Hausman Test
Posted: Fri Apr 28, 2017 9:40 am
Hi all,
I have a question regarding Hausman Test. I'm doing a research on systematic risk (dependent variable) in 1 country. The panel data consists of 40 companies and period of 10 years (2007-2016) - 400 balanced panel data.
I have 13 Independent Variables (IVs):
- 8 IVs are company-specific variables that vary between companies
- 3 IVs are dummy variables for financial crisis 2008, 2012, 2015.
- 2 IVs are macroeconomics variable (exchange rate & inflation rate) which values are different from year to year but constant for all companies.
I run POLS, Fixed Effect and Random Effect models. When I conduct Hausman Test, the output says: * Cross-section test variance is invalid. Hausman statistic set to zero. Since the chi-square statistics is set to zero, the probability is 1.0000
My questions:
What does this result mean? Is this result is invalid since it says *Cross-section test variance is invalid. Hausman statistic set to zero, and I should choose Fixed Effect Model?
Or is this result OK, and it simply means that I should choose Random Effects Model? But the Random Effect Model is very bad with low f-statistics and low adjusted R squared. The Fixed Effect Model is much better than Random Effect Model.
I have a question regarding Hausman Test. I'm doing a research on systematic risk (dependent variable) in 1 country. The panel data consists of 40 companies and period of 10 years (2007-2016) - 400 balanced panel data.
I have 13 Independent Variables (IVs):
- 8 IVs are company-specific variables that vary between companies
- 3 IVs are dummy variables for financial crisis 2008, 2012, 2015.
- 2 IVs are macroeconomics variable (exchange rate & inflation rate) which values are different from year to year but constant for all companies.
I run POLS, Fixed Effect and Random Effect models. When I conduct Hausman Test, the output says: * Cross-section test variance is invalid. Hausman statistic set to zero. Since the chi-square statistics is set to zero, the probability is 1.0000
My questions:
What does this result mean? Is this result is invalid since it says *Cross-section test variance is invalid. Hausman statistic set to zero, and I should choose Fixed Effect Model?
Or is this result OK, and it simply means that I should choose Random Effects Model? But the Random Effect Model is very bad with low f-statistics and low adjusted R squared. The Fixed Effect Model is much better than Random Effect Model.