Institutions, policy and quasi-experimental evidence

Sammanfattning: Political economics has been an active research area in the last decades. Lately, the number of empirical studies has caught up with the vast amount of theoretical work. However, the empirical studies that rely on natural experiments or exogenous sources of variation are relatively few. The importance of causal analysis in political economics cannot be overestimated, as pointed out by scholars. Persson and Tabellini (2002) state that "natural experiments" in political economics have not been sufficiently utilized. Moreover, we usually observe a bundle of components and an outcome. For example, in cross-country studies, we compare a very large bundle of components, denoted countries. Trying to unbundle specific components of a broad bundle and their relation to outcomes is of first-order importance if we want to further understand the mapping between institutions and outcomes, as pointed out by Acemoglu (2005). In the first three papers in this thesis, I try to respond to these two needs by using two strategies; first, by searching for exogenous variations in policies or institutions and, second, by using local governments in Sweden as a testing ground. Exploring exogenous variation is the key to causal analysis. Using local governments in Sweden is one plausible way of trying to unbundle specific components, since Sweden is a unitary state. This means that all local governments act in a world with similar institutions (or rules), which would not be true if Sweden were a federal state. Moreover, local governments do take care of a large part of Swedish GDP, since they are the main providers of social services. The first paper tests for a free-riding behavior by making use of an attractive boundary reform. The second paper rely on a regression discontinuity design to test for partisan effects in very small local governments. The third tests for the effect of representative democracy relative to direct democracy on fiscal policy relying both on a regression discontinuity design and panel data methods.