/output based programs, and these exceptions are
driven by purely pragmatic considerations which have nothing to do with the basic objectives of
Performance Pressure and Program Budgeting
Although the primary objective of program budgeting is improved expenditure prioritization,
this is not the only way in which program budgeting improves public sector performance. A
well‐designed program budgeting system also improves the effectiveness and efficiency of
expenditure by putting additional pressure to perform on ministries and other budget‐
dependent government organizations. Under program budgeting, each ministry’s budget
request to the MoF has to be accompanied by information on the performance of the ministry’s
programs (indicators and evaluation findings), as well as – where appropriate – performance
targets for the future. The knowledge that its budget request will be considered by the
government and, perhaps, the parliament in the light of its performance should then put
increased pressure on the ministry to improve its performance.
Program Performance Information
Program budgeting clearly requires not only budgeting in terms of programs, but also the
systematic use of program performance information. Only through the development of good
program performance information does it become possible to compare the budgetary cost of
each program with the results which the program delivers to the community.
This means, firstly, that program budgeting requires the systematic development of
performance indicators for each program (or sub‐program). Consistent with the expenditure
prioritization objective, these program performance indicators should as far as possible
measure program outcomes and outputs. The following provides an example of the types of
performance indicators which need to be developed for an effective program budgeting system.
Health Treatment Program
Outcome indicator Weighted average survival rate after treatment for selected
major life‐threatening conditions
Output quantity indicator Number of patients treated
Quality indicator Patient satisfaction rate,
Readmission rate after treatment,
Weighted average waiting time for selected major surgical
Efficiency indicator Weighted average treatment cost for selected basket of
One of the common deficiencies of program budgeting systems in many countries is that
program performance indicators are dominated by indicators of activities and inputs, such as
numbers of policies developed, numbers of meetings held and numbers of positions filled.
These types of indicators are very useful for internal management purposes, but they are not
the type of indicator needed to serve the objectives of program budgeting. The reason for the
dominance of this type of indicator is usually that they are readily available, whereas good
indicators of, say, outcomes and output quality are often not easy to develop. But using such
indicators simply because they are available is like the old story of the man who, having lost his
watch in a dark corner of a street at night, decided to concentrate his search under the street
lamp because he could see clearly there.
Program evaluation is at least as important an element of program performance information as
are program performance indicators. For the purposes of expenditure prioritization, the most
important performance consideration is the effectiveness of programs and their constituent
outputs. As noted in Section 6, performance indicators are rarely sufficient in themselves to
permit effectiveness of public expenditure to be assessed. Outcome indicators typically suffer
from many imperfections, one of the most important of which is that they rarely adjust, or
adjust fully, for the impact of “external factors”. Where this is the case, only evaluation can help
to distinguish the impact of external factors from that of the government’s own efforts.
Unfortunately, it is all too common internationally for governments to neglect the performance
information side of program budgeting and to believe that, merely by developing a program
classification of expenditure, they have implemented program budgeting. However, it cannot be
emphasized too strongly that merely knowing how much programs cost is of little use without
information about the benefits which programs deliver. Moreover, performance indicators alone
are insufficient for this purpose – particularly if they are mainly indicators of activities and
Program Titles and Objectives
Each program is defined by its title, overarching objective and the key outputs17 which it
includes. The program title should be short and informative. It should make as clear as possible
to the political leadership, parliament and the public what the program is. Depending on the
program, this might be done best by referring to the type of outputs, client, or objective of the
program. Examples of good program titles are: “nature conservation”, “crime prevention”,
“adult literacy”, and “curative health”.
Clearly defining the overarching program objective is important not only for clarity in program
definitions, but also to provide a framework for the derivation of program performance
indicators and targets. Program objectives should make clear the outcome which the program
as a whole seeks to achieve, as well as (if relevant) other defining characteristics of the program
such as the target client group or mode of intervention.18 The program objective must be a
17 Or, exceptionally, support services in the case of “administration programs (see Section 11).
18 Because the outcome is the core of the program objective, one might ask whether it would make things clearer
if spending ministries were told to develop statements of “program outcomes” rather than “program objectives”.
Some countries do exactly this. There is, however, one very good reason for referring to program “objectives’’,
statement of the objectives which are shared by all of the different services grouped together
within the program. Subordinate, more “operational” objectives may19 also be defined for each
program, and these will in general relate to some but not all of the services covered by the
program. The discipline involved in requiring spending ministries to define an overall objective
for each program helps to ensure that programs are – other than in the case of certain defined
exceptions – appropriately defined with reference to outputs and their common outcome.
Program objectives should be explicitly and briefly – ideally in a sentence – defined in the
budget documents. There are many cases internationally of program objectives which are not
well defined. It is, for example, not unusual to find program objectives which focus entirely on
the output (service) which the program delivers to the public, or on program
activities/processes, with no reference to the intended outcomes (see below). Sometimes,
program objectives are formulated in a wordy and unclear way, perhaps with excessive use of
The test of a common outcome which can be expressed in an overall program objective is an
important test of the integrity of the program definition. For example, if the army health service
has over time expanded its role from one of providing health services to soldiers to one of
direct health service delivery to the public – in competition with the health ministry – it would
not make sense to include the army health service in an “army” program, the outcome of which
should be defined in terms of national security.
Right and Wrong Ways of Defining Program Objectives
Examples of well‐formulated program objectives are:
“The conservation of biological diversity in healthy ecosystems” (Nature Conservation program),
“Maintenance of territorial integrity and national independence” (Armed Forces program),
“Increased foreign investment leading to technology transfer and a stronger economy” (Investment
“Reduced crime and greater security of persons and property” (Crime Prevention program).
rather than to program outcomes. This is that there are certain exceptional programs which do not deliver services
directly to the public (see below), and which do have aim to achieve specific outcomes, but rather to support other
programs which are outcome‐focused. Such programs still have objectives, but these objectives are usually not
19 In fact, in a fully developed program budgeting system, such more detailed objectives should definitely be
developed. However, this need not be done at the outset in the implementation of such a system, whereas
overarching objectives should be defined simultaneously with programs themselves.
Examples of the wrong approach to defining program objectives are:
“Provision of medical assistance to persons in an emergency” – refers only to outputs.
“Manage the development, implementation, evaluation and maintenance of national policy,
programs and systems for general education and quality assurance” – refers only to activities.
To achieve its aim of improved expenditure prioritization, an effective program budgeting
system must bring information on the performance of programs – that is, on their success in
achieving their intended results – together with information on their cost. Being able to see the
results achieved by programs alongside their cost helps budget decision‐makers to make better
judgments about whether programs should be cut, expanded, or maintained.
Parliament and the public should be kept informed via program statements presented with the
budget documents which accompany the annual budget legislation. Program statements should
include the following information for each program:
Title and objectives,
List of the main outputs (services) which comprise the program,
A brief narrative outline of program strategy, challenges and key new initiatives,
Key program performance indicators,
Program performance targets, if applicable,
Program expenditure estimates, preferably with medium‐term projections,
A breakdown of program expenditure by broad categories of economic classification (staff,
capital etc), for information purposes.
The specific content of the program statement in respect to a number of these elements is
discussed in more detail in subsequent sections.
One way of presenting this material is for each ministry to prepare a document containing all of
its program statements to be made available to parliament as an annex to the budget
documents. The MoF should determine a standardized format for these documents.
The Program Hierarchy
Programs are typically decomposed into constituent “sub‐programs”, and sometimes even into
lower levels (sub‐sub‐programs etc) – although it is suggested in later sections that it is better
to keep the number of levels limited. An example of this type of “program hierarchy” is shown
in the box below. Note that the terminology employed for these elements of the program
hierarchy varies greatly around the world (as does the terminology used for programs). The
following is an Australian example from a few years back showing how programs (in this case
called “outcomes”) are broken into sub‐programs (called “outputs”) which were further
decomposed into sub‐sub‐programs (called “sub‐outputs”).
From Program to Sub‐program: an Australian Example
Dept of Environment
Outcome (Program) 1
Environmental Protection and
Outcome (Program) 2
Antarctica are advanced
Output (Sub‐Program) 1.1
International Climate Change
Output (Sub‐Program) 1.2
Conservation of Land and Inland
Output (Sub‐Program) 1.3
Conservation of Coast and Oceans
Output (Sub‐Program) 1.4
Conservation of Natural, Indigenous
and Historic Heritage
Subprograms and other lower‐level elements of the program hierarchy are used mainly for
internal management within the ministry or agency concerned. For the most part, they
represent a level of disaggregation of expenditure which is too detailed for the central decision‐
makers to concern themselves with during the budget preparation process.
Number and size of programs
Programs are the level at which central decisions about expenditure priorities will generally be
made. This has two important implications when formulating the program structures for
Creating one big program covering all or most of a ministry’s expenditure is generally a
mistake. For example, it is better to have a number of programs such as primary education,
secondary education and tertiary education in an education ministry than to have a single
enormous education program. And rather than a single agriculture program, an agriculture
ministry should have a number of programs, perhaps along industry lines (i.e. separate
programs for crop industries, livestock industries, and fisheries). A program classification
based on giant programs will be too coarse to serve central decision makers in making
strategic spending reallocations such as, for example, shifting money from tertiary
education to primary education or from treatment health to preventative health. Expressed
differently, programs should capture each of the distinct key aspects of the role of each
ministry so as to permit central decisions about the broad lines of the ministry’s service
delivery focus. It is in no way inconsistent with this to note that, for small ministries with
narrowly‐focused missions, a structure with a single program may be appropriate.
Too many small programs should be avoided. Because central decision makers need to
concentrate primarily on the “big picture” of expenditure prioritization across government,
a proliferation of very small programs runs the risk of unnecessarily complicating the central
budget preparation process.
A consistent approach to the size and number of programs needs to be applied across
government. It would, for example, be inappropriate if the health ministry decided it would
have a single program, while the justice ministry decided to have twenty. Guidelines from the
MoF on this are essential (see further below in this section). Inevitably, such guidelines will
need to have recourse to rules which are arbitrary but necessary – in particular, a limit on the
number of programs any ministry have (perhaps something between 5 and 8, but this depends
on how many ministries the government has), and a minimum size for programs (e.g. not less
than a certain value, or alternatively no less than a certain percentage of the ministry budget).
The proper definition of programs and the program hierarchy is a task which must be got right
if program budgeting is to work properly. This means that the MoF cannot simply leave it to
ministries to determine what their programs will be. The MoF must, rather, issue clear
guidelines covering issues such as the formulation of program objectives, program size and
number, and whether there will be a single administration program in each ministry (see later
sections). Only in this way will the interests of central budget decision‐makers in having
programs which help them translate government priorities into budget numbers be respected.
On the other hand, it would be wrong for the MoF to dictate to each ministry, without
consultation, what its programs will be. Spending ministries know their business better than the
MoF does, and it would be a mistake not to benefit from spending ministry expertise.
Moreover, the program structure – particularly at the lower levels of the program hierarchy – is
intended to serve as an internal budgeting and management tool for the spending ministries
themselves. This will not happen if the spending ministries feel no sense of ownership of the
In summary, the development of the program structure must be a collaborative endeavour
based on a partnership between the relevant spending ministries and the MoF.
Integrating Strategic Plans with the Program Budget
Program objectives need to be explicitly linked to the objectives formulated in the
organization’s strategic plan and in any government‐wide strategic or national plans. More
generally, planning and programming should be seen as part of an integrated cycle.
A well‐formulated government‐wide strategy will define a small number of high‐level outcomes
upon which the government is focused. The clear specification of program objectives then
provides the natural means of linking programs to the government‐wide strategy. This is
because program objectives are generally “intermediate” outcomes through which the high‐
level whole‐of‐government outcomes are achieved. For example, the outcome of a Criminal
Policing program (“reduced crime”) would link it directly to a whole‐of‐government outcome
such as “a safe and secure society”. Similarly, the outcome of the High School Education
Program (“young people educated and socialized to take their place in the workforce or higher
education”) would link it directly to whole‐of‐government outcomes such as “rapid and
sustainable growth of GDP”.
This type of link between the programs and national strategic outcome can be represented
graphically as illustrated the figure below.
Linking Programs and National Strategic Outcomes
Strategy should also be linked with the program budget at the entity level. Canada, France and
the United States are amongst the leading countries which have fully integrated entity strategic
plans with the budget by including within the budget documents entity performance plans
Based on the program structure of the budget: in other words, performance objectives,
plans and targets are outlined for each of the entities programs,
Covers all entity spending, not just “new” spending,
Include only entity spending initiatives for which the government has approved funding in
These three countries also require that entities present annual performance reports which
report on success in achieving the objectives set out in the performance plans.
In the US, this approach has been mandatory since Congress passed in 1993 the Government
Performance and Results Act (GPRA). The Performance Plans which GPRA requires are not full
entity strategic plans, but GPRA stipulates that they must be consistent with the entity strategic
The structure of the French Projets Annuels de Performance (annual performance plans) is
slightly different in that it is based not on entities but on “missions”, which are groups of
programs which in some cases spread over more than one entity. Apart from that, the French
performance plan structure is very similar to other countries. It is broken into sections for each
budgetary program, with presentations of program objectives, strategy, indicators, and
Elements of this broad approach to integrating entity strategic plans with the budget are to be
found in many other countries. In New Zealand, for example, the government’s budget
documents include entity Statements of Intent which are essentially strategic plans structured
around entity‐specific outcomes (equivalent to programs). In South Africa – one of the many
developing countries which have adopted a similar approach – regulations also stipulate that
each entity’s “strategic plan must … be consistent with the institution’s published medium term
Robinson, M. and H. van Eden (2007), “Program Classification”, in Robinson (ed.) Performance
Kraan, D.‐J. (2008), “Programme Budgeting in OECD Countries”, obtainable at
9. Program Appropriation and Expenditure Control
Program budgeting involves not only planning, but also approving and executing the budget in
programmatic terms. In this section we focus on the nature of the budget authorization under
program budgeting and the degree of flexibility in respect to that authorization during budget
Under most program budgeting systems, the parliament votes the budget allocations in
programmatic terms, and this programmatic allocation constrains executive government in the
execution of the budget. The parliamentary authorization is in most countries only at the
program level. Only in a minority of countries does parliamentary appropriation of the budget
take place at sub‐program level, which tends generally to be thought of as areas of executive
In considering the nature of the enacted budget, however, we should have regard not only to
the form in which parliament approves the budget, but also the additional “control totals”
which the minister of finance or other executive bodies might impose. For example, in a few
countries, even though the parliament only approves the budget at the program level, the
minister of finance approves and controls the allocation of programs between sub‐programs.
A key question under any program budgeting system is therefore that of how detailed a control
to exert at the center (whether by parliament or central executive bodies). One approach to
this question is to recognize that the information available to central decision‐makers is never
going to be sufficiently strong to permit them to determine all of the detail of resource
allocation across the whole of government. Broadly speaking, the center is better of
concentrating on the “big picture” of resource allocation, while leaving considerable discretion
to spending ministries on the detail. From this perspective, allocation at the sub‐program level
may be better left to ministries themselves to determine.
Central approval of the budget in terms of programs raises the question of transfers
(sometimes known as virements) of expenditure between programs. This refers to transfers of
money between categories of funding approval – in this case, between programs. In many
program budgeting systems, parliament authorizes executive government to make limited
20 There are exceptions to these generalizations. One is the “global appropriations” system of countries like
Australia and Canada, under which the programmatic allocation is merely indicative and can be changed by
executive government at its discretion.
modifications to the approved program allocation of funds – for example, to move up to 5
percent of the funding of any program to other programs without having to come back to
parliament to modify the budget. This type of transfer arrangement is intended to create a little
flexibility to respond to unanticipated developments which might occur during the year.
The other key issue which arises under program budgeting is the reform of the traditional
budget appropriation framework. As mentioned earlier, traditional budgeting is based primarily
on “line‐item” approvals. This means legal allocations of funds to inputs – known formally as
“economic classifications” of expenditure – such as salaries, supplies and capital expenditure. In
traditional systems, the line‐item allocation is sometimes very detailed. For example, there may
be separate fund allocations to items such as office stationary supplies, books, internal training,
external training etc. This means that spending ministries are then unable ‐‐ at least without
central approval – to shift money between these categories during the year. In practice, some
flexibility is created by systems which permit some transfers between line‐item categories.
However, not only is the scale of such transfers typically greatly limited, but the process of
reviewing and determining transfer applications can involve a great deal of unproductive work
for the MoF.
Performance budgeting is opposed to his type of detailed control over the line‐item
composition of expenditure. The performance budgeting focus is on the results delivered by
agencies, not on the way on which they deliver those results. Performance budgeting calls for
ministries, and managers within those ministries, to be given greater freedom to choose the
best mix of inputs with which to efficiently and effectively deliver services. The corollary of this
is increased accountability for the results delivered.21
For this reason, performance budgeting has typically been accompanied by extensive reduction
of line‐item budget controls. For example, under the French system, controls on the line‐item
composition of program expenditure have been totally abolished with one (important)
exception – personnel expenses. It is not permitted to shift money from other input categories
to personnel expenses (although ministries can shift money from personnel expenses to other
21 Very detailed input controls have, moreover, a downside unrelated to performance budgeting issues – namely,
that they tend to aggravate considerably the problem of under‐execution of budgets which can be seen in many
countries. Under‐execution means that actual expenditure falls well short of the level of expenditure approved in
the budget. Detailed line‐item controls contribute to this because, when there is under‐spending in one line‐item
category (e.g. internal training), it is not possible to use the funds approved for other purposes.
The key practical question is then how much line‐item deregulation, and how fast. It should not
be thought that program budgeting demands the complete abolition of central controls on the
input allocation of program expenditure. In the French system, for example, control over the
allocation to personnel expenses is controlled for the very good reason that France is a
traditional civil service system where, once someone is hired, it is very hard to terminate their
employment. So if a ministry was to hire more people by shifting funds within its budget, it
would be creating commitments to future expenditure which could potentially greatly reduce
the budgetary flexibility of the government in the future.
In nearly all program budgeting systems, capital expenditure is protected – that is, spending
ministries cannot shift money from capital to current expenditure within programs. The
rationale for this is a concern about the temptation ministries would otherwise face to make
short‐sighted cuts to capital spending to ease their current budget constraints.
Particularly in developing countries, there are other types of line‐item controls which it may
make sense to retain, such as:
Line‐items for utility bills and other essential payments: this would be relevant if there was
a concern that, without such controls, ministries would spend their budgets on other items,
leaving insufficient to pay their electricity and phone bills, and would then demand
supplementary funding from the MoF to pay these bills.
Controls over travel, consultancy and other items particularly susceptible to abuse.
These examples highlight the more general point that the practical scope for line‐item
deregulation in any country depends significantly on the quality of governance and on the
degree of civil service discipline. Rather than simply following the extreme examples of France
or other countries which have reduced their line‐item controls to a minimum, each country
needs to make its own case‐by‐case decisions about what line‐item controls to retain. It will
also typically make sense to adopt a gradual approach to line‐item deregulation.
This being said, it makes no sense at all for any country adopting program budgeting to retain
the highly detailed traditional line‐item controls referred to above. Most such detailed controls
serve no useful function whatsoever. The best approach will be to accompany the move to
program budgeting with a review of line‐item controls which aims to abolish most detailed
controls, and to retain only those for which a clear rationale can be put forward.
Robinson, M. and H. van Eden (2007), “Program Classification”, in Robinson (ed.) Performance
10. Program Accounting and Costing22
This section examines the accounting, information technology (IT) and budget classification
requirements of program budgeting. The principal point is that if program budgeting is to work,
programs must be integrated into the accounting system, the way in which the budget is
structured, and the IT systems which support the budgeting process.
If a government is going to budget by programs, the budget documents and the accounting
system must be program friendly. In particular, it must be possible to follow budget execution in
programmatic terms – that is, to monitor expenditure program by program during the year so as
to make sure that program expenditure authorizations are not exceeded. Insofar as line‐item
authorizations (for spending by economic classification – see further below) are expressed in
program terms, it must be also possible to monitor these. For example, if limits are set on salary
expenditure for each program, then the accounting system must continuously monitor salary
expenditure by program. It is not sufficient, for example, to make an estimate of program
expenditure at the end of the year by some approximate means (which would be enough if
programs were intended only for reporting and not budget execution purposes – that is, to
provide information on the level of expenditure by policy objective).
