Heterogeneity and cross-sectional dependence in panels: Heterogeneous vs. homogeneous estimators
Akgun, Oguzhan
Heterogeneity and cross-sectional dependence in panels: Heterogeneous vs. homogeneous estimators - 2021.
73
This paper focuses on the comparison of homogeneous and heterogeneous panel data estimators, including partially heterogeneous ones, in the presence of cross-sectional dependence generated by common factors and spatial error dependence. Our specifications allow us to consider and contrast weak cross-sectional dependence and strong cross-sectional dependence in a general linear heterogeneous panel data model. An overview of the estimation procedures, including heterogeneous, homogeneous, and partially heterogeneous estimators, is presented. Then, an extensive Monte Carlo study is conducted using a general framework encompassing recent contributions in the literature, especially in terms of considering common factors and spatial dependence simultaneously. Our simulation results show that, even for small individual and time dimensions, heterogeneous estimators perform better in terms of bias, root mean squared error, size, and size-adjusted power compared to homogeneous estimators. Lastly, the superiority of the heterogeneous estimators is confirmed by an empirical application relating fiscal decentralization and government size in 22 OECD countries over the period 1973–2017. JEL classification codes: C13, C23
Heterogeneity and cross-sectional dependence in panels: Heterogeneous vs. homogeneous estimators - 2021.
73
This paper focuses on the comparison of homogeneous and heterogeneous panel data estimators, including partially heterogeneous ones, in the presence of cross-sectional dependence generated by common factors and spatial error dependence. Our specifications allow us to consider and contrast weak cross-sectional dependence and strong cross-sectional dependence in a general linear heterogeneous panel data model. An overview of the estimation procedures, including heterogeneous, homogeneous, and partially heterogeneous estimators, is presented. Then, an extensive Monte Carlo study is conducted using a general framework encompassing recent contributions in the literature, especially in terms of considering common factors and spatial dependence simultaneously. Our simulation results show that, even for small individual and time dimensions, heterogeneous estimators perform better in terms of bias, root mean squared error, size, and size-adjusted power compared to homogeneous estimators. Lastly, the superiority of the heterogeneous estimators is confirmed by an empirical application relating fiscal decentralization and government size in 22 OECD countries over the period 1973–2017. JEL classification codes: C13, C23
Réseaux sociaux