REndo: An R Package to Address Endogeneity Without External Instrumental Variables

Abstract

Endogeneity is a common problem in any causal analysis. It arises when the independence assumption between an explanatory variable and the error in a statistical model is violated. The causes of endogeneity are manifold and include response bias in surveys, omission of important explanatory variables, or simultaneity between explanatory and response variables. Instrumental variable estimation provides a possible solution. However, valid and strong external instruments are difficult to find. Consequently, internal instrumental variable (IIV) approaches have been proposed to correct for endogeneity without relying on external instruments. Our R (R Core Team 2017) package REndo implements various IIV approaches, i.e., latent instrumental variables estimation (Ebbes, Wedel, Boeckenholt, and Steerneman 2005), higher moments estimation (Lewbel 1997), heteroskedastic error estimation (Lewbel 2012), joint estimation using copula (Park and Gupta 2012) and multilevel GMM estimation (Kim and Frees 2007). Package usage is illustrated on simulated and real-world data.

Keywords: endogeneity, internal instrumental variables, multilevel models.