Budget Classification and Chart of Accounts
For a program budgeting system to work, the “Budget Classification” (BC) and “Chart of
Accounts” (COA) must incorporate programs. The BC and COA are classification systems, in
which there is a code number for each transaction which summarizes all of its characteristics
according to the classification system. These systems cover not only expenditure, but other
transactions such as revenue. But our focus here is exclusively on the classification of
expenditure. Expenditure on a particular item (say a salary payment, or a purchase of specific
supplies) will in these systems be coded with a sequence of numbers such as 1321‐325‐257‐3,
where “13” stands for the ministry, “2” for the department within the ministry, “1” for the unit
within the department, “”325” the three levels of the economic classification (i.e. the type of
input – whether, for example, salary, capital expenditure etc), “257” symbolizes the three levels
of the functional classification (more on this later) and the “2” the fund source (e.g. donor
funded). The classification categories used are not set in concrete, but may vary between
countries. Some countries might have additional ones, such as a geographic classification, in
22 This section relates to session 2.4 in the generalist course, and session 1.5.4 in the specialist course.
which case there would be an additional code representing the region in which the expenditure
The COA is a coding system used when recording transactions in the accounting system. This
means that for every payment made (or commitment made), ministry accounting staff would
enter the appropriate code when recording the transaction in the accounts (concretely, in the
computerized accounting system). The BC is the classification used in the budget itself – that is,
in describing in the budget documents the expenditure which is to take place in the coming
year.23 The COA and BC should be integrated in the sense that the former should subsume the
latter. In other words, the accounting system should record all the expenditure transaction
categories that are the basis of budgetary approvals or information, so as to enable expenditure
control and reporting in terms of those categories.
Under program budgeting, programs must be incorporated into the COA and BC. Not only the
program, but the subordinate elements of the program hierarchy (sub‐programs etc), should be
coded for. So if there is a three‐level program hierarchy, three digits will be required for coding
purposes. It will also be necessary that training and validation processes are put in place to
ensure that accounting staff record the program properly when entering expenditure
transactions into the accounting system.
Financial Management Information Systems
Financial management information systems (FMIS) refer to computerized systems for managing
budgeting and associated processes. Computerized accounting systems are the most basic type
of FMIS. However, many other processes are also computerized. Automated expenditure control
is an example of such a process, and refers to computerized systems which are used to
authorize commitments and/or payments by verifying that they are consistent with the
expenditure limits set in the budget. Other computerized system will cover other budget
execution functions such as cash management, debt management etc. There are also systems
which cover various aspects of budget execution.
Not only must the accounting system be program friendly, but other systems within the FMIS
must also be adjusted to work with programs if program budgeting is to succeed. The
23 Whether for the purpose of describing “control totals” (see the program accounting section) or providing
supplementary information on the expected composition of expenditure (e.g. details of the expected breakdown
of program expenditure by line item, even when this is indicative rather than binding)
expenditure control system is an obvious example – if the budget sets limits for expenditure by
program, it will be necessary that the expenditure control system as well as the accounting
system operates on a program basis. Concretely, this means that the expenditure control system
would prevent commitments being made for spending on program X in excess of the amount
voted by parliament for program X. If there are program transfer provisions, the expenditure
control system will also need to manage this – it should permit (subject to appropriate
authorization) transfers between programs up to the established limit (e.g. five percent), but
not in excess of this. If there is a computerized budget preparation system, this also will need to
change to recognize the fact that ministries will now be putting forward spending proposals
linked to specific programs.
The acronym IFMIS refers to an integrated financial management information system. This
means a single system which to a greater or lesser degree integrates multiple functions. With an
IFMIS, accounting, expenditure control, payments, budget preparation and a range of other
functions will be built into the same large computerized system. An IFMIS differs from a system
which is based on a series of separate systems which draw data from one another but are not
otherwise integrated – in the sense that they do not automatically “talk to” one another, so that
data entered into one system does not automatically modify relevant variables in other
The scope of an IFMIS – that is, the number of different budgeting and related functions which
it incorporates – is not fixed, but varies considerably internationally. To be considered an IFMIS,
a system must integrate at least a minimum set of “core” functions (typically including
accounting and reports, expenditure control, payments). It may also include certain “non‐core”
functions, such as budget preparation, cash management, debt management and even
performance management (i.e. recording of program objectives, indicators, monitoring against
The key issue for program budgeting is how to ensure that the financial management
information systems are program‐friendly. Some assume that this requires the introduction of
an IFMIS which incorporates all of the functions relevant to program budgeting, including for
example budget preparation on a program basis and performance management.
24 For example, there may be a distinct payments system which will be separate from the accounting system but
which will use data from the expenditure control system to determine whether a specific payment is based on a
commitment which was previously authorized by the expenditure control system as compatible with budgetary
This may be an inappropriate assumption for a number of reasons. Firstly, if program budgeting
is to be introduced in a country which lacks a functioning IFMIS – that is, a country which has a
financial management information system based on a set of separate interfaced systems – it
needs to be recognized that the replacement of such “legacy systems” (as they are often called)
with an IFMIS is not something which can be done overnight. IFMIS implementation takes, as a
rough generalization, a minimum of five years. Many countries, when introducing program
budgeting systems, wish to move more quickly than this. If they do, this will require them to
modify the legacy systems to make them program compatible, even if in the medium of longer
term they are planning to introduce an IFMIS. In France, the program budgeting reforms which
came into full operation in 2006 were in their early years made to work with the pre‐existing
financial management information system, modified in the most essential ways. This was
despite the fact that even at that time a massive project was already underway for the
introduction of the highly‐integrated “Chorus” IFMIS system. This is because the Chorus was
such a major project that it required many years preparation and implementation, and to wait
until this was completed would have inappropriately delayed the introduction of the
performance budgeting system.
Secondly, there is a close association between the comprehensiveness of the planned system
and the scale of the implementation difficulties and delays. As DFID (2001: 61) note,
“international expenditure suggests that there is a real risk of failure in the introduction of large
management information systems”. It is therefore a big mistake to assume that an IFMIS should
necessarily try to integrate everything into one gigantic system. Diamond and Khemani (2005: 3‐
4) also warn:
…the term “integrated financial management information system” can sometimes be
erroneously interpreted as describing a system that can capture all the functional
processes, and the relevant financial flows, within public expenditure management.
However, the complexity of information systems within the government sector is, to a
large extent, due to the multiplicity of functions and policy areas. In many functional areas
specialized information systems are in place and will still be required even with the
implementation of an FMIS.
French experience once again highlights the problem. It was just as well that performance
budgeting was not delayed to await Chorus, because Chorus implementation subsequently
experienced massive problems, to the extent that for a period in 2010 payments to suppliers
largely stopped and major payment arrears accumulated.
As noted in the section of performance information systems, it is also wrong to assume that
program budgeting requires that a performance management module be integrated into the
IFMIS. Program objectives, indicators and targets can be perfectly well managed via a much
simpler separate platform which might even be based on Excel spreadsheets. More generally, a
move to program budgeting should not be seen as requiring that one over‐complicate the
computerized financial management system in a way which is only likely to cause major
Program Cost Allocation
We now turn to a quite different issue: the methodology guiding the allocation of expenditure
between programs. Expenditure is carried out largely by the purchase of inputs25, so the
accounting for expenditure by program requires that accounting staff be able to determine the
program under which the expenditure on any given input should be recorded in the accounts.
There are two distinctly different aspects of this issue, concerning respectively direct and
In costing programs, a direct cost is expenditure on an input which contributes to that program
and that program alone. The same is true in respect to sub‐programs – a direct cost is a cost
associated with only the sub‐program concerned. Assume, for example, that there is an
agricultural “extension” service (i.e. a service providing technical advice to farmers) and that
this service constitutes a sub‐program within, say, a crop industries program. The wages and
other employment costs of field advisory staff that spend 100 percent of their time providing
advice to farmers would be a direct cost of that sub‐program. So would the operating costs of
any motor vehicles dedicated exclusively to that service. On the other hand, an indirect cost
would in this context be any item of expenditure which contributes partly to that sub‐program,
but also to other sub‐programs. For example, if a motor vehicle were sometimes used to
support agricultural extension service, and sometimes to support staff delivering other types of
services within the crop industries program (or even other programs), it would be an indirect
25 Of course, there are also “transfer payments”, such as income support benefits.
Consider this issue from the perspective of accounting for expenditure between programs
within a specific ministry. There are within any ministry a range of “overheads” – expenditures
on services which support the ministry as a whole. Such overheads include, for example, the
employment costs of top ministry management, the function of which is to manage across the
ministry as a whole. They also include the vital service functions within the department such as
finance and human resources. If we assume for the moment that there is no separate
“administration” program covering these overheads, then they are by definition indirect costs
which would have to be allocated to the ministry’s programs. For example, in the case of the
education ministry, it would be necessary to determine how much of the ministry’s finance,
human resources, top management etc expenditure contributed to the primary school
education program, how much to the secondary education program, and so on.
Direct costs are, in principle, the “easy” part of program costing. Because they are costs which
contribute to only one program (or sub‐program), all that has to be done is to ensure that they
are recorded against the correct program or sub‐program. (As noted below, however, even this
can be a significant challenge in some developing countries.)
The allocation of indirect costs is less straightforward. Take the example of the regional office
staff of an environmental agency, many of whom might be involved in a wide variety of different
services, ranging perhaps from nature protection to environmental pollution management.
What methods might be used to allocate their employment costs to specific outputs?
One method would be to require the staff concerned to fill in time sheets recording how their
time is allocated, in the manner that law firms typically use for billing purposes. Government
departments in some countries have implemented such time recording systems. This may be,
however, a costly process, particularly if the time sheets are to be reasonably accurate (which
will require careful monitoring by management to make sure that staff fill in their time sheets
accurately). This points to a key issue which arises in allocating some indirect costs: the expense
of the cost allocation process itself. Because of the expense of acquiring truly accurate cost
allocations, it is commonplace for shortcuts to be adopted, as a conscious compromise between
accuracy and expense. In the case of staff time allocations, for example, the compromise might
be to require staff to fill in time sheets for a specific period during the year (say one month) and
then assume that this sample data was representative of the whole year.
Information technology has in some cases greatly reduced the costs of obtaining accurate
allocations of indirect costs. Take the examples of telephone costs or electricity usage. It is a
relatively straightforward and cheap matter these days to generate accurate information
apportioning an organization's total phone bill to individual phone lines, or electricity usage
amongst different organizational units. In the past, however, this was impossible or very
difficult. Note, however, that in some developing countries, technology to record the use of
such services may not be affordable, and as a consequence such utilities remain in effect
Indirect cost allocation is a task of management accounting which, while widely employed in
the private sector, is not typically part of the arsenal of accounting tools of government. It
involves the definition and application of “cost drivers” to allocate expenditure. A cost driver is a
formula or principle which is used to determine the proportions in which an indirect cost is split
between two or more programs (or sub‐programs) to which it contributes. In the example
above, both the apportionment of telephone costs of a pro‐rated basis reflecting direct service
staff, and the apportionment is accordance with computerized usage records, represent
(different) cost drivers. One may be arbitrary and the other accurate, but they are both cost
drivers. Because cost drivers provide the basis for determining the proportions in which indirect
costs are allocated, the term allocation basis is sometimes used to mean the same thing as a
To keep things simple, accountants often allocate indirect costs using very simple cost drivers.
For example, the cost allocation formula might assume that all the support services in the
ministry contribute to programs in a manner proportionate to the staff numbers within the
major organizational units in the ministry. For example, if thirty percent of the staff of the
ministry of education work directly on the primary education program, it might be assumed that
thirty percent of the cost of the ministry’s support services is attributable to the primary
education program. However, this would be a very crude approximation which might be very
To allocate indirect costs in a reasonably accurate (as opposed to highly arbitrary) way is a
complex business, requiring not only additional highly skilled accountants, but more
sophisticated IT systems. Amongst the sophisticated indirect cost allocation methods available is
activity based costing (ABC). There are two defining features of ABC. The first is that it is a two‐
stage costing process as applied to program costing, in which inputs are allocated firstly to
“activities” (including support activities such as HR services or IT services) and activity costs are
then allocated between programs. The second defining feature of ABC is an insistence upon the
use of cost drivers which are not arbitrary but which provide an accurate reflection of the
contribution which inputs and activities actually make to outputs. Advocates of AOB view it as
requiring the use of cost drivers which constitute 'a model of organizational resource
consumption'. The ABC emphasis upon accurate cost drivers is laudable. However, the price of
accuracy can in many instances be unacceptably high. Even in developed countries it is not
clear that ABC has been a success in the public sector. In developing countries, it is generally
speaking far too demanding of skilled human resources and sophisticated IT systems to be
feasible or justifiable in cost terms.
The difficulty of accurate indirect cost allocation has led, for most government, to a pragmatic
modification of the program budgeting principle. This is the use of distinct “administration”
programs to cover the overhead (support service) of each ministry. Administration programs
are, clearly, inconsistent with the pure program budgeting principle of outcome/output based
programs, as outlined in Section 8. Internal support services do not provide services to external
clients, and are therefore by definition not outputs (see Section 3). Moreover, precisely because
they do not deliver services to the public, support services are not capable in isolation of
achieving outcomes for the community. Rather, they support those parts of the ministry which
do provide services directly to the public, and which through those services achieve outcomes.
Therefore, if all programs were defined in terms of outputs and outcomes, there would be no
administration programs and all ministry‐wide overheads would be indirect costs which will
need to be allocated between programs.
However, precisely because the allocation of indirect costs with an acceptable level of accuracy
is a demanding and technically complex undertaking which is not cheap, the majority of
countries choose to largely avoid it through the use of administration programs (also known as
“management”, “support” or “corporate services” programs)26. Administration programs cover
the costs of ministry support services which would otherwise be difficult to allocate. Countries
which make use of administration programs take the view, however, that it is better to
compromise on the program budgeting principle in this way than to use of an arbitrary cost
26 And, analogously, they commonly make use of administration sub‐programs within each program, covering
allocation approach which would yield “full cost” program expenditure measures which might
be highly inaccurate.
If the only justification for administration programs was the difficulty of accurate indirect cost
allocation, it might be appropriate to view administration programs as a temporary measure. In
other words, administration programs could be viewed as no more than a practical mechanism
to be used in the early days of a program budgeting system in countries where the capacity and
systems for accurate allocation of indirect costs do not yet exists. In the longer term, the
aspiration would be phase out management programs as and when the capacity to allocate
indirect costs is developed. This is a view which many people take.
It is, however, possible to argue the use of management programs to cover ministry support
services can be justified as a matter of principle even in countries which have the technical
capacity to allocate expenditure been programs.
To see why, keep in mind that program budgets are in most countries expenditure limits
(“control totals”), and not simply accounting measures of the cost of producing specific
“product lines”. In other words, as discussed in Section 9, when the parliament approves a
program budget, it is telling ministries that they are to spend no more than $x on program A, $y
on program B, etc. This means that if the expenditure of the ministry’s support services is
included in its outcome/output‐based programs, the government is in effect instructing the
ministry about the extent of the support the ministry’s support services should provide to each
of the ministry’s output‐based programs.
Suppose, for example, the education ministry has only two programs – primary and secondary
education – and no management program. Imagine, moreover, that the ministry’s
organizational structure consists of a primary school department and a secondary school
department – each of which is budgeted internally by the ministry to receive 40% of the
ministry’s overall budget – and a support services department which is budgeted internally to
receive 20% of the ministry’s budget. Supposed further that it is expected at the time the
budget is prepared that the support services department will provide equal levels of support to
each of the two main departments. Reflecting this, the program budget approved by parliament
will allocated one‐half of the ministry’s budget to each of the primary and secondary education
programs, in each case including that program’s share of expenditure on support services.
To ensure that neither program breeches it expenditure authorization, the education ministry
will then need to make sure that the support services department does not devote substantially
more than one‐half of its expenditure to supporting either of the two ministry programs. More
generally, because support services expenditure is included in the output based programs, it
becomes necessary in budget execution for the support services department to rigorously
manage the balance of support it provided to the other ministry departments.
Herein lies the problem. Why would the political leadership or the MoF wish to dictate the way
in which education ministry support services allocate their efforts between supporting primary
and secondary education during the year? If, for example, unexpected events mean that the
primary school program has a greater‐than‐anticipated requirement for the support of the HR
group in staff recruitment during the year, shouldn’t the ministry be left with the flexibility to
shift the disposition of its support services accordingly?
If one believes that ministries should retain the flexibility to allocate support services to where
they are needed during the year, then it makes sense in principle to have a management
program. That management program should included all those ministry support services which
are generic and which can be flexibly reallocated during the year from supporting one output‐
based program to another to meet shifting support requirements27.
Of course, it is still useful to know ex post where the support services are allocated – in order,
ideally, to have the best possible accounting measures of the full costs of delivering each
“product line”. For this purpose, additional cost analysis can be carried out to allocate the
management program expenditure between the output‐based programs.
This is, incidentally, exactly what the French do – even though France is perfectly capable of
carrying out the accounting allocation of support service expenditure between output‐based
programs, it nevertheless makes us of management (“support”) programs and supplements this
with additional cost analysis.
27 It is relevant in this context that in Australia, where there are no administration programs, programs are not
used by parliament as control totals (legal appropriation categories). This means that the allocation of funds to the
budget is purely indicative, and therefore does not in any way constrain the actual allocation of support services
between the main program areas of ministries.
Salary Costs and Programs
It was suggested above that the allocation of direct cost is a relatively easy task. In many
developing countries, even direct cost allocation between programs can present significant
challenges. It is not uncommon, for example, for ministry accounting systems in such countries
not to differentiate personnel expenditures by ministry organizational unit. This problem is
often compounded by systems in which, when staff from one ministry are posted to another
ministry, they continue to be paid by their original ministry, as a result of which their salaries are
not recorded as expenditure of the ministry in which they are actually working. This means that
the accounting system in, say, the education ministry might give no indication of the total wage
costs of staff working on primary education, making it essentially impossible to cost a primary
For these reasons, developing countries are often tempted to record all ministry personnel
expenditure under the “administration” program. While one can sympathize with the practical
reasons for doing this, such a course of action fundamentally undermines the logic of program
budgeting. Given that personnel costs represent a large portion of government expenditure in
most countries, the results of excluding such expenditure (at least on the current expenditure
side) is “programs” which only record minor expenditure (e.g. furniture, supplies) and which are
therefore of very little use as a tool of expenditure prioritization.
In countries which face this problem, the proper allocation of salary costs to relevant programs
should be the first priority on the accounting side in preparing for the introduction of a program
budgeting system. All personnel charges of staff members working exclusively on specific
programs (which will account for most personnel expenses) should be allocated to programs,
with only the salaries of staff in support services included in the administration program.
Cost Allocation and the Program Hierarchy
The cost attribution issue – in respect not only of indirect but also of direct costs – is relevant to
the decision of how many levels to have in the program hierarchy. The accounting system must
be capable of recording expenditure against each element in the program hierarchy. This means
that for each additional level added to the program hierarchy, the cost attribution issue
becomes potentially more challenging. If, for example, you have a program structure comprised
of four levels, it is necessary to be able to follow budget execution – that is, to monitor
expenditure via the accounting system during the year – right down to the lowest of those four
levels. Suppose that each program is comprised of four sub‐programs, each sub‐program of four
sub‐sub‐programs and, finally, each sub‐sub‐program or four sub‐sub‐sub‐programs. If this is
the case, then a four level program structure involves classifying and monitoring expenditure in
sixteen times as many categories as does a two‐level structure. This might greatly increase the
challenge of correctly recording expenditure as well as that of allocating indirect costs. Given
the magnitude of the difficulties many developing countries face with their accounting systems,
it will hardly come as a surprise to say that few succeed in successfully implementing program
hierarchies with three or four levels.
The simplest option – one which may be appropriate for many developing countries, for
example – is to have a program hierarchy with only one or two levels (programs, or programs
and sub‐programs). The appropriate approach to this issue will also need to take into account
the relationship between organizational units and program elements, which is discussed below.
Program Cost Estimation in Budget Preparation
The third and final issue addressed in this section concerns the estimation of program costs in
budget preparation. In other words, during the preparation of the budget, it will be necessary to
put a figure on each program’s expenditure for the coming year.28 Each ministry will therefore
need some method to make the estimate of planned expenditure on each program which it
includes in its “budget bid” presented to the MoF.
