Decentralization Measures Revisited

Abstract

For the last half of a century, economists in an ever growing literature have been trying to measure decentralization with the purpose of using that measure as either the dependent or an independent variable in their empirical models. The suggested measures essentially boil down to a few concepts: locally raised own revenues, autonomy on expenditure decisions, locally spent national grants, or the number and size of local units. These basic components have entered decentralization measures in various functional forms: additive, multiplicative, in powers, and various mixes of those. This paper revisits these measures and our main finding is that aggregating those distinct dimensions of decentralization into a single indicator inevitably leads to a loss of information. We conclude that in a multivariate framework the distinct aspects of decentralization should enter regression analyses separately, in the most flexible functional form possible. At the same time, while no single indicator can capture all aspects of decentralization, some indicators are more inclusive and informative than others. Therefore, if for some reason we have to use a single indicator, our study suggests which indicator would be more appropriate for the hypothesis at hand. We illustrate these points examining differences in the empirical behavior of different decentralization indicators in relationship to each other and in their statistical association with some relevant economic outcomes in the fiscal decentralization literature.