Arin Dube rebuts Casey Mulligan's contention that "supply is the primary reason that jobs are created during the summer" (Mulligan is *trying* to counter this post from Tim Duy):
Sweating over Seasonality: Is the Summer Surge all about Labor Supply? In a series of blog posts (most recently here), Casey Mulligan has argued rise in summer employment when school is out shows that labor supply always matters, even in a recession. Mulligan states that “[t]he economy creates jobs in the summer — even during the last several years, when our economy supposedly suffered from a lack of demand — because millions of people become willing and available to work,” and that even today “greater labor supply remains one route to higher employment.” However, as I show in this note, drawing this policy implication from seasonality, however, is a highly problematic exercise.
Most generally, production decisions over the whole year are planned knowing the greater availability of teens and young adults during the summer months. For example, some businesses and households plan to wait for the summer months to hire painters or landscapers. Second, demand for some types of work like camp counselors and babysitters rise during the summer months precisely because that’s when school is out. For both these reasons, employment changes in the summer can easily reflect higher demand during these months as well as an increase in labor supply. Some of this increased demand comes from inter-temporal substitution across months (the first example), while other part of this reflects a higher net demand (the second example).
Mulligan’s main evidence that summer is all about labor supply and not demand is that for teens and young adults, unemployment rises and average wages fall. However, this is not a great test – as there is likely to be systematic differences in the composition of both workers and jobs between the school year and the summer – precisely because a lot more youth work during the summer. Then there are data related challenges, like how to deal with unpaid (or barely paid) internships. For these reasons, looking at changes in average wages between summer and non-summer months for youth is problematic.
However, there is one simple way of assessing the claim that the summer surge in employment is only about labor supply – and that is to look at job vacancies. If summer is mainly a time of a big labor supply shock, we should not see any systematic seasonal patterns in job openings. Moreover, we should expect to see a lot of labor market slack – which economists typically measure using the “unemployment-to-vacancy” (U-to-V) ratio. The higher is the ratio, the more unemployed workers there are for a given job opening: if Mulligan is correct we should see the rise in employment (and unemployment rate) be matched by a parallel sharp rise in the U-to-V ratio, while vacancies should be stable.
To test this, I obtained the seasonally unadjusted monthly “job openings” rate from BLS between 2001 and 2010 (this comes from the JOLTS data). For the same years, I also obtained the seasonally-unadjusted monthly unemployment and employment-to-population (EPOP) rates between 2001 and 2010 for ages 16 or greater. To construct the average seasonality measure for a variable x, I first calculate the deviation of x from its annual mean, i.e., x(month,year) – x(year). Next I average the mean deviated values across the years. This is a standard measure – in technical terms, it measures the “month fixed effects” after taking out “year fixed effects.” It represents the “excess value of x” in a given month from its annual average.
Below I plot the monthly excesses of EPOP, unemployment, vacancy and U-to-V rates.
The top two graphs confirm that the summer months of June and July see an increase in employment (roughly 0.5 percentage point), accompanied by an increase in unemployment rate (roughly 0.2 percentage point). These two pieces of evidence would be consistent with a “labor supply only” story. However it turns out that job openings also increase over those months (by roughly 0.1 percentage point). This is particularly true for July, when the unemployment and vacancy rates move up by roughly equal amounts. As a result, the overall unemployment-to-vacancy rates show much less seasonality during the summer than the unemployment rate. Averaged together, June and July do not look like a period with particularly high amount of slack in the labor market as measured by the U-to-V ratio – which would have been the case if the only thing changing in the summer were labor supply. While there is a big “summer excess” in employment and also some in unemployment, it is much less true for the unemployment-to-vacancy rate – because job vacancies also tend to rise. That should be a big flashing warning sign to anyone wanting to attribute the summer surge to only labor supply factors.
So what does this tell us about the efficacy of Keynesian demand management? Nothing really, and that’s the point. For that, you’d have to look elsewhere – in whatever season.
Dept of Economics
University of Massachusetts-Amherst