The growing focus on labor-market power and antitrust enforcement has sparked important debates about both the empirical foundations and practical implementation of these emerging policy priorities. In a recent piece for ProMarket, Eric Posner argues that overwhelming academic evidence supports expanding antitrust scrutiny of labor markets—criticizing, in particular, the skepticism expressed by the Federal Trade Commission’s (FTC) Melissa Holyoak of labor-related provisions in the FTC’s originally proposed Hart-Scott-Rodino (HSR) rule changes.
The changes, proposed in 2023, would have added requirements for firms to collect and file information on employee characteristics, past workplace-safety issues, and local labor-market competition. The articulated intent was to allow the FTC “to screen for potential labor-market effects arising from the transaction.” Ultimately, every member of the commission—not just Commissioner Holyoak—voted for new HSR rules that eliminated the proposed labor-market-information obligations.
As we discuss in a recent paper (forthcoming in the DePaul Law Review, and cited by Holyoak), a careful review of the evidence suggests that Posner’s confidence in the evidentiary support for expanded labor-market enforcement is premature—both in terms of the underlying research on labor-market power and the proposed regulatory tools to address it.
The Research on Labor-Market Merger Effects Is Not Particularly Robust
Right off the bat, it is worth noting that, for all the confident claims some make about the need for more antitrust in labor markets, the direct evidence of the effects of mergers on labor outcomes consists of only a few papers. Posner admits that “a smaller number of papers have addressed the question whether mergers between firms with large shares of a common labor market result in wage reductions.” In fact, the two papers he cites—one by Elena Prager and Matthew Schmitt and another by David Arnold—constitute the bulk of the evidence. And even these provide a mixed bag when it comes to mergers.
Consider first Prager and Schmitt’s hospital-merger study. Posner writes that they:
found that mergers between nearby hospitals slow wage growth of medical professionals but not for employees who do not have hospital-specific skills—demonstrating that the mergers harm workers in concentrated labor markets but not unconcentrated labor markets, as theory predicts.
What Prager and Schmitt actually found is slightly different. Wage growth slows only when two conditions are simultaneously met:
- the merger causes a large increase in concentration (not a high level, necessarily); and
- workers have industry-specific skills.
In all other cases—including hospital mergers that increase concentration, but where the workers (such as, e.g., janitors) do not have industry-specific skills—they find no wage effects.
Commissioner Holyoak’s statement stressed those who were unaffected. Of course, it is worth noting both effects. But the absence of merger effects in some labor markets is surely relevant to the appropriateness of a broad rule that applies to all mergers.
The other main paper is Arnold’s working paper, which looks across industries. Posner correctly notes that Arnold finds that “mergers in highly concentrated labor markets reduce wages significantly.” In the most concentrated markets (top 5%), wages drop by 3%.
There are, however, a few things worth noting. First, by construction, Arnold is estimating only the effect on wages for the non-merging parties (i.e., the rest of the firms in the market). Second, Arnold actually concludes that “the evidence does not support… that lack of antitrust scrutiny for labor markets has been a major contributor to… stagnant wage growth.”
The third commonly cited paper in this literature (not discussed by Posner) is by Efraim Benmelech, Nittai Bergman, and Hyunseob Kim. They look at the manufacturing sector, which accounts for roughly 8% of the labor market in the United States. The authors do not directly regress wages on mergers themselves. Instead, they use mergers as an instrumental variable to measure employer concentration. Like Arnold, they find that mergers reduce wages.
While there may be other relevant research, these papers represent the core direct evidence on the causal relationship between mergers and wages in the academic literature. This thin empirical base should give pause to those advocating dramatic, economy-wide policy changes. Hospital mergers are crucial to understand and for the agencies to examine. But how much do they tell us about, say, grocery or manufacturing labor markets?
Which points to a further problem: the findings of even these few papers are far more qualified than Posner suggests. The hospital-merger findings suggest that effects are quite targeted, occurring only in cases where there is both a large concentration increase and industry-specific skills. This nuance matters enormously for policy. Surely, grocery workers have considerably fewer industry-specific skills than hospital workers.
Are findings from a paper dependent on the extent of industry-specific skills really so generalizable across the enormous diversity of U.S. labor markets? We would argue that what Prager and Schmitt’s findings counsel for most strongly is careful market definition and analysis of worker mobility—not broad presumptions about labor-market harms that would affect all mergers through expanded HSR requirements.
