We use the employment rate to define the cycle. As is evident in Figure 3.2, weekly hours leads the employment rate, a feature we attribute to a delay between an increase/reduction in product demand and the actual hiring/firing of workers. To capture the cycle in demand, we use the smoothed forward rate of employment as the indicator of the business cycle, and we compare peak to peak and trough to trough changes in the wage-effort offer curve. Using smoother year-ahead rate of employment again as our guide we set our trough years as 61-62, 70-71, 75-76, 81-82, and 91-92. The peak years were set at 58-59, 67-68, 72-73, 77-78, 88-89 and 93-94. The results of these peak and trough regressions are displayed in Table 3.7 review.
These regressions can be used to solve for the wage-effort offer curves like those displayed in Figures 3.5 through 3.7. When we do so the peak-to-peak and the trough-to-trough comparisons are completely in line with our first results that ignored the business cycle. In the 1960s the wage-effort offer curve moved up and to the right, in the 1970s it twisted, and in the 1980s it shifted sharply to the right. Additionally when we put consecutive peak and trough years next to each other such as 70,71-72,73 and 77,78-81,82 we find very little difference between the shapes of these curves. The trends we see in the wage-effort curves tend to be long term and appear to occur exogenous to cyclical effects.
Evaluation of Alternative Explanations of the Wage-Effort curve
We are excited by how well these results conform with the theory, but we need to be alert to the possibility that these findings are driven by some third factor that has nothing to do with effort. Our number one concern is that human capital is correlated with physical capital and with hours, and that what we are observing is not compensation for effort but compensation for skill and skilled workers choosing longer hours. Unionization is also a concern. Unions might be able to bargain for a wage-effort contract above the competitive market curve. A strong union effect might account for the outlying sectors observed in the 2-digit level data. Even without unions, profit sharing may help to explain the pattern of wages. The realized returns to capital vary widely across sectors and also across time. Firms in the less competitive sectors may collect positive rents and may share those rents with workers.
The Bartelsman data set does not include information on unionization or education. We have formed industry estimates of production worker education from the Current Population Surveys. To do this we had to match the 71 3-digit CPS manufacturing industries with the 448 industries in the NBER productivity database.13 Data from Kokkelenberg and Sockell(1985) on union status was used for the 1973 to 1981 period while data from Hirsch and Macpherson(1993) were used for the later years. These measures are also created from the CPS data, so they match with the NBER data in the same way that the education measures do.