The simplest and most common approach to estimating program costs in budget preparation is
to take the previous year’s expenditure on the program and adjust it for (1) input cost changes
or demand changes; and (2) policy changes such as an expansion of the program or a change in
It is sometimes suggested, by contrast, that under a program budgeting system – or under a
version of it which some call “output budgeting” – it is essential that the expenditure estimate
for each program be calculated by taking the planned outputs to be produced by the program
and multiplying them by the unit cost of the outputs concerned.29 In other words, unit costs are
supposed to be used in budget preparation to calculate program expenditure requirements as a
function of the quantity of services to be delivered to the public. Under this approach, the
program classification of expenditure would extend beyond programs and sub‐programs to the
28 As well, ideally, as the following three or so years, in a medium‐term budgeting framework – see Section 18. 29 Or, as a variation on that idea, as a calculation based on the number of planned program activities times the cost
of those activities (what is sometimes called “activity‐based budgeting”).
level of the specific individual types of outputs which comprise each sub‐program, and budget
estimation would proceed in a bottom‐up process based on unit costs and planned quantities
of each output.
This is essentially the application to program costing of the idea of formula funding, whether
based on outputs or activities. As discussed in subsequent sections on formula funding and
purchaser‐provider systems, unit costs are a powerful tool when selectively applied to the right
types of public services. The most outstanding international example of a sectoral performance
budgeting system based on unit costs in the “diagnostic related group” (DRG) hospital funding
system. In education, where costs per student at particular levels of schooling tend to be
relatively standard, unit costs are also a useful budgeting and performance management tool.
However, unit costs can never be an across‐the‐board budgeting instrument, because there are
many public services which do not have a stable unit cost. This is because of “heterogeneity”—
average costs vary because of differences in the effort required because of the circumstances of
particular cases and also because of what we in Section 12 as “contingent capacity services”.
Because many government outputs do not have stable unit costs, it is quite inappropriate to
seek to use unit costs as the method for estimating the required budgets or all programs across
government in a program budgeting system.
Moreover, the “output budgeting” version of program budgeting fails to recognize the
enormous complexity of calculating the unit costs each of the large number of types of services
delivered by government.
At its worst, the application of the unit cost approach to program costing can degenerate into a
justification of program expenditure in terms of planned quantities and unit costs of internal
activities such as meetings, internal training seminars, and the production of ministry reports.
Such an approach confuses activities with outputs, and activity‐based budgeting with output‐
based budgeting. A focus on activities (work processes) is far removed from the performance
budgeting emphasis on “outputs” (services delivered to the public) and outcomes (benefits
realized by the community).
Robinson, M. (2007), “Cost Information” in M. Robinson (ed.) Performance Budgeting: Linking
Funding and Results.
Diamond, J. and P. Khemani (2005), Introducing Financial Management Information Systems in
Developing Countries, IMF Working Paper, obtainable at http://blog‐
Khan, A. and M. Pessoa (2009), “Conceptual Design: A Critical Element of a Successful
Government Financial Management Information System Project”, IMF, PFM Technical Manual,
obtainable at http://blog‐pfm.imf.org/pfmblog/fad‐technical‐notes‐and‐manuals‐on‐public‐
Peterson, S. (2006), Automating Public Financial Management in Developing Countries, John F.
Kennedy School of Government, Harvard University, obtainable at
Robinson, M. and D. Last (2009), A Basic Model of Performance‐Based Budgeting, IMF Technical
Manual 1, obtainable at http://blog‐pfm.imf.org/pfmblog/fad‐technical‐notes‐and‐manuals‐on‐
Wynne, A. (2005), Public Financial Management Reforms in Developing Countries: Lessons of
Experience from Ghana, Tanzania and Uganda, African Capacity Building Foundation Working
Paper no. 7, obtainable at http://www.acbf‐
11. Special Topics in Program Budgeting
This section starts with a discussion of the relationship between program structure and
organizational structure, followed by an examination of the closely‐related question of the
distinctive design features of program structures in developing countries. The relationship of
programs to the functional classification of the budget is then considered. The final topic is the
relationship of program structures of different levels of government (in particular, national and
Programs and Organizational Structure
As discussed in previous sections, the program budgeting ideal is that programs should be
results‐based. That is, programs should be based on groups of services delivered to external
parties (“outputs”) which have common outcomes – “product lines” in the shorthand
terminology of the section on program budgeting fundamentals. It is by basing programs on
results in this manner that programs can serve as a useful tool for expenditure prioritization and
increasing pressure upon ministries to perform.
However, organizational structures are not always results‐based. In most government
structures, some organizational units exist which are not defined on the basis of a “product
line”. Differences between program structure and organizational structures can be a problem
because program budgeting will be most effective when there is clear organizational
responsibility for program management – that is, when a single organizational unit can be held
responsible for the results delivered by a program, and exercises the budgetary flexibilities
which program budgeting provides.
The question therefore arises as to whether, to what degree, and in what manner, program
structure and organizational structure can be brought together.
An organizational structure based on “product lines” would mean that all or most
organizational units would be defined in terms of distinctive categories of service delivered to
the public, with all services of that type delivered by the same organizational unit. Actual
government organizational structures differ from this in a variety of ways, including:
Organizational structures which are defined in terms of a specific client group (specific
industries, regions, types of individuals) and not in terms of a specific category of outputs,
and which might therefore provide a quite divergent range of outputs to their target client
group rather than the common product line. For example, agriculture ministries in certain
countries have taken on responsibility for HIV/AIDS prevention activities targeted at farmers,
on the theory that the agriculture ministry’s network of contacts with the farming
community places them in a good position to manage this service.
Organizational structures based on functional lines – for example, based on specific
professional competences and types of activity (e.g. an organizational structure for the
ministry of public works based on separate engineering, planning etc department, as
opposed to a structure based on the different types of infrastructure which the ministry
As a special case of the former category, support service units within ministries or agencies
extending their role to the provision of services to the public, and in doing so duplicate the
role of other ministries or agencies. A classic example of this is the army medical corps – the
original role of which is to provide medical services to the army itself – becoming a
significant provider of health treatment services to the public, in “competition” with the
Support service organizational units: organizational units within ministries which are
defined in terms of the provision of specific types of support services (e.g. human resources
management, IT, finance). These were discussed in the previous section of this manual.
Organizational units which contribute to several quite different product lines are particularly
problematic from a program budgeting point of view. Less difficult to resolve is the alternative
situation where multiple organizational units contribute to a specific product line, without there
being a person in hierarchical authority over all of those organizational units who is able to
provide the integrated management which a program comprising all of those units would
In a broader “managing for results” (MFR) approach to public sector reform, it often makes
sense to undertake restructuring to place organizational structures on more of a “product line”
basis than one tends to see in traditional civil services. For example, it could be appropriate to
phase out the role of the army in health services to the public, with the health ministry taking
full responsibility for this. And if there are two or more organizational units with a single
ministry which are involved in delivering the same “product line”, the units may either be
merged or someone can be appointed who has hierarchical authority over all of the units
concerned. Functional structures should be, in many cases, replaced with product line
structures. Such reorganization can be expected to considerably reinforce the results focus
which program budgeting aims to encourage.
Possible organizational restructuring alone cannot be expected to totally resolve the issue of
conflicts between organizational and program structure, for two reasons.
The first reason is that, even if a more results‐oriented organizational structure is highly
desirable, it may not be feasible or even desirable to extensively modify organizational
structures at the same time as moving to program budgeting. The magnitude of such a
restructuring task may be too great, or it may simply be impossible to obtain adequate support
for organizational restructuring at that time. An explicit decision may under such circumstances
be made to leave organizational restructuring to a later stage in the overall public sector reform
The second reason is that a purely “product line” organizational structure is not desirable.
Under a thorough‐going MFR reform, it may well be that some ministry‐wide support and
coordination service functions are taken over by the service delivery units of the ministry. For
example, MFR argues for greater managerial flexibility over human resources (hiring,
remuneration etc), and insofar as this occurs the power of ministry human resources
directorates (and civil service ministries) will be significantly reduced, with some of their
functions transferred to front‐line organizational units. Nevertheless, in any rational
organizational structure it will always make sense for a core of support services to be provided
centrally by functional units, rather than completely split between the service delivery units of
the ministry. There will always be a ministry finance unit, and usually some type of IT unit, HR
unit etc. Some ministries will also need to maintain regional offices which, for size reasons,
deliver multiple products rather than specializing in single product lines. So even if one wishes
to apply the “product line” organizational principle to the maximum degree to align programs
and structures, one cannot expect to achieve full alignment by this means.
This means that when introducing program budgeting, some conflicts between organizational
structure and a “pure” program structure based entirely on results are inevitable. Some of these
conflicts may be temporary, in the sense that they are capable of future resolution via
organizational restructuring. Some, however, will be permanent.
There are two approaches which can be taken to this conflict between organizational structure
and programs. The first is to define programs around organizational structures, even when this
involves significant departures from the principle of results‐based programs. The second is to
maintain the principle of results‐based programs and to accept that there will therefore be
difference between organizational structure and program structure – some of which may be
temporary, but others of which will be permanent. Management strategies will be put into
place to address the conflict between program and organization structure.
In practice, some mix of these two approaches is not only usual, but to some degree inevitable.
Ministry Boundaries and Internal Support Services
There are two broad areas where there is wide agreement that it is appropriate that there be
some compromise of the program budgeting principle by adapting programs to organizational
The first of these is in respect to ministry organizational boundaries. The great majority of
countries observe the principle that programs should not cross ministry boundaries. It might
appear attractive to say that if two different ministries are contributing to the same “product
line”, then a cross‐ministry program should be established based on that product line. For
example, the idea of a single “AIDS” program grouping together all of the HIV/AIDS prevention
and treatment activities of multiple ministries sounds very appealing. Against this, however, is
the fact that it is essential that each ministry has its own clear budget. It would be completely
unworkable for the budget to simply allocate funds to programs shared between two or more
ministries and then leave those ministries to determine how those funds were to be split
between them. If there were to be multi‐ministry programs, it would therefore be essential for
the budget to clearly indicate how much of the program’s allocation would be directed to each
of the ministries involved. Essentially, this would involve having sub‐programs which were
limited to single ministries, which would essentially take us back to the principle that programs
(or in this case sub‐program) should not cross ministry boundaries.30
30 The allocation of multi‐ministry programs between ministries at the sub‐program level can in principle work. But
it would have the undesirable consequence of forcing the parliament to approve the allocation of the budget at
the relatively detailed sub‐program level rather than, as is the usual international practice, at the broad program
Some countries attempt to bring greater coordination to areas of related service delivery by
designating super groupings of programs. The best example of this is France, where programs
are grouped together into large missions, some of which cross ministerial boundaries. The
principle that programs remain within ministry boundaries nevertheless continues to be
Cross‐Ministry Programs in Africa
In certain African countries which have introduced program classifications in their budgets, there has
been a deliberate decision to introduce programs which are shared by two or more ministries belonging
to the same sector. To make this work, the legal appropriations in the budget specifying the shares each
of the ministries concerned has of the overall budget of the program. The context of this approach is
government structures characterized by a far larger number of ministries than exist in most countries
around the world – for example, as many as 50 ministries. Under these circumstances, it is often the
case that multiple ministries are involved in pursuing the same outcomes. To ensure co‐ordination,
these countries not only employ sector‐based programs, but place much emphasis on sector‐wide policy
coordination. This is understandable, but a first‐best solution to the programs of fragmentation of the
policy effort between too many small ministries would be to radically reduce the number of ministries.
In many cases, the reason for the multiplication of ministries is the desire to create a larger number of
ministerial positions for politicians.
If programs are to be kept within ministry boundaries, one thing which immediately follows is
there will need to be programs which are exclusive to coordinating ministries, and that these
programs will for the most part not be defined in terms of outputs and outcomes delivered to
the public, but in terms of the type of support service provided to the government as a whole.
For example, a civil service ministry or commission might have a “Civil Service” program with an
overarching program objective defined along the lines of “a quality civil service able to provide
excellent policy advice and service delivery”. Clearly this objective is not an outcome, but it is an
important means towards delivering outcomes to the community.
The second area where the majority of countries with program budgeting systems compromise
the pure program budgeting principle is in establishing an administration program within each
ministry, which – as discussed in Section 10 and elsewhere above – groups together all of the
support services and other “overhead” costs of the ministry. They do this in order to avoid the
complexities of the alternative approach, which requires that the costs of all support services be
attributed to the results‐based programs which they support. As discussed in the section on
program accounting and costing, this alternative approach necessary involves quite complex
indirect cost allocation, which for many countries is either beyond their technical capacity or
which they cannot reasonably afford to undertake. Only a minority of OECD countries –
Australia being one example – have taken this latter approach.
Administration programs will in general have program objectives which do not refer to
outcomes. For example, suppose there is a government agency which has a program which
manages the provision of office accommodation to government ministries. The objective of
such a program would be something like “ensuring that the needs of government ministries
agencies for appropriate premises are met in a timely fashion”. In other words, the objective
would refer only to the output being provided. This is because the service involved is not
directly responsible for delivering outcomes.
Broader Conflicts between Program and Organizational Structure
What about conflicts between organizational and program structure which go beyond these
obvious areas for compromise just discussed? Here, the dangers of further compromise of the
program budgeting principle are much greater. If one simply modifies the program structure
wherever it conflicts with organizational structure, one ends up not with a program budget but
an organizational unit budget. The traditional budgeting model in quite a few countries
(including most of the “Latin” countries, South America, non‐Anglophone Africa, and much of
eastern Europe) was one in which budget appropriations were made not only by line‐item
(economic classification) but also to internal ministry organizational units. If the program
principle is to be simply made to fit the organizational structure, the result will be something
just like this type of traditional structure. And insofar as organizational structure is not result‐
based, the fundamental objective of making budgeting more results‐focused will be
If the program budgeting principle is not to be further undermined, the question arises of how
to manage conflicts between organizational and program structure. A common approach to this
problem – which France exemplifies – is to appoint program managers. When the organizational
unit corresponds to a program, the organizational unit and program managers would be the
same person. But where organization and programs diverge, they would be different people.
The program manager’s role would then be to maintain the product‐line focus and offset the
threat to that focus arising from a functional, sector or other organizational principle.
Specifically, the program manager would work with organizational unit managers to maintain
the focus on overarching program objectives and to promote coordination between the various
organizational units involved in delivering the program. If this approach works well, one would
not need to worry about divergences between program and organizational structure.
This approach has considerable attractions. However, it has to be recognized that it also raises
potentially important difficulties – in particular, that the existence of two cross‐cutting layers of
management can lead all too easily either to continual conflict between managers (which
undermines performance) or to the marginalization of one of the two managers (usually the
program manager). Conflict between program and organizational unit managers has apparently
been experienced in France.
France: Role of the Responsable de Programme (program manager)31
“The responsable de programme is designated by the relevant minister to assure the direction of the
program. He collaborates in the formulation of the strategic objectives of the program, under the
authority of the minister. He is responsible for the putting into effect of these objectives and their
realization. To this end, he translates the strategic objective of the program into operational objectives
relevant to each of the organizational units which has a role in the program, by means of dialogue with
the managers of those organizational units.”
Program Structures in Developing Countries
In the light of the above discussion, it makes sense for developing countries to avoid
unnecessary complications in their program structures by:
Avoiding multi‐ministry programs. If it is desired to recognize in the program structure the
need for cross‐ministry collaboration in overlapping areas of responsibility, it will be better
to use the French approach of multi‐program missions than to have shared programs.
Opting for administration programs rather than attempting the onerous task of indirect cost
Keeping their program hierarchies simple – in general, limited to just one or two levels
(programs, or programs and sub‐programs).
In some developing countries, organizational restructuring may be even more important in the
medium term. This is particularly the case in countries where, for example, there are far too
31 Source: France (2004), p. 13.
many ministries (sometimes forty or more). Where this is the case, adhering to the principle
that programs should not cross ministry boundaries does more damage to the program
budgeting principle than in countries with a smaller number of larger ministries. In some
developing countries, duplication of functions between ministries can also be a particularly
Functions and Programs
A question which tends to cause considerable confusion is the relationship between program
structure and the “functional” classification of government expenditure. The standard
international functional classification of the budget is that of the United Nation’s COFOG
(Classification of the Functions of Government) system. The COFOG classification was designed
in order to permit, via a standardized classification of government expenditure, international
comparisons of the allocation of resources between policy areas. Note, however, that a number
of countries have for historical reasons “functional” classifications which differ somewhat from
The COFOG Functional Classification
The COFOG functional classification consists is a hierarchical structure of three levels. The top level
(“divisions”) consists of four broad categories such as “defense”, “public order and safety”, “health” and
“education”. There are ten such divisions. Below this are two lower levels: “groups” and “classes”. Thus
in the education division, the groups include: “pre‐primary and primary education”, “secondary
education”, “tertiary education” and “subsidiary services to education”. And within, say, the “pre‐
primary and primary education” group, there two classes: “pre‐primary” and “primary”. Graphically, this
hierarchical structure can be represented as follows, taken the example of health (Division 7). (The
example shows only some of the groups under the health division, and just one example of the
decomposition of groups into classes).
The key reason why there is confusion about the relationship between functions and programs
is that the term “function” is somewhat misleading and the majority of the functional
categories are really categories of outputs – “product lines”, so to speak. The functional
classification is not a pure results‐based classification. Nevertheless, the functional classification
of expenditure looks so much like a program classification – particularly one with
administration programs – that it is natural to wonder why both are needed and whether the
two structures should be fused in some way. The most common way of linking the program and
functional structures is to insist that programs should all fit within one or other of the ten broad
COFOG divisions – with no programs permitted to cross the divisional boundaries. This is
illustrated in the graphic below.
Program A Program B Program C
The alternative view is that the functional classification should not constrain the program
structure and should not be formally incorporated into it. This point of view is based on the
recognition that the functional classification is only intended to be a statistical tool – for
international comparisons of resource allocation – and not a budget classification in terms of
which the budget should allocate resources.
The potential problem with making the program structure fit the functional structure is that the
choice of programs in any country should to reflect the specific policy challenges of that
country. Thus in a country where the fight against desertification is of crucial importance, it may
make sense to have a desertification program, whereas in many other countries no such
program will be required. To restrict the program structure to make it fit with a given functional
classification reducing flexibility in the definition of programs. This problem is not really a
significant one if all that is proposed is to keep programs without the boundaries of the COFOG
divisions (the top level of the COFOG hierarchy), although even there it is not clear why it is
appropriate to ban programs which cross divisional boundaries. However, it is a much greater
problem if the program structure is to be made to conform to a more detailed functional
categorization, as illustrated by the following case study.
Programs and Functions: a Case Study
One African country which decided to integrate its functional and program classification
encountered major problems as a result of the manner in which it chose to do so. The country
concerned had in place a pre‐COFOG functional classification, which like COFOG was a three
level hierarchy, as follows:
In this country’s structure, “sectors” (of which there are 10) were similar to COFOG divisions,
“primary functions” to COFOG groups; and “secondary functions” to COFOG classes. The fact
that the functional classification was somewhat different from COFOG was not, however, the
problem. The problem was that it was decided to make programs subordinated to the second
level of the functional classification, rather than the top level. This produced the following
integrated functional and program structure:
Program A Program B Program C
The effect of this was to require that no program should cross the boundaries the “primary”
(second level) functions.
The problem with this was that that the primary functions, of which there were approximately
seventy, were far too detailed for this purpose – and much more detailed than, say, the ten
divisions of the top level of the COFOG classification. Bear in mind that in most countries, the
total number of programs across government is of the order of 150‐200. Insisting that no
program should cross the boundaries of the primary functions proved to be extremely
constraining, and made it impossible to create certain programs which appeared important to
the government. Following advice, it changed its position, and chose to define the top
“function” level of the program structure in terms of the ten COFOG divisions.
Programs and Levels of Government
One question which arises in defining program structures is the relationship between the
program structures of different levels of government in a country. Should the national and
subnational governments adopt the same program structure? Or should the different levels of
government go their own ways in developing their own structures without any regard to what
other levels of government are doing?
This is an area where different countries take different approaches. In many federal countries,
where subnational government is constitutionally autonomous and is protective of this
autonomy, each level of government develops its own structure without paying attention to
what other levels of government are doing (this is true, for example, in Australia and Canada).
