Note to table 5 on page 29
Hausman test statistics are reported for each of my 5 models in table 5. I made mistakes when adding asterisks to the Hausman test statistics. In this note I will briefly explain the Hausman test and correct the mistakes.
The Hausman test
The Hausman test is used to determine whether Generalized Least Squares (GLS) estimation with random effects is consistent in its estimation of coefficients by comparing it to a GLS estimation with fixed effects. The Hausman test is based on the idea that when the random component of the error term is not correlated with any of the regressors, both fixed effects and random effects are consistent. The Hausman test tests the following hypotheses:
: Both random effects and fixed effects are consistent; the difference in coefficients between them is not systematic
: Only fixed effects is consistent; the difference in coefficients between fixed and random effects is systematic.
Hausman test results for my thesis
Conducting the Hausman test in STATA produces a test statistic, which is reported in table 5, as well as a Chi-square statistic or p-value. When the p-value exceeds the chosen value for α one fails to reject and can thus conclude that GLS estimation with random effects is consistent. Reported values for the test per model were:
Board Size Independent Directors Executive Directors Factor 1 Board Meetings Hausman 11.70 3.23 7.18 14.71 3.54 P-value 0.0690 0.7799 0.3042 0.0226 0.7389
