Benefits on the Margin: Observations on Marginal Benefit Incidence

Benefit incidence analysis has become a popular tool over the past decade, especially for researchers at the World Bank. Despite or perhaps because of the popularity of this method, recent research has pointed out many of its limitations. One of the most com- mon criticisms of benefit incidence analysis is that its description of average participa- tion rates is not necessarily useful in guiding marginal changes in public spending policies. This article considers a variety of methods for analyzing the marginal benefit incidence of policy changes. A key conceptual point is that despite the fact that the various methods measure "marginal" incidence, they do not measure the same thing— nor are they intended to do so. There are many possible policy changes and thus many margins of interest. Each method captures one of these and so is of interest for some analyses and inappropriate for others. Empirically, the precision of the methods dif- fers substantially, with those relying on differenced data or aggregations of households yielding standard errors that are quite large relative to the estimated shares. The past decade has seen a resurgence of interest in the relationship between poverty and public spending in developing economies. This resurgence has fos- tered the return of incidence analysis, particularly for the benefits of public spend- ing in the social sectors. Although analysis of tax incidence has a long and venerable history in economics, distributional analysis of the benefits of public spending—and public policy more generally—is more recent (Aaron and McGuire 1970; Brennan 1976; Meerman 1979; Selowsky 1979). Broadly stated, benefit incidence analysis assesses how the benefits of government spending are distrib- uted across the population. Though there are many ways to approach this issue, a fairly standard method has emerged, largely based on the work of researchers at the World Bank (Demery 1997; van de Walle and Nead 1995; Selden and Wasylenko 1992). This method takes "across the population" to mean "across the expenditure (or income) dis-