On Improving Public Expenditure Policies for the Poor
Major Informational Requirements

An improvement in the quality of public expenditures is needed in many countries, given binding macroeconomic and fiscal constraints, and poverty reduction and distributional objectives. This process involves a reassessment of methodology used for this purpose by countries and international agencies, and the data needed for improved decision making. This paper reviews methods that might be used by international agencies, such as the Fund and the Bank, jointly with a survey of data currently available, required improvements in the data currently available, and required improvements in the information base. The scope for improvements in governments' expenditure policy making is also considered.


An improvement in the quality of public expenditures is needed in many countries, given binding macroeconomic and fiscal constraints, and poverty reduction and distributional objectives. This process involves a reassessment of methodology used for this purpose by countries and international agencies, and the data needed for improved decision making. This paper reviews methods that might be used by international agencies, such as the Fund and the Bank, jointly with a survey of data currently available, required improvements in the data currently available, and required improvements in the information base. The scope for improvements in governments' expenditure policy making is also considered.

I. Introduction

Sustainable growth and sound macroeconomic policies are a major prerequisite for poverty reduction. Relative price or exchange rate adjustments are often involved in achieving appropriate macroeconomic balances and efficient resource allocation. While some of the existing poor may benefit directly from such measures, it is clearly necessary to protect the vulnerable and low-income groups from sharp real-income losses in the short term. Appropriately designed social safety nets are needed to facilitate this adjustment. 1/ Making the concern for the poor an operational reality for the government involves an assessment of the consequences of public policy choices for households in different circumstances, in addition to the effects on growth, incentives and overall fiscal balance. This approach also generates a need for data and informational flows which presents an opportunity for extremely fruitful inter-agency collaboration concerning both data generation and analysis.

In this note, we concentrate on two major sets of questions. The first relates to how distributional concerns may be incorporated into an assessment of public policy issues. The second highlights the data requirements needed for such an exercise.

In Section II, we briefly describe the methods of analysis, with particular reference to the distributional issues and concerns for the poor that should be central, but which are often ignored. In Section III, we review some of the data available pertaining to selected African countries, juxtaposed against the use that has been made of such data in a variety of contexts. We discuss improvements that are needed with respect to data and information flows in order to improve and clarify the focus on the poor, for government' policy formulation, as well as for the work of international agencies, for example, in the World Bank's Public Expenditure Reviews (PER). Fund staff would typically not address the same level of sectoral detail as in a PER (unless the items have major budgetary or macroeconomic implications, such as with the aggregate wage bill, subsidies and social safety nets) and would focus primarily on the principal tradeoffs given fiscal constraints. Section IV provides concluding remarks.

II. Evaluation Criteria and a Poverty Focus

The consequentialist approach described applies equally to adjustments in revenues or expenditures. The manner in which taxes are changed has to be assessed as much as the reform of public expenditures—such as generalized commodity subsidies. Often if distributional concerns can be effectively met through appropriate expenditure instruments, then taxation measures can be designed primarily with revenue or administrative concerns in mind, the weight of ensuring progressivity falls on the expenditure side of the budget. In this paper, we focus on the expenditure side to motivate the discussion of data and, in particular, on the effects of subsidy reform and price or exchange rate changes.

Improvements in the quality of public expenditures have to be assessed in relation to a number of factors. While “obvious mistakes” should clearly be eliminated, there are few examples of “white elephants” that would leap out in most countries. Thus improvements in the quality of public expenditures would typically involve some difficult tradeoffs. In Section I we discuss the scope for state intervention followed, in Section II, by a brief assessment of the role of aggregate ratios as a practical guide to public policy. We conclude with an indication of directions for reform.

1. Role of the state

A basic question that arises in relation to public expenditures is the role of the state. Should the state tie up scarce administrative resources to produce items which could be easily traded or produced by the private sector, even if there is shown to be a positive net present value to such outlays? Experience with public sector enterprises suggests that operations under civil service employment guarantees, with little threat of bankruptcy or market discipline, backed by government subsidies are not conducive to efficient management. What may be produced or provided by the private sector should be.

Defense and public administration are areas where public expenditures are typically justified. However, major issues relate to the level of the expenditure, as well as its transparency. Administrative expenditures are excessive in some cases, and it may be possible to provide better administrative services at lower cost. Efficient administration and economic management are needed to support other public sector activities. There are also concerns about excessive military expenditures in many countries, although assessments are complicated by difficulties in assessing appropriate levels of military expenditures and also by off-budget funds that are not subject to “normal” scrutiny and auditing. Such expenditures have a direct bearing on what is available for infrastructure or social objectives.

Public sector interventions are likely to be needed particularly in sectors where little private sector activity is feasible, and where there are substantial externalities or public goods. Infrastructure activities generally fall into this category. However, given other competing needs together with overall budgetary constraints, not all infrastructure projects are likely to be justifiable. Further, even if a project proposal is potentially viable, poor management and implementation delays may make for “unproductive expenditures.”

A major justification for public expenditures is to ensure that all citizens have reasonable access to essential goods and social services, particularly in the realms of education, health care, social security and protection against hunger. While private sector provision should not be excluded, it primarily caters for those who can afford to pay. Under these circumstances, the principal aim of public provision in these sectors should be to ensure that the poor and disadvantaged are not excluded.

