Bias-corrected instrumental variable estimation for spatial autoregressive models with measurement errors

Abstract

In this paper, bias-corrected instrumental variable estimation methods, specifically the bias-corrected two-stage least square (2SLS) estimation and the bias-corrected asymptotically best 2SLS estimation, are proposed for spatial autoregressive (SAR) models with covariate measurement errors, utilizing available information regarding the variance of the measurement error. Under mild assumptions, the consistency and asymptotic normality of the proposed estimators are derived. Simulation studies further reveal that the proposed methods exhibit robustness regardless of the presence of spatial dependence in the model. Additionally, a real data example is utilized to illustrate the developed methods.