Introduction: The Question That Split a Discipline
For most of the twentieth century, the effect of the minimum wage was considered settled. Draw a competitive labour market. Impose a wage floor above equilibrium. Quantity of labour demanded falls, quantity supplied rises, and unemployment appears in the gap. The diagram is clean, the logic is airtight, and it appears in every introductory textbook.
Then, in 1994, two Princeton economists compared fast-food restaurants in New Jersey — which had just raised its minimum wage — with restaurants across the river in Pennsylvania, which had not. They found no employment loss. If anything, employment in New Jersey rose slightly.
The paper detonated. Three decades later, the empirical minimum wage literature is among the largest in economics, David Card shared the 2021 Nobel Prize partly for this work, and the theoretical model that best explains the findings is not the competitive one but a model of employer wage-setting power that dates to 1933.
1. The Competitive Model: Why Economists Expected Job Losses
Marginal Revenue Product of Labour (MRPL) — the additional revenue a firm earns from employing one more worker. Equals the worker’s marginal physical product multiplied by the marginal revenue from selling that output. Because of diminishing returns, MRPL falls as employment rises, and is the firm’s labour demand curve.
In a perfectly competitive labour market, an individual firm is a wage taker. It faces a perfectly elastic (horizontal) labour supply curve at the market wage. It can hire as many workers as it likes at that wage, and none at all below it.
Now impose a binding minimum wage Wmin > Wequilibrium:
- Firms move up their MRPL curve and hire fewer workers. Employment falls from L* to LD.
- More workers want jobs at the higher wage. Supply rises to LS.
- The gap (LS − LD) is excess supply of labour — involuntary unemployment.
- A deadweight loss triangle appears, representing mutually beneficial employment relationships that are now illegal.
The employment loss depends entirely on the elasticity of labour demand. If demand is inelastic, job losses are small. If elastic, large.
Marshall’s rules of derived demand tell us labour demand is more elastic when: labour is a large share of total cost; the product’s own demand is elastic; substitute inputs are readily available; and the time horizon is long. Fast food fails the first three conditions in the short run — which is a hint about what Card and Krueger were about to find.
This model is not wrong. It is a special case of a more general one.
2. The Monopsony Model: When the Employer Sets the Wage
Monopsony — a market with a single (or dominant) buyer. The term was coined by Joan Robinson in The Economics of Imperfect Competition (1933). A monopsonist in the labour market faces the upward-sloping market supply curve rather than a horizontal firm-level one.
The critical difference is that a monopsonist faces the upward-sloping labour supply curve. To hire an additional worker it must raise the wage. And because it typically cannot wage-discriminate, it must raise the wage for every worker it already employs.
MCL rises faster than the supply curve and lies above it — exactly analogous to how a monopolist’s marginal revenue lies below its demand curve.
The monopsony equilibrium
The monopsonist hires where MRPL = MCL, at employment Lm. But it pays the wage read off the supply curve at Lm, which is Wm. Since supply lies below MCL:
- Wm < MRPL — workers are paid less than the value of their marginal output. This wedge is exploitation in Robinson’s technical sense.
- Lm < Lcompetitive — the monopsonist under-employs.
Students draw the monopsony diagram and then read the wage off the MCL curve. Wrong. The monopsonist pays the supply-curve wage; MCL is merely the internal cost calculation it uses to choose the quantity. Getting this backwards inverts the entire result and loses every subsequent mark.
The counterintuitive result
Now impose a minimum wage above Wm. The firm’s effective labour supply curve becomes horizontal at Wmin, up to where it meets the original supply curve. Over that range, MCL = Wmin — the firm can hire additional workers without raising the wage of existing ones.
The firm now hires where MRPL = Wmin. If Wmin lies between Wm and the competitive wage, both the wage and employment rise.
Set the minimum wage exactly at the competitive level and you replicate the competitive outcome perfectly. Set it above that, and employment falls — the competitive intuition returns.
Employment as a function of the minimum wage is therefore inverted-U shaped: rising, then falling. This dissolves the apparent contradiction between theory and the New Jersey evidence. Both camps were right about different regions of the same curve.
