Why it’s so hard to study the impact of minimum wage increases

Amazon announced this week that it would pay workers no less than $15 per hour.
Amazon announced this week that it would pay workers no less than $15 per hour.
Image: REUTERS/Jeffrey Dastin
We may earn a commission from links on this page.

If you’re like most college graduates, you took an intro to economics class that left you with at least this single and powerful concept: supply and demand are lines that cross on a graph at a magical center point known as price equilibrium. When our politicians and others say that “the market will sort it out,” a belief that verges on deity worship for some, this foundational tool of economics is often what they have in mind: buyers and sellers deciding, in the rough-and-tumble of free enterprise, how much of, and at what price, a good or service will change hands.

According to the sketch above, mandating a minimum wage above that market-determined sweet spot should put people out of work. The higher wage reduces the number of work hours employers are willing to fund, by an amount that depends on where along that demand line the minimum wage sits. Above a certain level, a minimum wage would reduce demand so much that income losses to people who lose jobs or hours would exceed the pay gains other workers get from the higher wage floor.

In the real world, however, mandated minimum wages don’t necessarily lead to job losses. There are theories as to how an innocuous jobs effect might be possible, and we’ll go into those more below.

But the research evidence of what actual minimum wage requirements do to job numbers goes both ways; many studies find that minimum wage laws reduce employment, and many other studies on the exact same laws find they have little or no effect on jobs. Some 60 years and hundreds of research papers from prestigious universities, government agencies and private organizations have created little consensus on the subject, academic or otherwise. Just last year, separate Seattle minimum wage studies by researchers at the University of Washington and the University of California Berkeley suggested polar opposite effects.

Sorting out the controversy is urgent.

Many newer state minimum wage laws include provisions for automatic increases over the next few years. By 2022, 17% of Americans will live in a city or state with a $15 minimum wage. (Change could come far more rapidly in the wake of Amazon’s decision to boost pay to $15 an hour for all its workers November 1. The move will affect 250,000 existing Amazon employees and another 100,000 or so seasonal workers this holiday season and beyond.) If the Fight for $15 movement continues to gain support, more cities and states will adopt higher minimum wage laws, even if the federal government and some states stay stuck at $7.25.

In recent years, the minimum wage debate has been reduced to pure economics. Do rate hikes cut job hours? hurt small businesses? boost consumer spending? cause prices to rise?

Contradictory findings, even by economists analyzing the same data sets, have been the frustrating norm in this field since David Card first proposed in the 1990s that a New Jersey minimum wage hike didn’t actually put people out of work. A pattern followed: a respected researcher publishes an important minimum wage and employment study, and another respected researcher follows with a direct take-down of its findings. The work has created protracted debate and dueling findings among the field’s top experts.

Edward E. Leamer, a UCLA Anderson economist, walked us through the controversies and confounding contradictions in this field, using his own research experience and talent for skepticism to find truths amid the fracas. He’s not a fan of minimum wage hikes, but he’s agnostic in his reviews of minimum wage research. “Ninety-nine percent of what economists believe is the theories they put forward,” he said. “That’s what leads most of them to ignore evidence. I’m a believer in evidence, not theory. The basic supply and demand model just doesn’t work for the labor market.”

In search of the perfect control group

Much of the controversy in this field boils down to disagreements over research study designs. In particular, the control group selection in any minimum wage study is likely to spark a hot debate, no matter what method the researchers used to devise it.

Ideally, a control group mimics the experimental group in every way except for whatever treatment is being measured. The standard approach for empirical economics calls for taking pay level and employment measurements before and after a minimum wage event, then comparing those measurements to an unaffected control group. In a perfect set-up, only the minimum wage differences could lead to different job and wage trends in the experiment and the control group. But in the real world, no two geographic locations have identical employers or workers or economies, and each of those factors affects jobs regardless of minimum wages.

The earliest studies compared data from a state that raised minimum wages to data from a nearby state that didn’t. But is the economy of California really comparable to that of nearby Utah or Idaho?

States with a $7.25 minimum wage (the floor set by federal law) are clustered in places with fewer unionized workers. Some of those regions have particularly high rates of migrant and immigrant workers. Employment trends vary by state and region, largely for reasons that have nothing to do with minimum wage events.

