Abstract
Unobservable individual effects in panel data models are employed to control for heterogeneity. These can be thought of as random variables that are uncorrelated with the regressors, thus generating a random effects model. Alternatively, these random individual effects are allowed to be completely correlated with the regressors, thus generating a fixed effects model. The choice between these two alternatives is usually settled using a Hausman (Econometrica 46:1251–1271, 1978) test. This article argues that one should interpret a rejection by the Hausman test as a rejection of the random effects model, not necessarily an endorsement of the fixed effects model.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
