The Uneven Geography of Employer Power
This chapter uses the Bureau of Labor Statistics’ 2023 QCEW evidence to answer a narrower question: how widespread is employer concentration in the local labor markets where workers search, and what does it appear to mean for pay and merger risk? The central judgment is two-sided. Concentration is common enough to matter, but its prevalence depends on how markets are counted; and its association with lower wages is important evidence, not a complete causal explanation.
A concentrated market is common, but not every worker faces the same exposure
Using employment shares in metropolitan-area-by-industry-group markets, the BLS article identifies 93,252 local markets in 2023. The average employment HHI is 3,734 and the median is 2,604. Under the 2010 Department of Justice and Federal Trade Commission taxonomy used in the article, both figures fall in the highly concentrated range. Fifty-one percent of the measured markets are highly concentrated, another 15 percent are moderately concentrated, and 13 percent are nearly or perfectly monopsonistic, with an HHI above 9,500.
Those figures are striking, but the unit of observation is a market, not a person. A small town with one dominant employer counts as one market, just as a large metropolitan industry with dozens of firms counts as one market. If concentrated markets tend to have smaller workforces, an unweighted average can be high even while most jobs are located in less concentrated markets. The BLS article finds exactly that pattern: the employment-weighted average HHI is 1,186, far below the unweighted average and median.
This difference is not an error in the data. It answers two different questions. The unweighted distribution asks how often a worker could encounter a highly concentrated local market if markets are treated as comparable units. The weighted distribution asks how many jobs are located in those markets. A policy discussion that uses only the first can exaggerate the share of all workers exposed to high concentration. A discussion that uses only the second can make the constraint faced by workers in small labor markets disappear.
The worker-weighted numbers put the exposure in concrete terms without turning it into a universal condition. Highly concentrated markets account for about 15 percent of private employment, or 17.4 million jobs, and about 18 percent of private payrolls. Moderately concentrated markets account for another 9 percent of employment and 9 percent of payrolls. Thus, a meaningful share of jobs is located in markets where employers have substantial measured concentration, even though most private-sector workers are not in a market classified as highly concentrated under this particular baseline.
The word “classified” matters. The thresholds are taken from the 2010 merger guidelines. The BLS article notes that the 2023 guidelines use lower HHI thresholds. Applying the older taxonomy therefore produces a conservative count relative to current guidance. That does not mean the 2010 categories are invalid for the article’s comparisons; they provide a consistent frame for the data and for earlier research. It means that a headline such as “51 percent of markets are highly concentrated” should always carry its year, market definition, denominator, and guideline frame with it.
Nor does a high HHI mean that every employer in the market can dictate every term to every worker. The measure identifies the distribution of employment among firms in a defined industry and geography. Individual workers may have portable skills, access to remote jobs, professional networks, or options outside the measured category. Others may face much tighter constraints than the average. The distribution tells us that the possibility of employer power is structurally present; it does not resolve each worker’s bargaining situation.
The near-monopsony figure makes the distinction especially sharp. Thirteen percent of measured local markets have an HHI above 9,500, meaning that nearly all measured employment is associated with one employer or an overwhelmingly dominant employer. That does not mean 13 percent of all workers are in such a market, because the statistic is a market count. It does mean that the local labor-market structure capable of producing extreme dependence is not a rare theoretical edge case. The worker-level implication must still be evaluated with the employment and payroll weights, the type of work, and the worker’s actual ability to search beyond the boundary.
The QCEW data give this exercise unusual breadth: the program is close to a census of U.S. employment and wages and covers private employers across the states and territories, although the BLS analysis excludes government employment. Broad coverage improves the descriptive picture, but it does not eliminate the conceptual problem. A near-universal record of the wrong market definition would still be a precise answer to the wrong question. Coverage and validity are related but distinct: one concerns how much of the economy is observed, the other concerns whether the observed units represent the alternatives workers can use.
This is why the BLS distribution should be read as a map of possible exposure rather than a direct map of individual vulnerability. A market may be highly concentrated because it is small and specialized, yet a worker with a portable occupation may have options outside that specialization. Another worker in a larger market may face a more serious constraint if most firms hire only a narrow set of skills or if switching requires a costly move. The HHI tells us where to look for employer power; it does not finish the diagnosis.
Lower associated pay is evidence, not the whole causal story
The BLS article also examines the relationship between concentration and average annual wages. Its baseline ordinary-least-squares estimate implies that a 10 percent increase in concentration is associated with about a 0.3 percent decrease in average wages. A shift from the upper threshold of an unconcentrated market, HHI 1,500, to the lower threshold of a highly concentrated market, HHI 2,500, is a 66.7 percent increase in concentration and is associated with about a 2 percent decrease in average wages, all else equal.
