How Does the Composition of Public Spending Matter?

Abstract

Public spending has effects on growth and distribution that are complex to trace and difficult to quantify. But the composition of public expenditure has become the key instrument by which development agencies seek to promote economic development. In recent years, the development assistance to heavily indebted poor countries has been made conditional on increased expenditure on categories that are thought to be “pro-poor”. This paper investigates the conceptual foundations and the empirical basis for the belief that poverty can be reduced through targeted public spending. While it is widely accepted that growth and redistribution are important sources of reduction in absolute poverty, a review of the literature confirms the lack of an appropriate theoretical framework for assessing the impact of public spending on growth as well as poverty. The dangers of policy decisions that are not well grounded in theory and supported by empirical evidence are indicated. With regard to the impact of any given type of public spending, policy recommendations must be tailored to countries and be based on empirical analysis that takes account of the lags and leads in their effects on equity and growth and ultimately on poverty. The paper sketches out such a framework and provides some evidence as the first step in what will have to be a longer-term research agenda to provide theoretically and empirically robust and verifiable guidance to public spending policy.

Notes

¹ The concept of human development (introduced by Mahbub al Haq and UNDP in the 1990s) was used to make the case that economic growth was the means to achieve human development, measured by a composite index based on life expectancy, adult literacy, school enrolment and GDP per capita. In 1994 a “human development 20: 20 compact” was proposed to earmark 20% of aid flows and 20% of government budgets to basic social services. This compact may have laid the groundwork for the subsequent HIPC emphasis on public spending on health and education as key to human development and, by association, with poverty reduction (see United Nations Development Programme (1996) Human Development Report 1996, Economic growth and Human Development (New York: United Nations Development Programme).

² The “decision point” refers to the stage when the Executive Boards of both the IMF and the World Bank formally decide on a country's eligibility for debt relief. At this stage, the international community also commits to reducing the country's debt to the sustainable threshold.

³ Foster et al. (2002) reviewed the experience with pro-poor budgeting in five sub-Saharan African countries and observed that the criteria for what constitutes poverty-reducing expenditure have generally focused on direct benefits to the poor. They suggested that the general tendency of equating poverty reduction with social service provision needs to be seen in historical context as a compensation for previous neglect of services to the poor.

⁴ A considerable amount of intellectual effort is consequently directed at debating and clarifying issues that would be better spent on real problems.

⁵ We are aware of the growing literature analysing how electoral politics, imperfect information and political incentives influence public spending choices. Some examples include Keefer & Khemani (2005), Mauro (1998), Milesi-Ferretti et al. (2001), Brender & Drazen (2005) and Robinson & Torvik (2005). There is less literature on the equally important incentives and moral hazards created by donors through aid, although Easterly (2003) provides a broad discussion. While these approaches are promising, the issue is far from settled. The focus of this paper is, however, on the more conventional economic explanations that complement the political economy explanations.

⁶ Perotti (2000) also noted the absence of a coherent model for evaluating the trade-offs between different expenditure programmes.

⁷ See Stiglitz (2000) for the six conditions when markets fail and create a rationale for public intervention. Devarajan et al. (1996) observed that “neither economic theory nor empirical evidence provides clear-cut answers to how the composition of expenditure affects economic growth”. It might be added that neither does theory indicate clearly how expenditure affects equity or poverty.

⁸ See, for example, Devarajan et al. (2001) and Pradhan (1996), as well as guidelines from the International Monetary Fund (IMF), the Asian Development Bank (ADB) and the Organization for Economic Co-operation and Development (OECD).

⁹ For example, while staff members of international organizations have recently developed practical guidelines for expenditure analysis, the theory of public expenditure has, as is well known, a very long history. Musgrave (1985) traced notable contributions in the development of principles of expenditure allocation, including those of Lindahl (1919) and Pigou (1928) in the application of the theory of marginal utility to government expenditures, Samuelson's (1954 a,b, 1955) path-breaking linkage of externalities and social goods and integration of social goods into principles of efficiency, and the development of cost–benefit analysis in the 1960s. Walker (1930) provided one of the first surveys of the theory of public expenditures.

¹⁰ A more detailed review of the recent literature appears in Appendix 2. See also Fozzard (2001).

¹¹ The World Bank's guidelines on public expenditures note that the rationale for public intervention could be either market failure (public goods, externalities) or redistribution. If the former, it suggests obtaining quantitative estimates for the degree of market failure to supplement arguments based on first principles. If the rationale is redistribution, it recommends analysis of the incidence of public expenditures.

¹² April 2002 Symposium, Journal of Public Economics, 86, pp. 311–360.

¹³ Budgets can be described using an economic classification, or a functional classification, or an administrative classification, or a programme classification. Atkinson & van den Noord (2001) also apply a classification that is consistent with a public economics approach, dividing up the budget into expenditure associated with pure public goods, merit goods, income transfers and economic services.

¹⁴ Dani Rodrik (2005) has argued that successful economies have employed pragmatic and non-unique institutional arrangements to achieve “first-order economic principles”—protection of property rights, market-based competition, appropriate incentives, sound money, etc. Thus, the appropriate role for the public sector and the composition of public spending is itself an issue of development strategy that must be defined by country circumstances and capabilities and is not uniquely predetermined. This framework would support a more creative, less doctrinaire, but ultimately more empirical basis for policy analysis and advice.

¹⁵ We recognize the problems introduced by this taxonomy of public spending. We explored other taxonomies, but found them even more unsatisfactory. For example, some distinguish between investment in “services” and investment in “growth.” Ferroni and Kanbur distinguish between social expenditure and “productive” expenditure. There are of course clear reasons to expect social spending to be just as productive as the so-called productive expenditures.

¹⁶ Other expenditure would include general administration expenses as well as law and order and security-related expenditure.

