Addressing Endogeneity Without Strong Instruments: A Practical Guide to Heteroskedasticity-Based Instrumental Variables (HBIV)

Abstract

This article provides an overview and guide to implementing heteroskedaticity-based instrumental variables (HBIV) in regression models with endogeneity, i.e., one or more of the regressors are correlated with the disturbance term. We discuss the problem of implementing standard instrumental variables (IV) solutions to the endogeneity problem when external instruments are either insufficient or not readily available, and when the disturbances are heteroskedastic, present a solution to the problem. We illustrate the implementation of HBIV in the presence of strong external instruments, weak external instruments, and no external instruments, using both traditional IV and HBIV. Finally, we discuss the pros and cons of using HBIV methods to address endogeneity.