2. Aggregate indicators: sectoral ratios

There is a temptation to use indicators based on international comparisons, such as the ratio of education or health expenditures to total expenditures, or GDP, as a guide to policy. While there is clearly some merit in knowing about public expenditure patterns and outcomes in different countries and contexts, such a method cannot be the basis for serious policy formulation for at least three major reasons.

The first difficulty is that the requirement for particular types of state expenditures, say, on preventive health care or primary education, depends on the existing level of provision and the needs within a particular context. It may be justifiable to criticize a group of low-income countries for spending 38 percent of the health budget on primary health care and other services, and 62 percent for hospitals. However China spent only about 5 percent of total health expenditures on preventive health care, and still achieved major success in dealing with infectious and parasitic diseases. 1/ As a consequence of its success and the changing nature of morbidity, with a greater requirement of curative care, it is unlikely that the relative proportion of expenditures on preventive health-care will be increased in the future (total health expenditures may rise as a percentage of all expenditures or GDP given the higher unit cost of curative expenditure.)

Similarly, the need for additional resources for primary education has to be related to current levels of provision and not to per capita income levels. An excellent illustration is from India, where the state of Kerala has already achieved universal primary education at a relatively low income level, whereas other states such as Bihar and Uttar Pradesh (at similar income levels) need considerable additional resources to achieve less ambitious goals for primary education. In general there need be no presumption that additional resources for primary education should be found within a “fixed” overall education budget, since secondary and tertiary education 1/ are also needed to raise overall skill levels and to ensure that primary education can be provided. 2/ Thus an examination of the ratio of expenditures on primary relative to tertiary education may not be particularly useful outside a given context.

The second difficulty lies in attempting to adhere to strict norms for ratios of expenditure for each sector. As the size of the government is reduced by fiscal reform, the proportion of funds spent on the social sector may have to increase. Maintaining a large civil service with a smaller public sector would not be justified either. In any case departments and ministries have an automatic tendency to hoard labor at the expense of non-wage recurrent expenditure, resulting in falling pay, low morale, moonlighting, and corruption. Thus, expenditure ratios are likely to change both across and within sectors.

Third, it would be quite surprising if aggregate expenditures, based on broad sectoral criteria or rules of thumb, are consistent with given overall revenue constraints. For example, the rules of thumb could specify absolute amounts for various sectors, including, inter alia, infrastructure requirements, basic administration, social expenditures, and minimized allocations for defense purposes. While broad aggregate norms may reflect some elements of intra-sectoral priorities, they do not provide a basis for choice across sectors in the presence of binding revenue constraints.

3. Directions of reform

Given the need for macroeconomic balances—including price stability—and appropriate relative prices, the reform options can be described as follows. The choice among competing demands for public resources is linked to the overall revenue availability, and the effects of the expenditures on incentives and growth, as well as the effects on households in different circumstances. The theory of policy reform 3/ integrates these elements, and determines welfare-improving directions relative to the status quo, with respect to a set of control policy variables, subject to the constraints posed by public supply and excess demands. In principle, the shadow prices generated by the policy reform problem are the same as those that are generated in the evaluation of projects, and may be defined as the opportunity costs defined with respect to the government's social welfare function. The directions of reform are given by the direct effects on social welfare, less the cost of net extra demands at shadow prices. The procedure yields a ranking of policy options in terms of marginal social costs, and can also be used, inter alia, to evaluate subsidies and tax policy, where these have been described as policy variables. It is noted that actual evaluation of such opportunity costs and policy alternatives is by no means a simple matter.

As was noted by the Development Committee report: 1/

“Both the level and the composition of public expenditures have important implications for poor groups. Many adjusting countries have significant room to reduce public expenditures by reorienting military expenditures, unproductive public investment projects and costly general subsidy programs. A reduction of these outlays would, other things being equal, permit greater outlays for effective and efficient social programs.”

When there is a high degree of concern for the poor, the directions of reform are greatly influenced by the direct impact of a project or policy change on the poor 2/. Under these circumstances, the evaluation of the direct impact on different types of households becomes crucial. In order to assess this impact one needs some measure of the “standard of living.” This could involve every aspect of direct consumption and non-consumption activities including quantification of, among others, the consumption basket of goods and the share of state-provided services received. In practice, the assessment could be based on readily available household expenditure data.

To the extent that there is an effective provision for the poor, to weather adverse policy shocks, through appropriate lump-sum transfers or subsidies, there will be a greater possibility of public acceptance of appropriate macroeconomic reforms. Transfers directed at categorical household characteristics, such as age or disability would be lump sum, as would a ration arrangement. General subsidies are not lump-sum, as individuals can vary purchases, and thus the transfer that accrues, at will. As seen in a number of cases, the rich benefit disproportionately through general subsidies, which can also be expensive in budgetary terms. Reforming subsidies and public sector pricing, in a manner which protects the poor, can in many cases release valuable resources for other public expenditures, as has been seen for example in the case of Jordan. 3/

Further, an appropriate safety net would permit a simpler tax system to be put in place, with less need for major differentiations and exemptions, and this would also have a positive effect on the overall revenue constraints. 1/ It is possible to protect segments of the economy without undermining the overall adjustment efforts by substantial additional spending. However it is important to set up the social safety nets to reinforce and not replace existing institutions such as the extended family and community based informal arrangements for old age, sickness, disability and life-cycle demands.

