在没有执行对穷人的拨款分散化的情况下，进行目标绩效监管

POLICY RESEARCH WORKING PAPER 2080 Monitoring Targeting Program funding and design choices by the central Performance When government can greatlyaffect Decentralized Allocations the targeting performance of decentralized social to the Poor Are Unobserved programs. The allocation to a province should depend on how successful it is at Martin Ravallion reaching the poor with the extra resources, rather than how poor it is. New measurement tools can help monitor performance with limited data. The World Bank Development Research Group Poverty and Human Resources March 1999  a POLICY RESEARCH WORKING PAPER 2080 Summary findings National antipoverty programs often rely heavily on in reaching the poor and to measure the relative provincial governments. The center targets poor contributions to the program's performance - before provinces in the hope that they will reach their own and after reforms - of the center's provincial poor. Without successful intraprovincial targeting, reallocation and decentralized targeting. however, even dramatic redistribution from rich to poor Funding and program design changes led to large gains provinces can have little impact on poverty nationally. for the poor, although with diverse performance across However, data for assessing performance at provincial provinces. level are often far from ideal. Can a centralized Program funding and design choices by the central government monitor the performance of decentralized government can greatly affect the targeting performance social programs in reaching the poor when their benefit of decentralized social programs. The allocation to a incidence is unobserved? province should depend on how successful it is at Ravallion shows that the poverty map and the reaching the poor with the extra resources, rather than corresponding spending allocation across geographic how poor it is. areas allow one to identify the latent differences in mean Design choices should provide incentives for provincial allocations to the poor versus the nonpoor. The national governments to target resources to the poor. Finding measure of targeting performance is also subgroup- feasible ways to monitor their performance and adjust decomposable. central government's efforts accordingly are then crucial Ravallion uses an application to an antipoverty to better outcomes for poor people. program in Argentina (Trabajar II) to assess performance This paper - a product of Poverty and Human Resources, Development Research Group - is part of a larger effort in the group to provide better tools for monitoring the impact on poverty of World Bank projects. The study was funded by the Bank's Research Support Budget under research project "Policies for Poor Areas" (RPO 681-39). Copies of this paper are available free from the World Bank, 1818 H StreetNW, Washington, DC 20433. Please contact Patricia Sader, room MC3- 632, telephone 202-473-3902, fax 202-522-1153, Internet address psaderCaworldbank.org. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/html/dec/PublicationsfWorkpapers/home.html. The author may be contacted at mravallion@worldbank.org. March 1999. (25 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center  Monitoring Targeting Performance when Decentralized Allocations to the Poor are Unobserved Martin Ravallion' Development Research Group, World Bank This paper is based on work done for the World Bank's Supervision Mission for the Social Protection II Project in Argentina. The paper draws on data produced by the excellent information system set up by the project office in the Ministry of Labor of the Government of Argentina. The support of the Bank's Research Committee (under RPO 681-39) is also gratefully acknowledged. Helpful comments on an earlier version of this paper were received from Francois Bourguignon, Emanuela Galasso, Jesko Hentschel, Polly Jones, Jennie Litvack, Lant Pritchett, Dominique van de Walle, and seminar participants at the World Bank and the 1998 Congress of the International Institute of Public Finance held in Cordoba, Argentina.  1. Introduction National anti-poverty programs often rely heavily on provincial governments. The center targets poor provinces, in the hope that they will reach their own poor. It is known, however, that without successful intra-provincial targeting efforts, even dramatic redistributions from rich provinces to poor ones can have little impact on poverty nationally.2 Outcomes will depend on the behavior of provincial governments.3 One can expect provinces to differ in relevant ways. Some will care more about the poor than others, or face different constraints in their efforts to reach the poor. Indeed, having a high incidence of poverty can result in worse targeting performance (Ravallion, 1999). While the decentralization of targeting decisions can allow local infonnation to be exploited, some provinces will be better able to secure such gains than others. Central monitoring of provincial performance, and use of the information to encourage better performance, can thus be crucial to success in reaching the poor. However, the data at hand can be very limited, either because no household surveys were done, or they did not ask the right questions at the right times, or because the surveys are not representative at provincial level. So benefit incidence between the poor and non-poor will often be unobserved (by either the center or the provinces). This can severely constrain contracts with the provinces for assuring maximum impact on poverty nationally. A poorly-informed center might also be inclined to assume the worst; imperfect information about outcomes can then lead to under-funding of anti- poverty programs. 2 For evidence see Ravallion (1993) (for Indonesia) and Datt and Ravallion (1993) (for India). 