Monopsony in Labor Markets

The mirror image of monopoly

A monopoly is a market with one seller. A monopsony is a market with one buyer. The word is borrowed Greek — mono (single) + opsōnein (to purchase provisions) — and was minted by Joan Robinson in 1933 in her Economics of Imperfect Competition to label a structure that classical theory had largely ignored. In labor markets, the buyer is the employer; the seller is the worker. When one employer is sufficiently dominant in a particular labor market, every analytical result from the theory of monopoly flips over and runs backwards.

The single most important consequence is the gap between the wage and the marginal product. Under perfect competition, every worker is paid exactly the marginal revenue product of their labor — the extra revenue an additional worker generates. Under monopsony, the worker is paid less. The size of that gap — sometimes called the rate of exploitation in the older literature — depends on how steeply the labor supply curve to the firm slopes upward. The steeper the supply curve, the more wage-setting power the employer has, and the larger the gap between what workers earn and what they produce.

The mechanics: why MC of labor sits above supply

To see the central trick, imagine an employer who already pays a wage w(L) to L workers and considers hiring one more. In a competitive market the firm is a wage-taker: it pays the going wage for the extra worker and does not affect the wages of anyone already employed. The cost of one more worker is simply the wage.

A monopsonist is not a wage-taker. It faces an upward-sloping supply curve — to attract one more worker it must raise the wage. And, crucially, in the standard model the employer cannot wage-discriminate: it pays the same wage to all identical workers. So when the marginal worker arrives at the higher wage w(L+1), every infra-marginal worker gets a raise to w(L+1) as well. The true cost of one more worker is therefore

MC_L  =  d(wL)/dL  =  w(L)  +  L · w'(L)

The first term is the wage paid to the new worker; the second is the raise paid to every existing worker. When w'(L) > 0 — that is, whenever the supply curve to the firm slopes up — the marginal cost of labor lies strictly above the supply curve at every L. If supply is linear, MC has the same intercept as supply but twice the slope. This is the cleanest graphical signature of monopsony: two lines starting at the same wage axis intercept, fanning out with the steeper one above.

The monopsony equilibrium

A profit-maximising employer keeps hiring as long as the extra revenue from one more worker (the marginal revenue product MRP) exceeds the marginal cost of that worker. It stops where they are equal:

MRP_L  =  MC_L     (profit-max condition)

This pins down the monopsony employment level L_m. To find the wage actually paid, the firm reads off the supply curve at that quantity: it pays the lowest wage consistent with attracting L_m workers. Because the supply curve lies below MC, the wage w_m = w(L_m) is strictly less than the marginal revenue product MRP(L_m).

Compare the competitive benchmark. A competitive labor market clears where the supply curve meets demand: w_c = MRP(L_c), with L_c > L_m. The monopsonist hires fewer workers at a lower wage. Two distortions stack: workers receive less than they produce, and fewer workers are employed than would be efficient. The deadweight loss is the triangle between MRP and supply over the range L_m to L_c.

The markdown formula

It is useful to convert the gap into elasticities. Let ε_S = (dL/dw)(w/L) be the firm-specific labor supply elasticity. Then the profit-max condition rearranges to

w  =  MRP · ε_S / (1 + ε_S)
   =  MRP · (1 − 1/(1 + ε_S))

The wage is the marginal product marked down by the factor 1/(1 + ε_S). Reading off two limiting cases sharpens intuition. As ε_S → ∞ (perfectly elastic supply, the competitive limit) the markdown vanishes and w → MRP. As ε_S → 0 (perfectly inelastic supply, pure monopsony) the wage can in principle be pushed all the way down to the reservation utility of the workers.

Recent direct estimates of firm-specific supply elasticities — from variation in wage policies across firms (Webber 2015) and from quit-rate responses to wages (Manning 2003; Bassier, Dube, Naidu 2022) — typically cluster between 1 and 5. Plug ε_S = 3 into the formula and you get w ≈ 0.75 · MRP: a 25-percent gap between productivity and pay, attributable to monopsony power alone. That is large enough to matter, and it is consistent with reduced-form estimates of the wage premium associated with switching to a higher-paying employer.