The issues to be addressed will dictate the data requirements of the exercise. For instance, in the case of subsidy reform, it would be important to know amounts that are spent on major commodities by various income groups, and how these amounts are calculated (budgetary subsidies may be low in the case of an over-valued exchange rate, although low consumer prices make for high effective subsidies). One then needs to relate changes in consumer subsidies (or for public sector enterprises) to price changes, and thus to the direct impact on households in different circumstances, and for this one needs suitably disaggregated information.

A relatively high weight to those at the low end of the expenditure scale would be consistent with a substantive concern for the poor, without necessarily defining the poor in terms of those falling below an arbitrary poverty line. The basic data requirement here then is a relatively disaggregated description of the consumption patterns of various types of households, using household income and expenditure data.

Alternately, raising income opportunities for the poor through infrastructure spending, credit programs, improvements in the functioning of markets, the provision of social services or enhancement of human capital all emphasize somewhat different data needs and requirements. For such interventions, one needs to bear in mind the interdependency of policies and an assessment of the fraction of those expenditures reaching the poor. It is important to be able to describe and utilize information on subsidies and other public expenditures, as well as outcomes. These aspects are discussed in Section III below.

III. Assessment of Data Requirements

As has been noted above, policy recommendations require a data base on household income and expenditure distributions in addition to the standard aggregate indicators of a country's macroeconomic performance. It is of concern then to find even these fundamental macroeconomic data are not available, in a reliable and timely form, in many African countries. Examples of the availability and use of household level data in the African context: are scarce.

The problem is often even worse than an absence of consideration of the impact on the poor; it is not uncommon to be unable to retrieve the most basic aggregate expenditure figures consistent with the GFS format. Table 1 shows the lag involved in the availability of aggregate fiscal figures, it is not uncommon to find data that are five years or more out of date. The Government Finance Statistics (GFS) data are collected with care and according to a well defined format. As Table 2 shows, nine sub-Saharan countries in 1991 had no GFS data, and expenditure data for some was over a decade out of date. It should be noted, however, that Fund and Bank staff rely extensively on data provided to them through consultation and PER missions. These are more up to date than GFS data (see below).

Table 1:

1991 Availability of Basic Fiscal Indicators 1/

article image
Source: African Economic & Financial Data, 1990; and Government Finance Statistics, 1990.

Figures refer to most recent year available in 1991 from either GFS or AEFD source (“G” indicates GFS data otherwise is AEFD), and illustrate the time lag that is to be expected with a standardized format. The 1992 GFS is now available (in Table 2).

Table 2:

Availability of Basic Fiscal Indicators in the GFS 1/

article image
Source: GFS, 1990 and 1992.

Figures refer to most recent year available in 1991, P means data in that year is provisional.

Countries not included in the GFS at all.

Figures based only on 1990 GFS.

Updates were reported for these countries in the 1992 GFS.

Many countries operate without reliable assessments of what is actually being spent. Unfortunately anecdotal recommendations on expenditure policies are not adequate. Without a basic fiscal reporting system, the government invites corruption, wastage and budgetary disorder. External advisors on the other hand face an uphill struggle to have even relatively minor changes ratified and implemented.

Many questions can be asked of the data that is produced. How far is the nexus between producers and users of the information realized? What evidence is there of an attempt at making the reporting framework as transparent as possible? To what extent does the flow of information facilitate efficient monitoring, which is after all a necessary (although not sufficient) prerequisite for control? Is there redundancy, wastage and replication of effort due to misinformation, poor coordination and confusion? The last point is something that should concern both the fiscal authorities and those who provide the data.

“All too frequently missions find themselves covering and recovering the same ground, asking the same questions of the same overstretched officials, and proposing funding studies that cover, to a large extent, the areas of previous studies.” 1/

It is, thus, important for the program designers to be able to use the information that is normally generated by the World Bank (e.g., through its Living Standards Measurement Survey, LSMS, or the Social Dimensions of Adjustment, SDA, programs), the UN Household Survey Capability Program, and by other specialized UN agencies.

The following summary of the primary sources currently available will examine the types of data available in the African context circa 1991. The figures per se are not the important issue here. Collection efforts must look deeper to the data's purpose of painting a comprehensive picture of fiscal activity and communicating the economic state of the country. There is evidence however that much of the collection procedure takes the path of least resistance. It is driven by supply conditions rather than recognizing and accommodating what is required.

1. Aggregate Data

The main sources of aggregate expenditure data that can be utilized for systematic studies are those of the standard national accounts produced domestically and the standard format of the GFS published by the IMF. A number of problems and shortcomings are, however, endemic in the observed data for economic policy design and implementation, and for the international donor community to assess policy.

