chapter  11
22 Pages

Handling Omitted Variable Bias in Multilevel Models: Model Specification Tests and Robust Estimation

ByJee–Seon Kim and Chris M. Swoboda

Whereas model specification tests have been one of the most important areas of research in econometrics for decades (Frees, 2004; Hausman, 1978; Wooldridge, 2002), specification issues have been often overlooked in multilevel analysis. We argue that concerns for omitted variables and other specification issues should be routine in multilevel analysis. An omitted variable at one level may yield severe bias at all levels in terms of regression coefficients as well as variance components (Kim, 2009).