The Question the Profession Reopened

For most of the twentieth century, the minimum wage was the easy question in economics. A price floor above the market-clearing level produces excess supply. In a labour market, excess supply of labour is unemployment. Introductory textbooks used it as the worked example, and surveys of economists showed strong agreement.

In 1994 David Card and Alan Krueger looked at what happened when New Jersey raised its minimum wage and neighbouring Pennsylvania did not.

They surveyed fast food restaurants on both sides of the border before and after. Restaurants in eastern Pennsylvania are a plausible control group for restaurants in western New Jersey: same regional economy, same weather, same customers driving between them, different policy.

Employment in New Jersey did not fall relative to Pennsylvania. If anything it rose slightly.

The reaction was ferocious, and some of it was fair. The original data came from telephone surveys, and a reanalysis using payroll records found different numbers, which Card and Krueger in turn disputed with administrative data of their own. But the design was the point, and the design spread. Border-pair comparisons, county-level matching and administrative payroll data have since produced a large literature, and the weight of it finds small or undetectable employment effects from the increases actually studied.

Card received a Nobel in 2021, substantially for this and for the method. Krueger had died in 2019.

So why does a price floor not produce the textbook result? Monopsony supplies an answer that is entirely orthodox.

Recall that a monopsonist restricts hiring, because taking on one more worker requires raising everyone's pay. A minimum wage removes that calculation for all wages below the floor: the firm can now hire an additional person at the legal minimum without giving everyone a rise, because everyone is already at it. The marginal cost of a worker drops to the wage itself, and hiring can increase.

The prediction is an inverted U. Starting from a low floor, raising the minimum increases both wages and employment. Beyond the competitive wage, further increases start costing jobs. Where the turning point sits is an empirical question, and it is the question that matters.

Which is why the relevant number is never the cash amount. What matters is the bite: the minimum as a share of the local median wage. A floor at 40% of the median binds on very few workers. At 70% it binds on a large share of the workforce. The same nominal figure can be irrelevant in an expensive city and severe in a poor region, which is why national minimum wages produce such different effects in different places and why cross-country comparisons in converted currency are meaningless.

Most of the evidence showing no disemployment comes from moderate bite ratios. That is the honest limit on what the literature establishes, and anyone citing Card and Krueger for a very large increase is citing it beyond its range.

It is also worth noting where the adjustment does show up, because "no employment effect" does not mean "no effect". Studies find reduced hours, slower hiring rather than layoffs, modest price increases in affected sectors, reduced turnover, and in some cases compression of the pay scale above the minimum. Firms have many margins and employment is only one.

The reason the profession's view shifted is methodological rather than ideological, and it is the more interesting story.

The older literature compared national employment before and after federal increases. That confounds the policy with everything else happening in the economy at the same time, and minimum wages tend to be raised when the economy is doing well, which biases the estimate in one direction, or during political windows, which biases it unpredictably. There was no control group.

What changed was the credibility revolution: the insistence that a causal claim needs a comparison that isolates the thing being claimed. Border discontinuities, matched county pairs, administrative data covering every worker rather than a survey sample. The same shift reshaped the economics of education, health and crime at the same time, and the minimum wage is simply its most publicly visible result.

None of this establishes that minimum wages never cost jobs. A floor high enough certainly would, and one set at twice the median in a poor region almost certainly does. What it establishes is that the textbook prediction failed in the range where the policy is actually set, and that the reason it failed was an assumption about competitive labour markets that turned out not to hold.