3. Case Study: Card and Krueger, New Jersey 1992
The policy: On 1 April 1992, New Jersey’s minimum wage rose from $4.25 to $5.05 per hour. Pennsylvania’s stayed at $4.25.
The study: Card & Krueger (1994), American Economic Review 84(4). Over 400 fast-food restaurants surveyed in both states, before and after.
The design: A difference-in-differences estimator — the change in New Jersey employment, minus the change in Pennsylvania employment over the same period. Pennsylvania served as the counterfactual for what would have happened in New Jersey absent the policy.
The finding: No evidence that the wage increase reduced employment in New Jersey relative to Pennsylvania. The point estimate on employment was, if anything, positive.
The controversy: Neumark & Wascher (2000), using payroll records rather than telephone surveys, reported negative employment effects for the same episode. Card and Krueger replied in the same issue using BLS data and reaffirmed their conclusion. The exchange is a case study in how measurement choices drive results.
The method. The natural-experiment framework — find a policy change, find a comparable place that didn’t get it, difference the differences — became the standard tool of applied economics. The 2021 Nobel Prize to Card, Angrist and Imbens recognised precisely this contribution to causal inference, not the minimum wage result itself.
4. The Modern Evidence: Bunching and the Missing Jobs
Cengiz, Dube, Lindner & Zipperer (2019), “The Effect of Minimum Wages on Low-Wage Jobs”, Quarterly Journal of Economics 134(3)
The innovation: They stopped asking “did total employment change?” and started asking “what happened to the distribution of wages?” Using 138 state-level minimum wage increases in the United States, they counted two quantities:
- Jobs that disappeared below the new minimum wage — the “missing jobs.”
- Jobs that appeared at or just above the new minimum — the “excess jobs,” or bunching.
The logic: If the competitive model held, missing jobs would exceed excess jobs, and the difference would be genuine job destruction.
The finding: The two numbers were close to equal. Jobs did not disappear; they moved up the wage distribution. The estimated employment effect was small and statistically indistinguishable from zero, while total earnings of affected workers rose substantially.
Why it matters: the bunching estimator is far more powerful than aggregate employment regressions, because it uses the entire wage distribution rather than a single count. It also makes the identifying assumption transparent: missing and excess jobs must be measured over a wage window wide enough to capture all reallocation.
Where the counter-evidence lives
The study: The Seattle Minimum Wage Study team (Jardim, Long, Plotnick, van Inwegen, Vigdor & Wething), examining Seattle’s phased increase toward $15 per hour.
The data advantage: administrative payroll records including hours worked — rare and enormously valuable. Most studies observe only headcount.
The finding: Reductions in hours worked in low-wage jobs, even where headcount employment held up. Total payroll for low-wage workers showed mixed effects depending on specification.
The crucial nuance: employment counts and hours worked are different margins. A firm can retain a worker while cutting their shifts. Studies measuring only headcount miss this entirely.
Neumark and Shirley (2022), reviewing the US literature, argued that the balance of published studies still points toward negative employment effects for teens and less-educated workers, and that the “no effect” consensus is partly an artefact of study selection.
At minimum wages up to roughly half the local median wage, employment effects are small. Above that, evidence thins rapidly and the profession genuinely does not know. Anyone claiming certainty in either direction about a $20 federal minimum is extrapolating far beyond the data.
5. How Widespread Is Monopsony Power?
The monopsony model was long dismissed as a curiosity applying to company towns. Recent research has overturned that view.
Azar, Marinescu & Steinbaum (2022), “Labor Market Concentration”, Journal of Human Resources 57(S)
The method: They measured labour market concentration using vacancy data from a major US job board, computing Herfindahl-Hirschman Indices for narrowly defined local occupation markets — the market for nurses in one commuting zone, not the market for “labour.”
The finding: A large share of labour markets — particularly outside major metropolitan areas — would be classified as highly concentrated under the very merger guidelines that antitrust authorities apply to product markets.
Why it matters: the same thresholds that would block a merger on product-market grounds are routinely exceeded in labour markets, and no antitrust authority acts. The paper reframed monopsony from a theoretical curiosity into a competition-policy question.
