Policymakers looking for ways to help low-income households have turned to minimum-wage increases in recent years. But a provocative new paper by the Seattle Minimum Wage Study Team at the University of Washington suggests that this may be a deeply counterproductive policy. This study was immediately met with skepticism from left-wing labor activists and several major media outlets, but it is important to understand why this paper is different and worth paying attention to.
The authors of the study are an official team commissioned several years ago by the city of Seattle to study the impacts of raising the minimum wage, in a move that I applauded at the time as an honest and transparent attempt at self-examination. The team includes researchers from a broad spectrum of political backgrounds, and its paper is one of the first studies of a very large, city-level minimum-wage increase. The researchers had unprecedented access to detailed administrative data. They compare wage levels, job counts, and hours worked between the city of Seattle and a “synthetic control,” composed of other parts of the state whose economies had been tracking Seattle’s reasonably well prior to the policy change. To simplify somewhat, the idea is that the two would have evolved similarly if the policy had never been introduced, and that differences following the minimum-wage increases can be attributed to that policy.
The group’s first study last year (of the policy’s first phase, increasing the minimum to $11 an hour) was a mixed bag, with fairly imprecise estimates. But the new findings, adding another year of data and including a second increase to $13 an hour, are unequivocal: The policy has caused serious damage to low-wage workers in Seattle. The number of hours worked by low-wage workers fell by a staggering 3.5 million per quarter. This is reflected both in thousands of job losses (or, more precisely, in jobs that would have been created but never were) and in reductions in hours worked by those who retained their jobs. These effects were so dramatic that total payroll accruing to low-wage workers fell by about $120 million per year, with workers actually losing $125 per month on average.
These estimates suggest that businesses respond to minimum-wage hikes much more strongly than previous research implied, which may have been expected given how much higher (and how localized) this minimum wage is relative to previously studied wage hikes. The impact also took some time to be reflected in the labor market, as predicted by several recent papers examining minimum-wage increases in other contexts.
This paper makes numerous valuable contributions to the economics literature and should give serious pause to minimum-wage advocates and policymakers. Of course, that’s not what’s happening. The mayor of Seattle went as far as commissioning another study, by a team at Berkeley whose findings on the minimum wage are so consistently one-sided that you can set your watch by them. Unsurprisingly, this study finds no effect of the policy, and its release was deliberately timed for several days before the UW paper came out, in order to muddy the waters. It appears that many journalists have obligingly played along, focusing on the caveats already included by the UW team and unquestioningly reporting others’ criticisms, including those from the union-funded Economic Policy Institute, which attacks the UW team’s paper on very shaky grounds.
Hand in hand with this is the assertion in many articles that the economics literature has converged on a consensus that the minimum wage does not affect employment. This is completely and utterly false. There are many papers, in top economics journals by respected researchers, showing negative impacts of the minimum wage on employment — including my own work with Jeremy West of the University of California at Santa Cruz, which showed that higher minimum wages slow job growth. This is an ongoing academic debate and is far from a settled issue.
Many of the criticisms of the UW study suggest a highly selective read of the minimum-wage literature, but they are worth addressing in detail. The first is that the results are somehow “too big” and inconsistent with prior research. Minimum-wage advocates would argue that any negative finding is “too big” and reflects errors in the research — an unscientific argument that borders on the tautological. But there are good reasons why the impacts are so large. This is by far the highest minimum wage that has ever been examined, and the wage increase occurred only in a single city. A bigger increase causes a bigger employment effect.
If your immediate reaction to this study was to dismiss this study, it is time to admit that your views cannot be swayed by science.
The other papers that found no changes in employment were limited to the restaurant industry and to analyzing raw employment levels, pooling workers of all pay and skill levels. When the UW researchers limited their study to restaurant workers and used the same estimation techniques, they found similar results. Thus, those other papers’ findings were likely driven by their data limitations: the lack of data on hours worked, the focus on a single industry, and an inability to separate out at employment effects by wage level.
The UW study’s primary limitation is that it excluded many employers with multiple establishments, because such employers might have some locations in Seattle and some elsewhere in the state. This is the sort of tricky decision that economists have to make all the time. However, the UW researchers conducted extensive surveys showing that multi-site employers in Seattle were more likely to report job reductions than the single-site establishments. To believe that the exclusion of multi-employer establishments negates the results, one would have to believe that multi-site employers somehow massively expanded their operations in response to the higher minimum wage, enough to make up for reductions and slowdowns elsewhere — and yet reported in a survey, en masse, that they had cut staffing.
The study also found an increase in work at jobs that pay more than $19 an hour, and some critics have pointed to this as evidence that something about the research design is off. This argument falls apart under scrutiny. First, “labor-labor substitution” is an obvious way for employers to adapt: For example, a store hires fewer low-wage clerks and the assistant manager puts in more hours to make up for it. Second, and more importantly, the UW team’s results hold under a wide range of definitions of low wages, including those higher than $19 an hour. This is strongly suggestive that the increase in the prevalence of these high-wage jobs is unrelated to the minimum-wage policy and simply reflects Seattle’s tech boom. And in future work, the UW team will investigate which industries are affected most, which will shed more light on this question.
EPI and others then claim that the results are overstated because Seattle has been booming. This is exactly backward: If Seattle is growing faster than expected, then the counterfactual comparison group is not keeping up as well at it should be, understating the extent of the job losses. It also seems strange to claim that low-wage work will do worse in good economic times, when the recent evidence of the Great Recession shows the opposite. More to the point, recent research shows that the negative impacts of the minimum wage are higher during economic downturns, not boom times. There is no cherry-picking here.
These job losses will only get worse as the minimum wage climbs and the effects fully phase in. The UW team is working on linking the workers in its data set to demographic data from other sources to examine the characteristics of the losers from this policy. Most likely, the losses are borne most heavily by low-income and minority households, high-school dropouts, those with criminal records, and others who are already most vulnerable — that is, those whom an employer is least likely to hire at $15 an hour.
I have conducted economics research on the minimum wage, though most of my work is on non-controversial topics like charitable giving, and I’ve been surprised how politicized the topic is even among academics. I don’t envy the backlash this team is facing for presenting results that are seen as heresy.
Of course, one paper, even a careful, well-documented one with superior data that’s written by a respected and apolitical group, does not fully establish the result. But critics who refuse to even consider that this policy may be hurting those it’s intended to help should take a step back and focus on the real question: How can policy best help low-income households? As a transfer program, this experiment has failed miserably. It destroyed three dollars’ worth of employment opportunities for every dollar that actually went to a low-wage worker. A genuine concern for the poor and working class is not well served by shouting slogans and ignoring real evidence.
If your immediate reaction to this study was to dismiss this study, it is time to admit that your views cannot be swayed by science. They might as well be religion.