This paper uses econometric models of public investment to investigate the institutional and political determinants of central vs. local government decision-making. I use a remarkable database from Bolivia's recent, radical decentralization program. I find that local government policy decisions are progressive both economically and in terms of need, and largely determined by a competitive interest group dynamic which provides poorer citizens, as well as private sector firms and civic

more »

... s, with political voice. This ensures that accountability is binding for elected officials. By contrast centralized investment -more insulated from grass-roots pressures -is regressive in both dimensions. The results suggest a healthy picture of local democracy in which voters are able to influence local government through both their civil institutions and the electoral mechanism. Where local government works well citizens have voice, providing an effective counterweight to the power of private firms and government's own politico-bureaucratic interests. level of analysis is often careful, nuanced and deep, such studies tend to suffer from a low level of generality, an excess of variables over observations -which in turn leads to a failure to control adequately for external factors, and in the worst case a conflation of causes and effects. Examples of large sample studies include de . Quantitative studies, on the other hand, tend to benefit from the high degree of generality, consistency and empirical transparency that statistical approaches provide. But they also suffer significant problems with the measurement of often abstract concepts, data comparability across diverse countries (or regions), and the possibility of omitted variables. Perhaps as a result of these methodological difficulties, neither side of the decentralization debate has been able to specify a complete model of how government -central or local -works. The effects they posit operate at the hazy margins of a governmental black box, and as a result competing claims from each side pass each other in the fog, failing to engage or move the conversation forward. Without a clear model of central or local government, both sides have difficulty analyzing the effects of switching from one to the other. And as a result, neither side can triumph over the other. An exception to this inconclusiveness is Faguet (2001) , who also examines the case of Bolivia and shows that decentralization did make government more responsive to real local needs. After decentralization, municipalities invested more in education, water & sanitation, water management and agriculture where illiteracy rates are higher, water and sewerage connection rates lower, and malnutrition a greater risk respectively. These changes were driven by Bolivia's smallest, poorest, mostly rural municipalities investing newly devolved funds in their highest-priority projects. Knowing that these things happened is important. But in order to comprehend decentralization and fully learn its lessons, we must also understand why they happened. We must unlock the black boxes of central and local government operation in order to unravel the workings of each, and how they differ. We need a micro-level approach that allows for complexity and nuance, examining policy outputs through the interplay of institutions, electoral competition and lobbying activity that produces them. This paper builds on Faguet's work in an attempt to do that empirically for the remarkable case of Bolivia. It employs