From company towns to motion-monopsony

The original textbook example was the company town: a single coal mine, textile mill, or lumber camp in a remote location, with workers' housing, store, and church owned by the same firm. Pullman, Illinois (1880-1894); Hershey, Pennsylvania; the asbestos camps of Quebec; the Welsh slate villages of Blaenau Ffestiniog — these are the canonical cases. In each, geographic isolation and high relocation cost made the local labor supply nearly inelastic, and wages sat far below what comparable workers earned in competitive markets. When Pullman cut wages 28 percent during the 1894 depression while keeping company-store prices constant, the result was the Pullman Strike that the federal government broke with troops.

Modern labor markets look nothing like a company town on the surface — but Alan Manning's Monopsony in Motion (2003) made the point that you do not need a single literal employer to generate the same wage-setting power. Any source of friction that makes the labor supply curve facing an individual firm upward-sloping does the job: search costs, idiosyncratic preferences over workplaces, geographic constraints, occupational specialisation, information asymmetries, contractual restrictions. Manning's empirical strategy was to back out the firm-specific supply elasticity from observed quit responses to wage differences, and to use it as a sufficient statistic for the markdown. The implied markdowns turned out to be substantial across a huge range of occupations, not just in remote settings.

Card-Krueger and the empirical revolution

For most of the twentieth century the monopsony model lived mostly in textbooks, dismissed in policy debate. A minimum wage above the competitive equilibrium must destroy jobs — that was the consensus. In 1994 David Card and Alan Krueger published a study that broke the consensus open. They surveyed fast-food restaurants in New Jersey and just across the Delaware River in eastern Pennsylvania before and after New Jersey raised its state minimum wage from $4.25 to $5.05 per hour in April 1992. The competitive model predicted that NJ fast-food employment would fall relative to PA, where the federal minimum still bound. They found the opposite: NJ employment grew slightly faster than PA's. Two years later they published the same result for federal increases in 1990-91. The findings were politically explosive.

David Neumark and William Wascher disputed the result, arguing that administrative payroll data (rather than the Card-Krueger telephone survey) showed the expected disemployment effect. The exchange continued for two decades. The modern reconciliation, set out by Doruk Cengiz, Arindrajit Dube, Attila Lindner and Ben Zipperer in a 2019 Quarterly Journal of Economics paper that exploits 138 separate state-level minimum-wage changes, is that moderate minimum-wage increases produce essentially no measurable disemployment effect on average. That is exactly the prediction of a monopsonistic labor market — provided the new minimum stays below the marginal product. For sufficiently large increases (recent debates about $20 and $25 minima) the data become noisier and the model predicts that the disemployment effect re-emerges once the minimum crosses MRP.

The minimum-wage flip in pictures

The graphical story is short. In the unconstrained monopsony equilibrium, the firm picks L_m where MC = MRP, then reads off the wage w_m on the supply curve. Now impose a minimum wage w* with w_m < w* < MRP. Two things happen:

• The MC-of-labor curve flattens at w* for all employment levels at which the supply curve sits below w*. The firm no longer has to bid up wages for infra-marginal workers — they already get the floor.
• The firm's optimal L is now read off where the new flat MC line meets MRP. That L is to the right of L_m. The firm hires more workers, at a higher wage. Profits fall (the rectangle that used to be monopsony rent is partly transferred to workers).

The result has the politically remarkable feature that w ↑, L ↑, and consumer welfare is unchanged (if output is sold competitively) — the only loser is the employer's profit. Once the minimum wage rises above MRP at L_m the standard disemployment story re-emerges; there is an interval over which a minimum wage is unambiguously beneficial, then a region where it starts to bite.

Measuring labor-market concentration

The most influential empirical paper of the 2010s on this question is by José Azar, Ioana Marinescu and Marshall Steinbaum (NBER 2017). They used vacancy data from CareerBuilder to compute a Herfindahl-Hirschman index (HHI) of vacancy share for each (commuting zone, six-digit SOC occupation) cell in the United States. The Department of Justice considers product markets with HHI > 2,500 highly concentrated. They found the average local labor market has an HHI of about 4,000; roughly 60 percent of commuting-zone × occupation cells exceed 2,500.

HHI band	DOJ classification (product market)	Share of US local labor markets (Azar et al. 2017)	Average wage markdown implied
< 1,500	Unconcentrated	~25 %	Small (~5 %)
1,500 – 2,500	Moderately concentrated	~15 %	~10 %
2,500 – 5,000	Highly concentrated	~30 %	~15 %
> 5,000	Highly concentrated	~30 %	20 %+

Subsequent work (Benmelech, Bergman, Kim 2018; Rinz 2018; Schubert, Stansbury, Taska 2022) has refined the measure with administrative wage data and shown that local concentration explains a non-trivial fraction of the long-run decline in the US labor share. Schubert-Stansbury-Taska in particular argue that conventional HHIs understate monopsony because they ignore the fact that workers' realistic outside options are limited to similar occupations — accounting for that pushes effective concentration higher.