Meanwhile, Arnold’s focus on non-merging firms and modest wage effects in the most extreme cases raises questions about magnitude. Indeed, Arnold’s own conclusion that inadequate antitrust scrutiny isn’t driving wage stagnation directly contradicts Posner’s claim that the evidence demands aggressive reform.
The Evidence’s Implications for Labor-Market Power Are Unclear
Rather than rely on a large literature of past mergers and their wage effects (which does not exist), Posner refers to the academic literature as providing clear evidence of widespread labor-market power. But as we discuss in our paper, the empirical case is far more nuanced and uncertain than Posner suggests.
Posner takes FTC Commissioner Holyoak to task for citing “numerous academic papers, most of them published in peer-reviewed journals or on their way” in support of her claim because, according to Posner, “nearly all of [these papers] show that labor market concentration, or similar frictions in labor markets, is a serious problem.”
Of course, Holyoak doesn’t contest the existence of some concentrated labor markets. What she contests is the assertion that the existence of concentrated markets is a sufficient basis to dramatically overhaul the HSR rules:
While concentration levels may have a role in antitrust enforcement (e.g., merger presumptions), general and imprecise observations of increased concentration are a slender reed upon which to base such a significant expansion of HSR authority. These limitations also apply in the labor context.
What’s curious is that Posner himself has previously reached the same conclusion. In a 2021 paper with Suresh Naidu in the Journal of Human Resources titled “Labor Monopsony and the Limits of the Law,” while arguing that “[a] growing body of empirical literature indicates that labor market monopsony is widespread,” Posner nevertheless concludes that:
[A] significant degree of labor market power is “frictional,” that is, without artificial barriers to entry or excessive concentration of employment. If monopsony is pervasive under conditions of laissez-faire, antitrust is likely to play only a partial role in remedying it, and other legal and policy instruments to intervene in the labor market will be required. (emphasis added).
Indeed, the assumption that increased concentration is a cause of antitrust harms is a tenuous one. As Holyoak notes (citing our paper, itself quoting Steven Berry, Martin Gaynor, and Fiona Scott Morton):
A main difficulty in [the monopsony power literature] is that most of the existing studies of monopsony and wages follow the structure-conduct-performance paradigm; that is, they argue that greater concentration of employers can be applied to labor markets and then proceed to estimate regressions of wages on measures of concentration. [S]tudies like this may provide some interesting descriptions of concentration and wages but are not ultimately informative about whether monopsony power has grown and is depressing wages (emphasis added).
Moreover, as our paper also notes, “[m]any factors other than concentration can affect wages, such as differences in firm productivity, local labor-market conditions (e.g., urban vs. rural), and institutional factors like unionization rates.” These are those “similar frictions in labor markets” that Posner now bundles in the same sentence as concentration, but that he previously recognized are not connected to antitrust or merger enforcement. Given these realities, dramatically expanding the scope and cost of the HSR-filing requirements is simply unjustified.
Indeed, the Extent of Labor-Market Power Itself Is Unclear
But it isn’t just a conceptual problem. Posner asserts that the literature conclusively demonstrates that labor markets are plagued by buyer (employer) market power and that mergers cause harm in labor markets. But neither of these is settled by the literature.
First, direct measures of labor-market power, such as labor-supply elasticity estimates, vary widely across studies: from 0.1 to 4.2, implying workers receive anywhere from 9% to 81% of their marginal product. This enormous range hardly suggests any consensus on the extent of employer-market power. And critically, even if we accept these estimates at face value, they necessarily measure a broad range of frictions in labor markets—search costs, job differentiation, occupational licensing, geographical preferences, and firm-productivity differences, among others—rather than solely market power from “artificial barriers to entry or excessive concentration of employment” that antitrust could address.
More fundamentally, these elasticity estimates rely on economic models and assumptions that may not translate well to antitrust analysis. As we note, “applying these estimates to a simple antitrust model of monopsony generates implications that go against the data.” For example, some estimates (including one co-authored by Posner) would imply implausibly low labor shares of income: around 8%, rather than the observed 60%. Such a reality check should engender caution.
Instead, in his ProMarket article, Posner claims that he is “aware of no academic paper that seriously contests the claim that thousands of labor markets are highly concentrated under traditional legal standards.”