A basic misgiving is the consistency across sources. Discrepancies are common even between the figures obtained by the country desks from national accounts and those provided for the GFS. 1/ The anomalies, however, are most visible in comparing the publications of different agencies (e.g., World Bank, IMF, or the other UN agencies). Users are faced with the question of which of the available variations to use, they may each lead to very different conclusions. Any choice between them will ultimately be arbitrary.

In some cases the required information is simply not produced. Often, however, the numbers are available, hidden in internal, administrative accounts. Expenditures must at least be catalogued at the department level, even if poor conduits of information prevent the data from flowing back to the central authorities. The disaggregated expenditures, once located and collected, could be extracted and presented in summary form to yield estimates of reported outlays. An example of this form of detective work, tracking down missing data, is the use of government payroll data to analyze civil service pay and employment in Uganda, although aggregate expenditure data are not available. 2/

In preparation of much of the aggregate data, in particular the GFS, a payments or cash basis of accounting is used. Although a preference for measuring, rather than estimating or imputing, gross flows seems desirable, the cash basis also has a number of drawbacks. Many of the government's roles are not clearly captured by mere flows: for example expenditure arrears have an impact that belies the reported current account outlays. Without representation on an accrual basis, one does not have an accurate picture, as would be given by amortization of a balance sheet item, of the flow of services over time from a capital account outlay. Consequently, the task of prioritization of expenditure is attempted in a vacuum. Payment accounting can be more than simply misleading. Arrears can result in overspending due to an inability to foresee future financial obligations, and lead to a subversion of the control function. The Ugandan example suggests 1/ that a focus on flows can lead to budgetary allocations that simply replicate historical shares, rather than looking more closely at past experience, current need and the rents flowing from the resources already in place. In fact, both cash and commitment based data are needed, as well as flow and stock data.

In presenting the data, both functional and economic classifications are desirable and the GFS does in fact attempt to produce both. 2/ In this context, it should be noted that purpose and function are related, but by no means the same thing. Supporting exports by smallholders, for the purpose of assisting the poor, through improving the transport network would be classified as transport and communication expenditures, while if a drainage or irrigation scheme is implemented for the same purpose, the classification can be agricultural Land management, multipurpose development projects or even water supply. Thus considerable “detective work” is involved in assessing the effects and consequences of particular projects or policy options. However, classification by purpose may prove subjective.

There often seems to be some difficulty encountered in drawing the line between maintenance and rehabilitation expenditures. This, in particular, is evidenced by the frequent complaints of insufficient funds to meet the operations and maintenance costs generated by public investment programs. Such programs all too often are viewed as one-time capital disbursements without explicit forecasts of the accompanying recurrent funding needs of projects. Wastage results then from underutilization and deterioration of capital projects; schools have no books or teachers, and hospitals are unable to treat patients without medicines or supplies.

One of the more conceptual problems of reporting government expenditure is, broadly speaking, how one values goods and services which, either by their nature have no market valuation, or for which the market price bears no resemblance to the social value attached to their provision. Thus cost-benefit analysis on public goods such as national defense expenditures is difficult to conduct. The World Bank's Public Expenditure Reviews do not generally address defense expenditures, although a few have included a forceful recommendation, as in the Ugandan PER 3/, for the government to investigate whether defense expenditure can be controlled and security provided more cost effectively. A basic problem with cataloguing all military expenditures remains in many countries.

It is also difficult to assess price and other externalities. For example it was noted of the Ghana Bauxite Company that the company's main constraint was not mining capacity but the inability of the railroad to carry additional bauxite. 1/ In this case, to determine the appropriate allocation to transport, one should also bear in mind the secondary effects on the country's industrial sector; the cross-sectoral synergies. Similarly, there are the external effects of education and health care. Here the individual costs and benefits do not reflect the value of human capital enhancement as a catalyst for increased aggregate growth potential. These questions have no easy answers but should certainly not be ignored.

It is important, for a balanced view of the fiscal program, to take into account expenditures undertaken at various tiers of government; concentrating solely on consolidated central government figures can distort the picture. For instance, perhaps the bulk of primary education and health care is carried out on a local basis, to ignore it may paint a pessimistic picture of health allocations. Often local authorities are a major source of wastage. In the case of Mozambique, which is by no means unique, local authorities face a lack of resources and definition of responsibilities due to a haphazard evolution of their role with reference to available resources. They are inefficient and absorb large sums in subsidies, without showing a reciprocal degree of accountability, either politically or in terms of reporting their activities to the central government.

The GFS tries to provide, where possible, a breakdown of activity at the local and state level, but often is unable to do so, since the local bodies do not produce any kind of timely data. As Table 3 indicates, the GFS has very little in the way of expenditures carried out at various levels of government in the SPA group of countries. For some sub-Saharan African countries this is not important as budgets are decided and recorded by the central authorities. However, for many countries to omit local or state allocations would give an outlook very different from reality. Finally one of the more basic capacities that is required of aggregate fiscal statistics is comparability. This need for comparison is both inter-temporal and cross-sectional. The desire to track the passage of an economy through time is important, in order to identify errors, ex-post, and to obtain parameters within which future policy recommendations can be set. One of the main advantages of the GFS data is precisely this facility for time series comparison, as well as for international comparability.