Five sources of monopsony power

• Geographic isolation. Single coal mine, rural hospital, military base, prison. Workers face high relocation cost; firm-specific supply elasticity is small. The classic case.
• Occupational specialisation. Nurses, professional athletes, electrical-grid engineers, certain academic specialties. Even in dense cities there may be only one or a few realistic employers. The NCAA prior to Alston was a near-pure monopsony for elite college athletes.
• Search frictions. Finding a new job costs time, money and stress. Even when many employers are nominally available, incumbent employers can shade wages below MRP up to the point where the gap exceeds the worker's cost of searching. This is the Manning channel.
• Information asymmetry. Workers often do not know what they could earn at competing firms, and pay-secrecy norms make it hard to find out. Pay-transparency laws (Colorado, New York) attack this channel directly.
• Contractual restrictions. Non-compete clauses, no-poach agreements between firms (the celebrated Silicon Valley anti-poach case of 2014), occupational licensing that constrains mobility across state lines, H-1B visas that tie workers to a sponsoring employer. The FTC's 2024 rule banning most non-compete clauses targets this channel; the rule survives in modified form even after partial judicial intervention.

Policy implications

The monopsony framing reorganises a lot of policy debate. Each of the items below has an unambiguous direction of effect under monopsony that is the reverse of, or strictly stronger than, the competitive prediction.

Policy	Competitive prediction	Monopsony prediction
Minimum wage (binding, below MRP)	Lower employment, higher wages for survivors	Higher employment and higher wages; profit transfer
Non-compete ban	Mild ambiguous effect via reduced training investment	Higher wages, higher mobility, lower turnover cost
Union recognition	Wage premium creates deadweight loss	Wage premium can be welfare-improving up to MRP
Pay transparency	Mild effect	Raises εS, shrinks markdown
Antitrust review of mergers	Focus on product-market concentration	Add labor-market HHI; block deals that consolidate hiring
No-poach agreement enforcement	Per-se illegal, but rarely prosecuted as labor case	Per-se illegal and high-priority

The FTC's 2024 non-compete rule and the DOJ-FTC 2023 merger guidelines (which for the first time formally incorporate labor-market effects) are direct policy applications of the monopsony framework. Whether these survive constitutional challenge is a separate matter; the analytical case for them rests on the empirical case for monopsony.

Counter-evidence and caveats

• Neumark-Wascher. Repeated finding using administrative payroll data that minimum-wage increases do reduce employment among teenagers and low-skill workers, especially in absolute level effects rather than ratios. The dispute hinges on choice of comparison group, treatment of bunching at the new minimum, and time horizon. Both Card-Krueger and Neumark-Wascher have evolved methodologically since 1994; the modern consensus is closer to small-or-zero effects on average with heterogeneity, not zero everywhere.
• Long-run versus short-run. Even if a minimum wage raises wages and employment in the short run under monopsony, in the long run firms substitute capital for labor, automate, exit, or relocate. Some of the disemployment effect attributed to competitive markets may reflect long-run adjustment that monopsony models with sticky technology omit.
• Skill heterogeneity. A binding minimum at $15 is far above MRP for some workers (under-19s, very low skill, agricultural labor in some regions) even where it is below MRP for most. The same nominal floor can be on opposite sides of MRP in different segments.
• Wage discrimination. The model assumes the firm pays a single wage to all identical workers. In practice firms wage-discriminate by tenure, by hire date, by negotiation, and informally by gender and race. Wage discrimination weakens the monopsony argument for minimum wages because the firm can absorb the marginal-cost wedge by raising only marginal-worker wages.
• Imperfect competition in the product market. If the employer is also a monopolist or oligopolist in the goods market, MRP is no longer a clean upper bound — the firm faces a markdown in labor and a markup in output, and the combined wedge can be substantial. The labor-share decline papers (Autor et al. 2020; De Loecker et al. 2020) emphasise this combined channel.

