Seattle’s decision to hike its minimum wage up to $13 an hour—on its way to $15—ended up costing its low-wage workers time on the job, hundreds of dollars of annual income, and a shot at a better livelihood.

That is a reasonable conclusion one could draw from a blockbuster, if not yet peer-reviewed, new study on the city’s famed minimum-wage increases. The research, performed by a group of academics from the University of Washington, looks at detailed data on the earnings and hours of workers affected by the hike of the wage floor from $9.47 an hour to $11 in 2015, and from $11 an hour to $13 an hour in 2016. It concludes that, for low-wage workers, that second wage increase reduced hours worked by nearly 10 percent and earnings by an average of $125 a month. The findings, though preliminary, call into question years of economic research and the decisions of dozens of states and cities to bump their wage floors up.

Granted, there are reasons to think that the study’s results might differ from the reality on the ground. The paper excluded a number of larger Seattle businesses from its analysis, for one, and could not cleanly disaggregate whether the changes in the structure of the labor market were due to the booming local economy or to the minimum-wage hike. Still, it underscores an uncomfortable truth for liberals pushing for wage hikes across the country: There is some minimum wage at which the costs outweigh the benefits, and Seattle might have found it.

The University of Washington researchers looked at a more detailed pool of data than some minimum-wage researchers before them, combing through numbers collected for the local unemployment-insurance program. When compared to data from other parts of Washington state, which represented a similar economy with no minimum-wage hikes, these numbers indicated that the hike had a major effect. The number of jobs paying less than $13 an hour dropped by 39 percent, they found, with the number of jobs paying less than $19 an hour falling by a smaller amount. At the same time, overall employment in Seattle increased by 13 percent, judging by headcount, and 15 percent, judging by hours. But many low-wage workers suffered from hours, job, and earnings losses. The wage hike pulled the bottom rung off of the ladder, in other words.

The study contradicts a paper put out last week by respected researchers at the University of California, Berkeley. It found that Seattle’s minimum wage boosted wages by roughly 1 percent in food services overall and 2.3 percent in limited-service restaurants, like fast-food chains, with little effect on employment. “This industry is an intense user of minimum wage workers,” the authors write. “If wage and employment effects occur, they should be detectable in this industry.” A body of other studies dating back for more than 20 years have found similar effects: Most wage hikes do not do much to change employment levels.

Given the University of Washington’s diverging results, researchers on the left argued for caution in interpreting them and applying them to policymaking. “The authors’ estimated employment effects stand as outliers in a large body of research on the employment effects of the minimum wage,” wrote Ben Zipperer and John Schmitt of the Economic Policy Institute, a left-of-center think tank based in Washington, D.C. “Rather than constituting an important new contribution to the research in this area, the findings are best seen as raising concerns about possible problems with their underlying data and statistical techniques.”

The University of Washington study excluded workers at companies with multiple locations—meaning McDonald’s, Starbucks, and the other big and small chains that account for about 40 percent of the overall workforce and a huge number of minimum-wage jobs—narrowing the scope of the results, Zipperer and Schmitt noted. The study also seemed to imply that the minimum-wage hike caused a boom in high-wage employment, a seemingly impossible feat. (It seems unlikely that a business would have reacted to a pay hike for a minimum-wage worker by paying many of them $19 an hour, after all.) It in addition had no way to tell if Seattle’s employers were switching to contractors, as opposed to employees, to avoid some provisions of the minimum-wage law; if that had happened, those workers would have dropped out of the data set.

Still, a number of prominent economists chimed in to say that the study’s findings merited consideration and further investigation. Even if its results were revised, it points to a discomfiting reality: Academics and policymakers are not exactly sure what the optimal minimum wage is in any given city or state, or across the country. Even the staunchest defenders of a high minimum wage know there is some level that is counterproductive: If businesses had to pay workers a minimum of $100 an hour, say, there wouldn’t be much employment in low-skill work, which would obviously hurt the teenagers who work at restaurant chains and the single parents who work in big-box stores more than it would help them.

It might take more studies to determine if Seattle hiked its minimum wage too much, too fast. Moreover, the city’s experiment does nothing to undercut the argument for a higher minimum wage for the millions of workers still earning $7.25 an hour, the federal wage floor, or just above it. But too high of a minimum wage would damage the prospects of the low-wage, low-skilled workforce. Seattle’s experiment might give some indication of how much is too much.