Bias Reduction in Nonlinear and Dynamic Panels in the Presence of Cross-Section Dependence, with a GARCH Panel Application

• Activity Number: 607
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Cavit Pakel*+
• Companies: Bilkent University
• Keywords: Incidental parameter bias ; Nonlinear dynamic panels ; Integrated likelihood method ; Composite likelihood method ; GARCH ; Hedge funds

Abstract

This study analyses the incidental parameter bias in non-linear and dynamic panel data models where the time-series and cross-section dimensions approach infinity at the same rate. The analysis focuses on the integrated likelihood method, allowing for dependence across both dimensions. I show that, although weak serial dependence leads to no extra bias, cross-section dependence generates a new type of bias, the magnitude of which depends on the strength of dependence. Likelihood-based analytical expressions for this bias term are provided under strong, weak and cluster-type cross-section dependence. Next, I consider the specific case of GARCH modelling using a panel of financial data. Simulation analysis reveals that the bias-corrected integrated likelihood method requires around 150 time-series observations to fit GARCH with little bias, compared to about 1,000 observations required for successful estimation by standard methods. Furthermore, the effect of cross-section dependence on bias is negligible, although it leads to higher variance. An empirical demonstration, analysing monthly hedge fund volatility characteristics is provided.