Statistics & Probability Letters | 2021
Weighted least squares estimation in a binary random coefficient panel model with infinite variance
Abstract
Abstract This article investigates the asymptotic properties of weighted least squares estimators (WLSE) for a binary random coefficient autoregressive (RCA) panel model with heterogeneous variances of panel variables. It is an extension of Johansen and Lange (2013) to a panel model, which is more practical for macroeconomic time series data. We develop asymptotic properties of the WLSE in cases of finite and infinite variances, respectively, as both sizes of panels and samples tend to infinity. In the latter case with infinite variance, the asymptotic for the WLSE β ˆ of the coefficient β is shown to be a curious result β ˆ ⟶ p β − 1 . It is proven by using the notion of a tail index and the stable distribution limit. In a Monte Carlo simulation, feasible WLSEs are computed iteratively and some evidences are given to verify our theoretical results.