Identification of causal mechanisms through an RD approach.

Causal Mediation Analysis has important implications in economics. It helps to deeply understand the policy makers' decisions and to better design policy strategies. However, the identification process is not an easy issue and analyzing causal mechanisms requires stronger assumptions than evaluating the classical average treatment effect. The main difficulty consists in the endogeneity of the mediator with the consequence that it is not possible to identify the effects of interest. Several methods have been developed, based on different set of assumptions and with different strategies for the estimation. I propose a new identification strategy for the estimation of the direct and the indirect effect, through an implementation of a Regression Discontinuity Design. I present two different models. The first one follows the traditional identification strategy based on linear equation models. The second model follows the most recent literature based on nonparametric identification procedures. I show the consistency of this last estimator, validating the results through a Monte Carlo simulation study.