Nonstationary heterogeneous panels with multiple structural changes

This article considers multiple structural changes in nonstationary heterogeneous panels with common factors. Kapetanios, Pesaran, and Yamagata (<a target="_blank" href="#CIT0038">2011</a>) showed that unobserved nonstationary factors can be proxied by cross-sectional averages of observable data.

This means that unobserved error factors can be treated as additional regressors, and different break points in slopes and error factor loadings can be considered as multiple breaks in linear regression models with panel data. We generalize the least squares approach by Bai and Perron (<a target="_blank" href="#CIT0009">1998</a>) to nonstationary panels and show that the break points in both slopes and error factor loadings can be consistently estimated for two important cases involving (i) nonstationary factors and (ii) nonstationary regressors.