Employment sometimes falls, and in theory the fall could occur because of a change in supply, a change in demand, or both. The best way to determine which is more important is to examine labor productivity: productivity is pro-cyclical when the employment cycle is a result of a demand shift, and counter-cyclical when it is a result of a supply shift.
Some economists have proposed using quits and layoffs to make this distinction. When labor supply causes the employment reduction, supposedly quits will be high and layoffs low.
However, as JOLTS measure quits and layoffs, the two are often POSITIVELY correlated. Thus, there are sometimes supply shocks that increase both quits and layoffs, and there are sometimes demand shocks that increase both quits and layoffs.
Consider, for example, the seasonals of layoffs and quits in the JOLTS data, which I measure in the monthly data as the residual from a regression of either layoffs or quits on a smooth polynomial in time. The nationally quit and layoff residuals are POSITIVELY correlated, even if I exclude Oct-Jan which is the seasonal most obviously associated with a demand shift.
I found the same pattern for the southern region only, where the changes May-July and July-Sep should be more about supply (more workers are available when school is not in session) than demand (perhaps in the north cold weather in the spring or fall affects demand).
I found the same pattern for each of the 25 industries tracked by JOLTS, with only two exceptions: mining and construction. Even in those two industries, layoff spikes seem to be associated with quit spikes, its just that quits are somewhat above average when there are several consecutive months of low layoffs.
Some economists have proposed using quits and layoffs to make this distinction. When labor supply causes the employment reduction, supposedly quits will be high and layoffs low.
However, as JOLTS measure quits and layoffs, the two are often POSITIVELY correlated. Thus, there are sometimes supply shocks that increase both quits and layoffs, and there are sometimes demand shocks that increase both quits and layoffs.
Consider, for example, the seasonals of layoffs and quits in the JOLTS data, which I measure in the monthly data as the residual from a regression of either layoffs or quits on a smooth polynomial in time. The nationally quit and layoff residuals are POSITIVELY correlated, even if I exclude Oct-Jan which is the seasonal most obviously associated with a demand shift.
I found the same pattern for the southern region only, where the changes May-July and July-Sep should be more about supply (more workers are available when school is not in session) than demand (perhaps in the north cold weather in the spring or fall affects demand).
I found the same pattern for each of the 25 industries tracked by JOLTS, with only two exceptions: mining and construction. Even in those two industries, layoff spikes seem to be associated with quit spikes, its just that quits are somewhat above average when there are several consecutive months of low layoffs.
No comments:
Post a Comment