IMPACT OF DECENTRALIZATION REFORMS IN PAKISTAN ON SERVICE DELIVERY—AN EMPIRICAL STUDY y GHAZIA ASLAM AND SERDAR YILMAZ * World Bank, USA SUMMARY By bringing decision making closer to people, decentralization is expected to improve governance and service delivery outcomes. Yet, the empirical evidence on the impact of decentralization on macroeconomic performance and public sector size presents a mixed picture. However, the findings of cross-country studies in the literature are sensitive to the way decentralization is defined, and how its extent and impact are measured. This article avoids the unwanted effects of incomparability and aggregation in cross-country analysis. We use a unique panel data set from 183 villages in Pakistan to analyze the impact of decentralization reforms implemented in 2001 on the provision of services—street paving, construction of water canals, sanitation sewer lines, and school facilities. Our results show that the magnitude of provision of all services increased significantly following decentralization reforms. We further show that the four services are impacted differently and service delivery improvement is not uniform, but not in ways that conform to the hypotheses of patronage theory. Copyright # 2011 John Wiley & Sons, Ltd. key words — decentralization; local governance; service delivery; Pakistan INTRODUCTION Decentralization has become the cornerstone of development reforms in many countries. The normative rationale for decentralization is deeply rooted in the political economy argument that decentralization leads to better service delivery. 1 Despite this widespread belief in the benefits of decentralization, empirical evidence of such effects in macro outcomes are notoriously hard to show. Ebel and Yilmaz (2002) and Rodden (2004) show that the empirical findings are sensitive to the way decentralization is defined, and how its extent and impact are measured. Cross-country analysis is the norm in the literature as time series data on smaller units within countries are not available. Thus, often countries that are not similar are forced into a single sample that is large enough for researchers to make econometrically reliable inferences (Rodden, 2004), but the result is usually no useful inference that could deepen our understanding. Conspicuously missing from the list of impact variables is service delivery—an important motivation behind decentralization reforms. Consequently, ‘the most discernible Leitmotiv in these studies [empirical studies of decentralization] is a growing disappointment with decentralization and federalism, especially among developing countries’ (Rodden, 2004: p. 481). This article uses a sample from a single country to quantify the effects of decentralization reforms on service delivery. Our comprehensive panel data on service delivery in one country—Pakistan—allows us to make inferences without the unwanted effects of incomparability and aggregation in cross-sectional data. The data set is large in terms of number of unit of analysis (183 villages) and time period (6 years each before and after public administration and development Public Admin. Dev. 31, 159–171 (2011) Published online 11 January 2011 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/pad.591 *Correspondence to: S. Yilmaz, Africa Public Sector Reform and Capacity Development, World Bank, 1818 H Street, NW, Washington D.C. 20433, USA. E-mail: syilmaz@worldbank.org y The findings, interpretations, and conclusions are entirely those of authors, and do not represent the views of the World Bank, its executive directors, or the countries they represent. 1 For example, see Tiebout (1956), Oates (1972), Cremer et al. (1996), Oates (1997), Raff and Wilson (1997), Bucovestsky et al. (1998), Oates (1999), and Mueller (2006). Copyright # 2011 John Wiley & Sons, Ltd.