Table 3:

Classification Details in GFS Data

article image
Source: Government Financial Statistics Yearbook, Volume XIV, (1990)Key: 1 data coverage0 no data coverageP provisional figures available

General refers to the presentation of consolidated accounts for central, state, local and supranational levels of government in the years the central accounts are provided.

No data for defense in Functional Classification of Expenditure table.

A reliable flow of aggregate data and the resultant ability to manage comes from an efficient government accounting system that is understandable and accurate. Such a system comes from careful design and is enforced by capable audits. Without such accounting, poor institutional capability leads to, first, bad budgetary allocations. Evidence for this tends to be widespread, for example in Madagascar in 1991 only 1.5 percent of public spending was allocated to food security, job training and social action, despite a majority of the population living in poverty. 2/ Second, without an effective reporting infrastructure observers have to contend with “creative accounting” and concealment of information. In Mozambique, as in many other countries, public enterprise subsidies were channeled as loans through the banking system rather than showing up as direct elements of the budget. 3/

Government's stated aims are often not reflected in actual allocations. In addition to poor information flows, this is a direct effect of poor monitoring and control facilities. In such cases, the budget takes a life of its own, and it is difficult to dislodge historical patterns of spending. The Fund is actively involved in providing technical assistance support in relation to budgetary monitoring and control as well as statistical data bases for the major macroeconoraic and budgetary items. However, detailed information on household budgets, labor markets and so on are not part of the Fund's data base per se and other international agencies have a comparative advantage in generating such information.

2. Disaggregated data

In order to bridge the gulf between macroeconomic constraints and microeconomic implementation of government outlays, one must be aware of the incidence of expenditure policies. Knowledge of aggregate health and military expenditures cannot reveal whether amounts spent on ambulances are really used to purchase military vehicles. Only alternative sources revealing the impact of spending can provide this information. The question of the gainers and losers from policy options should dominate the whole expenditure discussion. As mentioned above, data sources such as the GFS do not begin to address these issues, and there has to be a reliance on household survey and firm level data. There is an effort by the World Bank and other agencies to assist in generating the required data for most countries where the statistical base has been poor, and such data should be sought by Fund staff as needed.

Often, the a priori orientation of public policy evaluations allows macroeconomic elements to dominate with insufficient attention to the distributional issues that accompany adjustment. While growth is perceived as the most important determinant of welfare change, inadequate attention is paid to how this growth is to be achieved and the possible short-term costs and tradeoffs involved. There is insufficient thought given to the real impact of policies: who and where are the poor; how are they affected by fiscal policies and subsidy reductions; and how are they likely to fare under future changes.

Having identified the need for a more disaggregated data set, what properties should this information possess? First, the data should pick out the effects of a particular program from shifts in states of nature and, in particular, predict the impact of potential policies on various groups of the population—-such as by deciles, rural and urban households, and so on. The states of nature need to be taken into account when deciding the expenditure profile of a country; there is little point in building expensive dams or communication links in a a country in the throes of civil war.

Second, since the poor are not a homogenous group within a country, one should be able to attempt to identify a proxy for their socioeconomic characteristics. Once identified, policies can be designed to assist the typical profile of those in poverty. This form of characteristic targeting can be a powerful tool, but presupposes an adequate, disaggregated information set. Targeting transfers to individuals below a particular per capita income or asset level, such as with explicit means testing, is rather more demanding in terms of data requirements, is costly and open to abuse, and requires an administrative capacity that may not always be available.1/

Third, in designing policies aimed at changing the characteristics of the poor (e.g., job training programs, better health care, improved productivity through the introduction of new agricultural techniques…) national and international agencies must first pinpoint those initial shortcomings. This is closely allied to the needs outlined above. A coordinated flow of socioeconomic indicators is an essential prerequisite to designing policies aimed at efficient resource allocation.

Fourth, the figures should be able to deal with an analysis of the effects of changing relative prices, either from shifts in the terms of trade or direct fiscal intervention by taxes and subsidies. To obtain estimates of precise changes in welfare, as proxied by Hicksian compensating and equivalent variations, may not be feasible. However, an assessment of the direct impact of policy changes on the poor is quite feasible, and can be attempted within the context of a mission and its time constraints, as seen in the Jordanian example. 2/

Finally, the disaggregated data should pinpoint deficiencies in organizational and institutional areas in order that technical assistance can be provided in the most effective manner. The availability of this disaggregated data is imperative if policy attempts are to be monitored at the household level and questions are to be raised as to where policy implementation appears weakest.

In summary, the bridge between macroeconomic and microeconomic analysis needs to be built on quantifiable foundations involving disaggregated statistics (from household surveys to labor market, education and health data). These statistics should facilitate incidence analysis, assist policy design in pursuing equity in the most efficient manner, identify certain social groups, permit monitoring and control of spending (in particular, offer evidence of diverted resources, bottlenecks, corruption, and so on) and to mitigate the possible adverse effects of adjustment. The aim must be to design poverty-conscious social development and public expenditure reconstruction programs and offer supporting evidence to ensure their ratification.

a. Household survey data

Concentrating first on household income and expenditure data, one can highlight a number of ex ante considerations that must be tackled. The most useful format should be used in presenting the data, the commodity classifications and the breakdown of prices and quantities, and the trade-off of detail and cost (since the data is by no means costless). The format for household data should balance excessive detail, which can be difficult to interpret, and too much aggregation, which may obscure vital information.

