Decentralizing Anti-Poverty Program Delivery inDeveloping CountriesPranab BardhanDepartment of Economics, University of California, BerkeleyandDilip MookherjeeDepartment of Economics, Boston UniversityAbstractWe study effects on accountability in government service delivery of decentralizingadministration of an antipoverty program. While governments at both central and locallevels are vulnerable to anti-poor policy biases owing to political capture, centralizeddelivery systems are additionally prone to bureaucratic corruption owing to problemsin monitoring bureaucratic performance. Decentralizing the delivery system promotescost-effectiveness and improves intra-regional targeting at low program scales. Butinter-regional targeting may deteriorate, as central grants to high poverty regions shrink,owing to high capture of local governments by local elites in such regions.Keywords: bureaucracy, corruption, decentralization, poverty alleviation, targetingJEL Classification Nos.: D72, D73, H 41, H42, H77, I38, O17Corresponding Author: Dilip Mookherjee, Department of Economics, 270 Bay StateRoad, Boston MA 02215; tel no: 617-3534392, fax no: (617)3534392, [email protected] IntroductionDecentralization of authority over administration of redistribution programs to local com-munities has recently become widespread in the developing world. These initiatives havetransferred responsibility from central ministries to local governments or community rep-resentatives over procurement, selection of local projects and identification of beneficia-ries. Such experiments have been introduced since the 1980s in Armenia, Albania, Bosnia-Herzegovina, Brazil, China, El Salvador, Georgia, India, Mexico, South Africa, Uganda andUzbekistan (Conning and Kevane (1999), Jimenez and Sawada (1999), Mookherjee (2001)and World Bank (2000)). Other countries still characterized by traditional top-down deliv-ery systems are currently contemplating adopting similar reforms.The principal aim of these reforms has been to enhance accountability in the servicedelivery process by limiting problems inherent in the traditional centralized system. Oneof the most important of these is corruption among centrally appointed bureaucrats allo-cated authority over service delivery in local areas. While systematic empirical evidence oncorruption is scarce for obvious reasons, there are innumerable case studies of high rates ofcorruption among government bureaucrats in the process of delivering public services.1The presumed argument in favor of decentralizing the delivery system is that localgovernments will be subject to electoral pressures from local citizens who are able to monitordelivery better than a distant central authority. Large scale bribery, cost-padding anddiversions to the black market will be more noticeable by local residents that have to pay1Wade (1985, p. 474)) for instance describes the bribery among irrigation bureaucrats in the Indian stateof Andhra Pradesh, where he estimates illicit earnings from water bribes to be one or two times the annualsalary for an Assistant Engineer, and three to five times for an Executive Engineer. Rose-Ackerman (1999,Chapter 1) cites numerous reports concerning diversion of commodities to the black market, and charging ofbribes for delivering public services by government bureaucrats in diverse countries. These include diversionof coal and other producer goods to the black market in China at a profit of between 250 to 680%; payoffsneeded to obtain subsidized loans in Eastern Europe, Russia, Lebanon, Kenya, Fiji, Korea and Pakistan;bribes for certifying unqualified people for subsidized housing in Washington D.C.; corrupt payments forreceiving the best plots in land reform programs in India and Pakistan; fraud and corruption in processingof workmen’s compensation claims in Brazil; ‘routine’ payments that need to be made by citizens to receivetelephone installations, driving licenses, passports, electric connections in India, Russia and Ukraine.2these bribes and live in close proximity to local government officials. Residents are thenlikely to be motivated to eject such government officials from office, besides applying socialsanctions. This will conceivably enhance accountability in the delivery process.Yet this presumption of greater accountability under decentralization is frequently ques-tioned, since it is based on the assumption that local democracy will function effectively.It assumes a level of literacy and political awareness among citizens that may be unrealis-tic for many poor countries. Numerous case studies exist of development programs beingstymied by capture of local governments by powerful local elites, that distort and divertpublic programs to benefit themselves at the expense of poor minorities.2In the Feder-alist Papers (no. 10), such concerns about local capture led James Madison to argue forretention of powers at the federal level (Hamilton, Madison and Jay (1787[1937 pp 62–70]).Political scientists (such as Crook and Manor (1998) have accordingly studied patterns ofparticipation by the poor in local politics in a number of countries in South Asia and Africa,in order to predict the level of accountability that can be expected from local governmentsunder decentralization. These empirical studies, as well as our own theoretical analyses(Bardhan and Mookherjee (1999, 2000)) lead to no uniform or general conclusions aboutrelative vulnerability of different levels of government to special interest capture.3The purpose of this paper is to treat the extent of local capture (vis-a-vis capture of2Gaiha, Kaushik and Kulkarni (1998, 2000) conclude following a survey of implementation of rural de-velopment programs implemented by local government bodies (called gram panchayats) in the state of UttarPradesh (UP), India:”A survey of 3 districts in UP on the implementation of the 2 major anti-poverty programs, viz.JRY and IRDP in 1997 points to the pervasiveness of rent-seeking behavior in all the GramPanchayats. As a result, there were large-scale diversions of resources. The benefits of JRY,IRDP and the other rural development programs to the poor were minimum.... Regardless ofwhether the Panchayat chairperson belonged to upper or lower castes, there was little variationin the method and pattern of domination by the locally powerful groups”.In similar vein, Tendler (1997) describes how powerful local elites tended to be the most important roadblockin