3 This point is well recognized; "Since the central authorities do not control the local authorities, the effects of federal programmes need not be those intended; and the central authorities must take the reactions of the communities into account" (Atkinson and Stiglitz, 1980, p.55 1). For discussions of this 2  This paper proposes a measure of targeting performance for situations in which benefit incidence is unobserved. It is usually feasible to track spending across local government areas within provinces. Poverty indicators for those areas might then be based on large sample surveys or census data. 4 However, the poverty data do not include program participation. The proposed measure of targeting performance is the regression coefficient of spending on the poverty rate across areas within a government's boundaries. Conditions are identified under which this regression coefficient reveals benefit incidence. The paper also shows that the overall (national) measure of targeting performnance can be exactly decomposed into between-province and within- province components. This allows policy makers to assess the contribution of the center's provincial allocation - versus intra-provincial targeting - to overall performance. The following section defines the measures of targeting performance and the conditions under which they correctly reveal benefit incidence. Section 3 puts these tools to work in monitoring a World Bank financed anti-poverty project in Argentina. Section 4 concludes. 2. Measuring Targeting Performance without an Incidence Survey Consider a goveirnent that has a budget for a social program. This budget is to be allocated between the poor and non-poor within the government's boundaries. Those boundaries embrace a set of finer geographic areas; these might be provinces, if we are monitoring the central government, or municipalities if we are monitoring a provincial government. The government decides how much should go to each of the poor and non-poor within its and related issues concerning fiscal federalism see Brown and Oates (1987), Wildasin (1991), Bird et al., (1995), Inman and Rubinfeld (1997) and Keen and Marchand (1997). 4 The poverty map might be constructed from the census by extrapolating from a smaller living standards survey following the method of Hentschel et al. (1998). 3  jurisdiction. However, what we observe as data are not these allocations, but only the total expenditures across each of the geographic areas which lie within the government's boundaries, and the poverty rates across those areas, i.e., the geographic poverty profile. To fill in the missing data we will have to make some assumptions. The government's (unobserved) optimal allocation to a household can be assumed to depend on that household's level of welfare. That may in turn depend on where the household lives, but I assume that the poverty rate in the area where it lives does not matter to a household's allocation independently of its own level of welfare.5 In other words, there is no "poor-area bias" in that a poor person living in a poor local-government area expects to get the same amount from the program as an equally poor person living in a rich area of the same province. The same holds for the non-poor. This assumption can be thought of as a form of horizontal equity within provinces. The following observations can be made about this assumption: (i) While "poor-area bias" is ruled out, that does not mean that the government ignores the differences in poverty incidence between areas when making its disbursements. Indeed, that may well be the key information it uses. However, it is assumed that the government is not biased by those differences in making its allocations. A poorer area will still receive more from a government that is targeting the poor, but simply because there are more poor people in that area. (ii) It is an assumption about the behavior of the level of government that actually decided how the program's resources should be allocated between the poor and non-poor. Suppose that each province in a federal system allocates its budget between the poor and non- 5 Notice that I say " independent" which is a stronger assumption than "uncorrelated". Strictly, the method below requires that the deviations from the mean are uncorrelated with either the incidence of poverty or its squared values. That is implied by independence. 4  poor. - It is assumed that there is no "poor-area bias" within any given province. However, there may well be such a bias between equally poor people living in different provinces, depending on the inter-provincial differences in behavior, and the center's budget allocation. Horizontal inequity between provinces is still possible. I will return to this point. (iii) The assumption rules out effects of the programn on migration within provinces. If there is no difference in expected allocations for otherwise identical households then they will have no incentive to move. Migration effects may however be of concern between provinces. In the case of the Argentinean program to be studied later, its size and temporary nature make it unlikely that it affects residential locations, so one can abstract from this complicating factor in the analysis. However, this may be less plausible in other applications. Larger redistributive interventions at province level could result in migration and (hence) fiscal externalities; this raises a rather different set of issues for central government policy (Wildasin, 1991). Let us first consider how to measure the targeting performance of each province, under the assumption of horizontal equity in expectation. The central government allocates a total budget of G per capita across M provinces such that Gj per capita is received by province j. After that, each province decides how much should go to the poor versus the non-poor. The chosen allocation by province j is G,. per capita for the non-poor and GjP for the poor. Province j comprises Mj local government areas, which I call "departments". The per capita allocations to department i (=1,.., Mj ) within provincej can be written as: G' = G' +