Does the minimum wage cause unemployment? What theory and 30 years of studies say

This is one of the most studied questions in economics. The disagreement is narrower than the shouting suggests: almost everyone accepts the textbook logic, and almost everyone agrees small increases have hard-to-detect effects. The fight is over how big the effects are, where they show up, and which studies to trust.

A note on wording: strictly, "unemployment" means people looking for work who can't find it. Most studies measure employment, meaning jobs or hours. A worker priced out of a job may become unemployed, stop looking, or never enter the labor force, so this page talks mostly about employment effects.

The competitive model: why theory predicts job losses

Start with the simplest model of a labor market. Employers' demand for workers slopes downward: the higher the wage, the fewer workers it pays to hire. Workers' supply slopes upward: higher pay draws more people into work. Where the two curves cross, you get a market wage, W*, and a level of employment, L*.

Now set a legal minimum above W*. Employers move up their demand curve and hire fewer people, L_d. More people want jobs at the higher wage, L_s. The gap between the two is the surplus of labor the model predicts: some of it is people who lost jobs, some is people newly drawn in who can't find one.

This is Milton Friedman's price-control argument: a price set above the market-clearing level produces a surplus, and in the labor market it falls on the least skilled (see what Friedman actually argued). But the model gives only the direction of the effect. The size depends on how far above the market wage the floor sits and how sensitive hiring is to cost.

Why "binding" and "bite" matter

A minimum wage only matters if it is binding: above what some workers would otherwise earn. Put the line in the diagram below W* and nothing happens. That's why the federal minimum of $7.25, unchanged since July 2009, now directly affects relatively few workers in most of the country, while state and city floors do most of the work.

Economists measure how hard a floor binds with its bite, usually the minimum wage as a share of the local median wage. A floor that barely touches wages in a high-cost metro can cover a large share of jobs in a low-wage rural county. Our state guide explains the bite ratio in more detail.

Bite is the most useful idea for reading this literature. Most natural experiments involved modest bites; large ones are rarer and less studied. A result about a small increase does not carry over to a large one.

Elasticity: how big is the effect?

Economists summarize the size of the effect with an employment elasticity: the percent change in employment divided by the percent change in the minimum wage. An elasticity of zero means no effect. A negative number means job losses. As an illustration, an elasticity of −0.1 means a 10 percent increase in the minimum wage would reduce employment in the affected group by about 1 percent.

Two cautions. The number depends on whose employment you count: teens, restaurant workers and all low-wage workers give different answers, and counting many workers paid well above the floor dilutes it toward zero. And some researchers use an own-wage elasticity, which divides by the raise affected workers actually got rather than the change in the legal minimum. Check which definition a study uses before comparing.

For reference, Neumark and Wascher's 1992 state-panel study estimated teen and young-adult elasticities of roughly −0.1 to −0.2, which became a benchmark for the "conventional" view. Much of the later research is a fight over whether the true number for moderate increases is closer to that or closer to zero.

What 30 years of studies found

The table below pulls from our research tracker. The labels are our coarse summary of each headline finding, not a verdict on quality. Read the originals for methods and exact estimates.

The "little or no effect" results cluster around moderate increases and methods that compare nearby places. Negative results are more common for teens and the least-skilled, over longer horizons, and where the floor clearly binds. Newer work increasingly looks past headcount.

Each of the landmark episodes has its own deep dive: Card and Krueger's New Jersey study, Seattle's move to $15, where the University of Washington and Berkeley teams reached opposite conclusions, and the CBO's $15 estimates, whose uncertainty range for 2019 ran from roughly zero to 3.7 million jobs.

