Variable Selection in Panel Models with Breaks

We develop a Bayesian approach that performs variable selection in panel regression models that are subject to breaks. Our variable selection approach enables deactivation of pervasive regressors and activation of weak regressors for short periods. Allowing the coefficients on individual variables to change at each break point introduces a high-dimensional search problem even in settings with modest numbers of cross-sectional units and time-series observations. Our methodology is demonstrated on simulated data and in an empirical application to firms' choice of capital structure. We find that ignoring breaks to the data generating process typically results in overestimating the number of relevant regressors, but also leads to a failure to activate regressors that are only informative for short periods