Dr Milda Norkute, Lund University: The Factor Analytical Approach in Near Unit Root Panels
An Economics and Finance External Seminar.
In a recent study, Bai ("Fixed-Effects Dynamic Panel Models, A Factor Analytical Method". Econometrica, 81, 285-314, 2013a) proposes a new factor analytical (FA) estimation approach for stationary dynamic panel data models with cross-section-specific constants, or fixed effects. Our interest in this method originates with the fact it does not require explicit demeaning of the data, a practice that is known to cause problems of bias and low power in near unit root panels. The purpose is to study the properties of FA when applied to such panels with a possible linear trend. It is shown that the estimator is consistent with a well centered asymptotic normal distribution, leading to unit root tests with maximal achievable power. The exception is if the there is a linear trend in the model, in which case the asymptotic normality breaks down when there is an exact unit root. Hence, while it continues to be consistent, FA is not suitable for unit root testing in the linear trend case.