Asymptotically Unbiased Inference for a Panel VAR Model with P Lags

• Activity Number: 74 - Statistical Methods and Applications: Domestic and International
• Type: Contributed
• Date/Time: Sunday, July 28, 2019 : 4:00 PM to 5:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Luis Melo* and Juan Sebastian Cubillos
• Companies: Banco De La Republica and Banco de la Republica
• Keywords: Panel VAR models; Bias correction; Restricted OLS

Abstract

Panel dynamic estimators with fixed effects are biased due to the incidental parameters problem. At this regard, Hahn and Kuersteiner (2002) proposed an estimator to correct this issue. However, they only consider a panel VAR (PVAR) model with one lag. In this paper we extend this bias correction, its asymptotic and small sample properties for a more general case, a PVAR model with p lags. The simulation results indicate that the bias corrected estimator outperforms the OLS panel VAR estimator when sample size in time dimension is small, and when the persistence of the model is low. In these cases, the proposed estimator improves significantly in terms of both, the reduction of bias and mean square error.