First, it is unclear whether “thousands of markets” is a substantial, economy-wide concern. What is the denominator for this claim? How many labor markets are there? Given that there are 100,000 labor markets or more in the cross-industry studies, should we worry about 1%? Should we do so at any cost to employers and employees operating in the other 99%? There are a number of caveats in Posner’s sentence, but it is clearly meant to imply that Holyoak’s claim that labor-market concentration isn’t manifestly an enormous concern is false. In fact, however, it is Posner who is being coy with his evidence.
Given this discussion is about what information to obtain at the filing stage of a merger and not the trial stage, you would want evidence that labor-market concentration is pervasive, and that the relaxation of merger-review standards over time is to blame. Reasonable people can, of course, debate the relative level necessary to warrant additional scrutiny. But while some studies using online job-posting (vacancy) data find high concentration levels, more comprehensive administrative data tell a very different story:
- Kevin Rinz uses the Longitudinal Business Database, covering nearly all private-sector employers, to estimate labor-market concentration from 1976 to 2015. Overall, he finds that the average local Herfindahl–Hirschman Index (HHI), defined by commuting zones and four-digit NAICS industries, decreased from 0.16 in 1976 to 0.12 in 2015, indicating a shift toward less-concentrated local markets. Unsurprisingly, Rinz finds rural labor markets to be more concentrated than urban markets. Still, local concentration fell in all population quintiles.
- Ben Lipsius (in a working paper) also documents falling local concentration from 1976 to 2015, using alternative market definitions based on five-digit NAICS codes and urban areas. Despite these definitional differences, the average local HHI remains consistently low, ranging from 0.14 to 0.17, depending on the year and market definition.
- David Berger, Kyle Herkenhoff, and Simon Mongey further corroborate these findings with a different way of averaging HHI measures across markets. They estimate an average local HHI of 0.17 for the year 2014, with even lower concentration levels when analyzing individual sectors like manufacturing and services.
- Elizabeth Handwerker and Matthew Dey use governmental microdata to find an average HHI in the private sector of 0.03. When they look at the 26 occupations from José Azar, Ioana Marinescu, and Marshall Steinbaum’s vacancy study, they find concentration only one-tenth as large using the more comprehensive data.
The average local HHI levels documented in these studies are below the 1,800 (or 0.18) threshold associated with highly concentrated markets in the 2023 Merger Guidelines—in some cases, dramatically so.
These are not definitive, of course; that is not the argument. Indeed, a new Bureau of Labor Statistics study—using Quarterly Census of Employment and Wages data (the same as used by Handwerker & Dey) and covering nearly all private-sector employment—examines concentration at various market definitions, from national industries down to county-level industry groups. At their baseline definition (Metropolitan Statistical Area, by industry group), the authors find the average HHI is 3,734, with highly concentrated markets accounting for about 15% of private-sector employment. The results do, however, vary substantially based on market definition.
Sorting out these data implicates fundamental issues that go to the heart of whether we have reliable evidence of an antitrust-relevant problem. The models and measures used in labor-economics research were not developed with antitrust enforcement in mind and may not capture the complexities of real-world labor markets in ways that would support expanded enforcement. Holyoak is right to suggest caution, and Posner is wrong to take her to task for doing so.
We Don’t Even Know How to Define Antitrust-Relevant Labor Markets
One of the fundamental issues with studies of labor antitrust (glossed over by Posner) is market definition. Many inputs—including labor—are highly substitutable across industries, firms, and geographies. The skillset required to work at one retailer largely overlaps with that needed at another. Remote work has further expanded the geographic scope of many labor markets.
This makes defining relevant markets for antitrust analysis extremely difficult. For example, the FTC’s recent complaint against the proposed Kroger/Albertsons merger attempted to define a market for “union grocery labor” in local collective bargaining areas. But as we have written, this novel and narrow market definition assuredly fails to reflect the actual scope of competition for workers.
Posner chides Holyoak for “dismiss[ing] as an ‘obscure outlier’ the FTC’s labor-market challenge to the merger of Kroger and Albertsons, two of the largest grocery chains in the United States.” But she is right to do so. The merger of “two of the largest grocery chains in the United States” may be concerning for labor-market power if “union grocery labor” is a relevant market and if the relevant geographic scope is national. In reality, however, grocery stores’ labor-market power is enormously limited by the availability of nongrocery (to say nothing of non-union) employment. And the size of the companies across the United States is of little consequence, given that employment is almost always a local, not national, issue.
