The first sets of regressions consider the role of the exchange rates and other variables in moving industry labor market variables across pooled groups of industries. In addition to pooling across the full sample of manufacturing industries, we compare industry responses in High versus Low Markup industry groups. Specifically, this sample of industries is split according to the median level of average price-over-cost markup across the group of manufacturing industries.
The second set of regressions are for individual industries. While industry-specific regressions are ultimately what one would want, they each have too few observations to fully stand on their own merits. Here For each industry we will show the effects of exchange rates on it’s employment, hours, wages, and overtime activity.
Pooled Industry Regressions.
The estimated results from equations (13a) and (13b) are reported in Tables 1 and 2. Table 1 reports the estimated coefficients from using total industry wages, total hours employed, and total industry employment. For each dependent variable we report the results from pooling the full sample of manufacturing industries and from splitting the sample between high and low markup industry markups.
The specifications also include as regressors a set of industry dummies, the interest rate, the price of oil, the annual value of GDP, and the lagged values of our measure of employment for each industry. All variables other than lagged employment are expressed in log differences (except for the interest rate which is just in percentage differences). We allow for industry specific speeds of adjustment to shocks by letting the coefficient on lagged industry employment be industry specific (not reported).
Table 2 reports the estimated coefficients from analogous regressions using as the dependent variable the measures of industry overtime wage and overtime employment. These regressions are similar in format to those of Table 1 with two notable exceptions. First, we assume that since changes in the amount of overtime activity is a short-run, temporary practice, it is not subject to the same adjustment costs as changes in permanent employment.
Therefore, we estimate a static model instead of a partial-adjustment model in the overtime regressions. This implies that changing the size of the labor force through (limited) use of overtime has no cost other than the corresponding wage. We then drop from the estimation the lagged level of industry employment as a regressor. Second, since the model is static the amount of overtime activity is independent of whether a given exchange rate shock is permanent or temporary. We use actual (not just permanent) movements in exchange rates as the right-hand-side regressor.