How Widespread Is Labor Monopsony? Some New Results Suggest It’s Pervasive.

Last week, Ioana Marinescu, Jose Azar, and I released a working paper that seeks to answer the question, “How prevalent is labor market monopsony in the United States?” In short, we find that most labor markets (as defined by occupation and geography) are very concentrated, and that that concentration has a robust negative impact on posted wages for job openings. These findings are drawn from only one (large) dataset, so they are not the last word on the subject. The implication, along with the other papers cited above, is that the antitrust authorities should not operate under the assumption that labor markets are “naturally” competitive. Instead, they should consider whether existing policy adequately ensures that market power is not concentrated on the employer side, and their future policy should center the issue of monopsony power.

Data Analysis

To determine the breadth of labor market monopsony, we look at a unique dataset: the universe of posted job vacancies on the website CareerBuilder, the largest online job board, for 2010-2013. We define labor markets by occupation (at the “SOC-6” level) and geography, given previous research showing job search patterns are geographically confined. Specifically, we treat “commuting zones” as the empirical analog to a local labor market—in fact, the Census Bureau created the commuting zone classification for exactly that purpose, based on observed commuting patterns. Within each labor market, we calculate the Herfindahl-Hirschman Index (“HHI”) depending on the vacancy and application share by firm, in a given quarter.

We find that the average labor market is “highly concentrated” (HHI > 2500) by the standards of the federal antitrust authorities’ Horizontal Merger Guidelines.

This map plots the average HHI across occupations (and quarters) in each commuting zone. Commuting zones shaded orange and red have HHI greater than 2500, based on the CareerBuilder data. The only labor markets that are unconcentrated by the standard of the Horizontal Merger Guidelines are the largest metropolitan areas.

Furthermore, we find that labor market concentration in our sample is negatively correlated with posted wages: going from the 25th to the 75th percentile of the concentration distribution is associated with a 15-25 percent decline in wages.

We use fixed effects for both commuting zone and occupation, by quarter, thus estimating the relationship between concentration and wages solely off of the quarter-to-quarter difference in concentration, averaging out the level differences for each commuting zone and quarter. That prevents either long-standing geography- or occupation-specific differences in vacancy posting (for instance, in small versus large commuting zones by population) or national-level differences in vacancy-posting by quarter (for example, due to the changing state of the business cycle) from influencing the estimated relationship between concentration and wages. It also means that even if you observe all vacancies, not just the CareerBuilder ones, you might observe lower average levels of concentration, but the effect of varying concentration on wages would probably not be much different.

In order to be able to make a stronger, causal claim about the effect of market concentration on wages, we undertake several additional empirical exercises. We instrument for local labor market concentration with the market concentration for the same occupation in other geographic labor markets. The idea is that we want to make sure our results are not driven by local labor demand effects that are reducing both vacancy posting (hence increasing concentration) as well as reducing wages. We also independently measure labor market tightness using the observed ratio of vacancies to applications, since we can see the number of applications to each vacancy in our data. This “soaks up” a large chunk of the observed variation in vacancies, but even so, the market concentration as measured by vacancies is still robustly associated with a reduction in wages.

We also aggregate our data over longer time periods than a single quarter, which understandably reduces concentration—more vacancies, by a larger range of individual firms, are posted the longer you extend the time period. Two things are worth noting, however: the average job search is around 10 weeks, according to Bureau of Labor Statistics (BLS) data, meaning that workers looking for a job are generally in the market for around a quarter; hence, the quarterly concentration is what matters to them in determining how likely they are to receive one or more job offers. This is why it’s more meaningful to look at concentration among job vacancies than among employment. In other words, it may be the case that many firms employ people who do jobs similar to what a job-seeker is looking for, but if only a few firms are actually hiring (and especially if there’s less turnover because the job ladder has deteriorated), then that unconcentrated employment is irrelevant. Second, the rate of establishment churn in the labor market is high—even though it has been on the decline for the last two business cycles. This means that the longer any sample of vacancies is, the more different establishments and firms will be represented—but not necessarily at the same time. Thus, aggregating across time mechanically reduces measured concentration.

Our analysis has several shortcomings that might get in the way of external validity: we only have CareerBuilder data, which, although the largest online vacancy-posting website, is not necessarily representative of the entire U.S. labor market, and in particular there may be some occupational categories that are unlikely to be recruited online. In addition, only around 20 percent of online job ads post a wage, and that wage may or may not be the actual wage that the final candidate receives. That said, we do observe robust sample sizes for all of the largest occupations, and we validate our findings in the BLS wage data.

