The University of Washington minimum wage study and its critics

The University of Washington minimum wage study and its critics
This post was published on the now-closed HuffPost Contributor platform. Contributors control their own work and posted freely to our site. If you need to flag this entry as abusive, send us an email.
Seattle demonstrators for $15 minimum wage
Seattle demonstrators for $15 minimum wage
Peter Costantini

[Note: A team at the University of Washington is studying the Seattle minimum wage. A member of the Seattle City Council voiced some criticisms of the study in an article in the Seattle Times. I sent this e-mail to the UW team, Mayor Ed Murray’s office, the Seattle City Council, and the Seattle Times.]

From: Peter Costantini

To: 'Seattle Minimum Wage Study Team'; 'Jeff Reading'; 'Megan Coppersmith'; ''; ''; ''; ''; ''; ''; ''; ''; ''; 'Nick Licata'; 'Seattle Minimum Wage Study Team'; 'Daniel Beekman'; 'Gene Balk'; 'Rami Grunbaum'; 'Kate Riley'; 'Michele Matassa Flores'

Subject: The University of Washington minimum wage study and its critics

Sent: Monday, September 26, 2016 1:00 PM

The first installment of the longitudinal study of Seattle’s pioneering $15 minimum wage by the University of Washington team has begun to expand our understanding of the ordinance’s economic effects. As growing numbers of other localities and states pass similar increases, most recently California and New York, the Seattle study is being looked to by many as a first-draft evaluation of the law’s effects. And readers will find some useful information in it to draw on.

However, Councilmember Kshama Sawant has raised some issues that deserve to be addressed further by the study group. [ Beekman. “Sawant, UW researchers clash over impact of $15 minimum-wage law”. September 21, 2016. ] Constructing a “Synthetic Seattle”, a virtual control group against which to measure Seattle’s changes, applies a widely recognized econometric technique for studying the effects of minimum wages. But the particular methods used to define it in this study need to be examined more closely, as the devil in these details has raised hell with some other efforts. A paper by a group of economists at the University of California, Berkeley, and the University of Massachusetts, Amherst, has become the de facto authority on how to design and implement minimum wage research. [ Allegretto et al. “Credible Research Designs for Minimum Wage Studies.” September 23, 2013. ] If the Seattle team has not already integrated this model into their work, they should study it and perhaps request a review from the Berkeley / UMass researchers.

Bernstein and Reich’s critique

Sawant’s questions on the study are not the first ones raised. Jared Bernstein’s August 10 blog post in the Washington Post, “So far, the Seattle minimum-wage increase is doing what it’s supposed to do”, sheds more light on the strengths and weaknesses of the UW study. [ ] Bernstein is the former chief economist for Vice President Joe Biden.

The post highlights two important shortcomings of the study pointed out by economist Michael Reich of the University of California, Berkeley, one of the pre-eminent scholars of the minimum wage.

The most serious problem of the UW study, he asserts, is that the sample of firms it used for analyzing employment includes only single-establishment firms, those with one physical location. This excludes many of the biggest employers in food service, hospitality, health care services and retail, along with anyone else who operates in more than one location. Multi-location firms account for something like half of Seattle’s jobs. Using the single-establishment sample, low-wage employment increased in both Seattle and the control group, but slightly less in Seattle.

It turns out, however, that single-establishment businesses are not a good proxy for the whole Seattle economy. Critically, the small lag in employment (1.2 percent) relative to the control group, calculated by the study for single-location firms, is reversed and becomes an increase in employment relative to the control group when all Seattle businesses are included. Analyzing the full range of firms, Bernstein notes, “employment outcomes were relatively more positive in Seattle than in the control group, both for all firms and for lower-wage firms.”

In other words, the small negative employment effects so widely referred to by the study and media accounts of it apparently are not present when the entire Seattle low-wage job market is measured, and the overall impact of the minimum wage increase on employment was positive. If this is the case, it would seem that the study should have reversed its presentation: rather than relegating the broader results to the appendices, it should have put them front and center in the body of the study and done appendices on the single-establishment results (or not published them at all). This is a serious enough misinterpretation on a pivotal point that the team should consider republishing a corrected version of the study.

The other shortcoming of the UW report that Bernstein flags is that it does not show the standard errors of its estimates, even though this is a standard statistical practice. Without them, it is impossible to know how reliable the numbers are. In fact, as Bernstein and Reich point out, “[the study’s] calculations indicated that the employment effect was not distinguishable from zero.”