Variants and extensions

• Dynamic monopsony (Burdett-Mortensen 1998). Embeds the static model in a search-and-matching framework where workers quit and reapply over time. Firms post wages knowing higher wages reduce quits; equilibrium yields a non-degenerate wage distribution across firms even with identical workers. The canonical reference for modern labor-market modelling.
• Granular oligopsony (Berger-Herkenhoff-Mongey 2022). Hundreds of firms in each labor market compete strategically over wages. Gives firm-specific markdowns that vary with local concentration and can be estimated from US LBD data. Calibrated estimates imply a labor share roughly 9-10 percent below its competitive benchmark.
• Spatial monopsony (Manning 2003 ch. 6; Bhaskar-To 1999). Firms located at different points in geographic or characteristic space. Each firm has local market power even with many competitors elsewhere. Pure spatial differentiation is enough to generate observable markdowns.
• Wage posting versus bargaining. Wage-posting models (firm announces a wage, worker accepts or rejects) give classic monopsony markdowns. Bargaining models give surplus splitting — workers earn some fraction (typically a third to a half) of the joint surplus from the match. The two literatures produce structurally similar predictions for wage shading.
• Buyer-side cartels. Multiple firms colluding to suppress wages. Not technically monopsony (more than one buyer) but produces the same outcome. The 2010 Silicon Valley no-poach case (Apple, Google, Adobe, Intel) and the 2024 NFL coaches' antitrust class action are recent examples.

Worked example: a hospital in a small city

Suppose a hospital is the only employer of registered nurses in a commuting zone. The labor supply curve from local nurses is

w(L) = 30 + 0.5 L     (dollars per hour, L in nurses)

The wage bill is w · L = (30 + 0.5L) · L = 30L + 0.5L². The marginal cost of labor is the derivative:

MC_L = d(wL)/dL = 30 + L

Twice the slope of the supply curve, as advertised. Suppose the marginal revenue product (extra revenue per extra nurse-hour) is

MRP_L = 60 − 0.25 L

Profit-maximisation: set MC = MRP. 30 + L = 60 − 0.25L → L_m = 24 nurses. Wage paid: w_m = 30 + 0.5 · 24 = $42/hr. Marginal product at L_m: MRP = 60 − 0.25 · 24 = $54/hr. Wage gap: $12/hr below MRP — a 22 % markdown. Compare the competitive equilibrium: w = MRP on the supply curve → 30 + 0.5L = 60 − 0.25L → L_c = 40 nurses at w_c = $50/hr. So monopsony costs the city 16 nurses and pays each of them $8/hr less.

Now impose a minimum wage of $48/hr (above w_m but below MRP at L_m). The firm cannot pay less than $48; it will hire up to the point where MRP = $48 → L = 48 nurses. Hospital hires 24 more nurses, every nurse earns $48 instead of $42, and the only loser is the hospital's profit. Push the minimum to $55 and the firm is now constrained to L where MRP = 55 → L = 20 nurses, below L_m: now the minimum bites above MRP and disemployment re-emerges.

Common pitfalls

• Conflating monopsony with low-paying jobs. A market can be highly competitive and still pay low wages if MRP is low. Monopsony specifically requires that wages are below MRP — a gap, not an absolute level. Walmart cashiers may earn $14/hr because their MRP is $14/hr (competitive) or because their MRP is $20/hr and Walmart has wage-setting power (monopsonistic). The empirical question is the wedge, not the level.
• Assuming one big firm = monopsony, many firms = competition. Manning's central point is that frictions, not raw concentration, drive markdowns. A market with 50 firms can still be highly monopsonistic if quit costs are high and information is poor; conversely, a duopoly with perfectly informed mobile workers can approximate competitive wages.
• Forgetting the binding range of minimum wages. The "monopsony helps everything" result holds only for minima below MRP. Politically attractive proposals that push minima well above MRP for some workers recover the classical disemployment story. Card-Krueger does not generalise to arbitrarily large increases.
• Treating MRP as observed. MRP is a theoretical construct — economists rarely measure it directly. Estimates come from production-function estimation, structural job-search models, or experimental wage-setting designs. Numbers can vary widely depending on assumptions; the wedge interpretation is robust qualitatively but not always quantitatively.
• Ignoring monopsony power in the product market. A nominally competitive seller can still have product-market market power. When markups and markdowns coexist, both wages and prices distort, and the welfare arithmetic gets complicated. Single-channel monopsony stories often miss this.