Daniel Treisman (2007) has reminded us recently that there has been insufficient evidence to support theoretical advantages of decentralization. He made the point that more empirical analysis may be needed to strengthen the case for further decentralization. Of particular interest would be direct comparative study of quality of public service delivery in more and less decentralized regions. At the same time, there has been growing consensus in the literature that the results of decentralization depend critically on local conditions, in particular on comparative intensity of distortions in the incentive structure at different government levels.

In this paper we aim to contribute to the research agenda of assessing the effects of fiscal decentralization on quality of public service delivery using a large data set on performance of Russian regions and subnational governments in the late 1990s and early in this decade. Russian data could be of considerable interest for advancing our understanding of decentralization effects. This is because, on the one hand, Russian regions show a considerable variation in the degree of expenditure decentralization, and they differ considerably in their choices of expenditure and sectoral policies. On the other hand, there is a major cross-regional variation in efficiency of regional public spending in core sectors. Thus, the question is whether there is a link between regional decentralization and efficiency patterns.

We explore factors that may explain cross-regional variation in health and education indicators and investigate the role of decentralization factor in explaining such differences. We look at conventional measures of health and education quality, such as child and mother mortality rates or average grades at the standardized high school exams, and control for factors that usually influence such performance indicators (including per capita income, urbanization, education level, etc.), as well as for availability of sectoral inputs. In addition, we use in our models additional variables that reflect the incentive structure of subnational governments (grant and rent dependence, quality of subnational institutions).

We check if there is a link between decentralization and either the current level of social outcomes or their change since has initiated its decentralization program in the early 1990s. We also review the explanatory power of different measures of fiscal decentralization: a) general expenditure decentralization (share of municipal spending in the total spending of regional consolidated budget), and b) similar measures of sectoral expenditure decentralization (such as share of municipal health spending in the total health spending by consolidated budget).

Our framework for analysis is the following: the level of budget spending affects the quantity and quality of main observable inputs in each sector (such as schools, teachers, or computers in education) and then these observable inputs and spending both affect regional performance in the sector. The decentralization variable can, in principle, influence both stages of the above relationship between spending and outcomes.

Our estimation strategy would address a potential causality problem by using instrumental variables.