This also leaves us to wonder what bias the same partial dataset introduced into calculating the effect of the wage on total hours worked and worker pay. The study reports a tiny reduction in hours for low-wage workers in Seattle compared with the control group, 19 minutes (about 0.3 hours) per week. This would also be around one percent of total hours. But if the employment effect reversed when the full range of Seattle businesses and employees were included, would the relative reduction in hours similarly become an increase, given that they may be caused by related factors?

The question of net earnings

Another issue with some representations of the UW study by team members is that they seem to be implying that the minimum wage increase doesn’t produce much benefit for low-wage workers. For example, Professor Robert Plotnick, in an interview with Gene Balk of the Seattle Times, “discounted the idea that the city’s increased minimum made much of an impact. ‘I think that’s really unlikely,’ he said. ‘Weekly earnings went up maybe five or ten bucks, because there was an increase in wages but a decrease in hours.’” [“FYI Guy – Are low earners in Seattle moving up or moving out?” September 17, 2016. ]

Well, yes, one went up and one went down, but the relative magnitudes of the changes were very different. There was a substantial net increase in worker income and a very small decrease in hours. The study reports an overall pay increase for low-wage workers attributable to the minimum wage of seven percent: pay in Seattle went up 12 percent, while in the control group it rose 5 percent. And hours went up in both areas: the miniscule relative lag in hours calculated by the study may go away when all Seattle businesses are included. That all comes out, not to a near wash, but to a substantial gain in pay for workers.

That figure of “five or ten bucks” for the average weekly increase, with its implication that the minimum wage had little impact on most workers, does not square with the rest of the study’s data.

Over the range of affected workers, many workers got much bigger raises relative to the control group. But of course there’s a big variation in the wage increase between those with starting wages already close to the new minimum, those in the middle, and those at the bottom near the old minimum. The average increase is usually somewhere around one-half of the nominal increase, depending on the distribution of workers along this wage scale.

Let’s take the example of workers making $9.96 an hour, the median starting wage for affected workers according to the study. Their wage went up to $11.14, the median final wage for affected workers after the 2015 increase. At 35 hours a week, which is typically full-time for low-wage workers, that $1.18 hourly jump comes out to a $41.30 weekly raise. The study attributed about 62 percent of the total increase to the minimum wage increase, with the rest due to the strong local economy. So in this case, the study would credit roughly $25.61 of these workers’ increase to the minimum wage. Knock off the study’s estimates of around one percent for lowered employment and another one percent for reduced hours, and the minimum wage accounts for a weekly increase of $25.10 for this worker.

For those starting at the 2014 state minimum, $9.32, an hourly increase to $11.00 for 35 hours would yield a much larger weekly increase of $58.80. The 62 percent due to the minimum wage would amount to $36.46. Subtracting two percent for lowered employment and reduced hours leaves a weekly increase of $35.73.

Within the group of workers whose wages are affected, some of those on the statistical edges will bring the dollar average of the earnings increase down, but will still benefit from the increase. Many workers included in the study worked only part-time, so of course the absolute value of their pay increase would be smaller, because their total pay is smaller. Putting it in dollar terms is a bit deceptive: expressing it as a percentage would capture the effects of the minimum wage better. Likewise, for workers previously earning close to the new minimum wage of $11.00, the percentage and absolute increase will logically be small. For example, an employee whose wage rises from $10.75 to $11.00 will get only a 2.3 percent increase, which would amount to $8.75 for 35 hours. That’s just the intrinsic arithmetic of any minimum wage raise. For any increase in wages or benefits, those whose old level is already close to the new minimum will always get a smaller raise. This does not mean the policy is not working.

Clearly, Plotnick’s conclusions do not capture the experience of many affected workers. For more than half of all full-time workers (and many part-time ones as well), those starting from somewhere above the median wage down to the old minimum wage, the raise in the minimum fattens that paycheck significantly when the eagle flies, even in a strong economy. From workers’ perspectives, taking home a palpably larger paycheck for working slightly fewer hours is win-win. Besides the extra cash in the pocket, you also get extra hours to commute to the second job or take the kids to soccer practice or do homework for a night class.

These increases reflect the effective strategy of nearly all current minimum wage laws: phase in modest increases over several years. The raises are small enough that they have little if any effect on hours or employment, and on employers’ bottom lines, but big enough that they make a difference for low-wage workers. This has been the formula for nearly all of the new generation of local and state minimum wage increases.

