Seattle raised its minimum wage, and a city-funded study counted the hours

Let me begin with a sentence that ought to be on the wall of every legislature: a raise is a rate, and a paycheck is a rate times a quantity. A wage tells you what you get per hour. It does not tell you how many hours you get. If the rate goes up by 10 percent and the hours go down by 10 percent, you are slightly worse off, and the press release announcing the raise has told you only half of the equation.

Seattle is where that point was tested on a city of three-quarters of a million people, with a team of economists from the University of Washington paid by the city itself to look. The result was not what the city's leaders hoped to hear. I want to walk through what the law did, what the researchers found, what the rival study found, and how to think about the fight, because the arithmetic here is simple and the argument around it is not.

The terms

The minimum wage is the lowest hourly pay an employer may legally offer. Hours worked is the total number of paid hours. Total pay for a worker is the hourly wage times the hours. Elasticity is how much the quantity of something changes when its price changes; for labour, it is how much employment, or hours, fall when wages are raised by law. The debate over the minimum wage is almost entirely a debate about that elasticity.

Here is the logic in a hypothetical, labelled as such. A worker earns $11 an hour for 30 hours a week, which is $330. The wage is raised by law to $13, a rise of about 18 percent. If his employer cuts his hours to 25, he earns 13 × 25 = $325, which is less than before, even though the rate went up. The break-even point is easy to find: the raise is worth it to the worker only if his hours fall by less than 1 − 11/13, which is about 15 percent. If hours fall by more than that, he loses. The whole fight is over whether hours fall by a little, by a lot, or not at all.

Seattle's law

In June 2014, Seattle's city government enacted an ordinance, signed by Mayor Ed Murray, to raise the city's minimum wage in steps toward $15 an hour. The first step took effect in April 2015. The phase-in was staggered by the size of the employer: large employers had the fastest schedule, with the minimum going to as much as $11 in 2015 and as much as $13 in 2016, and smaller employers had a longer ramp, with the schedule crediting some employers for tips or health benefits. I describe the schedule in outline and would urge readers who need exact figures for a particular employer size to consult the City of Seattle's Office of Labor Standards, which administers the ordinance.

For context, the federal minimum wage has been $7.25 an hour since July 24, 2009, and the state of Washington's own minimum was among the highest in the country even before the city acted. So Seattle was raising the floor from an already high level to one of the highest in the nation, in a city with a booming economy and rising costs. That makes it a good place to test the proposition, and a hard one for opponents. If a higher floor were going to work anywhere, it would work there.

The University of Washington team

The city did something that, to its credit, not every government does. It contracted for an independent evaluation. The city hired a team of researchers at the University of Washington, working with the Evans School of Public Policy and Governance, to study the effects of the law over time, and the project later drew foundation grants as well. The team had access to an unusually detailed source of data: state administrative records that report the hours worked and wages paid for every job covered by unemployment insurance, which lets researchers track individual jobs and workers, something the usual survey data cannot do.

In June 2017 the team released a working paper, and the finding made national news. The study examined what happened when the minimum wage went to as much as $13 in 2016, the second step, after a first step in 2015 from $9.47 to as much as $11. It compared Seattle with similar places that did not raise their wages. The paper was revised in May 2018, and the figures below are the ones the National Bureau of Economic Research lists for the revised version:

• Hourly wages in low-wage jobs rose, by a modest amount, about three percent.
• Hours worked in those jobs fell, by a larger amount, about six to seven percent.
• The net result was that total payroll in low-wage jobs fell, an average loss of about $74 a month per job in 2016.

I give those numbers as the authors reported them and would urge you to read the paper itself rather than rely on my summary. Now check the arithmetic, which is the whole point. If wages rose about 3 percent and hours fell about 6.5 percent, then pay is 1.03 × 0.935, which is about 0.963. Total pay fell by about 3 to 4 percent. A raise on paper, a cut in the pay envelope. The study's dollar estimate and the percentages point in the same direction.

Why did the hourly wage rise so little compared with the legal increase from $9.47 to $13? Because the group of "low-wage jobs" in the study includes many jobs that were already paying more than the old minimum, and because employers adjusted by reducing hours and, to some extent, by hiring fewer new low-wage workers. Some workers did get raises and kept their hours, and the paper does not say otherwise. The finding is about the average, and an average includes winners and losers: some workers gained, and others lost hours or the chance of a job. But the average low-wage job, according to the researchers the city hired, ended up with less total pay.

The competing study

The University of Washington finding drew an immediate counterattack, and it is only fair to describe it. At the same time, researchers at the University of California, Berkeley, associated with the Institute for Research on Labor and Employment, issued a study of Seattle's restaurant industry. They reported that employment in food service was not affected and that wages had risen. Their conclusion, in effect, was that the minimum wage had done what its advocates promised.

Here, in numbered form, is the methodological fight, because it is the heart of the matter and it is worth understanding.

1. Who is in the data. The Washington team worked with administrative records that do not easily separate the employees of businesses with locations in many places from those of businesses with a single location. Critics noted that its main analysis excluded multi-site employers, which are a large share of low-wage workers, and said this could bias the result, since large chains might respond differently. The Washington team later revised its paper.
2. Which industry. The Berkeley study looked only at restaurants, one industry in which employers might respond to a wage floor by raising prices rather than cutting jobs. The Washington study looked at all low-wage jobs. Each answers a different question, and neither answers the other's.
3. Which comparison. Both studies compare Seattle with places that did not raise wages, but they build the comparison differently. In this kind of work, the choice of the control group can move the result, and honest economists on both sides acknowledge it.
4. What outcome. The Berkeley study focused on the number of jobs. The Washington study found that the number of jobs changed less than the number of hours. A study that counts jobs and not hours could miss the very margin on which employers adjusted.

