Keeping all these concerns in mind, the wage and hour regressions are re-estimated using the same functional forms as before but with three new variables; the percentage of employees with union status, the log of average education, and the measure of industry rents. The variable for percent female was not included in these regressions, as we believe that this is a supply side rather than a demand side variable. Also included were two interactive variables, between union status and capital intensity and between industry rents and capital intensity. These interaction terms allow for the greater market power that employees may have in capital-intensive industries. More info The inclusion of the interactive term between education and capital made the results highly unstable and increased the standard errors significantly, thus was dropped from the regressions.
We draw your attention to the set of supplemental regressions presented in Table 3.10. The adjusted R2 for each of the regressions has increased, although more so for the wage regressions than the hours regressions. The signs on the additional variables in the wage regression are mostly as would be expected, and mostly significant. The capital intensity variable has retained a significant positive impact on both hours and wages despite the inclusion of the other variables including education. When we include the respective interactive coefficients the magnitudes are quite similar to those of the initial regressions, although additional formal analysis will be performed to verify this.
The average education of production workers has a significant effect on average weekly wages that increases over time. There was a particularly sharp rise between 1980 and 1985, a result which conforms with the literature on changes in the returns to education over this period. Holding all else equal the average relative wage difference between a worker with two years of college and a high school dropout with 10 years of education increased from 31% in 1970 to 39% in 1980 to 55% in 1985 then dropping back slightly to 48% in 1993.
Table 3.10 Supplementary Log Wage and Log Hour Regressions
All variables in log form except union status.
Interestingly at the same time there was also a substantial change in the impact of education on effort. In 1970 the worker with 14 years of education worked on average 1.5 hours less than the high school dropout. By 1993 this had reversed such that the college educated worked 1 hour more per week on average than the high school dropout. This empirical fact sets us to thinking about possible explanations. One explanation we like is that new very-productive jobs emerged in the 1980s that required both high amounts of human capital and high inputs of physical capital (computers). But maybe it is the decline in the marginal tax rates, or the increasing cost of higher education, or something else entirely. Work is under way to answer this important question; why are the educated working so much harder today than they did 30 years ago?