Heterogeneous Structural Breaks in Panel Data Models: Application on Population Growth

This paper revisits the hybrid estimation method, GAGFL, which is a combination of the grouped fixed effects approach and the adaptive group fused Lasso, described by Okui & Wang (2018). This estimation method takes into account the heterogeneity of individuals as well as heterogeneity of coefficient estimates in panel data. It aims to model individual heterogeneity by estimating an underlying grouped pattern in the data, such that all the individuals within a group share the same slope coefficient estimates. Allowing for multiple structural breaks in the regression coefficients models the heterogeneity of the slope coefficients. However, the break date estimates as well as the number of breaks can differ among the groups. By means of a Monte Carlo simulation, it is shown that this method performs well in finite samples. Many studies have been conducted relating to panel data sets, however, the incorporation of heterogeneity in both the observations and the coefficient estimates have not been done before. These characteristics are, nonetheless, desirable as in practice it is possible that not all individuals are affected in the same way by some event. An empirical application concerning population growth relating to the rate of natural increase and international migrant stock illustrates this property.