Why studies disagree

Most of the disagreement comes from choices that sound technical but move the answer:

• The comparison group. States that raise their minimum wage differ from those that don't, often in ways that also affect jobs. Border-county comparisons (Dube, Lester and Reich) and regional-trend controls (Allegretto, Dube and Reich) find effects near zero. Neumark, Salas and Wascher argue those controls discard valid variation, and find negative effects with other comparison groups.
• The data. Card and Krueger used a phone survey of restaurants. Neumark and Wascher's comment used payroll records and found losses. Card and Krueger's reply used administrative data and did not. Same episode, three data sources, two answers.
• The outcome. Counting jobs misses cuts in hours. Seattle's hours data, rare in U.S. research, found the adjustment there.
• The time horizon. Meer and West find slower job growth over several years, which a short-window study can miss.
• The workers. Teens, restaurant workers and all low-wage workers are different groups. Cengiz and coauthors count every job near the floor. Clemens and Wither follow the specific workers the increase bound.
• Which results get published. Doucouliagos and Stanley found the published literature favors negative, significant results; critics of Neumark and Shirley dispute how their "preferred" estimates were picked. Selection is worth checking on both sides.

Monopsony: the main theoretical exception

The competitive model assumes workers can switch employers at no cost, so no single firm can pay less than the market rate. When that fails, because of commuting costs, scheduling constraints or few local employers, a firm has some power to set wages. Economists call this monopsony, an idea Joan Robinson developed in the 1930s.

A monopsonist hires fewer workers than a competitive market would, because hiring one more means raising pay for everyone. A floor near the competitive wage removes that incentive, so pay and employment can both rise. This is the strongest theoretical reason to expect small or positive effects, and may help explain near-zero estimates for moderate increases.

The limit is that the helpful window is narrow and local. Set the floor above the competitive wage and employment falls even under monopsony. Since that wage differs by town and occupation, one legislated number is unlikely to land in the window everywhere. We make that case, and steelman the other side, in monopsony and the minimum wage.

Beyond headcount: hours, hiring, benefits and prices

Layoffs are only one way employers respond, and often not the first. The others are harder to measure, so headcount studies can understate the cost.

• Hours. Employers can keep the same staff and schedule fewer shifts. In Seattle's step to $13, the University of Washington team found hours in low-wage jobs fell by roughly 9 percent, enough that total pay for those jobs fell even as the hourly rate rose.
• Hiring. Employers can stop replacing people who leave, which shows up as fewer new entries, not layoffs. The Seattle team's later work found experienced workers gained while new entries declined, and Meer and West's growth result points the same way. We explain why missing first jobs matters in the first job ladder.
• Benefits, training and scheduling. In principle, employers can trim training, perks or scheduling flexibility to offset a higher floor. These are real costs to workers but rarely appear in employment data, so the evidence is thin.
• Prices. Employers can raise prices. Harasztosi and Lindner found that most of the cost of Hungary's large increase was passed on to consumers. That limits job losses, but it shifts part of the bill onto customers, many of whom have low incomes too.
• Firms and machines. Germany's minimum wage moved workers toward larger, more productive firms as some small low-paying ones closed. Lordan and Neumark found increases reduce low-skilled workers' share of automatable jobs. See automation and the minimum wage.

Some margins are benign: moving workers to more productive firms raises output. Others fall quietly on the people the law is meant to help. Either way, "no change in headcount" is not "no effect."

Frequently asked questions

Does raising the minimum wage always cause job losses?

No. A floor below the going market wage changes nothing, and many studies of moderate U.S. increases find small or undetectable effects on headcount. Losses are more likely when the floor is high relative to local wages, and they often show up as fewer hours or slower hiring rather than layoffs.

What does it mean for a minimum wage to be binding?

It sits above the wage some workers would otherwise earn, so employers must raise their pay or stop employing them. How much it binds, its bite, is usually measured as the minimum wage divided by the local median wage. The higher the bite, the larger the expected effects.

Why do economists disagree about minimum wage job losses?

Mostly method: different comparison groups, data sources, outcomes (jobs, hours or new hires), time horizons and groups of workers. Each choice can move the estimate, and researchers disagree about which choices are right.

Can a minimum wage increase employment?

In theory, yes, if employers have wage-setting power (monopsony) and the floor is set near the competitive wage. Set well above that, it reduces employment even under monopsony, and the right level differs from one local market to the next.