While the alleged product-market definition in the case (on which the case was decided) aligned with the FTC’s approach in past supermarket mergers, the labor-market definition is radical and does not appear to have a direct precedent in prior cases. The FTC’s claimed labor-market definition in this case was, indeed, an “obscure outlier.”
The reality is that, for most workers, a great number of potential employers would remain following any particular merger. As economist Kevin Murphy has noted, when workers leave firms, they typically disperse widely: often, no more than 5% go to any one competitor, and many change industries or occupations entirely.
Fundamentally, the labor-economics literature has offered little guidance to date on how to define markets in labor-antitrust cases. As noted, concentration varies greatly, depending on the exact definition of the relevant market, especially the geographic market. It is virtually impossible to know what outside options to include in the relevant market, and it may not always be possible to identify even where such potential employers are located (e.g., are commuting zones better proxies for the relevant geographic labor market than metropolitan areas?).
These market-definition issues are far more acute in monopsony cases than in traditional monopoly cases, both because the intrinsic question of substitutes is more complicated and because there is far less precedent to guide parties and enforcers. It is unfounded to claim, in effect, that “the science is settled” here.
The Proposed HSR Rule Changes Were, Indeed, Premature and Problematic
Given these uncertainties in the underlying research, the proposed changes to the HSR merger-filing requirements to add extensive labor-market information were, contra Posner, appropriately rejected by the FTC as both premature and misguided. Our comments on the draft HSR rule discuss several key problems:
- The required information may not be analytically useful. The draft rule would have required data based on Standard Occupational Classification codes and U.S. Department of Agriculture commuting zones—neither of which were designed for antitrust analysis or necessarily aligned with actual labor markets.
- Much of this information is not collected by firms in the ordinary course of business, creating substantial new compliance burdens. The FTC estimated additional compliance costs of $350 million annually, while industry surveys suggest the true cost could exceed $1.6 billion.
- The draft rule would have required this information from all filing parties, even though less than 2% of reported transactions receive a second request from FTC staff. This would have imposed costs on thousands of benign transactions.
- The requirements would likely have exceeded the agencies’ statutory authority under the HSR Act to require only information that is “necessary and appropriate” for merger review.
Most importantly, the worker-safety and labor-violation information requirements were particularly questionable. The FTC asserted that past labor-law violations “may be indicative of a concentrated labor market where workers do not have the ability to easily find another job.” But, as Holyoak rightly responds:
No evidence, empirical or otherwise, was presented to support this assertion. And I am not aware of any supportive literature and have never seen a court opinion that suggests such evidence indicates competitive harm from a merger under Section 7 of the Clayton Act (or any other antitrust violation under the Sherman Act or otherwise). Instead, this proposal seems like an overt way to harass firms with any workplace failure under the guise of an antitrust investigation.
Indeed, according to the Occupational Safety and Health Administration, of the 10 most frequently cited violations, five are in the construction sector—not commonly a highly concentrated one—and five are in “general industry.” We also know of no literature showing a correlation between such violations and highly concentrated product markets (or even industries) or with highly concentrated labor markets, much less with anticompetitive mergers. And as Holyoak notes, the notice of proposed rulemaking (NPRM) accompanying the draft rule didn’t cite any.
A More Measured Approach Is Needed
None of this is to say that labor-market effects should never be considered in merger review. But the current state of both economic research and legal doctrine suggests we need a more careful and incremental approach that would include, e.g., further research to develop direct measures of labor-market power better suited for antitrust analysis; clearer frameworks for weighing potential labor-market harms against consumer benefits; and better analytical tools for defining antitrust-relevant labor markets.
Until we have made progress on these fronts, dramatically expanding merger-filing requirements around labor markets would risk imposing substantial costs, with unclear or nonexistent benefits. The HSR process is meant to help screen for clearly problematic mergers, not to serve as a research tool for developing novel theories of harm.
Posner is right that worker welfare matters. But effective antitrust policy requires both sound theoretical foundations and appropriate practical tools for implementation. The current push for aggressive labor-antitrust enforcement (like that encompassed by the original draft HSR rules) is getting ahead of the evidence and our ability to sensibly address these issues through existing enforcement mechanisms.