In many other countries, however, there is a belief that program structures should be in some
way standardized or harmonized. This is particularly true in developing countries where there
are single national plans which are intended to guide the entire public sector.
Even if one aims for the maximum degree of harmonization, one cannot have exactly the same
programs at national and subnational levels. Insofar as the levels of government have different
functions, they require different programs. Even in the same service area, their functions may
differ, and it might be thought best to reflect this in program titles. For example, it may be that
in school education, the national government is responsible for the development of national
standards (e.g. a common core curriculum), while subnational government actually delivers the
service. This may lead to the national program being titled, say, “education policy”, while the
subnational programs are simply labeled “education”.
Robinson, M and H. Van Eden (2007), “Program Classification”, in M. Robinson (ed.)
Performance Budgeting: Linking Funding and Results.
Diamond, J (2003) From Program to Performance Budgeting: The Challenge for Emerging
Market Economies, IMF Working Paper, obtainable at
IMF (2001), Government Financial Statistics Manual 2001, Washington: IMF (for detailed
information on the COFOG functional classification), obtainable at
12. Formula Funding and Purchaser Provider Systems
Formula funding – of which the purchaser‐provider mechanism is one type ‐‐ represents an
attempt to create a much tighter linkage between funding and results than characterizes
program budgeting. Under program budgeting, performance information is taken into account
in determining funding, but there is no automatic linkage. Under formula funding, the linkage
between funding and results is a very direct one.
When used as a form of performance budgeting, formula funding means that the funding
provided to a government ministry or agency is determined in whole or part as on the basis of
an explicit algebraic formula based on outputs or outcomes (“results” for short). The formula
might link funding to expected (i.e. future) results – as, for example, when school funding is
based on the number of students who enroll in a school at the beginning of the year. Or it could
be based on the actual results which the agency has delivered – which would be the case if
school funding was based on the number of students who actually attend the school during the
year (i.e. if a student enrolls at the beginning of the year, but then drops out, no funding would
be received). The latter type of formula funding, in which funding depends on results actually
delivered, is what is referred to as a purchaser‐provider system.
Many formula funding systems are cost‐based. For example, if schools are being largely funded
by formula based on the number of their students, then it is obviously necessary that the
amount of funding per student is based on an estimate of what it should cost to teach one
student. However, formulas are also used to provide performance bonus funding which are not
based (and could not conceivably be based) on cost estimates. For example, in some public
university systems, core funding based on per‐student amounts is accompanied by
supplementary bonuses linked to outcome measures such as percentages of graduates who
find employment. Nobody knows what the cost is of increasing the graduate employment rate
by, say, 5 percent, so the level of the bonus which is paid is in a sense arbitrary. This does not,
however, undermine its function are providing an incentive to improve quality and achieve
The purchaser‐provider mechanism has been used with considerable success in certain sectors
of government, most notably hospital services where is took the form of so‐called “diagnostic‐
related group” based funding.32 Broadly, the idea is that hospitals are paid by the government
for the outputs they deliver, and only for the outputs they deliver. Each different treatment
(e.g. treatment of a hip fracture patient) has a standard price. If it costs the hospital more to
provide that treatment than it is paid, it makes a loss. If, on the other hand, it succeeds in
32 For a detailed description of this system, see Robinson and Brumby (2005), section V.
delivering the service for less than the price paid, the hospital makes a profit. In this way, a
powerful incentive in introduced for the enhancement of hospital efficiency (cost‐
containment). When this system was originally introduced (in the USA), there were enormous
concerns that it would lead to serious adverse consequences such as the erosion of the quality
of services and the refusal to treat higher‐cost patients. In practice, these fears turned out to be
greatly exaggerated, in part because of other complementary quality‐control systems and in
part because of the considerable professional pressures for the maintenance of patient service.
Overall, experience has demonstrated the DRG funding system to be a great success in keeping
costs down without unacceptable adverse consequences. For this reason, an increasing
number of countries have adopted it as the funding model for their public hospitals.
The purchaser‐provider system is not, however, something which is suitable for all or most
public services. This was demonstrated clearly when Australia and New Zealand attempted
during the 1990s and early 2000s to reform their entire budgeting systems to place them on a
purchaser‐provider basis, under what was known as “accrual output budgeting”. The
experiment failed badly.33
Limits to Formula Funding and Purchaser Provider
Formula funding and purchaser‐provider arrangements require a stable relationship between
output quantity and cost – expressed differently, a stable output unit cost. These types of
funding arrangements therefore suit what have sometime been called “production”‐type
agencies: that is, agencies which produce large volumes of a limited number of fairly standard
services. Motor license testing is a good example. The output delivered by the agency which
issues motor licenses to drivers is standard in the sense that every applicant for a license has to
undertake the same test. The cost per license issued is therefore stable, and there would be no
difficulty in principle in basing funding on numbers of licenses issued.
There are, however, many services produced by government which are not at all like this, and
which are unsuited to funding on this type of basis. Heterogeneous and contingent capacity
services require particular mention here.
As discussed earlier in this manual, output heterogeneity arises when the activities required in
the production of one unit of an output may differ significantly from those involved in the
production of another unit of the same type of output because of differences in case or client
33 See “Purchaser Provider Systems” in Robinson, Performance Budgeting, for details on the failure of these
characteristics. Output heterogeneity means that the cost of one unit of an output will differ
from the cost of another unit of the same output for reasons which have nothing to do with
efficiency. Other than in the quite special circumstance where the cost effects of heterogeneity
average out over large volumes of production of the output concerned34, significant output
heterogeneity can undermine the foundations of an output‐driven funding model. When a
service is characterized by serious heterogeneity, its unit cost will be unpredictably variable,
and it will therefore not be possible to base funding on unit cost. For example, the costs of
treating accident victims in hospital emergency centers are too variable to permit them to be
included within the DRG funding systems referred to above.
The fire brigade is a good example of a contingent capacity service. The output delivered by the
brigade is putting out fires. But it would be impossible to fund the fire brigade on the basis of
the numbers of fires it puts out. The fire brigade is like an insurance policy – government is not
so much funding the output as the readiness of the brigade to provide the output quickly when
and if it is needed. Deliberate spare capacity is built into such a service.
One of the reasons why attempts to apply the purchaser‐provider system to the whole of
government failed was that so many services provided by government ministries were unsuited
to this method of funding. For these reasons, formula funding and purchaser‐provider systems
should be seen not as the basis for government‐wide performance budgeting models, but
rather as systems which can be selectively applied to selected government services – those
which are relatively standardized, and produced in large volumes.
Robinson, M. (2007), “Purchaser‐Provider Systems”, in Robinson (ed) Performance Budgeting.
Smith, P. (2007), “Formula Funding and Performance Budgeting”, in Robinson (ed) Performance
34 Which is the case for many types of health treatments.
13. Targets and Performance Budgeting
Performance targets are an important managing‐for‐results tool. When targets are
systematically linked in some way to budget funding, we can talk about a specific model of
performance budgeting. The UK Public Service Agreement (PSA) system, in the form in which it
operated between 1998 and 2007 under the Labour Government, is the most important
example of this model of performance. Under the PSA system, several hundred high‐level
targets were set every three years as part of a “spending review” process in which multi‐year
ministry budgets are set. The spending review in fact constituted the core of the budget process
– it was the context in which most ministry funding was decided. So the PSA system was one in
which budgeting and target‐setting were part and parcel of the same process.
Before considering the nature of the relationship between funding and targets, it is useful first
to consider performance targets and some of the key issues which they raise.
A performance target is a quantitative goal with a timeline (usually explicit, but sometimes
implicit) for the achievement of that goal. Targets may be set for outcomes, output (quantity,
quality or efficiency), or even for activities or inputs. The distinction between indicators and
targets is an important one. Targets are always based on specific performance indicators, which
provide the yardstick for measuring target accomplishment. But indicators do not include
quantitative objectives, nor timelines. Thus the percentage of HIV/AIDS infected persons in the
population is a performance indicator. Cutting the rate of HIV/AIDS infection by 10 % over 5
years is, by contrast, a performance target.
It is also important not to confuse targets and objectives. An objective states the type of
improvement sought without necessarily indicating the measure to be used, quantitative
objectives or timelines. “Drastically reducing HIV/AIDS infection rates” is, for example, an
objective rather than a target.
Performance Target: Some Examples
Increase adult literacy from 60% to 70% by 2015 (an outcome target),
Vaccinate the whole population against polio by 2012 (an output target),
Ensure that all monthly accounting reports are completed within 15 days of the end of the financial
year (an activity target, with implicit timeline of “immediately/this year”),
Fill all vacant agricultural extension officer positions with suitably qualified persons during this
financial year (an input target).
There is a widespread assumption that performance targets should be set for all programs and
perhaps even for all performance indicators. However, it can be argued that selectiveness and
gradualism in the setting of performance targets is preferable.
In the first place, target‐setting requires a firm basis of good, timely and verified performance
indicators. In leading OECD countries such as the UK, good performance indicator systems took
decades to develop. Most countries – and in particular most low and middle income countries –
do not have such systems. From this perspective, it seems to make sense in developing
countries to focus initially on building a solid base of core performance indicators, and only at a
later stage engage in extensive target‐setting.
Secondly, setting appropriate and credible performance targets is quite difficult. Selecting
which measures to turn into targets is not necessarily a straightforward matter, and the
challenge of setting the quantitative targets which are neither too difficult nor too easy is quite
Thirdly, there are many who take the view that targets should only be set for performance
indicators which are relatively controllable. The concern underlying this view is that setting
targets for highly uncontrollable variables is more likely to de‐motivate than to motivate
agencies and their staff. If one takes this view, the most obvious implication is that targets
should generally not be set for outcomes, and particularly not for high‐level outcomes, because
these tend to be greatly influenced by uncontrollable “external factors”. From this perspective,
setting targets for outputs will generally make more sense.35 At the same time, however,
outcomes are what matters most, and high‐level outcomes are the outcomes which matter
most of all. So many governments do in fact routinely set targets for outcomes over which they
have quite limited control: e.g. rates of economic growth and unemployment, or rates of
growth of tourist numbers.
A fourth and final point is that there is no point setting more targets that government can
monitor and act on. In many countries, there is little follow‐up of agency performance against
target and, as a consequence, the targets seem not to have been taken very seriously. The UK
experience under the PSA system was a major exception to this. There the government was
very serious in monitoring and acting on performance relative to the PSA targets. Special
attention was given to monitoring and following up on – including by means of managerial
intervention when necessary – performance on the fifty most important PSA targets, via a
“service delivery unit” reporting directly to the Prime Minister. The British experience also
underlines that, if targets are to be taken seriously, it is crucial that they are target for
35 With the qualification that targets may not be appropriate for outputs which are highly affected by
indicators which really matter to the public and politicians and not, for example, targets relating
to purely internal activities within ministries.
Because it makes little sense to set performance targets if one does not follow‐up on
performance against them, the more limited the capacity of central government to monitor and
follow up on performance, the more selective central target‐setting should be.
Gaming and Perverse Effects
The wisdom of a target‐setting approach to improving public sector performance in contested
by a school of opinion which regards targets as more likely to damage than improve
performance. This school of opinion focuses on the danger that, to the extent that agencies and
individuals within them are motivated to achieve targets, the targets may be met without a real
improvement in performance because of:
“Gaming” (manipulation or falsification of performance indicators, in this case of the
indicators upon which targets are based) – for example, when a hospital falsifies its waiting
time data in order to appear to have met a timeliness target which it has in fact failed to
“Perverse effects” (deteriorating performance) arising from the use of targets based upon
imperfect performance measures.
The problem of gaming requires careful verification and auditing of indicators, and sanctions for
falsification. The problem of perverse effects, on the other hand, should in significant measure
be minimized by careful design of, and the use of the right combination of, target measures.
However, all performance measures are imperfect to a greater or lesser degree, so that even
the most careful design of the set of targets cannot eliminate the potential for behavioral
Concerns about perverse effects are based in significant measure on theory. The most familiar
theoretical point is there are some key dimensions of performance – such as quality – which are
notoriously hard to measure, and tend therefore not to be captured in targets. The fear then is
that what is not measured will be sacrificed to what is measure. In addition, theory points to
the possibility that agents might pursue the easiest means of fulfilling their targets, with
undesirable consequences. Theory is all very well, but empirical evidence on these questions is
arguably more important. It is striking here that the empirical evidence of serious perverse
effects under the UK PSA system – where the pressure to achieve targets was very considerable
– is very limited and the overall impression is that the critics have greatly exaggerated the
Judging on the basis of reported performance against targets, the UK PSA appears to have
worked very well as an instrument for improving public sector performance, although the
research on the subject has not been as in‐depth as one might have wished for. The fact that
targets were set in the budget process gave them real weight, as did the fact that performance
against target was followed up by the Prime Minister and the finance minister, and was an
important focus on the budget decision‐making process. At the same time, it is clear that the
connection between funding and targets was a very loose one. This is particularly the case
because so many of the PSA targets were outcome targets, including some quite high‐level
outcome targets at that (e.g. targets for reductions in cancer rates and childhood obesity, over
which the government clearly has quite limited control).
Audit Commission (UK) (2005), Target Setting – A Practical Guide, obtainable at
Smith, P. (2007), “Performance Budgeting in England: the Public Service Agreements”, in
Robinson (ed) Performance Budgeting.
Social Market Foundation (2005), To the Point: a Blueprint for Good Targets, London: SMF,
obtainable at http://www.smf.co.uk/to‐the‐point‐a‐blueprint‐for‐good‐targets.html.
House of Commons Select Committee of Public Administration (2003), On Target? Government
by Measurement, obtainable at http://www.bercy.gouv.fr/lolf/downloads/1400_target.pdf.
Gay, O. (2005) Public Service Agreements, House of Commons Library, obtainable at
36 A paper by Kelman and Friedman (2007) on the responses to a 4 hour target for treating patients in UK hospital
accident and emergency rooms provides an antidote to fears about targets. Kelman and Friedman conclude that
“that waiting‐time performance improvement was dramatic and that dysfunctional responses, as far as we can tell,
entirely absent.” They add that “none of the hypotheses predicting effort substitution or gaming in connection
with attaining this target has been confirmed”, and that in fact dimensions of performance not captured in the
targets appear to have improved. Bevan and Hood (2006) and Hood (2006) have undertaken important research
which provides some limited examples of perverse effects, but which at the same time suggest that targets have
worked well in raising performance (in Hood’s words, the evidence “strongly suggests that targets made a marked
difference in reported performance”).
14. Public Financial Management Reform Foundations
Performance budgeting must be seen in the context of broader reform of public financial
management (PFM) systems. Prior to any consideration of the adoption of performance
budgeting is it, in particular, essential that the PFM system is able to deliver on two basic
requirements. The first is that it is capable of supporting sound macro‐fiscal outcomes. The
second is that it broadly assures probity – that is, that money is spent only on public purposes
duly authorized, and not widely diverted to private purposes by corrupt politicians and civil
servants. If the PFM system and broader governance framework do not deliver these basic
requirements, the initial focus should be on reforms in these areas, and performance budgeting
should be deferred to the future.
In respect to sound macro‐fiscal outcomes, fiscal sustainability is particularly important,
because if budgetary policies are not sustainable it is likely that there will sooner or later be a
major crisis in which public expenditure will have to be cut severely. Under the unstable
circumstances which then arise, efficient and effectiveness management of expenditure
inevitably suffers. It is only through the pursuit of sustainable fiscal policies that a sufficiently
stable environment can be created in which public managers can successfully focus on results.
Although the principle causes of unsustainable fiscal policies are usually political, weaknesses in
the PFM system can contribute significantly to the problem. Areas of PFM reform which can
help greatly in promoting fiscal sustainability include:
Expanding budget coverage so as to ensure that the budget is comprehensive – i.e. that it
covers all expenditure (and revenue) which impact on deficits and debt.
Ensuring in budget preparation that an aggregate expenditure limit compatible with fiscal
sustainability is established and then respected.37
Ensuring in budget execution that ministers and ministries respect the expenditure limits
imposed upon them, and do not either overspend or accumulate payment arrears which
will lead to future overspending.
37 There has in this context been great interest in recent times in the development of more “top‐down” budget
processes. The reason for this is concern about the tendency of completely “bottom‐up” budget processes – that
is, processes in which spending ministries make unconstrained bids for extra resources during each annual budget
process – to undermine respect for aggregate expenditure limits. In its extreme form, top‐down budgeting
counters this via unilateral central determination of the budgetary spending authorization of each spending
ministry. In practice, a completely top‐down process is neither feasible (other than in the very short term) nor
desirable. However, fiscal sustainability may be strengthened by making the budget process more top‐down in
important ways, including – when medium‐term budgeting is well‐developed – the entirely top‐down setting of
envelopes for existing programs (as distinct from new expenditure initiatives).
Assuring probity is, of course, a matter of degree. No country is entirely free of corruption in
the use of public monies. The question is therefore whether the system is capable of ensuring
an acceptable minimum level of probity. Particularly important here is the quality of the
parliamentary and other controls which are limit executive discretion in choosing the purposes
for which public monies may be spent – that is, which are designed to ensure that monies can
only be authorized for public purposes. The legal framework of budget approval is crucial here.
The other aspect is safeguards during budget execution to ensure that money is spent in the
manner intended. This requires sound expenditure control processes, as well as good internal
and external audit.
For both sustainability and probity, a strong system of commitment control is important.
Commitment is the stage where an obligation to make a future payment is entered into, such as
when a contract to purchase a good or service is signed with a supplier. Once a commitment is
legally entered into, it becomes essentially impossible to avoid the subsequent payment. A
weakness in the PFM systems of many countries is the absence of sufficiently strong discipline
at the commitment stage, with the result that commitments are entered into which are either
or both excessive and for purposes inconsistent with the budget authorization. Sound
commitment control should therefore be regarded as one of the foundational elements of good
PFM. However, approaches to commitment control vary greatly. In countries where civil service
discipline in weakest, the MoF will need to directly control commitments itself – that is, to
require that no commitments be entered into without its approval. In more advanced systems,
commitment control is decentralized to ministries and managers within those ministries,
backed by sanctions which are applied if excessive or inappropriate commitments are entered
It is therefore important, before deciding to adopt a performance budgeting system, to take
stock of the overall PFM framework and to determine whether there are weaknesses in these
fundamental areas which should be addressed first. Particularly useful in this context is the
PEFA framework – a standardized international diagnostic framework for public financial
There are a number of areas of PFM and related systems which are co‐requisites (rather than
prerequisites) of performance budgeting. This means that while it is not essential that reforms
in these areas are in place prior to the start of the performance budgeting implementation
process, these reforms must accompany performance budgeting if it is to succeed. The
importance of such accompanying reforms is demonstrated by the unfortunate examples of
many countries which have implemented key elements of performance budgeting (in particular,
program classification of the budget and the development of program performance
38 See the PEFA website (www.pefa.org) for more details.
information) but found that this in itself did not change budget outcomes: in other words,
expenditure prioritization did not improve and there was little discernable impact on efficiency
Other sections of this manual discuss two critical co‐requisites: namely, improved expenditure
prioritization processes in budget preparation, and civil service performance management
reform to give civil servants the incentives and freedom to focus more on results.
The other area of PFM reform which is critical is the reduction of expenditure inflexibilities.
Expenditure inflexibilities are obstacles to the reallocation of resources. Because government‐
wide performance budgeting is critically concerned with the reallocation of resources from low‐
priority and ineffective programs to high‐priority ones, it will never succeed if expenditure
inflexibilities make such reallocation impossible or very slow and difficult. Amongst the
inflexibilities which can be of particular concern are:
Extra‐budgetary fund arrangements which bind certain revenues to be used only in
designated program areas.
Civil service employment inflexibilities which, in some countries, make it impossible to shift
or terminate the employment of civil servants who work on programs which the
government wishes to eliminate or cut back.
Creating more flexible civil service employment conditions is a task which must be approached
in different ways in different countries, depending upon their governance arrangements and
traditions. In certain OECD countries, traditional civil service guarantees of job security have
been largely abolished and the civil service made to operate under general labor law. Such a
radical approach may not, however, be appropriate for countries where there is great concern
about the scope for abuse of civil service employment flexibility for political or nepotistic
reasons. However, even if much more limited reform is adopted, it is essential at least to make
it possible to re‐assign workers and to make use of strategies such as voluntary departure
packages to downsize when necessary.