I will tell you where I come down, and then give the other side its chance. I think the Washington study is the more informative, for three reasons. It used administrative data on hours, which is the margin where the adjustment shows up. It measured the quantity that actually matters to a worker, which is total pay. And it was commissioned by the city that had passed the law, which means that its authors had no incentive to produce a result unwelcome to the officials who paid them. I do not say that Berkeley's work is worthless or that its authors are insincere. I say that the dispute is a real scientific dispute and that anyone who tells you the question is settled, in either direction, has not read both papers.

The steelman

Here is the case for the minimum wage at its best, because the arithmetic above is not the whole story. First, the labour market is not a perfectly competitive one: employers often have some power to set wages, because workers face costs when they change jobs, and in that setting a higher floor can raise pay with little loss of employment. Second, a higher wage reduces turnover and raises morale, which can pay for part of the increase. Third, some of the cost can be passed on to customers in small price increases, so that employment need not fall. Fourth, many studies of smaller increases in the minimum wage, in other places and times, have found little effect on employment. A law that works at $9 or $10 may or may not work at $15, but there is no obvious reason to think it must fail.

These are serious points, and the fourth is probably the strongest. A floor set modestly, near the level at which few workers are affected, usually produces small effects, because most workers already earn more. I accept that. My reply has four numbered parts.

1. The effect depends on how binding the floor is. A floor that touches few workers changes little; a floor that touches many changes a lot. Seattle's was a high floor, relative to the local labour market's old pay structure, and for that reason it is the case in which one should expect to see effects.
2. Hours are an easy margin. An employer who cannot lower the wage can lower the hours, by trimming shifts, closing earlier, or letting scheduled breaks stand in for paid time. This reduces total pay without a single layoff, and it does not appear in a count of jobs.
3. Price increases are not free either. If the restaurant raises the price of a meal, customers buy fewer meals, and a worker whose pay depends on the meals sold feels it. Passing on the cost is another way of saying that someone else pays it, and that someone includes low-income customers.
4. Small effects in some places do not justify large changes everywhere. Even if a floor of $12 does little harm in a high-wage city, a floor of $15 or more in a low-wage county is a different thing altogether. This is where national proposals get into trouble.

What it means for a national $15 or more

Now bring the lesson to the national debate. Proposals in Congress to raise the federal minimum to $15 an hour have come up repeatedly. The Congressional Budget Office has analysed such proposals in reports in 2019 and 2021, and I would urge readers to read the actual reports at the CBO's website, where the agency states its estimates of the pay gains and the employment effects and the wide range of uncertainty around them, instead of relying on my paraphrase, and to see for themselves how the agency frames the tradeoff; our own guide to reading such documents is at Read a CBO cost estimate yourself before you believe the headline.

The relevant point is that Seattle, a city with high wages and high costs, is among the places best positioned to absorb a $15 minimum, and still the study that the city itself hired found an average loss of hours. A national floor would apply to places with far lower wages and costs, where $15 would be a much larger share of the typical pay. The arithmetic of binding floors says that the effects there would likely be larger. A conservative who believes in federalism sees the point at once: the right minimum wage in Seattle is not the right one in rural Alabama, and a single number from Washington cannot be right for both.

The history behind the policy

None of this is new. For the history of how the federal floor began, and who the early advocates expected to lose their jobs, read the minimum wage's 1938 birth and the workers it was meant to shut out. And for a related case in which a well-intended rule changed the labour market in ways its authors did not expect, see California's AB5 put freelancers out of work, and voters pushed back.

I would add one thing about the political economy. The people who benefit from a high wage floor are visible: they are the workers who get the raise, and they show up at rallies. The people who pay for it are scattered: the worker who never got the shift, the teenager who could not find the first job, the customer who paid a little more, the small business that never opened. A policy whose benefits are visible and whose costs are invisible will always seem to be a good idea. That is precisely why a careful study of the costs, such as the one Seattle paid for, is worth more than a rally.

The Founders and the humility of data

There is a Founding point here too. The Constitution reserves to the states and the people powers that it does not delegate to the federal government, and the reason is not only legal. It is that a country as large and varied as ours cannot be well governed from one place by one rule. Seattle tried its own experiment, a state or a city can try another, and the results can be compared. That is how knowledge is gained in a federal system. A national floor ends the experiment before the results are in. See the Constitution for the Tenth Amendment, which says as much.

The case in brief

1. A paycheck is a wage times hours, so a raise in the rate is a gain only if hours do not fall by more; in the hypothetical above, the break-even is a cut of about 15 percent in hours.
2. Seattle's 2014 ordinance phased in toward $15, and the University of Washington team the city funded found for the step to as much as $13 in 2016 that hourly pay in low-wage jobs rose about 3 percent while hours fell about 6 to 7 percent.
3. Those figures imply total pay fell by roughly 3 to 4 percent, consistent with the researchers' estimate of an average loss of about $74 a month per low-wage job.
4. A rival Berkeley study of restaurants found no effect on food-service employment; the dispute is about which workers, which industries, which comparisons and whether hours or jobs are counted.
5. A high floor in a high-wage city is a favourable test, and a national floor would bind harder in low-wage places, which is the reason to leave the number closest to those who live with it.

What you can do: read the executive summary of the Congressional Budget Office's most recent report on raising the federal minimum wage, and send your representative one question about the range of its job estimates.