A New Study of Labor Market Concentration

March 5, 2018

Labor economists have traditionally focused on worker-side characteristics, such as education, as the crucial causal variable for explaining outcomes like earnings, unemployment, and inequality. But that point of view depends on labor markets remaining competitive, so workers can earn their marginal product of labor—because if they earned less, they’d leave for another job. What a growing body of literature shows is that when labor markets aren’t competitive, employer power—monopsony—not a lack of education, is what holds workers back–from switching jobs, migrating, or from starting a new business. There’s a strong circumstantial case to be made that monopsony is behind a litany of labor market problems at the center of public debate in recent years.

In a paper we released last December, economists Ioana Marinescu, Jose Azar, and I showed that labor markets in the U.S. are highly concentrated, and that in a given labor market, higher concentration causes a reduction in posted wages. That paper was based on data from the online job-posting website CareerBuilder, which covers only about a third of job postings. Furthermore, we only used the occupations that appear most frequently in the dataset, potentially leaving large swathes of the labor market out.

In a new paper, “Concentration in U.S. Labor Markets: Evidence from Online Vacancy Data,” we extend our estimates of labor market concentration to cover the near-universe of online job postings in the United States for the year 2016. To do that, we use a dataset from the online labor market data company Burning Glass, which scrapes job posting information from thousands of websites to populate a vacancies database with rich information for the characteristics of each vacancy. In this new paper, Ioana, Jose, and I are joined by Burning Glass Chief Economist Bledi Taska.

Our latest results confirm our findings in “Labor Market Concentration”: The average labor market is very highly concentrated. By way of background, in antitrust policy, defining the relevant market is a crucial piece of the analysis. Market definition matters for both buyers and sellers (of labor, in our case): If you’re a worker looking for a job, where are you likely to look, and how many different employers, in which locations, are viable options? Alternatively, if you’re an employer, whom are you likely to consider when hiring?

We define the relevant labor market as a commuting zone (essentially, a metropolitan area defined by commuting patterns) by “SOC-6” occupation by quarter. “SOC” is a hierarchical occupational classification system created by the U.S. Bureau of Labor Statistics. The average level of concentration (the Herfindahl-Hirschman Index) in a market defined that way is 3953, well in excess of the threshold for a “highly concentrated” market in the antitrust authorities’ Horizontal Merger Guidelines, which is 2500.

When we weigh labor markets by their level of employment, as measured by the BLS, we find that 17 percent of U.S. workers work in highly-concentrated markets, and a further 6 percent in “somewhat concentrated” markets. We provide concentration estimates for a number of different permutations of labor market definitions.

A map of the average concentration of the 200 occupations that appear most frequently in the Burning Glass data, by Commuting Zone.

We also address the question of market definition in labor markets, as well as its correspondence to key issues in the theory of labor market monopsony. Previous literature that allowed for firm-level wage-setting finds that labor supply elasticity to the individual firm is low and hence that firms have the power to set wages below the market product of labor. Using traditional market definition techniques from Industrial Organization, we conclude that defining labor markets in terms of commuting zones and SOC-6 occupations is “conservative,” in the sense that it likely underestimates the degree of market power enjoyed by employers by over-estimating the likely alternative options functionally available to workers.

Since we released our last paper, two others have explored its ramifications for antitrust policy. Hovenkamp and Marinescu (2018) focus on how to strengthen merger enforcement to protect against the threat of labor market monopsony. Naidu, Posner, and Weyl (2018) offer a comprehensive intellectual history of labor market monopsony and provide scenarios estimating its substantial impact on aggregate welfare, given estimates of firm-level labor supply elasticity from the literature. Those authors also delve into its implications for antitrust, including both merger control and anti-competitive conduct. Finally, Benmelech, Bergman, and Kim (2018) offer independent estimates of labor market concentration and its wage effect, covering manufacturing industries over a far longer time period than our data in either this or our previous paper covers. These authors show that monopsony is on the rise.

That literature, and a growing chorus of commentary on it in the media, among policy organizations, and in the financial sector, links monopsony to a wide range of worrying trends in the labor market, which we touch on in our own paper as well. Among these are declining ‘business dynamism,’ labor mobility, and entrepreneurship. The causal arrow between monopsony power and documented concentration likely goes in both directions, through several different mechanisms. First of all, most directly, concentration creates monopsony power on the part of incumbent firms, which means lower demand for labor. That reduces outside job opportunities for employed workers and hence labor mobility, and it encourages the casualization of the labor force–the so-called ‘gig economy’ or Fissured Workplace. Second, fewer small and fast-growing businesses hiring in a given labor market means fewer posted jobs, and therefore higher measured concentration. Third, by attenuating the job ladder, monopsony power also means a given job is vacated less frequently, thus increasing concentration as measured by posted vacancies. Fourth, monopsony reduces the economic security of workers who might consider forming their own businesses, if their opportunities for re-employment were secure. Instead, a concentrated labor market means they risk losing everything if they need to return to it. Fifth, independent contractors and gig economy workers can (and should) themselves be considered small businesses, and yet the antitrust laws are being used to undermine any attempt on their part to improve their situations–contributing to more powerful de facto employers’ control over their lives and earnings. Instead, given the balance of power between small and large businesses, antitrust law should scrutinize those powerful tech platforms and their monopsony power.

Altogether, the concerns elevated by this growing literature focus public and academic attention on the potential for concentrated market power to harm workers, small business formation, labor market dynamism, and the economy as a whole.