What happened when Seattle went to $15

In June 2014, Seattle's city council voted unanimously to phase in a $15 minimum wage. At the time, the federal floor was $7.25 and no major American city had tried anything close. Supporters called it the start of a national movement. Skeptics called it a reckless experiment. Both sides agreed on one thing: it was an experiment, and someone ought to measure the results.

Someone did. Two teams, in fact, and they came back with answers that looked almost impossible to reconcile. The story of how that happened is more instructive than either headline. It shows what the data can and can't tell us about wage floors, and why the most careful study of the episode ended up pointing toward a concern Milton Friedman raised decades earlier: the people with the least experience are the ones a wage floor is most likely to leave behind.

The experiment

Seattle did not jump to $15 overnight. The ordinance set a staggered schedule that depended on employer size and on whether an employer paid toward health benefits or counted tips. Large employers, those with more than 500 employees nationwide, moved to $11 in April 2015 and then to $13 at the start of 2016. Smaller employers followed a slower track and reached $15 years later.

That design matters for what came next. The cleanest window researchers had was the jump to $11 and then $13 for large employers. By the standards of the time, $13 was already a very high floor, well above anything in the surrounding state and a large share of the local median wage for many low-paid occupations.

The city itself commissioned an evaluation from a team at the University of Washington's Evans School of Public Policy and Governance. The group, which included economists Jacob Vigdor and Ekaterina Jardim among others, had an unusual advantage: access to Washington State's unemployment-insurance records, which, unlike most states' data, record hours worked as well as earnings. That let the team compute hourly wages directly and see not only whether jobs disappeared but whether hours did.

What the UW team found

The team's first report, released in 2016 and covering the move to $11, found modest effects. Wages rose a little, employment effects were small, and the authors were cautious about drawing strong conclusions from the early phase.

The second paper was the one that made news. Released in June 2017 as an NBER working paper, it examined the increase to $13 and reached a striking conclusion. Hourly wages in low-wage jobs rose, as intended. But the hours worked in those jobs fell by enough to more than offset the raise. In the working paper's headline estimate, hours in jobs paying under $19 fell by roughly 9 percent, while wages in those jobs rose by around 3 percent. Put together, the authors estimated that the average low-wage job in Seattle paid about $125 less per month than it would have without the increase.

In other words, the policy appeared to have lowered total earnings for the group it was meant to help. Workers were paid more per hour, but they got fewer hours.

The implied responsiveness of hours to wages was much larger than most of the prior literature had found, large enough that even the authors treated it with care. The paper went through further revision and was eventually published in the American Economic Journal: Economic Policy in 2022. The published version is more cautious than the 2017 draft. It reports aggregate employment elasticities across a wide range (−0.2 to −2.0), warns that the aggregate analysis likely overstates employment effects because some jobs simply moved above the low-wage cutoff, and finds the clearest effect in individual-level data: significant reductions in hours, concentrated among less experienced workers.

The Berkeley study

A week before the UW paper appeared, a team at the University of California, Berkeley's Center on Wage and Employment Dynamics, led by Michael Reich, released its own study of Seattle. It reached close to the opposite conclusion.

The Berkeley team focused on food service, the sector with the largest share of minimum-wage workers. Using public employment data and a synthetic-control comparison, they found that restaurant wages rose with the minimum and that food-service employment did not fall in any detectable way. On its own terms, the Berkeley paper read as a clean success story for the $15 policy. Reich later co-authored the Berkeley studies of California's $20 fast-food minimum wage.

The timing was awkward. The effect was that Seattle's experiment arrived in public with two competing verdicts, each from a credible university, each easy for one side of the debate to quote.

The methods fight

Within days, the two camps were picking apart each other's designs. The arguments are technical, but they come down to three questions any reader can follow.

Who was left out? Washington's data could not reliably assign workers at multi-site employers to a specific location. A chain with stores both inside and outside Seattle reports its payroll as one entity. So the UW team excluded multi-site employers and analyzed single-site businesses only. Critics, including Reich and economists at the Economic Policy Institute, argued that this was a serious gap. Chains employ a large share of low-wage workers, and if workers moved from small independent shops to chains, the UW design could count that as lost hours when it was really a shift between employers. The UW authors replied that their single-site results were consistent with checks they could run, and that the multi-site question cut both ways. Neither side could fully settle it with the data available.

