The last two columns in Table 2 present the estimates of the random effects
model. We argued in Section 2.6 that endogenous managerial effort implied that
the random effects model was likely to be misspecified. The Hausman 1978 test Ž .
indicates that the assumption that area effects are not correlated with the other
regressors cannot be rejected and so the random effects model is appropriate. Even
though random effects is the preferred specification we also report the results from
the fixed effects specification for comparison.