This paper introduces an unbiased estimator based on least squares involving time-specific cross-sectional averages for a first-order panel autoregression with a strictly exogenous covariate. The proposed estimator is straightforward to implement as long as the variables of interest have sufficient time variation. The number of cross-sections (N) and the number of time periods (T) can be large, and there is no restriction on the growth rate of N relative to T. It is demonstrated via both theory and a simulation study that the estimator is asymptotically unbiased, and it can provide correct empirical coverage probabilities for the 'true' coefficients of the model for various combinations of N and T. An empirical application is also provided to confirm the feasibility of the proposed approach.
Keywords: C22; C23; C33; first difference least squares (FDLS); fixed effects; panel autoregression; pseudo-panel data; time-specific average (TSA).
© 2021 Walter de Gruyter GmbH, Berlin/Boston.