The two-way Hausman and Taylor estimator

 This paper reconsiders the two-way Hausman and Taylor (1981) estimator suggested by Wyhowski (1994). The two-way HT estimator allows some but not necessarily all the regressors to be correlated with the individual and time effects. It also allows the estimation of the effects of time-invariant as well as individual-invariant regressors which are wiped out by the two-way fixed effects estimator. Hausman type tests are proposed for this two-way HT regression to test the over-identification conditions implied by the choice of the uncorrelated regressors. This should prove useful for empirical work in this area.