Manning (2003), in Monopsony in Motion, made the broader theoretical argument that monopsony power does not require literal concentration. It requires only search frictions. If it is costly for workers to find and switch jobs, every employer faces an upward-sloping labour supply curve, however many competitors it has. Under this view, monopsony power is the normal condition of labour markets, not the exception.
- Search and information costs — workers do not know what other firms pay.
- Geographic immobility — commuting constraints, housing, family.
- Non-compete agreements — a growing share of low-wage contracts.
- Occupational licensing — restricts movement across state lines.
- Firm-specific human capital — makes leaving costly.
- Differentiated job amenities — jobs are not homogeneous, so workers are not perfectly mobile.
6. Beyond Employment: The Other Adjustment Margins
A firm facing a mandated wage increase has more than two options. Understanding the full menu is where evaluation marks are earned.
| Margin | Mechanism | Evidence |
|---|---|---|
| Employment | Hire fewer workers | Small effects at moderate levels (Cengiz et al. 2019) |
| Hours | Cut shifts, retain headcount | Found in Seattle (Jardim et al.) |
| Prices | Pass cost to consumers | Substantial pass-through in restaurants |
| Turnover | Higher wages reduce quits, saving hiring costs | Dube, Lester & Reich (2016), JOLE |
| Effort | Better-paid workers work harder | Shapiro–Stiglitz (1984) efficiency wages |
| Profits | Absorb cost from monopsony rents | Predicted directly by the monopsony model |
| Automation | Substitute capital for labour | Long-run margin; Lordan & Neumark (2018) |
| Non-wage benefits | Cut scheduling flexibility, training | Growing literature; hard to measure |
Dube, Lester and Reich (2016) showed that minimum wage increases substantially reduce both separations and hires in low-wage sectors. Firms in high-turnover industries spend heavily on recruiting and training. If a higher wage cuts quit rates, a meaningful portion of the wage increase pays for itself.
This is a mechanism the competitive model, which assumes costless hiring, cannot represent at all. It is not a modification of the standard model. It is outside its vocabulary.
7. Distributional Effects and Poverty
Q: If the minimum wage raises earnings for low-wage workers, why is it a poorly targeted anti-poverty tool?
A: Because the unit of analysis differs. The minimum wage targets low-wage workers; poverty is a property of households. Many minimum wage earners are teenagers or second earners in non-poor households. Many poor households have no earners at all and are untouched by any wage floor.
The US Congressional Budget Office has repeatedly modelled this trade-off, finding that a substantial minimum wage increase would raise earnings for millions while causing some job losses, with the poverty reduction concentrated but not perfectly targeted.
Earned Income Tax Credit-style wage subsidies target household income far more precisely — but they are financed by taxpayers rather than employers, and Rothstein (2010) found evidence that a portion of the EITC is captured by employers through lower pre-tax wages.
A wage floor prevents employers from capturing the subsidy. The subsidy targets the households the wage floor misses. Arguing for one against the other misunderstands what each does. This is a strong evaluation point and few students make it.
8. Exam Technique
“Discuss the likely effects of introducing a national minimum wage.” A top-band answer will:
- Draw and explain the competitive model with the unemployment gap. Analysis marks.
- Draw and explain the monopsony model showing wage and employment rising. This is what separates top bands from middle.
- Argue that the outcome depends on which model describes the market — and therefore on the degree of employer wage-setting power.
- Introduce empirical evidence: Card and Krueger, Cengiz et al., Seattle.
- Discuss other margins: hours, prices, turnover, automation.
- Evaluate targeting: does the minimum wage reach poor households?
- Conclude conditionally, naming the level of the minimum wage relative to the median as the decisive variable.
- Draw the monopsony diagram accurately. MCL must lie above and be steeper than the supply curve.
- The wage is read from the supply curve, never from MCL.
- Identify the exploitation wedge (MRPL − Wm) at the monopsony employment level.
- Know that a minimum wage can raise employment in monopsony but never in perfect competition.
