Question I
The file CEOSAL2.wfl contains data on 177 chief executive officers. The variables are defined as follows: 1. salary – 1990 compensation, $1000s 2. age – in years 3. college – =1 if attended college, 0 otherwise 4. grad – =1 if attended graduate school, 0 otherwise 5. comten – years with company 6. ceoten – years as CEO with company 7. sales – 1990 firm sales, millions 8. profits – 1990 profits, millions 9. mktval – market value, end 1990, mills. 10. lsalary – ln(salary) 11. lsales – ln(sales) 12. lmktval – ln(mktval) 13. comtensq – comtenˆ2 14. ceotensq – ceotenˆ2 15. profmarg – profits as % of sales We would like to investigate how CEO compensations are determined. Use lsalary as dependent variable to conduct the analysis. Include a constant, age, college, grad, comten, ceoten, lsales, profits, lmktval as independent variables. Please note that a one unit change in ln(z) can be interpreted as a 100% change in z. 1. Assess the overall goodness of fit the model. (You could imagine how “bad” this model should be given your prior belief that CEO compensations were “out of whack.”) Answer: The regression output is as follows. As can be seen, the R2 is very low, 0.355378, which means only about 35% of the sample variation in lsalary is explained by the model; so the model does not fit well.
1
Dependent Variable: LSALARY Method: Least Squares Sample: 1 177 Included observations: 177 Variable C AGE COLLEGE GRAD COMTEN CEOTEN LSALES PROFITS LMKTVAL R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) Coefficient 4.781665 7.72E-05 -0.065452 -0.094189 -0.010608 0.016701 0.191536 6.41E-05 0.083520 0.355378 0.324681 0.498046 41.67239 -123.1536 11.57722 0.000000 Std. Error 0.514250 0.005289 0.233712 0.079397 0.003724 0.005746 0.040237 0.000149 0.062977 t-Statistic 9.298328