Finally, our work on concentration as the potential cause of monopsony power differs markedly from the treatment of monopsony prevalent among labor economists, which sees it as arising from workers’ inelasticity of labor supply to the individual firm where they work. The idea is that since searching for a job is time-consuming and costly, relative to an employment relationship that may have “match-specific surplus,” workers have relatively low labor supply elasticity. From this perspective, firms have the power to push the wage they offer below the worker’s marginal product of labor, without him or her leaving for another job that offers to pay the full marginal product. That conception of monopsony arises from search theory, which is a newer theoretical apparatus than is market structure and concentration—yet the two are consistent with the same observed set of facts. Our story of why firms have wage-setting power is that since they have few competitors among would-be employers, their workers receive few outside job offers and hence can be forced to accept a lower wage.

Implications for Policy

The issue of market power in the labor market plays into a range of policy questions, including for the minimum wage, unionization, and macroeconomic stabilization, as well as what we focus on in our paper: antitrust. The treatment of the labor market in antitrust is somewhat inconsistent: On the one hand, merger guidelines and agency decisions to enforce the Sherman and Clayton Acts against monopsonistic conduct in labor markets imply that the potential for employer market power is substantial and therefore regulators must guard against it. On the other hand, enforcers have tended to view the labor market as “naturally” competitive, and hence that the greater threat to competition arises not from employer market power, but from the potential for rent-seeking on the part of workers—through collective bargaining, for example.

A recent DOJ-FTC brief arguing that antitrust scrutiny should be applied to collective bargaining by ridesharing drivers is a case in point: The authorities evidently believe that Uber’s coordination of prices and wages across thousands of independent contractors is pro-competitive, while the idea that drivers might bargain collectively on their own behalf ought to be exposed to scrutiny under the Sherman Act—and not just scrutiny, but in fact should be subject to a per-se finding that the conduct is illegal on its face. The fact that Uber has the market power to set wages in the status quo should be evidence enough that the labor market is monopsonized. And yet, rather than analyze which side of the ridesharing labor market in fact exercises market power, the authorities directed their attention to the drivers’ effort to bargain collectively—and to regulation from the City of Seattle granting them the right to do so.

This speaks to the larger problem of antitrust enforcement against labor monopsony: notwithstanding that the merger guidelines and other policies recognize the potential for its existence, the authorities evidently do not consider it a major threat. They have been far more active in attempting to combat state-level “occupational licensing” through litigation and moral suasion of state legislatures—which a recent paper shows is responsible for a mere 2 percent of the decline in labor market mobility, while at the same time it serves both to increase wages and reduce racial- and gender-wage gaps.

The core explanation for the antitrust authorities’ lax attitude toward labor market monopsony is the consumer welfare standard, which holds that the purpose of antitrust law is to ensure competitive outcomes for consumers, which in practice is equated with short-term price reductions. Just last week, the Senate held a hearing on the standard, with a range of past antitrust officials in agreement that it is doing a good job of ensuring the economy remains competitive.

But there are numerous reasons why the consumer welfare standard is inadequate, but among the foremost is that it neglects the issue of market power in supply chains and, specifically, in the labor market. Senator Cory Booker sent a letter to the authorities asking them how they enforce their mandate in the labor market, given the scant record of merger challenges based on monopsony theories of harm and the relatively narrow prosecutions of no-poaching agreements in the tech and healthcare sectors—both of which were settled for relatively light penalties. Ultimately, the mandate for antitrust enforcement ought to, and does, embrace both consumers and producers—ensuring all economic agents have access to the market, regardless of structural disadvantages like race and geography.

Indeed, there’s a strong argument that antitrust enforcement ought to be more stringent in the labor market than the generic product market. Consumers can probably afford to search for a washing machine over a longer period than they can afford to search for a job, and the market for washing machines is national, whereas job searches happen locally. This is a longstanding insight in labor economics—it is one of the central reasons why labor economists have long felt that canonical models of perfect competition do not apply to labor markets. And indeed, we find additional evidence that this empirical insight is correct: The most concentrated labor markets, and the ones where the effect of concentration on wages is largest, are the rural ones. Hence, market definition in antitrust must account for the greater necessity that workers find work quickly, because they cannot afford to remain unemployed for long, and also that their practical geographic scope for finding a job is a good deal smaller than for the typical consumer good.

In summary, the findings in this paper lend further support to the idea that monopsony power in the labor market is a practical economic problem that the antitrust status quo is not doing enough to solve. The prior papers by Barkai, Gutierrez and Philippon, and De Loecker and Eeckhout document declines in the labor share of national income and locate the culprit in lax antitrust policy and economy-wide concentration, but their focus is on concentration in the output market and at the industry level. We show that labor market concentration, by occupation and geography, exerts its own, independent, negative effect on wages, and thus warrants scrutiny under antitrust law—over and above whatever the consumer welfare standard might allow.