There are two breakdowns to consider, the household group decomposition and the commodity classification. The former attempts to construct groupings, neither mutually exclusive nor collectively exhaustive, to supply the most appropriate conditional representation of the empirical joint distribution contained in the survey data. The latter then paints the picture of the tastes, endowments and characteristics of each particular grouping. As a general guideline one should aim for the households within a division to be relatively homogenous, maximizing inter-group variations while minimizing intra-group deviations. The categorization should lend itself to the identification of particular “groups” known to be vulnerable and subject to poverty, such as those unsupported by an extended family network, rural small-holders, female-headed households or the unemployed. Boxes 1 and 2 identify the more useful decompositions that can be demanded of survey data.

It is important to highlight some of the difficulties encountered with this type of disaggregated data. One should be aware of a loss of intra-household detail in taking the household as the basis of measure. One cannot identify those unsupported by an extended family network, which poses a problem since often poverty is inextricably linked with a lack of kin or friends. Identification of individual-specific features may be impossible.

To display the data by income deciles may prove misleading, with income traditionally being unreliable and often downwardly biased. Per capita expenditures may be more appropriate, often reflecting “permanent income,” and being less subject to reporting biases.

It would be a mistake to say though that the problem lies solely in the sketchy availability of data. In making policy recommendations, few of the Bank's Public Expenditure Reviews (with some notable exceptions) have made attempts to use the household data available. In general, the demand and potential use for household data such as that generated by the LSMS and SDA programs is evident. The gap between provision and demands however remains

b. Information on characteristics of poverty

The availability of additional data sets is summarized in Table 4. The FAO is probably the most comprehensive source for agricultural statistics 1/. With agriculture being fundamental to the economic well-being of SPA countries, accurate and informative data is very important. The FAO produces, among other things, breakdowns of irrigation, land usage, production by crop and livestock, methods of production, use of pesticides, machinery and fertilizer and food supplies (by nutrients). From this information one can ascertain much about the rural sector to complement the sectoral reports. Bottlenecks or unrealized productivity potential may be dealt with by fiscal incentives and one might even draw international parallels from the uniformly presented data.

Household Categorization

  1. Individual level characteristics: examine the position of certain target groups such as women, the old, the very young and other dependent groups that prevail amongst the poor.

  2. Household characteristics: total income or expenditure (broken down by fractile) or classification according to characteristics of household head.

  3. Factors of production or sources of income:

    • (a) rural small-holder perhaps subdivided into domestic and export crop farmers;

    • (b) landless agricultural worker, dividing into estate workers, tenants and other;

    • (c) formal government sector;

    • (d) formal private sector;

    • (e) non agricultural self-employed;

    • (f) informal sector;

    • (g) the unemployed;

    • (h) other.

  4. Locational divisions:

    • (a) geographical area (e.g., north, south, …);

    • (b) climatic zone;

    • rural-urban classification distinguishing between the capital, provincial centers others;

    • (c) main agricultural land usage.

Productive Inflows:

  1. Wage income

    • (a) public sector

    • (b) private sector

    • (c) self-employed income

  2. Gross and net agricultural income broken down

    • (a) by main foodstuffs (wheat, rice etc.)

    • (b) into tradeables and non-tradeables

  3. Income from capital and interest (esp. farm assets)

  4. Government transfers and social security

  5. As far as is possible imputed rental income and service flows from other durables

  6. Estimated other income—likely to include informal sector, windfall transfers etc.

  7. Divisions by fractile

Expenditure Outflows

  1. Food spending

    • (a) by main foodstuffs (both prices and quantity)

    • (b) tradeables and nontradeables

  2. Clothing and footwear

  3. Recreation and entertainment

  4. Fuel - by type and usage

  5. Housing and household facilities

  6. Transportation

  7. Medical

  8. Education

  9. Durable consumer goods

  10. Capital and investment goods

  11. Divisions by fractile

Table 4:

Breakdown of Data Available from Main Alternate Sources

article image
article image

Most sources cover all countries but the LDC and African publications are, obviously, limited to a subset of countries.

In Ghana agriculture accounts for about half of GDP, 75 percent of foreign exchange earnings, a substantial part of fiscal revenue and employs two-thirds of the labor force. As a result, directing public expenditure to infrastructure and support services for this sector, together with appropriate exchange rates and agricultural pricing policies, are methods of reducing poverty and easing internal and external imbalances. Improving investment, research and introduction of new farming methods and small-scale irrigation projects can only be done efficiently if one can identify the current shortfalls and areas with the prospects for the highest returns. This in turn requires a timely, relevant data base. Details from microsurveys (e.g., from IFAD) may be particularly useful on aspects of income sources of the poor.

