Panel vector autoregression under cross-sectional dependence

This paper studies estimation in panel vector autoregression (VAR) under cross-sectional dependence. The time series are allowed to be an unknown mixture of stationary and unit root processes with possible cointegrating relations. The cross-sectional dependence is modeled with a factor structure. We extend the factor analysis in Bai and Ng (2002, Econometrica 70, 91-221) to vector processes. The fully modified (FM) estimator in Phillips (1995) is used for estimation in panel VAR and we also propose a factor augmented FM estimator. Our simulation results show this factor augmented FM estimator performs well when sample size is large.