Bayesian treatment effects due to a subsidized health program: The case of preventive health care utilization in Medellin (Colombia)

We analyze the treatment effects due to patients' status, covered or uncovered by the subsidized health program in Medellin (Colombia), on the number of preventive health care visits to physicians. We use a Bayesian endogenous switching model that allows interaction effects as well as endogeneity due to patients' status. This framework allows the calculation of the posterior distributions of heterogeneous treatment effects and presents Bayesian learning of the covariance between the two potential outcomes, a relevant policy-maker parameter, even though we do not observe individuals in both states at the same time. We found that there are self-selection effects as well as moral hazard, which may imply adverse consequences for the health care system in Colombia.