¹⁷ Some have argued that B is also a function of income inequality (I); thus, B = f(S, K, O, Y, I). In particular, some papers have provided evidence that more equal societies are healthier (Wilkinson, 1992; Kawachi et al., 1999). However, Deaton (2003) has been critical of this literature. The main conclusions that follow from our framework hold, regardless of whether or not social indicators are a function of income inequality.

¹⁸ We distinguish between the impact of S and B, to allow for the direct impact of selected social expenditure items (such as transfers) on I that is distinct from the impact of B on I in the medium to long term (e.g. reduction in income inequality via human capital formation among the poor).

¹⁹ This effort, a collaborative effort led by the World Bank, also known as the “Operationalizing Pro-poor Growth Research Program”, is described in detail in World Bank (2005). The studies are posted on the World Bank's web site: http://www.worldbank.org/poverty. See also Wilhelm & Fiestas (2005) for a summary of the case studies. It should be noted that while many of these country case studies provide insights into the links between expenditure composition, economic growth and poverty, many of their empirical frameworks suffer the same shortcomings as described in Section 3: the tendency to equate social spending with pro-poor spending; a heavy reliance on static or dynamic benefit incidence analysis; the failure to account for immediate and lagged effects; direct and indirect effects; and complementarities. Nevertheless, the cases provide some empirical evidence of the trade-offs in public spending.

²⁰ See Heller & Gupta (2002) for a review of these macroeconomic issues.

²¹ On the other hand, the expansion of education may be associated with greater wage inequality; for example, expanding the supply of primary education may lead to lower returns to primary education for older workers. A recent study by Duflo (2002) of education expansion in Indonesia through a large programme of school construction (the Sekolah Dasar INPRES programme) initiated in the 1970s found that an increase in the proportion of primary school graduates in the labour force decreased wages of older cohorts. Duflo observed that in Indonesia's case, physical capital did not adjust to the increases in human capital in the regions where schools were built. She was unable to explain why the stock of physical capital failed to adjust despite the public announcement of the programme and its gradual implementation over 10 years.

²² These statements of effects are meant to be illustrative and not intended to overstate the extent to which policy can determine such outcomes.

²³ There will be wealth effects due to increase in the value of land, due, in turn, to better infrastructure. The wealth effects will accrue largely to the landowning groups. This is likely to increase inequality in the distribution of wealth.

²⁴ Equivalently, if a vector of income quintiles represents income distribution, the scheme increases the income of the lowest quintile, Y ₅, while reducing the income of the top quintile, Y ₁, thus reducing the poverty headcount P.

²⁵ While the proposed framework uses a broad characterization of “social” and “productive” expenditure, it would work equally well with a more detailed functional classification of expenditure.

²⁶ Appendix 3 provides a brief description of various tools and techniques that can be used for the analysis of public expenditure links with poverty reduction. They can be categorized into five groups: (i) reduced form regressions; (ii) general equilibrium models; (iii) investment appraisal methods; (iv) incidence analysis; (v) other quantitive tools.

²⁷ Literature comparing the relative efficiency of redistributive expenditure policy and redistributive tax policy is rare (see Chu et al., 2004). Warr (2003) is one exception. He compared the impact of expenditure reform with tax reform on poverty. He also considered the likely significant effects of simultaneous changes in both the composition of taxes and the composition of revenues.

²⁸ Admittedly, there is yet further work that investigates other important facets of the link between public expenditure and poverty. As it is not the objective of this paper to give a complete account of all this work, we are not discussing this further. None the less, there is an abundant literature on the role of governance on the effectiveness of public expenditure.

²⁹ There are many other examples of this type of complementarity documented in the literature. For example, Deininger & Okidi (2003) used micro-level survey and panel-data evidence from Uganda spanning 1992–2000 and found that the benefits of education and health care for growth and poverty reduction depend on complementary investments in electricity and other infrastructure, and reductions in civil strife.

³⁰ See, for example, Toye & Jackson (1996).

³¹ See Piggott and Whalley (1987) on the improper measurement of consumer surplus. In addition, as typically implemented, net fiscal incidence analysis focuses only on the static incidence, not on net benefits accumulated over time (see, e.g. World Bank, 1990).

³² Devarajan et al. (1996), using data on 29 countries over the period 1970–90, suggested that public capital spending has a negative but insignificant effect on growth, while current spending has a positive and significant impact on growth. They also found that transport and communication have a negative and significant impact on growth, while health and education have a negative but insignificant impact on growth. In contrast, a study of 39 low-income countries over the period 1990–2001 found that higher public expenditure on capital raises economic growth (Gupta et al., 2005). A recent study found that investment in education is the only expenditure outlay significantly associated with growth in a sample of 30 developing countries (Bose et al., 2003).

³³ Mosley et al. (2004) estimated the correlation between expenditure components and poverty using cross-country data. Based on this exercise, they calculated a “pro-poor expenditure” (PPE) index or the ratio to GDP of spending components associated with poverty. In addition, results from recent cross-country regressions have been used to quantify the poverty elasticity of growth.

Additional information

Notes on contributors

Erwin R. Tiongson

We received valuable comments and suggestions from: Sudharshan Canagarajah, Shanta Devarajan, Lionel Demery, Alan Gelb, Maureen Lewis, Humberto Lopez, Sanjay Pradhan, Sudhir Shetty, Vera Wilhelm, participants at the 2004 PREM Learning Week session, “Pro-poor Expenditures: A Debate” (4 May 2004), participants and discussants at the ODI workshop, “Current Issues and Debates in Aid & Public Expenditure Management” (13–14 May 2004) and the Banca d'Italia Public Finance Workshop (31 March–2 April 2005), and two anonymous referees. The usual disclaimer applies.