Cette thèse porte sur la prise en compte des interactions dans l’estimation de l’effet d’une politique publique. En effet, les méthodes standards d’estimation s’appuient généralement sur des hypothèses fortes excluant partiellement ou totalement les effets d’interactions aussi bien dans le processus de sélection des bénéficiaires que dans la performance de la politique. Cependant, la validité de ces hypothèses est discutable surtout dans le cadre d’une politique non-expérimentale, car un individu selon qu’il est bénéficiaire ou non de la politique ne vit pas en autarcie, il est en interaction avec son milieu. Ainsi, l’effet estimé de la politique par les méthodes traditionnelles est susceptible d’être biaisé (surestimation ou sous-estimation), avec pour conséquences des prises de décisions basées sur des résultats non fiables.Dans cette thèse, nous supposons que le traitement reçu par un individu peut affecter les autres individus de son voisinage à travers différents mécanismes de diffusion. Nous proposons un cadre méthodologique d’évaluation basé sur les méthodes de l’économétrie spatiale. Nous développons quatre extensions spatialisées de méthodes d’évaluation prenant en compte les interactions dans les processus de sélection et au niveau de la performance de la politique, avec une application empirique pour chaque méthode. / This thesis focuses on the integration of interactions in the estimation of the public policy’s effect. Indeed, the estimation standard methods usually rely on strong assumptions that exclude partially or completely the interaction effect both in the selection process of beneficiaries and in the performance of the policy. However, the validity of these assumptions is questionable, particularly in the context of a non-experimental policy because an individual according as it is the beneficiary or not the policy, is not isolated: he interacts with his neighborhood. Thus, the estimated effect of the policy by traditional methods is likely to be biased (underestimation or overestimation) therefore induce decision-making based on erroneous conclusions. Therefore, in this thesis, we assume that the treatment received by an individual can affect other individuals in his neighborhood through different diffusion mechanisms. Using spatial econometric methods, we propose a methodological framework to take into account both in selection process and in the outcomes the interactions in the estimation of the policy effect. Four extensions of the evaluation methods are developed with an empirical study for each method.

Submitted by Vinicius Peçanha (viniciusrpecanha@hotmail.com) on 2015-08-21T15:06:47Z No. of bitstreams: 1 dissertacao_ViniciusPecanha_biblioteca_final2 Copy.pdf: 2020205 bytes, checksum: cc440ff198ec3e166a25a3db500c5c89 (MD5) / Approved for entry into archive by BRUNA BARROS (bruna.barros@fgv.br) on 2015-08-24T13:07:46Z (GMT) No. of bitstreams: 1 dissertacao_ViniciusPecanha_biblioteca_final2 Copy.pdf: 2020205 bytes, checksum: cc440ff198ec3e166a25a3db500c5c89 (MD5) / Approved for entry into archive by Marcia Bacha (marcia.bacha@fgv.br) on 2015-08-25T19:49:01Z (GMT) No. of bitstreams: 1 dissertacao_ViniciusPecanha_biblioteca_final2 Copy.pdf: 2020205 bytes, checksum: cc440ff198ec3e166a25a3db500c5c89 (MD5) / Made available in DSpace on 2015-08-25T19:49:14Z (GMT). No. of bitstreams: 1 dissertacao_ViniciusPecanha_biblioteca_final2 Copy.pdf: 2020205 bytes, checksum: cc440ff198ec3e166a25a3db500c5c89 (MD5) Previous issue date: 2015-04-29 / The paper quantifies the effects on violence and police activity of the Pacifying Police Unit program (UPP) in Rio de Janeiro and the possible geographical spillovers caused by this policy. This program consists of taking selected shantytowns controlled by criminals organizations back to the State. The strategy of the policy is to dislodge the criminals and then settle a permanent community-oriented police station in the slum. The installation of police units in these slums can generate geographical spillover effects to other regions of the State of Rio de Janeiro. We use the interrupted time series approach proposed by Gonzalez-Navarro (2013) to address effects of a police when there is contagion of the control group and we find that criminal outcomes decrease in areas of UPP and in areas near treated regions. Furthermore, we build a model which allows to perform counterfactuals of this policy and to estimate causal effects in other areas of the State of Rio de Janeiro outside the city.