A Princeton professor investigated the effect of computer skills on wages (Alan B. Krueger, How Computers Have Changed the Wage Structure: Evidence from Microdata, 1984-1989 Quarterly Journal of Economics, 108(1), 1993). He had no direct measure of computer skills, but he could observe whether a person used a computer at work. The data came from the 1984 and 1989 versions of the Current Population Survey, a nationally representative survey conducted by the Census Bureau.
Two tables from the paper are reproduced below. Table I reports descriptive statistics for workers who use a computer at work for each of the sample years. Table II gives the results of OLS regressions of log(hourly wage) on a dummy for computer usage, and other variables. Each column in Table II reports the results from a different regression. The unit of observation is an individual.
a)
Table II, Column (1) reports the regression of log(wage) on the computer dummy (Uses computer at work (1=yes)). What is the interpretation of the coefficient? Column (4) reports a similar regression for 1989. What happened to the importance of computers between the two sample years in explaining expected log(wage)?
b)
Why does R increase in regressions (2) and (3) compared to regression (1)?
c)
Under what conditions could the computer coefficients in regressions (3) and (6) be interpreted as causal? What are some specific reasons that would bias these coefficients?
d)
Interpret the coefficient on Married*Female.
e)
What is the marginal effect of a year of experience? When would an additional year of experience imply a decrease in the expected log wage?