Who is the comparison? To estimate what would have happened without the increase, both teams built a "synthetic Seattle" from other places. The UW team drew on other regions of Washington State; Berkeley drew on counties from around the country. This is the hard part of any city-level study. Seattle in 2015 and 2016 was one of the hottest labor markets in America, driven by a tech boom. Critics argued no blend of other Washington regions could mimic that, and that a booming city might see low-wage jobs vanish simply because employers were already paying above $13 for reasons that had nothing to do with the law.

What counts as a "low-wage job"? The UW team defined low-wage jobs by a fixed threshold, under $19 an hour. Critics pointed out that in a fast-rising labor market, many jobs would cross that line on their own, which could make hours below the line appear to fall. The UW team ran versions with different thresholds and argued the finding held up, but the concern was legitimate and never fully dissolved.

None of these critiques proved the UW team wrong. They showed that a city-level natural experiment, in an unusual city, during an unusual boom, has wide error bars. That cuts against overconfidence in either direction, including on our side of the argument.

Who gained, who didn't

The most interesting work came later. In 2018, the UW team released a follow-up paper that stopped asking what happened to jobs and started asking what happened to people. Instead of comparing aggregate hours, it tracked individual workers over time and sorted them by how much work experience they had before the increase.

The results split cleanly. Workers who were already employed and had more experience saw their hours trimmed but their hourly pay rise by more, and on net they came out modestly ahead, on the order of ten dollars a week by the authors' estimate. Some of them also picked up hours at jobs outside the city. For this group, the policy did roughly what its supporters hoped.

Less experienced workers told a different story. The paper found no significant earnings gain for them, and the evidence pointed to reduced entry: fewer new, inexperienced workers getting their first low-wage jobs in Seattle than the comparison suggested would otherwise have done so.

The workers who gained were the ones already on the ladder. The ones who lost out were the ones trying to climb onto it.

This finding is quieter than the 2017 headline, and in some ways more important. It reconciles part of the dispute. A study that looks only at incumbent workers in established restaurants could plausibly see higher wages and stable employment. A study that looks at entry into the workforce could see fewer people getting started. Both can be true at once, and the people bearing the cost are, almost by definition, the ones hardest to find in the data: people who would have gotten a job and didn't.

What it means for abolition

It would be easy to treat Seattle as a scoreboard and count the UW paper as a point for our side. That is not the right lesson. Seattle's economy was booming, its experiment was messy, and the strongest negative finding came from a dataset that left out a big part of the low-wage market. Anyone claiming that Seattle settled the minimum-wage debate is overselling it.

But the episode does sharpen the case for abolition in three ways.

First, it shows how much the measurement matters. For decades, the empirical debate centered on headcount. Seattle's hours data revealed a margin that headcount studies largely miss. When a wage floor binds, employers have many ways to adjust: hours, scheduling, hiring pace, training, automation. A law that "doesn't cost jobs" may still cost work.

Second, it locates the cost where theory predicted. The 2018 follow-up is close to a direct test of the first-rung argument. The gains went to people with experience. The losses, to the extent they appeared, fell on people without it. That is exactly what a price floor on low-productivity labor should do: protect insiders and make it harder for outsiders to get in.

Third, it shows how hard the cost is to see. The worker whose hours were trimmed knows it. The teenager who never got hired doesn't know what she missed, and neither does anyone else. A policy whose benefits are visible and whose costs are mostly counterfactual will always be politically popular. That is not evidence it is good policy.

None of this means low-wage workers in Seattle, or anywhere else, should simply be left to the market with nothing. It means the wage floor is a blunt instrument that seems to help incumbents at the expense of entrants, and that tools like the Earned Income Tax Credit can raise take-home pay without closing the door on the next worker in line. Seattle's experiment didn't prove that case on its own. It did add one of the most careful pieces of evidence for it, and it raised a question that every city considering a big increase should be made to answer: who isn't getting hired, and how would we know?