Kang, Byunghoon (2018) Higher Order Approximation of IV Estimators with Invalid Instruments. Working Paper. Lancaster University, Department of Economics, Lancaster.
Abstract
This paper considers the instrument selection problem in instrumental variable (IV) regression model when there is a large set of instruments with potential invalidity. I derive higher-order mean square error (MSE) approximation of two-stage least squares (2SLS), limited information maximum likelihood (LIML), Fuller (FULL) and bias-adjusted 2SLS (B2SLS) estimators with allowing for local violation of the instrument-exogeneity conditions. Based on the approximation to the higher-order MSE, I propose instrument selection criteria that are robust to potential invalidity of instruments. Furthermore, I also show the optimality results of instrument selection criteria in Donald and Newey (2001, Econometrica) under faster than N^(-1/2) locally invalid instruments specication.