Diamond, J. (2007), “Challenges to Implementation” in Robinson (ed.), Performance Budgeting.
15. Expenditure Prioritization and Performance Budgeting
Improved expenditure prioritization and increased performance pressure on ministries and
agencies are the two channels by which government‐wide performance budgeting aims to
improve public sector performance. However, as noted previously, merely producing
information on the benefits and costs of programs does not ensure that this information will be
used to improve prioritization and hold ministries to account for performance. There need also
to be formal routines for the reconsideration of spending priorities integrated into the budget
process, and these routines need to be designed so as to make maximum use of available
information on program performance.
The key point of contact between performance budgeting and expenditure prioritization
processes during budget preparation is spending review. Spending review refers to the
systematic scrutiny of existing expenditure to identify, in particular, options for cuts. Spending
review draws on evaluation (see Section 6). More specifically, it draws on both program
evaluations (the review of specific services provided by government) and efficiency reviews
(which focus on reducing the cost of delivering services). However, spending review also goes
beyond evaluation to include systematic priority analysis – in other words, the systematic
identification of programs or elements of programs which could be cut because they are low
priority. This is a completely different matter from the evaluation of ineffectiveness or
inefficiency. A program might be highly effective and efficient, but still be very low priority
because the outcomes which it aims to achieve are not very important to the community, or
are not rated as such by the government of the day.
Without spending review, the risk is that programs which are ineffective, low‐priority or which
have outlived their usefulness will continue to command public resources. It is in the process of
spending review that performance indicators and evaluations can be systematically employed,
in conjunction with the measure of program costs which a program budgeting system
Spending review is an area where the budgeting systems of many countries are weak. In such
countries, the budget process is overwhelmingly about new spending, and ongoing expenditure
is not generally speaking seriously scrutinized. “Incrementalism” is a term coined by budgeting
writers in the 1970s to describe this tendency of budgeting to take spending on existing
programs for granted.
Spending review is critical to good aggregate fiscal outcomes and to the capacity of the
government to respond to new spending needs. If substantial room is to be created for
important new spending initiatives, it will almost always be necessary to cut existing spending.
This is important also for aggregate fiscal discipline, because if such cuts are not identified, the
danger is that new spending will simply be added on to the budget, pushing up aggregate
spending at a higher rate than is consistent with keeping the budget deficit at sustainable levels.
Good spending review also puts increased pressure to perform on spending ministries, because
it greatly increases the probability that poorly‐performing programs or areas of inefficiency will
be identified by the center and result in either budget cuts or sanctions being applied to
ministry management. Ministries which wish to protect their budgets will as a result be
motivated to lift their performance. In this context, spending review should also be linked with
processes for management improvement and program re‐design. This is because, if a program is
identified as ineffective, it will not necessarily follow that its funding should be cut: a change in
program design or management may be more appropriate.
Spending review should be integrated with the budget process. In most countries, this will
mean that some spending review is undertaken every year as part of the annual budget
process. In countries (such as the UK) where fixed medium‐term expenditure ceilings are set for
spending ministries, spending review is a process which is carried out only every three or four
years (see Section 18). In most of the following discussion, it is assumed that spending review is
an annual process.
Advocates of zero‐base budgeting favor the comprehensive review of all expenditure every
year, or at least at regular intervals. Experience shows, however, that this is not practicable –
the analytic task is simply too great. Spending review needs instead to be selective and strategic
in its focus. The question then arises as to what approach should be taken to selecting
programs for review. One approach is the discretionary targeting approach, in which spending
review is focused on programs which officials and ministers believe to be most likely to yield
savings. The alternative is a regular program review cycle in which all programs are reviewed
over a regular multi‐year cycle of, say, five years. Canada is one example of a country with such
a system (see Box), and the U.S. Program Assessment Rating Tool until recently involved a
similar five‐year cycle in which 20 percent of US federal programs were subject to review each
Spending Review Cycle: The Canadian Example39
The Government of Canada introduced a new expenditure management system in 2007 as part of an
ongoing commitment to better manage government spending. This system aims to ensure value for
money for all government spending. A key pillar of this system is the ongoing assessment of all direct
program spending, or strategic reviews. Through the Strategic Review process, the Government
systematically assesses the relevance and performance of every program on a cyclical basis. The
Strategic Review process requires government organizations to review 100 percent of their direct
program spending and the operating costs of their major statutory programs on a cyclical basis. From
this review, the organization identifies five percent of spending for reallocation from its lower
performing, lower priority programs. This system is combined with a system of targeted strategic
reviews based on the selection by ministers and the Treasury Board Secretariat of selected programs for
The scope of expenditure covered by the spending review should be the whole of general
government, and should certainly not be limited to “discretionary” expenditure. It is, in
particular, crucially important to include social entitlements and other mandatory spending.
Spending review involves three levels of activity. At the lowest level are reviews of individual
program, efficiency and other topics. The next level is the preparation, on the basis of these
reviews and other analyses, of coordinated advice to the political leadership identifying the
best options for expenditure cuts. The final level is decision‐making about which cuts to make,
which is the role of the political leadership.
Processes for the political leadership to make prioritization decisions should therefore be well‐
integrated into the budget process. The nature of the most appropriate mechanisms for political
decisions about priorities will vary greatly between countries. In some countries, a process
involving a Cabinet committee may be the best approach. In other countries, the political
system may dictate a more centralized approach, focused on the president and/or prime
minister in conjunction with the finance minister. The role of the parliament is discussed in
Section 17, reflecting the fact that there is great variation internationally in the extent of the
power which parliament exercises over the allocation of resources in the budget.
39 Source: Treasury Board Secretariat, http://www.tbs‐sct.gc.ca/sr‐es/faq‐eng.asp#q1
Cabinet Committees and Expenditure Reallocation: The Australian Example
The use of a cabinet committee as the instrument for key decisions about expenditure reallocations has
been a distinctive feature of the budget process of the Australian national government since the 1980s.
Throughout this entire period, there has been a cabinet “Expenditure Review Committee” (ERC) charged
with examining major new spending proposals put forward by spending ministries, and also with
discussing major cuts which could be made to finance new spending or reduce the deficit. The ERC is
comprised of the Finance Minister and the Treasurer (who share responsibility for the budget in the
Australian system), together with the Prime Minister and one or two powerful spending ministers. A key
function of the ERC is to strengthen the hand of the guardians of spending discipline within the Cabinet,
thus countering the tendency of spending ministers to form coalitions to push spending up. The analytic
support of finance ministry staff has been crucial to the success of this mechanism. MoF officials provide
ERC members with so‐called “Green briefs” for each spending ministry, which not only critically evaluate
the ministry’s new spending proposals, but identify savings options and alternative strategies. They are
also present at meetings of the ERC to be able to advise ministers on the spot, and to act as a
counterweight to spending ministries. The Labor Government elected in 2007 modified the ERC process
to some extent by instituting a Strategic Budget Committee (SBC), comprised of the Prime Minister, the
Deputy Prime Minister, the Treasurer and the Minister for Finance, which met at the very beginning of
the budget process and assumed the function of overviewing spending ministers’ new spending
proposals. The ERC was, however, retained, and pursues its work in the light of the parameters
established by the SBC. The Government also reverted in 2007 to practice previously employed in times
of significant fiscal consolidation – namely, of holding meetings of the ERC throughout the year, rather
than only during specific stages of the budget process.
The provision of advice to the political leadership on options for spending cuts, arising from
spending review, should – depending on the institutional structure of the country – be the
function of the MoF, potentially in collaboration with the president’s or prime minister’s office.
This advice needs to be based on policy analysis as well as financial considerations. Only civil
servants can carry out this function on an ongoing basis—external advisers do not, in general,
have a sufficiently in‐depth knowledge of government, and are not able to provide the
continuity required. Nevertheless, external advice has its place, and in carrying out the
spending review function, the MoF may draw in part on program and efficiency reviews carried
out by external consultants or commissions.
Spending review should also in part be carried out by the spending ministries themselves,
which should be encouraged to develop their own internal program evaluation. But spending
review cannot be left to the spending ministries alone. The MoF must lead it, and in doing so
should not allow itself to become dependent on program evaluations carried out under the
exclusive control of spending ministries.
Expenditure Prioritization Processes More Generally
As noted above, spending review is the crucial point where performance budgeting interacts
with expenditure prioritization processes. To work well, however, the expenditure prioritization
process requires other key elements.
Good processes for determining priorities for new spending are crucial. Two elements are
particularly important here. The first is the existence of an effective strategic phase in the
budget preparation process – that is, of a stage early in the budget preparation process where
the political leadership sets the priorities which will guide the process. A budget strategy paper,
presented to cabinet of ministers for consideration at the start of the budget preparation
process, can be one good vehicle to facilitate the determination of government priorities during
the strategic phase.
The second key element of importance in relation to the prioritization of new spending is
rigorous processes for central review of these proposals. Particularly useful in this context can
Strict requirements about the manner in which proposed new spending initiatives are
presented: what type of information is to be provided about the proposal (e.g. a clear
statement of the objective), a requirement for medium‐term estimates of the cost of the
initiative, minimum amounts of notice to be given, a requirement of circulation of the
proposal to a specified list of central agencies who will analyze and comment upon the
A requirement that the cost estimates for each new initiative be agreed by the MoF, so as
to prevent spending ministries deliberately underestimating the costs of their proposals.
Proposals for new spending should to the maximum degree be integrated with the budget
process. In some countries, a large portion of proposals for new spending by spending
ministries are put forward and decided by the government outside the budget preparation
process. As a consequence, during budget preparation, the focus is not so much on making
decisions about new spending options as on trying to fit within the budgetary resource limits
the spending required by new laws passed prior to the budget. An obvious danger under such
circumstances in that the new spending initiatives adopted outside the budget process will cost
more than the government can afford. The other adverse consequence is weaker expenditure
prioritization. This is because, with new spending proposals being put forward at any time
during the year, they tend to be considered in isolation rather than being compared with other
possible new spending options in the way that would happen if new spending proposals were in
general considered jointly during the budget preparation process.
Countries are sometimes advised to set firm ministry (or sectoral) budget ceilings right at the
start of the budget preparation process, in an entirely “top down” process prior to any budget
bids or other bottom‐up input from spending ministries. This advice is based on the proposition
that a completely “bottom up” budget process, in which ministries are able to bid for whatever
they like with no indication of resource constraints, gives ministries no incentives to identify
savings, and inevitably results in bids so far in excess of available resources that they are
impossibly difficult for the MoF to handle.40
The concern about purely bottom‐up budget processes is a valid one. There is, however, a real
danger that, unless the ceilings are formulated very carefully, setting early and firm ministry
ceilings will aggravate allocative rigidity, greatly limiting the possibilities for improved
expenditure prioritization. Neither ministers nor central agencies such as the MoF have the
detailed knowledge of the program areas to unilaterally determine where new spending should
be focused. Moreover, the amounts of new money which ministries or sector should be given
for new spending can only properly be decided after the examination of concrete program
proposals. It makes no sense to say something like “health is a top priority sector, so we’ll
increase its funding by 15 percent without considering what concretely the health ministry will
spend the extra money on”.
Few OECD countries in fact set ministry ceilings before considering bottom‐up proposals for
new policy. Countries like Canada and Australia manage to avoid the dangers of a purely
bottom‐up budget process while nevertheless retaining considerable scope for bottom‐up
ministry proposals for new spending initiatives. They do this by requiring spending ministries to
submit major new policy proposals separately from their core budget proposals, with the latter
based on the continuation of pre‐existing policies and programs. They then subject the new
spending proposals to very rigorous and systematic central scrutiny. Chile did the same thing
some years ago with its well‐known “Bidding Fund”.
Sweden is sometimes seen as a model for purely top‐down ministry ceiling‐setting. In theory
the ministry ceilings are all set at a cabinet retreat one month into the budget preparation
process, prior to any bottom‐up input. However, in reality, discussions on new policy initiatives
which impact on the ceilings continue for months afterwards.
40 See Schiavo‐Campo & Tommasi (1999) and Potter & Diamond (1999).
Good prioritization requires the right blend of top‐down and bottom‐up elements in the budget
preparation process. It certainly will not be facilitated by an entirely bottom‐up process.
Equally, however, a totally top‐down process is neither desirable nor workable, other than
temporarily under extreme circumstances (e.g. an acute fiscal crisis requiring quick and drastic
expenditure cuts). Defining precisely the right balance between top‐down and bottom‐up
processes during budget preparation arguably depends very much on country circumstances.
Planning and Prioritization
The prioritization of expenditure is, of course, is the key function of planning. Planning systems
are, nevertheless, often weak at prioritization. This is because they are often better at
identifying purposes for which the government should spend more, than at identifying where
spending could be cut back to make room for these new spending priorities.
Moreover, where the planning process is institutionally separated from the budget process,
budget decision‐makers may not take the priorities identified in the plan seriously. The problem
tends to be particularly serious in countries which think of planning and budgeting as sequential
processes. That is, the plan is prepared first and is where all the priorities are formulated. The
budget is then supposed to merely give financial expression to the plan’s priorities. Often, this
leads to the plan being formulated out of the context of budgetary constraints, with the result
that the plan doesn’t really prioritize in the sense of making the hard resource allocation
There are two key lessons to be learnt from this: firstly, that planning and budgeting must be
fully integrated. Expressed differently, planning should take place as an integral part of the
budget process, rather than being something which is undertaken only prior to the preparation
of the budget. Secondly, good planning has to be as much about identifying cuts as about
selecting areas for new spending – that is, it has to be designed to prioritize expenditure, not
just to dream up “blue skies” ideas for new spending.
Expenditure Prioritization at the Sector and Ministry Levels
One influential approach to the prioritization of expenditure has been the so‐called “sector‐
wide” approach. In this approach, much of the task of expenditure prioritization is delegated to
sector groups of ministries. The centre of government would initially determine ceilings for
sectors such as, say, agriculture, and then all of the ministries with a role in that sector would
get together to decide the optimal allocation of that ceiling between sector programs (after a
discussion of sector objectives and priorities).41 This system appears to have been inspired by
the Canadian Policy and Expenditure Management system (PEMS) which attracted considerable
favorable attention in the early 1980s.42
This system has certain advantages. However, Canadian experience suggests that multi‐ministry
committees are not necessarily a good forum for spending review. Moreover, it can also be
argued that the sector‐wide approach delegates too much of the prioritization task. The center
is supposed to decide the sectoral allocation, but it is hard to decide how much money sectors
should get except in the context of specific major measures or savings which are proposed.
The sector‐wide approach is correct at least in its recognition of the need for significant
decentralization of prioritization decisions. However, some would suggest that this may be more
appropriately done at a ministry rather than sectoral level.
Given that it is neither desirable nor feasible to centralize all prioritization decision, a key
question is how to create incentives and pressures for spending ministries to allocate the
resources at their discretion in the most effective manner. Two important mechanisms can be
mentioned: (1) a system whereby ministries are expected to fund minor new spending
initiatives through internally‐identified savings, and are therefore given an incentive to find such
savings, and (2) the pressure imposed by good central spending review (as discussed earlier).
A good expenditure prioritization process, integrated fully in budget preparation, is critical to
the success of a government‐wide performance budgeting system. Without good prioritization
processes, performance information may not be actually used in budget decision‐making, and
the results may be that all the effort of developing indicators, evaluations and program costings
may end up having little impact on the allocation of resources.
41 An outline of this approach in the MTEF context can be found in the World Bank (1998) Public Expenditure
Management Handbook, pp. 47‐8. In this version, the sectors also undertake a range of other related performance
planning and management roles, including the explicit definition of sector objectives, performance indicators and
42 See D. Good (2008) The Politics of Public Money (University of Toronto Press).
Kim, J. M. and C‐K. Park (2006), “Top‐down Budgeting as a Tool for Central Resource
Management”, OECD Journal on Budgeting, 6(1): 93‐94.
Ljungman, G. (2009), Top‐Down Budgeting – An Instrument to Strengthen Budget Management,
IMF Working Paper, obtainable at
OECD (2005), Reallocation: the Role of Budget Institutions, OECD: Paris, can be ordered at
Robinson, M. (2011), “Keeping the Lid on Aggregate Expenditure during Budget Preparation:
Enforcing Aggregate Expenditure Ceilings while Preserving Allocative Flexibility”, in proceedings
of Banca d’Italia Fiscal Policy Workshop 2001, available at
Robinson, M. (2010), Ministry Ceilings under Top‐Down Budgeting, at
Robinson, M. (2010), Keeping the Lid on Aggregate Spending, at
16. Accrual Accounting and Performance Budgeting
This section focuses on the relationship between accrual accounting and performance
budgeting. There are some who say that accrual accounting, and perhaps also accrual
budgeting, is essential for performance budgeting. We consider in this section whether this is
correct. The conclusion is that it depends on the form of performance budgeting. There are
certain forms of performance budgeting for which accrual accounting is indeed essential. But
this is not true of most forms of performance budgeting. It is, in particular, not true for program
budgeting. Although program budgeting may arguably be more effective in an accrual
environment, it will work perfectly well under traditional “cash” (or “commitment”) budgeting
The section starts with an explanation of accrual accounting (AA) and accrual budgeting, before
moving on to consider their relevance for performance budgeting.
What are Accrual Accounting and Accrual Budgeting?
AA is a financial reporting system. In other words, it is a specific methodology by which
organizations report their financial transactions and position. AA is the standard accounting
methodology of the private sector, where it arose primarily as a method for properly measuring
enterprise profit. AA may be defined as a methodology by which entities may properly measure
their costs of production and, where relevant, revenue earned.
In a private sector context, it is only through the proper measurement of costs of production
and revenue that profit may be correctly measured. The correct measurement of revenue
earned and profit or less is also relevant to government business enterprises (GBEs), because
they also earn revenue from selling products to consumers, and make a profit or loss on their
operations which needs to be measured. It is precisely for this reason that GBEs in most
countries have therefore long employed AA.
By contrast, the concepts of revenue earned and profit are largely irrelevant to government
bodies – that is, to ministries and other organizations which are primarily dependent on tax
revenues for their financing. For them, the potential benefits of AA arise largely from its use to
measure their costs of production – that is, the costs of the outputs which they deliver. For this
reason, in explaining AA, we focus firstly in on the accrual treatment of costs, rather than of
Traditionally, governments around the world have used so‐called cash accounting and
budgeting (coupled, in some countries, with so‐called “commitment” accounting and
budgeting) in their budget sectors. Only relatively recently have some governments moved to
replace cash accounting with AA. Under cash accounting, the principal focus of financial
reporting is upon payments and receipts of money during the financial year. At the ministry
level, the main form of financial reporting is the reporting of ministry expenditure, with
expenditure defined as payments to external parties43. Under cash budgeting, the spending
budgets given to each government ministry are also expressed in cash expenditure terms. In
other words, each ministry is given an annual budget formulated as a limit on the payments it is
permitted to make during the year.44
The easiest way of understanding AA is to contrast it with cash accounting. In other words, we
can best understand the nature of accrual accounting by explaining why it is that cash
accounting is inappropriate as a basis for measuring the costs of production, and how it is that
AA overcomes the weaknesses of cash accounting in this respect.
If a ministry wishes to work out the costs of production of the services it has delivered to
external clients this year (i.e. of its outputs), it will need to take into account the costs of all
resources used to produce those services. If it tries to do this by looking at this year’s cash
expenditure – that is, on the basis of its cash accounting – it will face a number of difficulties.
One is that some of the resources used to produce outputs this year will have been paid for in
previous years (e.g. buildings, equipment and supplies purchased in the past). Equally, some
resources used this year may not be paid for until sometime in the future (e.g. bills for supplies
delivered and used at the end of this financial year, which are not payable until the beginning of
43 More precisely, non‐repayable payments: in other words, payments other than repayments of public debt and
loans made by government to external parties. 44 In countries with commitment budgeting, there would also be a limit on the value of commitments to make
future payments which the ministry could make during the year. A “commitment” refers to a legal commitment to
make a payment, whether at the time the commitment is made or at some subsequent point of time (potentially
even some years in the future, in the case for example of multi‐year contracts with, say, construction companies
for major public works). In a commitments budget, quantitative limits are imposed on the quantum of new
commitments which may be entered into during the financial year in question.
the next financial year). A focus on cash expenditure will overlook these ‘non‐cash’ costs, and
to that extent will understate the costs of production.