With the main resource of the poor being their own human capital, assessments of broad employment groups is often helpful in dealing with those in poverty. Labor market indicators of employment, hours worked and wages by level, industry, sex, age and occupational grouping are presented by the ILO 1/. This can give a view of labor's participation, returns and share of value added. The unemployed are often those that suffer most from tough adjustment measures, it is the poor who are hit first and hardest by downturns in the labor market. Consequently monitoring and an appreciation of labor market conditions is invaluable; policy proposals need to be set within the context of this information. However, one of the major sources of information on transfers, The Cost of Social Security, published under ILO auspices, was in 1991 only available for the 1981–83 period and is out of date. More recent input is needed for effective policy formulation.

The public sector wage bill is something that should be a direct concern to the fiscal authorities. The civil service is often one area where substantial savings can be made if a duplication of functions and resource wastage can be avoided. In Uganda civil service registration and crosschecking led to the elimination of 20,000 ghost employees, a saving of U Sh. 100 million per month. In many countries, corruption, overstaffing, low pay and lack of motivation are serious problems. Civil service reviews and the introduction of more efficient payroll systems could yield results within a very short period. Also the hardship associated with civil service retrenchment can only be mitigated with well-conceived compensatory measures. Details that can be made available (e.g., through the ILO or the World Bank) would be particularly useful to the Fund.

In connection with the importance of human capital the social sectors are always, justifiably, areas for attention. Health and education are important in enhancing and improving the status of the population in order that they might fully take advantage of opportunities offered by general improvements in the economic situation. UNESCO 2/ provides detailed coverage on all aspects of education. There are descriptions of the systems in place, school enrollment by education level, age and sex, pupil-teacher ratios and coverage of book, film and radio demand supply and trade. This may go some way to see if there are books, and indeed students, in the schools catalogued, solely from a capital aspect, in government PIPs.

The Ugandan education sector report highlights the “chronic lack of provision of recurrent funds in education for salaries, materials and books” as being one of the biggest handicaps to human capital enhancement. Similar stories can be told in many SPA countries. There are disturbingly repetitive themes of inadequate provision of primary education services; the same goes for health care. Poor information and deficient prioritization however leads to systematic misallocation of expenditure.

Health and nutrition data is probably one of the most important indicators of social well-being and development. UNICEF 1/ and WHO 2/ provide comprehensive coverage of the relevant indicators. Access to water, health services and immunization, child and infant mortality, incidence of low birth weight and malnutrition, the distribution of health care and doctors, causes of death and frequency of disease by age and sex, calorie, cholesterol and nutritional intake and other average vital statistics are all reported in a standard format. This can tell the user where the malnourished are, and who are the groups that have inadequate facilities for a basic, healthy standard of living.

These sources are all potential alternatives that could be used in an assessment of policies for the poor. Each source should be viewed not in isolation but as a component of the information that allows an economist to paint a detailed picture of the situation in a particular country, not just the outline provided by government produced figures. The use of these sources requires a little detective work in establishing links, causes and most importantly effects. The effect of fiscal policy may be traced through a combination of many of the available sources. The status and attainment of particular groups, children, women or the old are of particular interest bearing in mind their proven vulnerability in many SPA countries. For example, if WHO data identifies rural smallholders as those suffering most from malnutrition and high infant mortality, while UNESCO corroborates that the same socioeconomic types have little access to education, then one can begin to frame a picture of precisely who are those in poverty. Perhaps FAO production figures will establish the production and exports of products by these groups. One can take advantage of these and other figures to give a quantitative guide of the cost and effect of appropriate exchange rates and pricing policies for the rural smallholders and other groups. This process of identification, corroboration, policy suggestion, and objective impact quantification (to pick out the most attractive alternative) can be expedited by not one but a synthesis of multiple sources of information.

IV. Concluding Remarks

Improvements in evaluation criteria could be based on the principles of policy reform, that utilize the information generated in the detailed sectoral analyses, subject to overall revenue constraints, and explicitly placing a high weight to the welfare of the poor. Aggregate expenditure ratios based on international experience are of indicative usefulness, but cannot be used to provide operational assessments.

While the collection of detailed household data needs to be tailored to what a country can reasonably afford, the information needed to incorporate distributional evaluations can be fairly cheaply obtained. Improvements in data collection and utilization are essential as part of improved public sector management. Additional information collected by the World Bank and other UN agencies in identifying vulnerable groups and their characteristics would be invaluable for the Fund's work in assessing poverty questions. Fund staff would utilize the information from diverse sources to identify major fiscal and distributional issues to assess the scope for sustainable growth with equity in a given country.


  • Ahmad, E. (1991), “Jordan: Reforming Public Expenditures and Protecting the Poor”, processed, International Monetary Fund.

  • Ahmad, E., (1991a), “Social Safety Nets in Transition Economies”, in V. Tanzi (ed.), Fiscal Issues in Economies in Transition: The Institutional Dimension, International Monetary Fund, Washington D.C.

    • Search Google Scholar
    • Export Citation
  • Ahmad, E., and Stern, N.H. (1990), “Tax Reform and Shadow Prices for Pakistan”, Oxford Economic Papers, 42, 135159.

  • Ahmad, E., and Stern, N.H. (1991), The Theory and Practice of Tax Reform in Developing Countries, Cambridge University Press.

  • Ahmad, E., J. Drèze, J. Hills and A.K. Sen, eds (1991), Social Security in Developing Countries, Clarendon Press: Oxford.

