On Wed, 21 Apr 2010, Allin Cottrell wrote: > On Wed, 21 Apr 2010, Summers, Peter wrote: > > > I'm still not sure about the df in the Hausman test though. In > > the last model it seems to me that this should be 2, not 1... > > I will take another look at that.
You still have perfect collinearity in the relevant respect. The issue is that in the Verbeek schooling dataset the experience values are "fake" (as is quite often the case in labor-market datasets); they are calculated numbers representing "potential experience", as in exp76 = age76 - ed76 - 6 This means that when you run the unrestricted regression for the Hausman test one of the "hat"-variables gets dropped automatically, leaving just one degree of freedom for the test. Allin