The problem, moreover, cuts the other way as well. A measure of this year’s cash expenditure
will include all capital expenditure undertaken during the year. To treat this year’s expenditure
as a measure of this year’s costs of production therefore involves an implicit assumption that all
capital expenditure which takes place this year contributes to production in this year and this
year alone. By definition, however, capital expenditure is expenditure on assets which
contribute to production over a number of years. This means that to count all of this year’s
capital expenditure as part of this year’s costs of production is to overstate costs.
AA solves the above problems by replacing the cash expenditure measure with a measure of
the cost of the resources used in production, known as expenses. The expenses of an entity in a
given financial year measure the costs of all of the resources the entity uses to produce outputs
in that financial year, irrespective of when those resources where actually paid for. The
difference between the expenses concept and the expenditure concept therefore concerns the
financial year in which payments for inputs used in the production process are recorded.
Expenditure measures when the inputs are paid for, whereas expenses measures when they are
used in the production process.
Take the example of capital expenditure. Rather than recording all capital expenditure as a cost
of production in the year the expenditure takes place, AA treats a portion of the capital
expenditure as a cost of operation (an expense, in accrual language) in each of the years in
which the assets concerned contribute to production. Roughly speaking, if an asset’s life is, say,
ten years, AA might count one‐tenth of the price paid for the asset as a cost of production in
each of the ten years of its life.45 This is what is known as “depreciation”.
The difference between the cash accounting concept of expenditure and the AA concept of
expenses does not arise only in relation to the treatment of capital. Another example relates to
bills (accounts) presented to a department by its suppliers. Under cash accounting, such bills
are recognized in the department’s financial reports only in the financial year when the
45 This simplifies the concept of depreciation greatly. In practice, the accrual treatment of capital expenditure is
considerably more complicated, when one takes into account things such as inflation and the possibility that the
asset has some scrap value at the end of its life, let alone the more complicated asset valuation and depreciation
methodologies accountants have developed. However, the idea of the accrual treatment of capital as cost
allocation over time illustrates the essence of the accrual concept of expenses as applied to capital assets.
expenditure actually takes place (i.e. when the bill is paid). By contrast, under AA bills are
recognized as an expense in the financial year when the relevant supplies are delivered and the
bill therefore become payable (i.e. when the department incurs the liability to pay the
suppliers). This distinction can make a difference because, as noted above, payment for some
supplies may not take place until the financial year after the relevant they have been received
and used in production.
Civil service superannuation – pension costs – offers another example of the accrual treatment
of costs. Most governments operate their own superannuation schemes. In other words, they
promise retirement benefits to their civil servants. This makes the retirement benefits
entitlements which civil servants build up during their careers obligations which the
Government itself must meet. Consequently, from the Government’s point of view, these
entitlements represent a major financial liability. Moreover, they are an important part of the
cost of delivery of government services, because during each year a civil servant is employed
they build up additional superannuation entitlements. The question therefore arises as to how
should one measure the superannuation component of the Government’s costs of production
in any particular year. Cash accounting is no good for this task, because it is not possible to
measure the annual cost of superannuation by looking at annual superannuation expenditure.
To see why not, consider the position of an individual public servant, Jim. Each year during
Jim’s career, he builds up a growing entitlement to superannuation benefits. However, the
actual payment of those benefits (i.e. the expenditure) takes place only after the conclusion of
Jim’s career. If one employed cash accounting, we would have the nonsense result that:
No superannuation ‘cost’ would be recorded for Jim during the years when he was working
and providing services to the public,
There would be a substantial ‘cost’ recorded only when Jim was retired and manifestly
making no contribution at all to production.
What AA does is to recognize the fact that during each year of Jim’s career, there is a
superannuation cost attached to his employment. This cost is, of course, a non‐cash expense
(i.e. it involves no payment at that time), but this makes it no less real. How is the
superannuation cost of public service employment measured under AA? Essentially, the idea is
that instead of measuring the expenditure on superannuation benefits, the government as a
whole measures its annual superannuation costs by counting the gross increase in the total
amount of superannuation entitlements which government ‘owes’ its employees.46
It will thus be clear that only by using AA can public sector agencies accurately measure the cost
of producing the outputs which they deliver to the public. Of course, what AA generates is a
total figure which measures the aggregate cost of all of the outputs produced by an
organization in a particular financial year. AA per se does not provide information on the cost
of producing a specific output or group of outputs, nor does it provide information on the unit
cost of outputs. To obtain such information requires accrual accounting measures of expenses
to be brought together with output information. It is a function of so‐called management
accounting to split the total accrual expenses of an agency between its various programs (or, at
a more detailed level, between specific types of services).
The discussion to this point has focused on accrual accounting. Accrual budgeting is the use of
accrual accounting as the basis for budget allocations – more precisely, the use of accrual
concepts to specify budgetary control totals.
47 As discussed in the section on program
appropriations, control totals are the quantitative spending limits imposed on spending
ministries, whether as appropriations in the annual budget law or by administrative directive
(e.g. from the minister/MoF). Accrual budgeting must therefore not be confused with accrual
accounting, which refers only to the recording and reporting of financial operations of
government in accrual terms. Governments may budget in cash terms while accounting and
reporting in accrual terms48.
What does using accrual concepts to specify budgetary control totals mean? The most basic
type of accrual control total is an expenses control total—that is, a centrally‐imposed
quantitative limit on the expenses each spending ministry is permitted to incur within the
budget year.49 An expenses control total means that current year expenses are counted as the
46 “Gross” means here the increase in superannuation entitlements ignoring the quantum of the entitlements of
already‐retired civil servants which the government pays off during the year. 47 More formally, accrual budgeting has been defined (Robinson, 2009) as the specification of budgetary
expenditure authorizations and revenue estimates in terms of accrual accounting measures. 48 Note here that accrual accounting reports not only on expenses etc, but also on cash flows, and may therefore
be said to incorporate cash accounting. 49 As explained later, it is not essential that an expenses control total cover all categories of the ministry’s
expenses. The imposition of an expenses control total covering most categories of each spending ministry’s
expenses should, nevertheless, be regarded as a core feature of any accrual budgeting system.
use of the budgetary funding, irrespective of the timing of any associated cash payment. For
example, the following would be counted against the expenses control total:
Bills payable: amounts which are owed for goods and service delivered and used during the
financial year, even though payment may not be made until the next financial year.
The additional entitlements for future pension payments which civil servants accumulate
during the financial year.
Stocks (of supplies etc) actually used in the production process during the year, irrespective
of when those stocks were purchased.
Depreciation of the ministry’s fixed assets.
The Complexity of Accrual Accounting and Budgeting
The core case for AA in respect to performance budgeting rests on the proposition that, the
better the measure of the cost of outputs, the more useful this will be for performance
budgeting. As discussed below, this is – broadly speaking ‐‐ true. However, a judgment about
whether AA should accompany performance budgeting cannot be made without considering its
costs as well as benefits. In this context, it is important to recognize that AA has significant
downsides, the most important of which is that it is more complex and, as a result, costs more
in skilled labor and accounting system terms to operate. These costs and skilled labor demands
make the introduction of AA particularly challenging in low‐income (and even middle‐income)
countries. The question which each country must therefore ask is whether the benefits of
introducing AA into government outweigh the costs. In respect to performance budgeting, it is
not sufficient merely to demonstrate that program budgeting or some other performance
budgeting system can in theory work better in the presence of AA. What needs to be
demonstrated is that the improvement is sufficiently great (taken in conjunction with other
benefits of AA) to justify the costs and complexities of the move to accruals. It also needs to be
borne in mind that both performance budgeting and AA and demanding reforms in their own
right. It may not necessarily make sense to attempt to introduce them both at the same time.
This is even truer of accrual budgeting, which is considerably more complex that AA alone. The
additional complexity not only means that an accrual budgeting system is highly demanding of
skilled human resources (accountants and others) and supporting IT systems, but that it may
weaken expenditure control in countries where accrual concepts are not well understood by
managers (see Robinson, 2009).
Accrual Accounting and Program Budgeting
As repeatedly emphasized, the primary objective of program budgeting is improved expenditure
prioritization. Program budgeting improves expenditure prioritization by making it possible for
budget decision‐makers to compare (usually in an informal way) the costs and benefits of
The discussion above makes it clear that cash accounting is imperfect for this task. Because cash
accounting does not properly measure the costs of service delivery, it can distort program cost
comparisons by making some programs look less costly than they really are – and less costly
relative to other programs than they really are. Suppose, for example, budget decision‐makers
wish to compare a health treatment program and a preventative health program. Health
treatment is very capital‐intensive because a great deal of expensive equipment and buildings
are used. Health prevention relies on information campaigns, and makes little used of costly
infrastructure. If one compares these two programs on a cash accounting basis – that is, by
looking at the cash expenditure which each undertakes – during a period when there is not
much new investment in health treatment facilities underway (i.e. not many new hospital being
built etc), then the comparison will be distorted. The health treatment programs costs will look
artificially low, while this will not be the case for the preventative health program.
Other distortions can arise when making comparisons between programs as a consequence of
using cash accounting. For example, if we are comparing a program which is delivered primarily
by civil service staff to another which relies largely on contracted private‐sector companies and
their employees, the considerable deferred employment cost component of civil service
employment may make the former program look relatively cheaper than it actually is.
It is clear, then, that it is in principle better to base a program budgeting system on AA. AA
measures of program costs facilitate much fairer comparisons of the costs of alternative
programs. Accrual budgeting would be even better, because then the full cost of programs
would be charged to ministry programs, giving them a powerful incentive to base their
prioritization decisions on the real cost of programs, and not just on the level of cash
expenditure that they require in any given year.
The Link between Accruals and Performance Budgeting
“One of the key elements of the new resource based [i.e. accruals] approach is that it requires ministries
to undertake more accurate costing of activities, with expenses and income allocated to each of a
ministry’s objectives. This will assist the Government in ensuring that resources are allocated to priority
services in line with the Government’s objectives.” (HM Treasury, 2002)
“The introduction of cost [i.e. accruals] principles into government budget planning and accounting has
to do with improving government effectiveness through visibility, transparency, and cost‐conscious
behavior. … Transparency of costs provides politicians and administrative leadership a better basis for
prioritizing use of resources.” (Danish Ministry of Finance, 2006)
However, even though cash accounting creates some distortion in measures of comparative
program costs, it cannot be said that these distortions are on the whole so serious as to prevent
program budgeting from achieving its basic objective of improved expenditure prioritization.
Program budgeting has worked well in some countries with cash (or cash and commitment)
accounting and budgeting systems. Moreover, amongst the factors which determine whether
program budgeting works or not, the precision of the accounting system in measuring program
costs is in most countries a much less important consideration that whether the budget process
is designed to facilitate good prioritization decisions (see the Section 15). So in terms of making
program budgeting work well, the introduction of AA is usually not the most important priority.
For developing countries in particular this is good news. They should not feel that they must
combine the introduction of performance budgeting with a move to AA, let alone accrual
budgeting. They should bear in mind that most program budgeting systems – including most
recently the well‐designed system introduced in France in 2001‐2006 – have operated in a
traditional cash or cash/commitments accounting and budgeting framework.
Accrual Accounting, PurchaserProvider and Formula Funding Systems
By contrast to program budgeting, a purchaser‐provider (PP) system absolutely requires AA. As
discussed Section 12, PP requires that government agencies operate like businesses, with a
focus on their financial bottom line (profit or loss). There is only one way of measuring profit
and loss – through the use of AA. The use of AA is essential for this purpose not only because it
is necessary to measure production costs accurately, but also because the accurate measure of
revenue earned is absolutely imperative.
The AA approach to the measurement of revenue applies exactly the same principle as is
applied in the measurement of expenses. That is, accrual revenue is recorded in the year it is
actually earned, as opposed to the year when the cash happens to be received. The difference
between cash revenue and accrual revenue is then, once again, about the financial year in
which the revenue is recorded. Suppose, for example, that a business received a large payment
this year for delivery of services to be made over the following, say, three years. Cash
accounting would record this all as revenue this year. But this would be absurd, as the revenue
concerned will in reality only be earned progressively over the next three years as and when the
product is delivered. Reflecting this, AA would recognize the revenue in stages over those three
Similarly, if an enterprise sells an asset, cash accounting would treat this as revenue. By
contrast, accrual accounting would see that the cash received is not really revenue earned,
because it was received only by sacrificing the asset concerned. From the accrual perspective,
to treat such receipts as revenue would be rather like someone who sells their car treating the
payment they receive as if it were equivalent to salary.
Hence the most important of the suite of financial statements generated in an AA context: the
operating statement, commonly referred to in the private sector as the Profit and Loss
Statement (P&L). The operating statement reports total revenue and total expenses in the
financial year concerned (identifying the key elements of both). The 'bottom line' of the
operating statement is the operating result, which is total revenue minus total expenses. The
operating result is, of course, a measure of accounting profits.
Any hospital or other agency which is funded via a PP mechanism rather than traditional budget
funding needs to measure its operating result, just like a private business. Under the PP system,
the agency concerned needs to be sure that it is not making a loss, and AA provides the only
means of assessing this. It is precisely for this reason that when Australia and New Zealand
attempted (unsuccessfully) to apply the PP principle to the whole of the government budget,
they coupled it with a move to accrual budgeting – hence the term “accrual output budgeting”.
In respect to formula funding (FF), the story is a little more complex. As explained in the
relevant section, some FF systems are based on costs, and it is in relation to these that AA has a
potential contribution to make. Cost‐based FF is generally based on an output (rather than
outcome) funding formulas, of which the simplest version is funding = quantity x output unit
cost. Because accruals measures output costs better, it will immediately be obvious that FF will
operate better with AA. However, this does not mean that all forms of cost‐based FF necessarily
require AA. Particularly if the funding formula funds variable costs (i.e. does not cover capital
costs), some funding formulas may potentially work fine in a cash accounting and budgeting
context. As a generalization, however, it may be said that cost‐based funding formulas are likely
to require accruals.
Robinson, M (2007), “Cost Information” in M. Robinson (ed.) Performance Budgeting.
Robinson, M (2009), Accrual Budgeting and Fiscal Policy, IMF Working Paper, obtainable at
Diamond, J (2002) Performance Budgeting – Is Accrual Accounting Required? IMF Working
Paper, obtainable at http://www.imf.org/external/pubs/ft/wp/2002/wp02240.pdf.
17. Parliament and Performance Budgeting
The role of parliament in a government‐wide performance budgeting system varies greatly
depending on the political system, and in particular upon the role of the parliament vis‐à‐vis
executive government in the formulation of the budget. This is therefore an area in respect to
which we need to be very cautious in making generalizations. Broadly speaking, however, we
can distinguish between two extremes. At one end of the spectrum are parliaments which have
a major direct influence on the allocation of resources in the budget. At the other end of the
spectrum are parliaments which, although they must legally approve the budget, have in reality
little influence over its content, which is essentially determined by executive government.
The US Congress is a notable example of the first type of system. Although the American
president presents a budget to Congress for consideration, there are few limits on the changes
which Congress can (and does) make before voting its own final budget. In respect to the
second type of system, the British parliament is the most striking example. In Britain, not only is
the parliament usually tightly controlled by the cabinet via the party system, but any
amendment by the parliament to the budget presented by the executive is considered to be a
motion of no confidence, the passage of which would lead to the government’s resignation.
Parliament therefore essentially accepts the budget or sacks the government. The
parliamentary systems of many countries lie in between these extremes, with the parliament
exercising some degree of direct influence over the allocation of resources in the budget.
To the extent that the parliament has substantial independent budgetary power, a
government‐wide performance budgeting system will only work to insofar as the parliament
links its funding decisions to results. If, however, the parliament has little or no independent
budgetary power, this is essentially irrelevant. In either case, however, the performance
accountability (as opposed to budgetary control) role of parliament is a potentially important
part of a successful performance budgeting system. We therefore start by considering aspects
of this accountability role.
Parliament and Performance Accountability
In successful government‐wide performance budgeting system, performance reporting – that is,
reporting to parliament on the objectives and results achieved by government agencies – is a
key element of the system. A growing number of countries have adopted an approach originally
adopted in the US under the 1993 Government Performance and Results Act, under which
parliament is presented with annual performance plans and annual performance reports for
each ministry or sector within government.
Attempts to encourage parliament to focus on and debate performance in plenary session (i.e.
when the parliament meets as a whole) have generally met with limited success (see box for
Disappointing Dutch Experience in Parliamentary Plenary Performance Review
In order to increase the political relevance of [the debate on annual performance reports], [the
government] proposed in 2004 to hold a plenary debate on the annual reports on the third Wednesday
of May in the presence of the entire cabinet (Blok Resolution 2004). During the May 2005 debate, the
entire Cabinet was present. However, the floor leaders of different large parties were not present at the
debate. The number of MPs present at the debate fluctuated between 35 and 60 (of 150) despite the
fact that there were no parallel committee meetings . … The Blok Resolution …has led to an increased
political attention for the annual reports and increased media coverage. The attention for the plenary
debate unfortunately went hand in hand with a diminishing attention for the commission debates. The
attention of members of parliament for the annual report remains rather low. Interest tends to be
confined to the parliamentary finance commission.50
It is parliamentary committees which tend to be the main forum for effective parliamentary
performance accountability. Approaches to enhancing the role of parliamentary committees in
this area vary greatly internationally. One approach is to expand the role of the finance/budget
committee to cover performance as well as financial issues. Building up the role of sector
committees (i.e. committees covering specific sectors of government such as health) can,
however, be very useful. As is true in other areas of parliamentary committee work, the degree
of success of parliamentary committees in building their role in performance accountability
tends to be closely linked to the support resources at their disposition – particularly in respect
to professional research support staff. Particularly useful in this context is the creation of
professional analytic services which support the parliament as a whole, of which the US
Congressional Budget Office is one of the best‐developed examples.
The supreme audit institution (e.g. national audit office, court of accounts) is in many political
systems intended to serve as an accountability instrument of the parliament, independent of
50 Source: Sterck & Bouckaert (2006), p. 14.
executive government. In this context, the extension of the role of the SAI from purely financial
audit to performance audit (as discussed in the section of performance auditing) is potentially a
useful means of reinforcing the performance accountability role of the parliament. With
performance audits being presented to it by the SAI, the parliament is provided with a stronger
information base with which to hold executive government to account for the results achieved
with taxpayer resources.
Parliament and Expenditure Priorities
When parliament exercises substantial independent budgetary power, it becomes essential to
the success of government‐wide performance budgeting to encourage and facilitate it to use
results information when considering the budget. Philip Joyce (2007) identifies certain
important steps which can be taken to this end, including the presentation of performance
information in a form meaningful and readily usable by parliamentarians and the establishment
of a regular routine for program review and reauthorization. Beyond this, the scope for
generalizations about how to promote performance budgeting in parliament is a matter which
needs to be examined in a country‐by‐country context, with regard to key constitutional
variable (e.g. does the constitution prevent the parliament from increasing aggregate
expenditure – if so, this tends to force it to be more focused on prioritization) and political
variables (e.g. how strong is the party discipline – to the extent it is weak, it becomes very hard
to prevent budgeting fragmenting into un‐prioritized expenditures favoring particular
parliamentarians constituencies). Experience certainly suggests that, the greater the
fragmentation of budgetary power in the political system – both between executive
government and the legislature, and within the legislature – the more challenging it can
become to implement an effective performance budgeting system.
Joyce, P. (2007), “Performance Budgeting under the separation of powers”, in Robinson (ed),
Lienert, I. and M.‐K. Jung (2004), The Legal Framework of Budget Systems, Paris, OECD,
obtainable at http://www.oecd.org/dataoecd/3/28/43487903.pdf.
Mordacq, F. (2008), “Budgetary Reform and Parliament: The French Experience”, paper
delivered at International Symposium on The Changing Role of Parliament in the Budget
Process, Afyonkarahisar, Turkey, 8‐9 October 2008, obtainable at
18. MediumTerm Budgeting and Performance Budgeting
Performance budgeting has usually seen as being closely linked to medium‐term budgeting. In
this section we consider why this is the case. The starting point is a discussion of the nature and
different versions of MT budgeting, the key technical challenges it involves, and the experience
with MT budgeting in developing countries.
