Page 1 of 1

hausman test

Posted: Wed Nov 15, 2017 8:08 am
by enilorac
Hello everyone,

This is my task:
Run the Hausman test for FE vs. RE, report estimation results (print screen etc.) and
analyze them; in particular, which model is favored.

The RE is no problem, but when I wanna do it with the FE, this message appears:

Test requires equation estimated with random effects.
Please re-estimate equation.

What do I have to change?

Thanks in Advance

Re: hausman test

Posted: Wed Nov 15, 2017 8:31 am
by enilorac
And my second question:

which random effects do I have to choose?


Cross section or period?

Thanks a lot!!

Re: hausman test

Posted: Wed Nov 15, 2017 10:01 am
by bfluff
In my limited knowledge of econometrics, the random effects model assumes the error term is uncorrelated with the explanatory variables. Fixed effects allows for endogeneity. The Hausman test checks this assumption. Thus it follows that the you can test the null hypothesis that there is no endogeneity in the random effects model and if it is found is thus rejected. It would be a bit more conceptually difficult to test a system which assumes endogeneity and prove it doesn't contain any.

Does this make sense? This is how I understood it.

As for the options for testing, I would run Hausman tests with random effects on each dimension individually and see what the result is. That should then tell you what you need to know.

Just so you know, I don't really know what I'm talking about, this is just how I've worked through it in my head.

Re: hausman test

Posted: Thu Nov 16, 2017 12:37 am
by enilorac
Thanks for your answer.

Unfortunately, I still don't know which one I should use.
Maybe some words to the data: Panel Data, 48 States during the time period 1970-1986 were observed.