For many years, researchers have manipulated these control groups to weed out factors that don’t equally affect employment in both locations. But there are huge differences among experts over how the data can justifiably be manipulated.

Synthetic controls

As researchers struggle to find a universally accepted model for these studies, some econometricians are building artificial control groups to substitute for real-life geographic areas.

A UC Berkeley study that looked at the impact of Seattle’s minimum wage, for instance, relied on a sample of localities with stagnant minimum wage rates in recent years.

When data from those counties is combined and weighted variably, it creates an economy very similar to Seattle’s, according to the researchers who developed this synthetic control group. In other words, employment trends of Walla Walla-Carbon-Snohomish-and-42-other-counties combined should be the same as Seattle’s would have been without recent minimum wage hikes to $13 an hour, and as it moves to $15 an hour.

Using this comparison, the study concludes that Seattle’s minimum wage did not lead to job losses.

Of course, synthetic control groups have their critics too, including Leamer. “This is ridiculous,” he says. “I suggest that a synthetic cohort for myself would be Roger Federer and Stephen Hawking.”

Imperfect data sets

The University of Washington’s Seattle minimum wage study—the one that said Seattle lost lots of jobs from the rate hikes—illustrates another obstacle to credible findings in this field.

These researchers used individual pay records to see if low-wage workers lost jobs when the minimum wage rose.

Theoretically, this was an improvement over data sets from past studies, which, for lack of anything better, usually relied on some overall wage and payroll data that lumped together employees earning minimum wage with other workers.

But the UW data set had shortcomings too: it did not include data from employers with multiple locations that are believed to employ some 40 percent of Seattle’s low-wage earners. So, mom-and-pop retailers were counted, but McDonald’s franchises and the like were not.

Self-serving advocates

Opinionators in this field often attack the credibility of studies they don’t like by pointing out biases of the institutions that support them.

Job-killing results are applauded by business groups, such as the National Restaurant Association, that tout the higher unemployment effect as bad for the economy. Their members set wages for hundreds of thousands of Americans at the lowest legally allowed minimum.

Labor organizations, like Service International Employees Union, promote studies that find minimum wage hikes create positive changes for workers and the economy. This union was a driving force in getting $15-an-hour minimum wages passed in cities that include Los Angeles and Seattle. But per SIEU request, the new laws made its own dues-paying members ineligible for the higher minimum wage, arguing that they might want to negotiate lower-than-minimum-wage pay in exchange for better benefits.

Monopsony vs. that classic labor curve

Why wouldn’t minimum wage hikes kill jobs, as that classical labor curve predicts? Recent research points to monopsony, a sort of monopoly whereby the business has buying power over labor rather than selling power over buyers. Monopsony could be keeping wages lower than an absolutely mobile work force would produce, meaning employers forced to pay a higher minimum wage don’t feel as much pain as theory would suggest.

Monopsonists can set lower wages because they have little or no competition from other employers for the workers they need, like in a rural town where a prison is the only major employer. Monopsony also thrives in fields where non-compete agreements keep workers from leaving one job for a better paying job at a competitor. In cities, employers gain monopsony power when they don’t compete for labor via wages, that is, they pay the same low rates. Economists agree that monopsony can significantly reduce the wage floor.

Too High, Too Low, or Just Right?

Regardless of any employer’s unnatural wage-setting power, most economists believe the classical laws of supply and demand take over at some level. Theoretically, for example, a $100 minimum wage in Kansas would kill businesses there, because many don’t have profits enough to pay it, or the ability to pass on the costs to customers.

The ideal minimum wage counteracts the power of monopsony, leaves prices and employers’ profits relatively stable, and employs the workforce, while avoiding that doomsday scenario. Although there are questions about the rate of increases―perhaps sharp minimum wage increases cut jobs but incremental changes don’t―the fundamental disagreement here is over the amount. Is the market’s true equilibrium at $15 an hour? $7.25 an hour? Or is it some amount above or below either of those?

Perhaps one day a consensus on the research will decide. But municipalities aren’t waiting. Last week, the Port Authority of New York and New Jersey agreed to lift minimum wages for 40,000 airport workers to $19 over the next four years.

This article was adapted from Through the Minimum Wage Looking Glass, a guide by UCLA Anderson Review.