Summary
The minimum wage debate is not a debate about whether demand curves slope downward. They do. It is a debate about the shape of the labour supply curve facing an individual firm.
If that curve is horizontal, the competitive model applies and wage floors destroy jobs. If it slopes upward — because workers face search costs, geographic constraints, and limited employer choice — then firms hold wage-setting power, wages sit below marginal product, and a well-set minimum wage can raise pay and employment simultaneously.
Three decades of natural-experiment evidence suggests the second description is closer to the truth of most low-wage labour markets, at least at the minimum wage levels observed so far. What no one knows is where the inverted U turns over.
Exercise 1 — Setting a Minimum Wage You Cannot Observe
The monopsony model predicts that a minimum wage raises employment up to the competitive wage and reduces it thereafter. This implies that the “correct” minimum wage is exactly the competitive equilibrium wage — a number that is, by construction, unobservable in a monopsonistic market. We only observe Wm, not W*.
If policymakers cannot observe the competitive counterfactual, on what basis should a minimum wage level be chosen? Consider whether the ratio of the minimum wage to the local median wage is a defensible proxy, what assumptions that proxy embeds, and what evidence would falsify it.
📄 Read: Cengiz, D., Dube, A., Lindner, A., & Zipperer, B. (2019). “The Effect of Minimum Wages on Low-Wage Jobs.” Quarterly Journal of Economics, 134(3), 1405–1454. Focus on the bunching estimator, and consider precisely what assumptions are required for “missing jobs” and “excess jobs” to be counted over a comparable window.
Exercise 2 — Headcount, Hours, and Welfare
Card and Krueger measured headcount employment. Jardim and co-authors measured hours worked, and found effects the headcount studies missed.
Suppose a firm responds to a minimum wage by keeping all its workers, cutting each one’s hours by 10%, while simultaneously reducing turnover so that workers stay in the job longer. Has the policy made workers better or worse off? Construct the welfare calculation explicitly — including the value of leisure, the cost of job search, and the returns to tenure — and identify what data you would need that neither study collected.
📄 Read: Dube, A., Lester, T. W., & Reich, M. (2016). “Minimum Wage Shocks, Employment Flows, and Labor Market Frictions.” Journal of Labor Economics, 34(3), 663–704.
References
- Azar, J., Marinescu, I., & Steinbaum, M. (2022). Labor Market Concentration. Journal of Human Resources, 57(S), S167–S199.
- Card, D., & Krueger, A. B. (1994). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. American Economic Review, 84(4), 772–793.
- Cengiz, D., Dube, A., Lindner, A., & Zipperer, B. (2019). The Effect of Minimum Wages on Low-Wage Jobs. Quarterly Journal of Economics, 134(3), 1405–1454.
- Dube, A. (2019). Impacts of Minimum Wages: Review of the International Evidence. HM Treasury and BEIS, United Kingdom.
- Dube, A., Lester, T. W., & Reich, M. (2016). Minimum Wage Shocks, Employment Flows, and Labor Market Frictions. Journal of Labor Economics, 34(3), 663–704.
- Lordan, G., & Neumark, D. (2018). People Versus Machines: The Impact of Minimum Wages on Automatable Jobs. Labour Economics, 52, 40–53.
- Manning, A. (2003). Monopsony in Motion: Imperfect Competition in Labor Markets. Princeton University Press.
- Neumark, D., & Shirley, P. (2022). Myth or Measurement: What Does the New Minimum Wage Research Say About Minimum Wages and Job Loss in the United States? Industrial Relations, 61(4), 384–417.
- Neumark, D., & Wascher, W. (2000). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment. American Economic Review, 90(5), 1362–1396.
- Robinson, J. (1933). The Economics of Imperfect Competition. London: Macmillan.
- Rothstein, J. (2010). Is the EITC as Good as an NIT? Conditional Cash Transfers and Tax Incidence. American Economic Journal: Economic Policy, 2(1), 177–208.
- Shapiro, C., & Stiglitz, J. E. (1984). Equilibrium Unemployment as a Worker Discipline Device. American Economic Review, 74(3), 433–444.