Seattle's $15 Minimum Wage: 2 Studies, 2 Answers

Seattle is the most argued-over test of the $15 minimum wage. It passed the ordinance in 2014, phased it in, and drew two well-funded research teams who looked at the same city and published conflicting headlines. Anyone who cites the Seattle minimum wage study without saying which one is picking a side. This article lays out both, why they differ, and what holds up.

What did Seattle actually do?

In 2014, the Seattle City Council passed a plan to raise the city minimum wage to $15 in stages. It moved to $11 in 2015, then $13 in 2016 for large employers, and $15 for those employers in 2017. Smaller employers and those that provided tips or health benefits had longer schedules. At the time, Washington State's minimum was below $10, and the federal floor was $7.25, as it still is.

Seattle was a good test case for a few reasons. It had the highest city minimum in the country by a wide margin, a strong local economy, and a research community close at hand. It was also a risky test case, since a booming tech-driven labor market gave employers more room to absorb costs than a weaker market would.

What did the University of Washington Seattle minimum wage study find?

A team from the University of Washington's Evans School (Jardim and colleagues) examined state administrative data on hours and earnings. Their headline result for the move from $11 to $13: hours worked in jobs paying under $19 an hour fell by about 9%, while wages in those jobs rose around 3%. The net effect, they calculated, was a loss of roughly $125 a month for the typical affected worker, because the hour cuts outweighed the pay rise.

The study drew national attention and got cited widely by opponents of $15 minimums. It also drew criticism. Its dataset excluded workers at employers with multiple locations, because the state data could not tie hours to a specific site. That left out a large share of low-wage jobs, including chain restaurants and retailers.

What did UC Berkeley find?

A UC Berkeley team led by Michael Reich studied Seattle's restaurant industry, using data that included multi-site employers. They compared Seattle with similar counties and found that wages for restaurant workers rose, with no measurable job losses. Restaurants are a labor-intensive, low-margin sector, which makes them a useful place to look for job losses.

The Berkeley study had its own critics, who noted it covered one industry and relied on a smaller sample of firms. A fair reading is that each study answers a different question: what happened to all low-wage hours among single-site employers, and what happened to restaurant jobs across all firm types.

Why do the two studies disagree?

Mostly through choices about whom to count.

• Who is in the sample. UW dropped multi-site employers. Berkeley kept them but narrowed to restaurants.
• What counts as a loss. Hours versus jobs are different outcomes. A worker can keep a job and lose shifts.
• Comparison groups. Each team built a different picture of what would have happened without the raise.
• Time period. The UW study covered the step to $13. Later phases to $15 were less studied at the time.

Economists call this the problem of the counterfactual. Nobody can observe Seattle without the law. Each study estimates it, and the estimates differ. Later research has kept the argument going without a clean resolution. The honest summary is a modest effect on hours in some jobs and no clear collapse in employment.

What does Seattle tell us about the national debate?

Less than either side claims. Seattle shows that a city with a strong labor market can raise its floor sharply without an employment crash. It does not show that every region can do the same overnight. A $15 minimum is a different policy in a high-wage metro, where it reaches a minority of workers, than in a low-wage region where it reaches half of them. The larger body of research, including a review of the job-loss studies, finds the risk grows as the floor approaches the local median.

That is why economists often favor phasing in increases and indexing them to local conditions. It is also why the national conversation tends to fail on facts: a single city has been asked to settle a question about fifty different labor markets. See also what a raise does to small business and who pays for a wage increase. The wider case sits in why we should raise the minimum wage.

What should readers watch for in the headlines?

Seattle's results are often quoted in one line, either "hours fell 9%" or "no job losses," and both lines are accurate descriptions of different analyses. A few cautions apply when you see them.

The first is scope. The UW headline refers to hours in low-wage jobs at single-site employers during the step from $11 to $13. The Berkeley headline refers to restaurant workers, a narrower group. Neither covers every worker in the city.

The second is time. Wages and hours can move at different speeds, and a city's economy can shift for reasons unrelated to the policy, including a tech boom that raised demand for service workers.

The third is the comparison. Both teams built statistical stand-ins for Seattle without the law. A study is only as good as that stand-in. Reading both side by side is better than reading either alone.

What is the takeaway?

Seattle did not prove a raise is free, and it did not prove a raise is ruinous. It showed a real policy can produce a real mix of outcomes: higher pay for most low-wage workers, some reduction in hours for some, and no sign of economic collapse. The cost to the workers who lost hours is worth taking seriously, and so is the gain to the many who earned more.

The deeper issue sits in Washington, not Seattle. A city had to act alone because the federal wage has been frozen at $7.25 since 2009 (U.S. Dept. of Labor), and states and cities fill the vacuum with a patchwork of experiments. That leaves workers in one place with a living wage and workers a few miles over the border without. The pattern reflects why the dream feels out of reach, a theme in why the American dream is broken.