What Is a Minimum Wage? The Argument That Changed Sides

In a competitive labour market with many employers and many workers, the wage settles where the number of people willing to work equals the number employers want to hire. Imposing a higher floor means employers demand fewer hours while more people want to supply them, and the gap is unemployment concentrated among exactly the workers the policy is meant to help, since they are the ones whose productivity is below the new floor. The prediction is specific and testable, which is the reason the debate has been so productive. It also generates secondary predictions that matter: employers should respond by cutting hours, reducing training, removing non-wage benefits, raising prices, or substituting capital for labour, and any of these is a cost to workers even where the headcount does not fall.

The turning point was a 1994 study by David Card and Alan Krueger comparing fast food employment in New Jersey, which raised its minimum wage, with neighbouring Pennsylvania, which did not, using the border to construct a natural experiment. They found no employment loss and a slight increase, a result that was contested vigorously including with payroll data reaching a different conclusion, and which generated a research programme rather than a settled answer. Three decades of subsequent work using border comparisons, bunching estimators, and the many state and city increases in the United States has produced a central finding that moderate increases have small or negligible employment effects, with a persistent minority of studies finding negative effects concentrated in particular groups or in larger increases. Card received a share of the Nobel Prize in 2021 partly for this work and for the methods it popularised. The honest summary is that the effect is close to zero over the range actually tested and that nobody claims it stays zero at any level.

The theoretical explanation that has gained ground is monopsony, meaning employer power in the labour market. If workers cannot costlessly move between employers, because of location, childcare, information, non-compete clauses or the simple friction of job search, then an employer faces an upward-sloping supply of labour rather than a flat one and can set wages below the competitive level, employing fewer people than it would if it had to pay the market rate. In that situation a minimum wage can raise both wages and employment, because it removes the employer's incentive to restrict hiring to hold wages down. Evidence for meaningful employer wage-setting power is now substantial, including studies finding that a small number of firms dominate hiring in many local labour markets and that wages respond weakly to productivity. Other explanations for the null result include reduced turnover, since replacing a worker is expensive, efficiency wage effects on productivity, price pass-through to customers, and compression of profit margins.

The design details do most of the work and are usually ignored in the argument:

• •The level relative to the local median wage is the meaningful measure, since the same nominal figure is trivial in one region and binding in another, and international comparison uses the ratio rather than the amount
• •Indexation to inflation or to median earnings, without which the real value erodes, which is what happened to the American federal minimum, unchanged in nominal terms since 2009
• •Regional variation, used in several countries and rejected in others on grounds of simplicity and fairness
• •Age bands and training rates, which lower the floor for younger workers on the argument that they are most at risk of exclusion
• •Coverage and exemptions, including tipped workers, agricultural labour, apprentices and the self-employed, which is where enforcement problems concentrate
• •An independent commission recommending the level from evidence, as in the British Low Pay Commission, which is designed to depoliticise the decision and has been unusually successful at it

Even its supporters note the limits. A minimum wage reaches people in work and does nothing for the unemployed, the retired or those outside the labour force, which is why it is a poor instrument for reducing poverty overall; a substantial share of minimum wage earners are second earners in households that are not poor, and many poor households contain nobody in paid work. It interacts with the benefit system, since a pay rise can be partly or wholly withdrawn through reduced transfers and higher taxation, which is the effective marginal rate problem. It does nothing about hours, and someone paid more for fewer hours is not better off. The complementary instruments are earnings subsidies and tax credits, which target households rather than jobs and which economists have generally favoured, with the counterargument that such credits subsidise low-paying employers and that a wage floor prevents that. The two are usually better together than either alone, which is broadly the arrangement most developed countries have arrived at.

Standard theory predicts that a wage floor above the market rate causes unemployment among exactly the workers it targets, and three decades of empirical work beginning with a 1994 border comparison have found employment effects close to zero for moderate increases. The leading explanation is that employers hold wage-setting power because workers cannot move costlessly, in which case a floor can raise both wages and employment. The design matters more than the headline, particularly the level relative to local median pay and whether it is indexed, and it does nothing for households with nobody in work.