Abstract: In this paper we use detailed job vacancy data to estimate changes in skill demand in the years since the Great Recession. The share of job vacancies requiring a bachelor’s degree increased by more than 60 percent between 2007 and 2019, with faster growth in professional occupations and high-wage cities. Since the labor market was becoming tighter over this period, cyclical “upskilling” is unlikely to explain our findings.
1 Introduction
The yearly wage premium for U.S. workers with a college degree has grown rapidly in
recent decades: from 40 percent in 1980 to nearly 70 percent in 2017 (Autor, Goldin, and
Katz 2020). Over the same period, the share of adults with at least a four-year college
degree doubled, from 17 to 34 percent (Snyder, de Brey, and Dillow, 2019) (Digest of
Education Statistics, 2020). In the “education race” model of Tinbergen (1974), these two
facts are explained by rapidly growing relative demand for college-level skills. If the
college premium grows despite a rapid increase in the supply of skills, this must mean
that the demand for skills is growing even faster.
The education race model provides a parsimonious and powerful explanation of US
educational wage differentials over the last two centuries (Katz and Murphy 1992; Goldin
and Katz 2008; Autor, Goldin, and Katz 2020). Yet one key limitation of the model is that
skill demand is not directly observed, but rather inferred as a residual that fits the facts
above. How do we know that the results from the education race model are driven by
rising employer skill demand, as opposed to some other unobserved explanation?
We study this question by using detailed job vacancy data to estimate the change in
employer skill demands in the years since the Great Recession. Our data come from the
labor market analytics firm Burning Glass Technologies (BGT), which has collected data
on the near-universe of online job vacancy postings since 2007.
Our main finding is that skill demand has increased substantially in the decade following the Great Recession. The share of online job vacancies requiring a bachelor’s degree
increased from 23 percent in 2007 to 37 percent in 2019, an increase of more than 60 percent. Most of this increase occurred between 2007 and 2010, consistent with the finding
that the Great Recession provided an opportunity for firms to upgrade skill requirements
in response to new technologies (Hershbein and Kahn 2018).
We present several pieces of evidence suggesting that the increase in skill demand is
structural, rather than cyclical. We replicate the findings of Hershbein and Kahn (2018)
and Modestino, Shoag, and Ballance (2019), who show that skill demands increased more
in labor markets that were harder hit by the Great Recession. However, when we extend
the sample forward and adjust for differences in the composition of online vacancies, we
find that this cyclical “upskilling” fades within a few years. In its place, we find longrun structural increases in skill demand across all labor markets. In fact, we show that
the increase in skill demand post-2010 is larger in higher-wage cities. We also find much
larger increases in the demand for education in professional, high-wage occupations such
as management, business, science and engineering.
Previous work using the BGT data has found that it is disproportionately comprised of
high-wage professional occupations, mostly because these types of jobs were more likely
to be posted online (e.g., Deming and Kahn 2018). As online job advertising has become
more common, the BGT sample has become more representative, and the firms and jobs
that are added later in the sample period are less likely to require bachelor’s degrees and
other advanced skills.
We adjust for the changing composition of the sample in two ways. First, we weight
all of our results by the employment share of each occupation as well as the size of the
labor force in each city in 2006. This ensures that our sample of vacancies is roughly
representative of the national job distribution in the pre-sample period. Second, our preferred empirical specification controls for occupation-by-MSA-by-firm fixed effects. This
approach accounts for compositional changes over time in the BGT data.
Our results suggest that increasing demand for educated workers is likely a persistent
feature of the U.S. economy post-recession. Recent work has documented a slowdown
in the growth of the college wage premium since 2005 (Beaudry, Green, and Sand 2016;
Valletta 2018; Autor, Goldin, and Katz 2020). Yet this slowdown has occurred during a
period of rapid expansion in the supply of skills. We find rapid expansion in the demand
for skills, suggesting that education and technology are “racing” together to hold the
college wage premium steady.1
Posts
No comments:
Post a Comment