The direction is consistent with a bargaining-power mechanism. When workers have fewer realistic employers to choose among, a firm may be able to offer a lower wage without losing as many applicants or incumbents. The result also fits the broader question posed by this series: extra UI time may be more valuable when it gives a worker a way to avoid the low-paying offer, but the value of avoidance is limited if the other employers in the market are not credible substitutes.
The regression does not, however, prove that concentration caused every wage difference. Local markets differ in productivity, unemployment, union membership, industry composition, and other characteristics that can affect both the number of employers and their pay. A market may be concentrated because it specializes in a low-paying industry, or low pay may affect which firms enter and survive. The observed negative relationship is compatible with employer power, but it can also contain other pathways.
The article therefore reports an instrumental-variables exercise. It uses the concentration of related local markets in the same industry and year as an instrument for a market’s own concentration. The motivation is that related markets may be correlated in their concentration while not directly determining the productivity of workers in the focal market. Under that exclusion assumption, the estimate is larger: a 10 percent increase in concentration is associated with about a 1 percent decrease in wages, and the move from HHI 1,500 to 2,500 corresponds to about 6.8 percent lower average wages.
That larger estimate should not be read as a settled causal number. The article calls the exclusion assumption plausible but unverifiable. It also notes that QCEW wages are average compensation measures that do not fully account for hours worked and can be skewed by high earners. The comparison between OLS and IV is useful because it shows how conclusions depend on identification, but it does not remove the need to state what the instrument assumes.
The choice of outcome also matters. Average annual pay is relevant to the economic consequences of employer concentration, but it is not the same as an hourly wage, a worker’s reservation wage, or a complete measure of match quality. A market can have the same average pay under different distributions of hours, benefits, stability, and worker composition. The BLS evidence establishes an important wage relationship; it does not tell us precisely how a given worker would use an additional month of UI.
The sequence clarifies the order of inference. Administrative reports are aggregated into a constructed market measure. That measure can be used in a wage regression, which produces an association or an IV-based estimate under additional assumptions. It can also be used to calculate what a market’s HHI would be after a hypothetical merger. Neither path is the same as observing workers’ wages before and after a completed merger.
Local competition can weaken while national concentration rises
The BLS time series adds a scale distinction. From 2002 to 2023, the average local MSA-by-industry-group market became about 7 percent less concentrated, while the average nation-by-industry-group market became about 46 percent more concentrated. At first glance, those movements seem contradictory. They are not, because a large nationally dominant firm can enter many local markets. Its national employment share rises, increasing national concentration, while its entry gives local workers another employer and lowers local concentration.
For a worker deciding whether to accept an offer, the local effect is usually the more immediate one. A national trend can matter for corporate power, but it does not replace the question of which employers are within reach. This divergence is another reason not to use national concentration as a shortcut for local bargaining conditions.
The article then simulates a merger between the top two employers in each eligible MSA-by-industry-group market in 2023. Among markets with at least two employers, the simulated merger meets the article’s 2010-guideline presumption for enhancing market power 60 percent of the time. Those markets contain more than 25.5 million workers. Restricting the exercise to markets with at least two employers and at least 1,000 workers still produces a 26 percent rate. The result says that local merger review could matter in thousands of relatively large labor markets, not only in tiny isolated towns.
The article also applies its baseline wage elasticity to the simulated changes in HHI. It estimates an employment-weighted average expected wage decline of 1.6 percent for a merger of the top two employers in the eligible markets, roughly $1,008 per worker in the average market. This calculation is useful as a way to translate concentration into an economic magnitude. It remains a counterfactual projection. It assumes the estimated elasticity applies to the simulated change and does not observe the wages that would follow an actual merger.
That boundary is especially important for the search-time question. A simulated merger can show why a worker’s outside option might deteriorate if two major employers become one. It cannot show that every merger will reduce wages by the projected amount, nor can it show that UI would offset or fail to offset that loss. The BLS article itself identifies analysis of completed mergers as a route to stronger causal evidence.
The evidence therefore establishes the scale and stakes of the structural constraint, while leaving the policy interaction open. Local concentration is widespread in the measured market distribution, covers millions of jobs, and is negatively associated with average wages. But the next question is different: when additional UI time is assigned in a way that separates it from the bad conditions that normally trigger it, does it change workers’ outcomes—and does it do so differently across concentrated markets?