Better indicators of overall economic gains and losses

Besides correcting these serious flaws and gaps in the study and its interpretation, the UW study should also investigate a related question: Is some measure of the change in total yearly income for each low-wage worker, and its variance across demographic dimensions, a more useful indicator of the effects of a minimum wage ordinance than simply jobs, employment, and hours?

If the total job market is reduced by one job or one hundred, that does not mean that this number of workers is out pounding the pavement for months or sitting at home pounding six-packs. This is not a situation where the mill or the mine shuts down and some families have to move on down the road to find new work.

The reality is that a “job” does not mean the same thing in modern low-wage labor markets as it did in the manufacturing economy of past decades. For much of the service economy, especially in the most-affected sectors such as restaurants and hospitality, jobs and hours are volatile. Businesses fail and start up, lay off and hire workers, and lower and raise weekly hours continuously. Turnover is also high in many low-wage sectors. Workers frequently change jobs voluntarily or involuntarily, and their hours fluctuate often and sometimes without notice (although the new Seattle ordinance may help with this). Unemployment insurance, for those who qualify, doesn’t usually amount to much or last long, and few have enough savings to serve as a buffer. But there are often plenty of other low-wage jobs, so most workers can’t and don’t have to stay out of work for long.

This means that, for low-wage workers, job loss doesn’t usually translate into the kind of life-changing event we’ve seen in the Rust Belt when industrial firms shut down. More often, a layoff or quit means a hiatus of a few days or weeks until the next job. If employment is lower and jobs are fewer, it may mean a bit more time between gigs.

As economist Dean Baker puts it, “… the story of job loss due to the minimum wage is not a story of people losing their jobs and going without work for the rest of the year. It's a story of people taking longer to find jobs when they lose or leave their job.” [ “Serious Confusion on the Minimum Wage and Job Loss”. July 6, 2016. ]

Given these realities, the total number of hours worked over a year should incorporate these fluctuations of hours, employment and jobs. When you multiply those total hours by the average wage to produce total yearly earnings, you should get a better measure of the up and down sides of a minimum wage change for a given worker.

Beyond averages, though, effectively measuring the wide variance in workers’ total income and hours over different demographics might require collecting statistically significant samples of individual paychecks over the long run. This approach could create a more granular and real-world measurement of the aggregate effects of a minimum wage increase on workers’ economic well-being. However it is done, what we should be trying to quantify is not just whether any jobs, employment or hours are “lost”, but to what extent workers end up better off overall. The UW study, with its five-year time frame, is in a good position to collect this sort of data.

Some of these arguments are developed in more detail in my paper, “Maximizing Minimum Wages”. It is based on my coverage over the past few years of the local, national and international movements to raise minimum wages and the economic debates underlying it.[ Peter Costantini. “Maximizing Minimum Wages”. Seattle: January 20, 2016. ]

I am an analyst and journalist who has lived and worked in Seattle for the past 43 years. I write mainly for Inter Press Service, a global non-profit newswire based in Rome. For twenty years, I made my living as a blue-collar worker in construction, shipyards, light manufacturing and office-equipment repair.

Peter Costantini

Seattle Correspondent, Inter Press Service –

IPS Writers in the Blogosphere –

Twitter – @petercostantini


Sylvia Allegretto, Arindrajit Dube, Michael Reich and Ben Zipperer. “Credible Research Designs for Minimum Wage Studies.” Berkeley, CA: IRLE Working Paper No. 148-13, September 23, 2013.

Dean Baker. “Serious Confusion on the Minimum Wage and Job Loss”. Washington, DC: Center for Economic and Policy Research, July 6, 2016.

Gene Balk. “FYI Guy – Are low earners in Seattle moving up or moving out?” Seattle Times, September 17, 2016.

Daniel Beekman. “Sawant, UW researchers clash over impact of $15 minimum-wage law”. Seattle Times, September 21, 2016.

Jared Bernstein. “So far, the Seattle minimum-wage increase is doing what it’s supposed to do”. Washington Post, August 10, 2016.

Seattle Minimum Wage Study Team. Report on the Impact of Seattle’s Minimum Wage Ordinance on Wages, Workers, Jobs, and Establishments Through 2015. Seattle: University of Washington, July 2016.

Popular in the Community