CMStatistics 2018: Start Registration
View Submission - CFE
Title: A flexible Bayesian model for treatment effects on panel outcomes Authors:  Helga Wagner - Johannes Kepler University (Austria) [presenting]
Abstract: Identification and estimation of treatment effects is an important issue in many application fields, e.g. to evaluate of the effectiveness of social programs, government policies or medical interventions. In contrast to randomized studies, unobserved confounding due to endogeneity of treatment selection has to be taken into account for data from observational studies. In the Bayesian approach this is accomplished by specifying a joint model of treatment selection and the potential outcomes. For the estimation of dynamic effects of a binary treatment on a continuous outcome observed over subsequent time periods two models, the switching regression model and the shared factor model, have been suggested so far. We show that both impose restrictions on the joint correlation structure of treatment selection and the two outcomes sequences that can result in biased treatment effects estimates. To achieve more flexibility we propose a new model that allows us to separate longitudinal association of the outcomes from association due to endogeneity of treatment selection. We employ this model to analyse the effects of a long maternity leave on earnings of Austrian mothers, where we exploit a change in the parental leave policy in Austria that extended maternal benefits from 18 months since birth of the child to 30 months. This analysis is based on data from the Austrian Social Security Register which contains individual employment histories since for all Austrian employees.