The table above shows results from a regression of log wages on a dummy for whether a job has pay linked to performance (e.g. salespeople paid on commission) and other variables. The data are panel data on workers. In addition to the reported coefficients, the regressions include industry, occupation, and year dummies; county unemployment; and marital status, race dummies, and union status. Standard errors are in parentheses.

Based on the third column, is the return to education higher at performance pay jobs or non-performance pay jobs? What is the difference and is it statistically significant?

According to the third column, what is the return to having a performance pay job for somebody with a college degree (16 years of education), 20 years of experience, and 10 years of tenure?

The fourth regression includes worker-level fixed effects. The coefficient on years of education falls from .0637 in third column to .0167 in the fourth. Is this a large change in economic terms? Explain.

Give an explanation for the difference in the coefficients discussed in question 3 (.0637 vs. .0167). The explanation should be concrete.

Consider three possible ways to compute standard errors for the regressions in the table: homoskedasticity-only; heteroskedasticity-robust; and clustered at the individual-job level. Which is the most appropriate method, and why?