Nature and Benefits of MT Budgeting
The core elements of a MT budgeting are:
A clearly stated aggregate MT fiscal framework (MTFF), including a statement of desired
medium‐term outcomes for key fiscal aggregates (especially deficits and debt)—that is,
medium‐term fiscal targets—and of the aggregate expenditure ceiling and aggregate
revenue levels consistent with that fiscal policy.
A “top‐down/bottom‐up” reconciliation process to ensure that expenditure and revenue
policy are consistent with the MTFF—that is, to maximize the probability that the
government’s expenditure and revenue policies will, over the medium‐term, result is levels
of aggregate expenditure and revenue consistent with the fiscal targets set in the MTFF.
Resulting from this process, a set of medium‐term sectoral or ministry expenditure
projections or ceilings which both embody government expenditure policies and are
consistent with its MT fiscal policy. In most countries, these are indicative (i.e. the
government makes no advance commitment to maintaining sectoral/ministry expenditure
at the projected levels), but in some countries the government commits itself to funding
sectors or ministries at these projected MT levels.
The key benefits of an effective MT budgeting are:
Improved aggregate fiscal discipline: MT budgeting significantly reduces the chances of
sound fiscal policy being undermined by higher‐than‐expected expenditure arising from
expenditure policies and commitments which the government made in the past without
fully considering their future implications.
Reduced spending ministry uncertainty about their future levels of funding, leading to
better planning and management. This reduced uncertainty is greatest when government
makes commitments about funding over the MT time horizon. However, even where the MT
sectoral/ministry expenditure projections are only indicative, effective MT budgeting can
significantly reduce uncertainty. This is because, to the extent that the MT budgeting
process ensures that all expenditure policy decisions are fully consistent with aggregate
fiscal policy over the medium‐term, it will substantially reduce the need for government to
make ad hoc spending cuts because of unexpected fiscal developments.
Improved expenditure prioritization: the process of reconciling expenditure policy with top‐
down aggregate expenditure ceilings encourages a more systematic consideration of
expenditure priorities. For this reason, effective MT budgeting is closely linked with
improved expenditure review and prioritization processes, of the type which are discussed
in the section on expenditure prioritization. These processes provide an excellent means by
which to feed performance information more systematically into budget preparation,
providing an important link between the MT budget framework and performance
A fundamental tool of MT budgeting is what is often referred to as “forward estimates”.51 These
are MT estimates of expenditure and revenue on a “current policy” basis—that is, projections
which indicate what expenditure and revenue will be in each of the next three or four years if
there are no new spending initiatives, no changes to tax laws, and all explicit and clear
commitments made to future expenditure (including political promises) are taken into account.
Armed with good forward estimates, ensuring the consistency of “bottom up” expenditure and
tax policy with “top down” aggregate expenditure ceilings driven by fiscal policy is a two stage
Firstly, an assessment via the forward estimates of whether unchanged expenditure policy is
consistent with desired levels of aggregate expenditure. If the forward estimates indicate
that existing expenditure policy will result in an excessive level of aggregate expenditure
over the medium‐term horizon, then adjustments are made to current spending policy (or,
51 Also known by a range of other names, such as “annual reference level update” in Canada.
potentially, to tax policy). If, on the other hand, current expenditure policy implies
expenditure below the aggregate envelopes, then there is room for new spending initiatives
(or tax cuts).
Secondly, a similar assessment of the impact of potential new spending initiatives, using
estimates of their MT costs to determine whether they can be accommodated without
breaching the aggregate expenditure envelope or whether they must either be scaled back
or room found for them by cutting elsewhere.
Success in reconciling expenditure and revenue policy with sound fiscal policy over the MT
depends critically on the quality of the forward estimates—that is, on the extent to which the
forward estimates provide an accurate forecast of the levels of aggregate expenditure and
revenue which will result over the medium term from current expenditure and revenue
policies. Only to the extent that the forward estimates are accurate will they provide advance
warning that changes in expenditure policy are needed to meet aggregate fiscal constraints.
Poor quality forward estimates will tend to undermine the entire MT budgeting process.
Countries which have attempted to introduce MT budgeting without investing significant effort
in the forward estimates process tend, unsurprisingly, to have been disappointed with the
results. In the absence of a system and capacity to produce quality forward estimates,
projections of medium‐term aggregate spending and revenue tend to be prepared on the basis
of the crudest techniques (e.g. updating based only on the application of a general inflation
factor) which fail to capture the dynamics of current policy. This makes the top‐down/bottom‐
up reconciliation process pointless, and it is therefore not surprising that such countries have
also tended not to invest much effort in the enhancement of this process.
MT Ceilings: Fixed vs. Indicative?
MT budgeting produces a set of MT expenditure projections, often referred to as “ceilings”. As
noted above, however, the status of these ceilings differs significantly between countries.
At one end of the spectrum are countries which set fixed MT ceilings for ministries. That is, the
government makes commitments to ministries about the level of funding they can expect to
receive over the coming, say, three years. One of the most well developed examples of this is
the UK, which since the late 1990s has operated a system based on fixed three‐year budgets for
spending ministries (recently extended to fixed commitments lasting the life of the five‐year
parliament). A number of other OECD countries – including France – have more recently
followed the British example.
At the other end of the spectrum are countries where the MT expenditure projections are
indicative and imply no commitment to ministries in respect to the MT funding they will
receive. In such countries, government retains and exercises the right to change ministry and
sectoral allocations each year, in order to progressively improve its expenditure prioritization.
Australia is an example of this approach.
The Australian Approach to MT Budgeting
In the Australian system, the budget which each spending ministry receives is based essentially on the
expenditure forward estimates — with modifications only for such new policy proposals as the
government accepts (or, conversely, for cuts arising from explicit decisions to eliminate some programs
inherited from the past). In preparing the budget for, say, 2011, the finance ministry first takes the
estimates of 2011 expenditure which were prepared in 2009. It updates those estimates for any changes
which have impacted on the costs of delivering existing policy, as well as any policy changes which might
have happened during 2010, and then uses this as the starting point of the budget preparation process.
Spending ministries are, approximately speaking, told not to bother debating the budget they need to
deliver ongoing programs in the coming year – these are given by the updated forward estimates. Any
requests for additional funding which the spending ministries make must, instead, be based on
proposals for new policy initiatives. Thus the budget bids lodged by spending ministries early in the
budget process – the so‐called Portfolio Budget Submissions – are entirely focused on proposed
expenditure policy changes. Using the expenditure forward estimates in this way creates a powerful
incentive for the spending ministries to collaborate with the MoF in developing reliable forward
estimates methodologies, because getting the estimates wrong could adversely affect the spending
ministries budget position. In Australia, forward estimates were developed some years before they
started to be used for budget preparation in the manner described above. It was only after they started,
in the early 1980s, to be used as the basis for determining ministry budgets, that it became essential to
substantially upgrade the quality of the estimates. This underlines the fact that this system can only
work well once the quality of forward estimates has become sufficiently good that the estimates provide
a broadly adequate assessment of the future costs of expenditure on an “unchanged policy” basis.
Both of these approaches represent legitimate models of MT budgeting. As noted above, even
when ceilings are indicative only, spending ministry uncertainty about the levels of MT funding
they will receive is reduced, because the compatibility of the indicative ceilings with aggregate
fiscal policy will (unexpected shocks aside) have been verified via the MT budgeting process.
The technical requirements for setting fixed MT ceilings are considerably greater than for
indicative ceilings. They include:
Excellent macroeconomic forecasting,
Excellent forward estimates of both revenue and expenditure,
Reasonable macroeconomic stability,
A strong expenditure review and prioritization mechanism,
Sustainable aggregate fiscal policy settings.
Without a capacity to make good medium‐term revenue forecasts, there is the danger that
unrealistically high multi‐year expenditure ceilings will be set. Down the track, when revenue
turns out to be below forecast, the government will be faced with the difficult choice between
withdrawing its budget commitments to spending ministries or finding additional revenue.
Without good processes to scrutinize and reprioritize expenditure, making medium‐term
budget commitments to spending ministries will simply increase expenditure inflexibility, and
make it harder to reallocate spending over time to the areas where it is likely to deliver the
greatest benefits. With only limited capacity to review expenditure, the MoF may find itself
forced to recommend to the government multi‐year ceilings largely based on the status quo
rather than on critical analysis of the status quo. Under such circumstances, ministries which
should have the budgets reduced could easily find themselves protected from scrutiny for a
number of years at a time. It is highly relevant that in the UK’s case, these expenditure ceilings
are set only after thorough spending reviews (discussed in the section on expenditure
MT Budgeting and Performance Budgeting
A MT budgeting perspective can greatly reinforce the effectiveness of a government‐wide
performance budgeting system. This is first and foremost because expenditure prioritization –
the central objective of program budgeting systems – works best when carried out from a MT
perspective. In other words, when deciding how best to allocate budget resources between
programs, it is highly advantageous to know the cost of those programs not just for the next
financial year but also for the following years. Program costing on an annual basis may be
misleading. This is particularly true for new programs, for which the first‐year costs are often
significantly different from the ongoing costs (due, e.g., to substantial program set‐up cost, or to
the fact that the program may not be fully operational until part way through the financial
year). It is therefore essential that new program proposals be accompanied with MT cost
estimates. The reduced uncertainty about future funding levels which MT budgeting provides
helps support performance budgeting because decisions about program allocations can be
made with greater confidence that they will be sustainable – in other words, that it won’t be
necessary to revisit them two or three years down the track because they can’t be afforded.
Conversely, performance budgeting helps MT budgeting. Most fundamentally, program
budgeting will enable the best quality expenditure forward estimates to be developed, because
it will facilitate the modeling of program‐specific cost drivers (that is, demand or cost factors
which affect the evolution of the ongoing costs of specific programs). In a system of fixed MT
ministry ceilings, performance budgeting is virtually a necessity because, as outlined earlier,
excellent expenditure prioritization mechanism are critical. Thus in the UK system, good
performance information (especially, but not only, excellent performance measures) has been
an essential tool of the triennial spending reviews.
The MediumTerm Expenditure Framework (MTEF)
To this point, the discussion has been about MT budgeting in general, with no mention of the
familiar concept of the medium‐term expenditure framework (MTEF). The MTEF concept has
been the cornerstone of the approach to MT budgeting advocated by international
organizations including the World Bank and the International Monetary Fund. The problem,
however, is that the term “MTEF” has been used in so many different ways by different authors
that it can no longer be said to have a clear, widely‐agreed definition. In particular, the term is
very often used in a sense much wider than multi‐annual budgeting – for example, to refer to a
whole system of budget preparation, often incorporating also policy formulation and planning.52
This extremely broad definition was promoted in the World Bank’s 1998 Public Expenditure
Handbook, in which the MTEF was defined as “a whole−of−government strategic policy and
expenditure framework … The MTEF consists of a top−down resource envelope, a bottom−up
estimation of the current and medium−term costs of existing policy and, ultimately, the
matching of these costs with available resources”. While the second part of this definition is
essentially the same as the concept of MT budgeting presented above, the italicized words
52 Indeed, the Public Expenditure Handbook could itself be seen as having contributed to this much broader
interpretation. Notwithstanding the crisp definition of an MTEF quoted above, there are a number of comments
elsewhere in the manual which suggest a much more expansive notion of the MTEF.
could be read to imply a definition which covers not only the whole of the budget preparation
process, but the planning process as well. On the other hand, one can find some authors who
use the term MTEF in a much narrower sense – for example, refer only to detailed forward
expenditure estimates.53 The term “MTEF” is not one which is used much by OECD countries,
where the terms “multi‐annual” or “medium term” budgeting are much more common.
The broad conception of an MTEF as a total budgeting, planning and policy system means that
the MTEF concept has often been seen as including within it, amongst other things,
performance budgeting and the expenditure prioritization mechanisms. The problem with this
is that it is really not possible or desirable to specify as part of the definition of an MTEF specific
forms of performance budgeting and specific types of expenditure prioritization mechanism
which are appropriate to all countries and which will be agreed for all time to be the best
approaches. Thus, for example, many (including the Handbook) say that a sector‐wide approach
to planning is an essential element of the expenditure prioritization mechanism under an MTEF.
However, as discussed in the expenditure prioritization session, there are good grounds to
dispute the effectiveness of this approach to prioritization.
The same point applies to performance budgeting. Some proponents of MTEFs have seen it as
necessarily incorporating program budgeting. Others have taken the view that some form of
“output budgeting” or “activity‐based budgeting” – by which they mean a formula funding
approach to estimating budget requirements as a function of outputs or activities to be
delivered – is a necessary part of an MTEF. This type of “output budgeting” (which is critiqued in
the section on program costing and accounting) was fashionable in the late 1990s and early
2000s, but is much less so today. Again, it is not clear that it makes sense to incorporate such
specifics into the definition of an MTEF.
While discussing the MTEF concept, two other widely used acronyms – Medium‐Term Fiscal
Framework (MTFF) and Medium‐Term Budgeting Framework (MTBF) – should be mentioned.
The term MTFF, to which reference was made at the outset of this section, has a clear meaning.
It is widely understood to refer to the aggregate fiscal policy element of an MTEF – clear
medium‐term fiscal objectives, and a set of fiscal projections consistent with them. The term
53 This appears to be the version of the concept used by Allen & Tommasi (2001), p. 180. For Allen and Tommasi,
an MTEF so defined is merely a component of a broader ”medium‐term budget framework”.
MTBF has, by contrast, been used in so many radically different ways that it is pointless to
attempt define it.54
Overall, it seems to make little sense to approach MT budgeting based on definitions – in other
words, to start with a definition of what an MTEF is supposed to be, and then work on
introducing an MTEF based on that definition. It is much more important to focus on the core
elements of MT budgeting, and the optional extras (e.g. fixed MT ministry ceilings) and work
out on a country‐by‐country basis what elements will work best.
At the time of writing, considerable effort was underway in international organizations to re‐
think the whole issue of MT budgeting, and the MTEF in particular, and it appeared likely that a
new paradigm would emerge within the subsequent years.
Developing Countries and the MTEF
The MTEFs introduced in developing countries have differed greatly, particularly in terms of the
“add‐on” elements of prioritization mechanisms and performance budgeting. The core element
has been the development of MT ministry expenditure “ceilings”. While there have been
different approaches on the question of whether these should be regarded as fixed or indicative
– and whether they would be rolling or fixed – it has been universally assumed that the MT
ceilings should at the very least guide annual budget preparation.
In practice, the experience with MTEF has often been quite disappointing. The biggest problem
has been that the funding allocations in the annual budget have often had little or no relation to
the MT ceilings. Expressed differently, the MTEF has been a document with little impact on the
real budget. This has particularly tended to be the case in countries where the MTEF was
essentially prepared by consultants, with limited ownership by the authorities, so that the
expenditure priorities expressed the MTEF were not necessarily priorities to which the
authorities are really committed.
Poor quality forward estimates have been perhaps the most widespread obstacle to the
development of effective MT budgeting in developing countries. As a South African
commentator (Fölscher, 2007) noted in respect to the experience with MT budgeting in African
54 See, for example, the different concepts of an MTBF in the IMF’s Fiscal Transparency Manual (2007: 129) and in
Allen and Tommasi (2001), p. 180.
…the quality of forward estimates is poor. They consist far too frequently of the proposed
budget for the first year of a multi‐year framework, followed by inflation adjusted
projections of cost for the outer year ...they pay little attention to, for example, the likely
phasing of policy implementation, changes in demand that will effect spending unevenly
or the impact of once‐off capital spending on the base‐year estimates. ...A key aspect of
embedding a medium‐term perspective therefore is deciding what the rules are for rolling
over and adjusting and determining the forward estimates.
A key reason for this is that the preparation of forward estimates is, at the technical level, a
more difficult task than is often assumed.
With all the focus on MT budgeting, it seems sometimes to be forgotten that many developing
countries have major difficulties with the quality of their annual budgeting. Many are not able
to prepare good quality annual projections of the funding requirements of ongoing services.
Some question, in this context, whether it makes sense to be pushing countries to introduce
MTEFs when the quality of their annual budgeting needs such considerable improvement. In
this context, the findings of a recent review by the World Bank’s Independent Evaluation Group
(IEG) are highly relevant. The IEG (2008: xv) noted the importance of “dealing with the basics
first, such as ensuring that the government is executing a one‐year budget reasonably well
before launching sophisticated multiyear budgeting.”
It may therefore make more sense for developing countries to think of the task not so much as
one of “introducing an MTEF”, but rather as one of gradually introducing more of a MT
perspective into the budget, starting initially with efforts to define clearer medium‐term fiscal
objectives and gradually improve the quality of expenditure and revenue forecasts.
If it is the case that the introduction of full‐blown MT budgeting is a difficult task which not all
developing countries can aspire to in the short term, what does this mean for performance
budgeting in developing countries? It would be difficult to argue that a full‐blown MT budgeting
system is an essential prerequisite for performance budgeting. Even in the context of a strictly
annual budgeting system, introducing the systematic use of performance information can help
to significantly improve expenditure prioritization and to put more pressure on ministries to
improve the effectiveness and efficiency of their spending. From this perspective, one can think
of performance budgeting and medium‐term budgeting as two dimensions of PFM reform in
developing countries which have important synergies, but which should not be thought of as
absolutely essential to one another.
In conclusion, introducing a MT perspective into budgeting can offer great benefits in terms of
improved fiscal policy implementation and reduced funding uncertainty – and consequently
better management – at the ministry level. MT budgeting has considerable synergies with
performance budgeting: each reinforces the other in major ways. There is no single “correct”
model of MT budgeting, nor even a single model of the MTEF. The form, and degree, of MT
budgeting should be tailored to the circumstances of each country.
World Bank (1998) Public Expenditure Handbook, obtainable at
Kasek, L. and D. Webber (2009), Performance‐Based Budgeting and Medium‐Term Expenditure
Frameworks in Emerging Europe, Warsaw: World Bank, obtainable at
Le Hourerou, P. and R. Taliercio (2002), Medium Term Expenditure Frameworks: From Concept
to Practice: Some Preliminary Lessons from Africa, World Bank, Africa Region Working Paper
Series, No. 28, obtainable at
Wynne, A. (2005), Public Financial Management Reforms in Developing Countries: Lessons of
Experience from Ghana, Tanzania and Uganda, African Capacity Building Foundation Working
Paper no. 7, obtainable at
19. Implementing GovernmentWide Performance Budgeting
This section summarizes key implementation issues which arise in implementing a government‐
wide system of performance‐informed budgeting. In other words, the focus is on the
implementation requirements of a program budgeting system which aims to make systematic
use of performance information in budget decisions to improve expenditure prioritization and
increase pressure for performance. The special implementation requirements of more complex
systems of performance, such as purchaser‐provider, are discussed elsewhere in this manual.
The development of a performance information base is, obviously, one of the first steps in the
development of a basic system of performance‐informed budgeting. The development of
program performance indicators should not be thought of as a single‐step reform which can be
accomplished in a short time‐frame such as a couple of years. Rather, it is a gradual process
which – as experience in leading OECD countries shows – will proceed over decades. This does
not, however, mean that it is necessary to wait decades before a workable performance
budgeting system can be introduced. Even a highly selective set of program indicators can
enormously improve the expenditure prioritization decisions made by the government with the
support of the finance ministry. The initial aim should therefore be to develop a relatively small
set of useful program indicators, along with relatively simple program evaluation which is
designed to be usable in the budget process.
It is very important not to neglect evaluation in a single‐minded effort to develop indicators.
Indicators in isolation are sometimes of limited value, and evaluation can be crucial to
interpreting indicators. In the early stages of the developing of a government‐wide
performance budgeting system, the focus will tend to be on the development of simple – in
many cases “desk based” – reviews of a small set of key programs during each budget cycle.
The development of a program structure for the budget is a reform which takes some years.
The first step is the definition of appropriate programs in spending ministries – which means
programs which are defined in terms of outcomes and outputs (with limited exceptions such as
administration programs). This can usually be expected to take an absolute minimum of two
years (and usually longer than that). In this process, spending ministries need extensive support
from the MoF, in the form of technical instructions and guides and a major training program.
The actual development of the programs needs, moreover, to be an interactive process in
which the spending ministries prepare drafts and these are then reviewed and approved by the
finance ministry. It is, on the one hand, undesirable that the MoF simply impose its choice of
programs upon spending ministries – not only does it not have sufficient knowledge to do this
properly, but the lack of “ownership” of the resultant program structure by the spending
ministry will tend to undermine the reform. On the other hand, because the programs are
intended to be an instrument of central expenditure prioritization, spending ministries cannot
be left to just define programs in any way they wish.
