Thèse de doctorat
Résumé : In the past decades, decentralization has become increasingly important in both developing and developed countries. Based on the standard decentralization theorem, policy makers believe that local governments are closer to citizens and know more about local contexts and needs than the national governments. Consequently, they can design policies leading to a more efficient allocation of public goods. Moreover, when accompanied by empowerment of local decision-makers, decentralization is supposed to make local politicians more accountable to voters in a way that national politicians are not. Yet, the effective implementation of decentralization policies may heavily rely on local institutions. This thesis contributes to a rising literature analyzing the political economics of decentralization, that is the extent to which local political dynamics may reinforce or jeopardies decentralization reforms, ultimately affecting the citizens' well-being. I consider three countries that have devolved power to local politicians to a different extent, and tackle three obstacles that may undermine the beneficial effects of decentralization reforms: first, strong political competition at the local level; second, the need for coordinating the provision of a local public good; the interaction between political competition and coordination needs on accountability and eventually the quality of a local public good. Political competition may hinder the beneficial effects of decentralization on stability in a post-war country like Burundi. After a long-lasting and devastating civil war, in 2010 Burundi organized the first local elections, with the hope of establishing political stability through democratic means. However, together with two co-authors, I show that such political decentralization partly failed. We use a unique dataset and geographic fixed effects to show that violence was higher in municipalities characterized by fierce political competition and acute polarization between demobilized rebel groups. The former protagonists of the civil war used the elections as another stage to engage in a stiff struggle for power, and used “specialists of violence” to illicitly steer the electoral outcome.Political dynamics may undermine the coordinated management of the local sewerage networks in Brazil. I use geospatial data to proxy for the scope of coordination between neighboring municipalities: municipalities that are close “enough” are those more exposed to the spreading of water-borne diseases, which ultimately justifies cross-boundary coordination of local sanitation networks. By exploiting a Regression Discontinuity Design in close municipal elections, I show that political alignment between neighboring mayors may lead to lower access to sanitation networks of households. I argue that mayors co-managing a public good have a stronger incentive to monitor each other and ensure effective coordination when they come from different political parties, essentially because of political competition. Mayors from the same party would tend instead to be more lenient to each other, ultimately undermining the quality of the local public good co-provided.Finally, I study the effect of decentralization on the accountability of local politicians co-managing local police in Belgium. In 2005, one of the regions of Belgium introduced the direct election of mayors, while in the rest of the country mayors remained appointed by the local city councils. Together with a co-author, I exploit this reform to show that crime incidence in municipalities affected by the reform decreased faster than anywhere else in the country. We argue that the direct election of mayors increase the accountability of mayors and their incentives to fight criminality. However, we find that the effects of the reform decrease when an increasingly larger number of neighboring mayors has to coordinate the local police. The need to coordinate the local police blur accountability, mitigating the effects of the reform. To conclude, the overarching message of the thesis is that local institutions matter for the implementation of decentralization policies. Policy-makers redesigning the distribution of power between levels of government need to take into account pre-existing political and institutional dynamics that could jeopardize their policy initiatives. In particular, policy-makers could envision decentralization “at different speeds”, enabling local actors to identify objectives of development together with the right tools to pursue them, and finally decentralize accordingly.