  • Development Committee Report (1991), “Development Issues: Presentations to the 39th Meeting of the Development Committee”, Report No. 26, World Bank, Washington D.C.

    • Search Google Scholar
    • Export Citation
  • Drèze, J. and Sen, A.K. (1991), “Public Action for Social Security: Foundations and Strategy”, in E. Ahmad et al. (1991).

  • Drèze, J. and Sen, A.K., and Stern, N.H. (1987), “The Theory of Cost-Benefit Analysis”, in A. Auerbach and M. Feldstein (eds.), Handbook of Public Economics: Vol II, Amsterdam: North Holland.

    • Search Google Scholar
    • Export Citation
  • Food and Agriculture Organization of the United Nations (1990), “FAO Production Yearbook”, Rome.

  • Glewwe, P. (1990), “Improving Data on Poverty in the Third World”, The World Bank, Living Standards Measurement Survey, Working Paper No. 416, Washington D.C.

    • Search Google Scholar
    • Export Citation
  • International Labor Office (1990), “Year Book of Labor Statistics”, Geneva.

  • International Labor Office, (1991) Cost of Social Security, Geneva.

  • International Monetary Fund (1986), “A Manual on Government Financial Statistics”, Washington D.C.

  • Tait, A. (1988), Value Added Tax: International Practice and Problems, Washington D.C. International Monetary Fund.

  • UNICEF (1991), “The State of the World's Children”, Oxford University Press.

  • United Nations Development Program (1990), “The Social Dimensions of Adjustment (SDA) Project: An Interim Evaluation”, New York.

  • United Nations Development Program, (1991), “Proposals for a Program on Structural Adjustment and Social Development in Africa”, Memorandum.

    • Search Google Scholar
    • Export Citation
  • World Bank (1989), “The Mozambique Public Expenditure Review”, Report No. 7615-MOZ, Washington D.C.

  • World Bank (1990), “Ghana: Public Expenditure Review 1990–92”, White Paper, Washington D.C.

  • World Bank (1990a), “Malawi: Growth Through Poverty Reduction”, Report No. 8140-MAI, Washington D.C.

  • World Bank (1990b), “Malawi Public Expenditure Review”, Report No. 7281-MAI, Washington D.C.

  • World Bank (1991), “Jordan Public Expenditure Review”, Report No. 8986-JO, Washington D.C.

  • World Bank (1991a), “Public Choices for Private Initiatives: Prioritising Public Expenditures for Sustainable and Equitable Growth in Uganda”, Report No. 9203-UG, Washington D.C.

    • Search Google Scholar
    • Export Citation
  • World Bank (1991b), “The Madagascar Public Expenditure Review”, White Paper, Washington D.C.

  • World Health Organization (1990), “World Health Statistics Annual”, Geneva.


The views expressed in this paper are those of the authors and should not be ascribed to the International Monetary Fund. This paper is based on work by the authors to review Bank Public Expenditure Reviews for the Special Program of Assistance for Africa in 1991. It was presented to an International Workshop on Poverty Monitoring in International Agencies, held under the aegis of the ILO and UNICEF in Santiago, Chile, in September 1991. Mr. Chalk, a doctoral candidate at the University of California (Los Angeles), was a summer intern in 1991 in the Fiscal Affairs Department. Helpful comments from Ke-young Chu and assistance from Tarja Papavassiliou are acknowledged, but all errors are ours.


This position, taken by Fund staff, is reported in “The Fund and Poverty Issues: A Progress Report,” published in a Development Committee pamphlet 26 in 1991. For a discussion of the World Bank approach, where safety nets are seen as an adjunct to a two-pronged strategy encompassing investments in human and physical capital and labor-intensive growth, see World Development Report (1990), also summarized for the Development Committee pamphlet 26.


Fixed costs tend to be higher for secondary and tertiary education than primary education, which ensures that the former will be more expensive than the latter. Similar cost patterns hold with respect to health care. There may however be considerable waste in connection with all types of education and health care … such as primary schools without books and universities producing unemployable lawyers; or primary health clinics without personnel, and hospitals without drugs but with fancy equipment that cannot be used.


For example, educated teachers are needed to staff primary schools.


The Development Committee, # 26, (1990), p. 30.


See E. Ahmad and N.H. Stern (1991) for a theoretical discussion as well as empirical applications, and A. Tait (1988) for a description of a rate structure of indirect taxes which would be desirable from an administrative point of view.


Based on Fund missions for Article IV Consultations and Use of Fund Resources. Fund Recent Economic Development reports collect and report fairly detailed and up-to-date fiscal information, although formats tend to vary and are often less detailed than the GFS requirements. This information is not publicly available.


See International Monetary Fund (1986).


See Ahmad (1991a) for a discussion of some of the options.


See “FAO Production Yearbook” and individual country profiles.


World Bank 1991a. See e.g., ILO, 1990.


See the UNESCO statistical yearbook.


See e.g., UNICEF (1991).


WHO (1990).

On Improving Public Expenditure Policies for the Poor: Major Informational Requirements
Author: Mr. Nigel A Chalk and Mr. Ehtisham Ahmad