The modification of the accounting system and the financial management information system
(FMIS) to make them program‐friendly is a quite major implementation step. Although
changing the chart of accounts is the first step, modifying the FMIS is usually the biggest task
here. It may not make sense to wait till the introduction of a completely new IFMIS to introduce
a program structure to the budget, so modifications to existing systems are often what is
needed. Extensive training in new accounting procedures and related systems changes will be
Once a program structure has been developed and the accounting system and FMIS have been
made program friendly, it is possible to start approving the budget on a programmatic basis. In
the run‐up to this, most countries first develop indicative program budgets which are presented
to the parliament as an annex to the traditional budget. Such an indicative program budget
shows the parliament and public what the budget appropriations would look like if they were
approved in programmatic terms.
In moving to program appropriations, it is crucial to make a clear decision about the manner
and extent to which traditional budget controls will be reduced. Most important here is the
question of the degree of line‐item control simplification. It is a serious mistake to simply
impose program appropriations on top of a highly detailed traditional budget – this will not only
deprive agencies and managers of the increased managerial freedom which is an essential part
of performance budgeting, but will in fact tie their hands even more than was previously the
case. In reducing line‐item controls, most countries will not wish initially to go as far as OECD
countries like France, Australia and the UK. But they will usually need to significantly reduce the
line‐item controls which they have previously enforced.
As discussed in the sections on PFM reform and expenditure prioritization, an important part of
the implementation of a successful government‐wide performance budgeting system is the
reform of the budget process and the reduction of expenditure rigidities so as to make it
possible to more readily reallocate limited public resources to the sectors and programs where
it can deliver greatest social benefit.
The MoF will necessarily play a pivotal role in the implementation and ongoing operation of the
performance budgeting system. Changes in the modus operandi and skill set of the MoF are
therefore essential if performance budgeting is to be a success. The MoF can no longer be
simply a controller of expenditure. It must play a central role as an expenditure policy adviser to
the government. In particular, it must develop its capacity to advise government on where
expenditure can be cut – e.g. via cuts to unsuccessful or low priority programs or efficiency
improvements – in order to make room for new priorities or to assist fiscal consolidation. Such a
role will in most countries requires a broadening of the skill base of the ministry, with the
reinforcement of policy analysis skills alongside accounting and economic analysis capacity. This
will normally require a combination of retraining and new recruiting. The increased expenditure
policy role of the MoF will require additional resources, which raises the question of overall
staffing levels. However, part of the answer to this can be found in the systematic elimination of
unnecessary traditional control functions, such as the review and approval of large numbers of
transfer requests arising from excessively detailed line‐item controls.
In discussing here the role of the MoF, it should be recognized that in some countries the
expenditure policy role described above will appropriately be shared with other central
agencies (e.g. the prime minister’s or president’s office).
Broadly speaking, countries may choose between a “big bang” and a more gradual approach to
the implementation of performance budgeting. The “big bang” approach – of which Russia
provides one example – means implementation within a couple of years. Generally speaking, it
will require that the country concerned already has not only the major PFM prerequisite in
place, but that the FMIS works well and can be relatively easily adapted to programs. The great
advantage of the big bang approach is momentum. By contrast, a gradual approach may be
more realistic in many countries. Equally, however, the great danger of too gradual an
approach is the loss of momentum, as a result of which the reform may never be completed.
The example of France is a good one to have in mind when determining the implementation
timetable. After having spent some years determining the broad parameters of its new
performance budgeting system, France deliberately chose a five year implementation period
between the 2001 passage of the law mandating the new system and its coming into full force
in 2006. This five year period proved to be an extremely busy one, give the scale of change
which the new system involved, and one could not say in retrospect that the reform could have
easily been implemented in a shorter period of time.
The management of the implementation process is critical to success. Performance budgeting
cannot, firstly, be implemented successfully without strong support from the political
leadership. It will usually make sense for the MoF – perhaps in conjunction with other central
agencies – to create an implementation task force. Spending ministries need to be closely
involved in the process at a senior level, perhaps via a consultative committee comprised of
senior ministry representatives. Throughout, the MoF will need to provide strong technical
guidance to spending ministries to assist them in properly developing the new systems (e.g.
properly defining programs and selecting the right types of program performance indicators).
Developing countries often face more serious implementation challenges arising, for example,
from greater capacity constraints, weaknesses in their overall PFM systems and, sometimes,
from governance problems which increase expenditure rigidities. These have to be taken into
account when determining the target timeframe for implementation. The resource and capacity
constraints which face developing country MoFs make it particularly important to simplify
unnecessary traditional controls, in order to free staff time for new responsibilities. Above all,
however, developing countries can ease the implementation burden by keeping their
performance budgeting systems simple. In this context of program budgeting, this means, for
example, avoiding overly complex program structures and unnecessary “add‐ons” such as
activity‐based costing. More generally, it usually means deferring to the future the possible
application the more sophisticated performance budgeting mechanisms such as formula
funding and purchaser‐provider. Performance targeting should also be approached in a gradual
and highly selective manner. Overall, the initial focus should be simply on making the budget
more performance informed: that is, on introducing a systematic consideration of program
results when determining the allocation of resources in the budget.
In conclusion, performance budgeting should not be seen as a simple reform to be introduced
overnight. Even in its least complex form, it is a demanding multi‐faceted reform which takes
years to put in place.
Diamond, J. (2007), “Challenges to Implementation”, in Robinson (ed.), Performance Budgeting.
20. The Managing for Results Context
Performance budgeting uses the budget as a tool for making public management more results‐
focused. It is, however, only one of a range of “managing‐for‐results” instruments which share
this broad objective. In this final section we look at some key areas of MFR reform which need
to be pursued together with performance budgeting if the full benefits of the latter are to be
Civil service reform is one of the most crucial accompanying reforms in many countries. There
are a range of reasons why, to varying degrees in various countries, the civil service may not be
sufficiently client and results‐focused. MFR seeks to change this by focusing employment
arrangements – hiring, pay, promotion and termination – more on results. Formal performance
planning and monitoring is a key part of this, and makes increasing use of tools such as formal
performance “contracts” or agreements.
It is in respect to the setting of clear objectives for organizational units and individuals within
them, and the development of performance measures to assess the degree of achievement of
these objectives, that civil service performance management needs to be linked to the broader
strategic planning and performance budgeting systems. In particular, the objectives and
indicators established for work unit and individuals should be linked to program and ministry
objectives and indicators. Generally speaking, objectives and indicators should “cascade”
through organizations, linking those for the organization as a whole to those set for individuals
and work units.
The question of performance incentives is an important one for performance budgeting. In
traditional civil services, the material incentives for good performance are often weak, and it
follows that strengthening these – for example, through performance pay arrangements – can
be an important element of MFR. However, we need to be aware of the potentially perverse
effects which too large a component of performance pay may create as a results of the
inevitably imperfect measurement of individual or work unit performance. It is also a major
mistake to view performance motivation as exclusively a question of material incentives.
“Public service motivation”—the altruistic commitment of public workers to client service (e.g.
doctors and nurses to patients) or to a cause (e.g. the desire of environment ministry staff to
save the environment) – is very important. An important part of good performance
management is the reinforcement and harnessing of public service motivation.
Managerial flexibility in human resource management is a very important part of MFR.
Traditional civil service arrangements tend to give managers little scope to reward good staff,
or to sanction or terminate ineffective staff. In countries which have gone furthest along the
MFR path, managers have acquired much greater responsibility and authority in these areas.
However, the scope for movement in this direction has to be viewed in a country‐specific
context, taking into account the overall status of governance – for the same reasons as were
discussed in respect to employment flexibility in the section on PFM reform.
Finally, the success of MFR depends completely on the political circumstances of the country
and, more specifically, on the degree to which the politicians and the electorate are themselves
performance‐focused. If, for example, the political system is one where continuing electoral
support for government has little relationship to their capacity to deliver efficient and effective
expenditure – but is more related to, say, group loyalties or rent‐seeking behavior –
performance budgeting, and MFR more generally, are not likely to make much headway.
21. Concluding Comments
Performance budgeting is a challenging but potentially important reform in the ongoing
struggle to make government more results‐oriented. Coupled with other “managing for results”
reforms, it can help substantially improve the effectiveness and efficiency of public
expenditure. The objective of this manual has been to provide an outline of key forms of
performance budgeting, their differing performance information requirements, the key design
and implementation issues which they raise, and their pre‐requisites and co‐requisites. The
diversity of performance budgeting models has been emphasized. It is important for each
country to make a clear decision about what type of performance budgeting which is
appropriate to its own circumstances. This decision should explicitly take into account the
specific capacity and other characteristics of the country concerned. Another key point
emphasized has been the importance of a realistic approach to implementation strategy and
timetable. Successful performance budgeting is not something to be introduced in a rush. It is,
rather, a reform which can be expected to take at very minimum several years to implement in
a preliminary way, to reach the point where budgeting starts to become more systematically
Accrual accounting System of accounting in which a key role is played by the recording of
expenses, assets and liabilities.
Accrual budgeting Budgeting system in which budget approvals take the form principally
of authorizations to ministries to incur expenses.
Accrual output budgeting Purchaser –provider budgeting model applied to the whole
government budget, as introduced in New Zealand and Australia in the
Activities Types or categories of work undertaken in the production and delivery
Activity‐based costing A costing methodology in which input costs are allocated to categories
of activity, using an allocation basis which reflects, as closely as
possible, the real consumption of resources by those activity
categories. When used for program costing, the activity costs are
allocated between programs in a second stage.
Aggregate fiscal policy The government’s overarching objectives for the budget deficit, debt
and other relevant fiscal aggregates.
Appropriation Amount of expenditure legally authorized by the parliament to be
Administration program Programs which cover overhead costs of a ministry or agency, such as
central management and personnel services. Also known as, e.g.,
administration programs or support programs.
Allocation basis Formula or principle used to allocate specific expenditure item
between two or more “cost objects”—in the context of program
costing, between two or more programs.
Alternative budgeting A variation of zero‐base budgeting in which decisions are focused not
on a zero base but on the margins near the current budget base.
Usually three or more alternative budgets have to be submitted for
each program. Generally, at least one of the alternatives has to be less
than the current budget. Often a specific percentage reduction is
Bottom‐up budgeting Budget preparation process which is based primarily on spending
ministries making bids for resources, with central consideration of
those bids then determining the overall shape of the budget.
Budget classification Categories of expenditure used in the budget, particularly for the
approval of expenditure.
Budget preparation Stage of the budget process the government decides how much
funding it will provide to which agencies and for which purposes, and
prepares the budget law which will be presented to parliament for its
Budget execution Implementation of the expenditure plan developed in the budget –
including the entering of contracts and expenditure of funds.
Cash accounting A system of financial accounting in which the focus is upon money
actually paid out (expenditures or outlays) and money actually
received in the financial year.
Chart of Accounts System of classification of transactions – revenue, expenditure,
financing etc – used for accounting purposes.
Classification of the
Functions of Government
Functional classification of expenditure developed by the United
Nations and incorporated into the International Monetary Fund’s
Government Financial Statistics methodology.
Commitment Contractual obligation to make a future payment.
Control total A centrally‐imposed quantitative limit on some category of
expenditure which can be undertaken by the spending ministry during
the budget year. May be imposed by the parliament or by a central
executive body such as the cabinet of ministers or the finance ministry.
Cost‐effectiveness The achievement of intended outcomes at lowest possible cost.
Cost driver In the context of program budgeting, a formula or principle which is
used to determine the proportions in which an indirect cost is split
between two or more programs (or sub‐programs) to which it
Direct cost In the context of costing programs, expenditure on inputs which
contributes to only one program.
Effectiveness The degree of success of an output in delivering its intended outcomes.
Efficiency Production of an output at minimum cost while holding quality
constant, given prevailing input prices.
Evaluation Analytic assessments typically addressing the cost‐effectiveness or
appropriateness of public policies, organizations or programs.
Expense Costs attributable to the present financial year in the accrual
accounting framework. Represents a consumption of resources,
regardless of whether it is paid for immediately or results instead in a
liability or a reduction in assets.
External factors Factors outside the control of government which influence the
outcomes achieved by public programs. External factors may be either
client/case characteristics or aspects of the context in which the
program is delivered. Sometimes also referred to as “contextual
information system (FMIS)
Computerized systems for managing budgeting and associated
processes. See also IFMIS.
Formula funding When used as a performance budgeting tool, formula funding is a
system in which funding provided by government to a public sector
agency is an explicit (that is, algebraic) function of measures of
expected and/or actual results—that is, of measures of outputs and/or
Forward estimates Projections by a central budget agency of aggregate expenditures over
a fixed term (usually budget plus three years) on a no policy change
basis. Forward estimates are the basis for medium term fiscal planning
and may also be used as the basis for imposing expenditure limits.
Functional classification Classification of expenditure by “functions”, a somewhat imprecise
term which represent a mix of output types and activity types.
Heterogeneity The deliberate variation in the amount and/or types of activities
delivered to different clients/cases receiving the “same” service,
particularly in response to difference in client/case characteristics. For
example, more intensive teaching activity directed to students with
High‐level outcomes The more indirect outcomes of outputs, which arise as a consequence
of the achievement of proximate outcomes. For example, in education,
the proximate outcome of higher numeracy and literacy contributes to
the higher‐level outcome of better economic performance. Also
sometimes known as “end” or “ultimate” outcomes.
A computerized system which to a greater or lesser degree integrates
multiple functions into the same system, so that accounting,
expenditure control, payments, budget preparation and a range of
other functions will be built into the same big computerized system.
Impacts Term used by some to refer to longer‐term or higher‐level outcomes.
Incrementalism Budgeting is characterized by “inattentiveness to the (budgetary)
base”—in other words, that budgetary decision‐makers take the
budgetary base more or less for granted as the starting point in budget
formulation, and focus their attention primarily on the size of the
increment (or, occasionally, decrement) in agency or program budgets,
mainly by a process of adjusting budgets for cost changes.
Indirect cost In the context of costing programs, costs of inputs or activities which
contribute to more than one program. More generally, shared costs
which need to be allocated between a number of “cost objects.”
Inputs Resources used in the carrying out of activities to produce outputs (for
example. labor, equipment, buildings).
Intermediate outcome The more direct or immediate impacts of outputs. For example, in
education, the student knowledge (for example, higher numeracy and
literacy) is a key proximate outcome.
Line‐item budgeting Budgeting in which agencies are provided with budget appropriations
specified in terms of input categories (that is, by economic
Managing‐for‐results The use of formal performance information to improve public sector
performance across the board, including in human resources
management, in strategic planning and budgeting.
Not a concept with a clear generally‐accepted definition. However,
defined in World Bank Public Expenditure Handbook (1998) as “a
whole−of−government strategic policy and expenditure framework …
The MTEF consists of a top−down resource envelope, a bottom−up
estimation of the current and medium−term costs of existing policy
and, ultimately, the matching of these costs with available resources”.
Explicit policy on desired medium‐term outcomes for key fiscal
aggregates (especially deficits and debt)—that is, medium‐term fiscal
targets—possibly supported by approximate medium‐term forecasts of
aggregate expenditure and revenue.
Outcome Changes brought about by public interventions upon individuals, social
structures, or the physical environment.
Outcome variable The characteristics of individuals, society or the physical environment
which public programs seek to change.
Output A good or service provided by an agency to or for an external party.
Performance audit Evaluation of the efficiency and effectiveness of public expenditure
carried out by a supreme audit authority or other group of auditors.
Performance indicator See performance measure.
Performance measure Ratings or quantitative measures which provide information on the
effectiveness and efficiency of public programs.
Performance target Level of measured performance to be achieved by a specific date.
Perverse effects Behavioral distortions or other adverse consequences of performance
Processes The processes by which inputs are transformed into outputs. Same as
Rating Tool (PART)
Technique developed by the US under the Bush presidency in order to
rate program performance for Tool use in the budget process.
Programs rated in five categories ranging from “effective” down to
“ineffective” and “results not demonstrated.”
Program budgeting The systematic use of performance information to inform decisions
about budgetary priorities between competing programs, based on the
program classification of expenditure (see programs).
Program hierarchy Multi‐level program‐based classification of expenditure, in which
programs are broken into one or more lower‐level categories (sub‐
Programs Categories of expenditure based on groups of outputs (or support
services) with a common objective, which is usually an outcome.
Public Service Agreements
System developed in the United Kingdom from the late 1990s until
2004 under which high‐level performance targets are set for each
ministry as part of a biennial spending review process which defines
multi‐year agency budgets. Targets have evolved over time from being
primarily output focused to primarily outcome focused.
Purchaser‐provider Funding systems under which agencies are paid “prices” for systems
the results (usually outputs) which they deliver.
Quality The extent to which the characteristics of an output—in the case of a
service output, the activities delivered and their timeliness— are such
as to increase its potential capacity to achieve its intended outcome.
Not to be confused with the outcome itself.
Spending review The systematic scrutiny of existing expenditure to identify, in
particular, options for cuts. It involves both program reviews (the
review of specific services provided by government) and efficiency
reviews (which focus on reducing the cost of delivering services).
Standardized outputs Outputs where every client receives essentially the same level of
service – that is, outputs which are not characterized by heterogeneity.
Strategic phase (of the
Stage early in a well‐designed budget preparation process where the
political leadership sets the priorities which will guide the process.
Supreme audit institution Generic term used internationally for bodies such as the UK National
Audit Office, the US Government Accountability Office, and the French
Cour des Comptes, which are independent (to varying degrees) of
executive government and designed to hold it to account.
Transfers Shifting of funds between appropriation categories.
Top‐down budgeting The unilateral setting of expenditure ceilings by the center (president,
cabinet, finance ministry etc) without “bids” by the spending ministry.
Virement See transfers.
Zero‐base budgeting A system of performance budgeting in which expenditure is broken
down into, and analyzed in terms of, “decision packages” (also known
as “service increments”) which constitute a series of optional funding
levels from the presumed base of zero to and beyond the current level
of service. Priority rankings are attached to these decision packages,
and these rankings are used to ensure that the available level of
revenue funded those decision packages which are of highest priority.
Allen, R. & D.Tommasi (2001), Managing Public Expenditure: a Reference Book for Transition
Countries, Paris: OECD.
Bevan, G and C. Hood (2006), “What’s Measured is What Matters: Targets and Gaming in the
English Public Health Care System, Public Administration, 84(3).
Danish Ministry of Finance, 2006, Introduction of a Cost‐Based Appropriation System,
Copenhagen: Ministry of Finance.
DFID (2001), Understanding and Reforming Public Expenditure Management, version 1,
Fölscher, A. (2007), Are We Asking the Right Questions? Embedding a Medium‐Term Perspective
in Budgeting. Proceedings of a CABRI seminar on medium‐term budgeting held in Ghana
France (2004) Démarche de Performance : Stratégie, objectifs, indicateurs. Guide
méthodologique pour l’application de la loi organique relative aux lois de finances du 1er août
2001, Paris: Le Ministre d’État, Ministre de l’économie, des finances et de l’industrie, et al.
HM Treasury (2002), Resource Accounting and Budgeting Green Book, London: HM Treasury
Hood, C. (2006), “Gaming in Targetworld: the Targets Approach to Managing British Public
Services”, Public Administration Review, July/August.
Independent Evaluation Group (2008), Public Sector Reform: What Works and Why? Washington: World
Kelman, S. and J. Friedman (2007), Performance Improvement and Performance Dysfunction: An
Empirical Examination Of Impacts Of The Emergency Room Wait‐time Target In The English
National Health Service, Working Paper, Kennedy School of Government, Harvard University.
Potter, B. & J. Diamond (1999) Guidelines for Public Expenditure Management, Washington:
Robinson, M. and J. Brumby (2005) Does Performance Budgeting Work? Washington: IMF.
55 Note: this list is confined to materials cited in the manual, and does not cover the key readings identified for
Schiavo‐Campo, S. & D. Tommasi (1999), Managing government expenditure. Manila: Asian
Sterck, M. & G. Bouckaert (2006), “The impact of performance budgeting on the role of
parliament: a four‐country study”, Paper presented at the 2nd Transatlantic Dialogue, Leuven,