Detection of structural breaks in linear dynamic panel data models

A break detection testing procedure for the well-known AR(p) linear panel data model with exogenous or pre-determined regressors is developed. The proposed method can accommodate a structural break in the slope parameters as well as in the fixed effects. Breaks in the latter are not constrained by any type of cross-sectional homogeneity and are allowed to be correlated with all past information. Monte Carlo simulations indicate that the test performs satisfactorily even in the type of panel datasets with short time-dimension often encountered in practice. As an empirical illustration, the paper implements the test to detect the effects of the 1997 Asian crisis on the investment decisions of Asian companies.