Abstract: During the last three decades, decentralization reform has been a leading trend in policy-making reform. The recurrent argument is that decentralization is linked to increased economic development and improved public service provision. This dissertation attempts to evaluate the validity of this assumption. To do so, it uses a three-paper approach to evaluates the outcomes of decentralization reform from different perspectives and on different dimensions. The first paper uses a geographic regression discontinuity design to estimate that on average a higher level of decentralization at a border implies an average decrease in annual GDP growth. These findings evidence a negative relationship between decentralization and annual GDP growth which are both statistically and economically significant at the one percent level. These results invite a rethinking of decentralization's effects and its implementation process. The second paper focuses on the impact of decentralization reform on Colombia's agricultural extension services. The paper argues that decentralization reform introduced significant changes in the sector. Borrowing from Evans, Hirschman and Snyder, it proposes a framework to explain why the institutional reform led to heterogeneous effects on different products and types of farmers within the sector. Integrating these changes into traditional narratives leads to a better understanding of the last twenty-five years of decline of the Colombian agricultural sector. The third paper is an impact evaluation of a public bicycle sharing system. Public transport systems, like bicycle sharing schemes, are typically under the control of local authorities --due to decentralization designs. Hence, this impact evaluation assesses if a private actor can provide a quasi-public good in the absence of action by the local authority. This paper presents an impact evaluation with a difference-in-difference quasi-experimental design that allows identification and estimation of the treatment effect of a bicycling encouraging intervention. Furthermore, in an attempt to evaluate the effect that a bicycle sharing intervention can have on workers' productivity or well-being, this paper presents an instrumental variable design informing future research on the effect of active commuting by bicycle. Approaching decentralization from these very different perspectives allows a better understanding of the intricacies of decentralization and the way these processes contribute to economic development or better public service provision.