Behind the Numbers: Cooked to Order

May 30, 1996

John Schmitt
May-June 1996, The American Prospect 

This is the story of how the fast food industry and its conservative allies sought to discredit two distinguished economists, and how the attack backfired. The economists in question committed the sin of conducting research that challenged the conventional view of the minimum wage. Their attackers may have committed rather cruder sins.

Almost every introductory economics textbook warns that raising the minimum wage will cost jobs. Assuming a standard model of the labor market, the reasoning is a straightforward variant of the law of supply and demand: If you raise workers’ wages, you increase the price of labor—and firms will naturally hire fewer workers.

But in fact recent data provide little support for the theory. Over the last decade or so, as the after-inflation value of the minimum wage has fallen close to a 40-year low, most economic research has found no connection between minimum-wage increases and levels of employment.

In 1994, two well-respected Princeton economists, David Card and Alan Krueger, published an unusually convincing study. The Card-Krueger study, later included as a chapter in their book, Myth and Measurement, used the 1992 increase in New Jersey’s minimum wage as a natural experiment. There had been no such increase in neighboring Pennsylvania. Standard theory would predict that New Jersey’s 19 percent increase would cost jobs. But nothing of the sort happened.

Like several earlier studies in Texas, California, and the United Kingdom, the New Jersey data suggested that minimum-wage increases, in some circumstances, might actually increase employment. How? Labor markets are not exactly like product markets. Wage levels can influence how hard people work and how often they quit. Pay people more, and they might work harder or display lower rates of turnover. So a higher minimum wage might raise productivity and lower recruiting and training expenses that would offset its costs.

For proponents of the minimum wage, the Card-Krueger study provided a novel argument in support. Traditional arguments for raising the minimum wage cited several benefits: increasing the incomes of the working poor, boosting the wage floor for all workers, stimulating demand in the economy, and rewarding work. If there were mild job losses, these could be made up in other ways. But now, evidently, raising the minimum wage (within limits) might even be a job-creation program.


The clash between theory and the real world set off nothing short of a brawl within the economics profession, especially because the Clinton administration cited the research as evidence to support its proposed increase in the federal minimum wage from $4.25 to $5.15 per hour. Opponents of the minimum wage, led by conservative economists and low-wage employers, responded with a no-holds-barred assault against the new research in such influential forums as the Wall Street Journal, the Washington Post, and Business Week.

No one took more public abuse from the anti-minimum-wage forces than Card and Krueger did. The Card-Krueger study drew attention both because of the stature of its authors and because of the meticulous care of their research. Both are prolific, highly respected economists with tenure in a top economics department. In his early thirties, Krueger took a leave from Princeton to serve as chief economist in Robert Reich’s Labor Department. In 1995, the American Economics Association (AEA) selected Card to receive the John Bates Clark medal, given every two years to the best economist under the age of 40.

The study itself was innovative and careful, and exemplified high professional standards in research design and implementation. Prior to the New Jersey study, the usual practice in minimum-wage research was to compare statistical variations in employment with variations in the real value of the minimum wage over time, an approach fraught with technical problems. While economists usually rely on government data collected for other purposes, Card and Krueger undertook their own major telephone survey of fast food restaurants in New Jersey and Pennsylvania.

Card and Krueger surveyed 331 fast food restaurants in New Jersey and 79 restaurants in eastern Pennsylvania, just before the April 1992 increase took effect. They then re-interviewed the same restaurants about eight months later, after the minimum wage rose in New Jersey. Using the Pennsylvania restaurants as a “control” group, Card and Krueger were able to estimate the employment impact of the minimum-wage hike by comparing employment changes in the two state samples. They found that employment actually grew more in New Jersey than in Pennsylvania, though the difference was not statistically significant using their best measure of the effects.

In early 1995 a vicious counterattack began, substantially underwritten by the fast food industry. In March the Employment Policies Institute, the leading employers group opposing the minimum wage, unveiled a counter-study. Richard Berman, a lobbyist for restaurant companies and the executive director of the Employment Policies Inst itute, wrote a piece for the Wall Street Journal‘s editorial page. He claimed to possess evidence from the payroll records of 71 fast food restaurants in New Jersey and Pennsylvania that directly contradicted Card and Krueger’s telephone survey data. He alleged that Card and Krueger’s data were “grossly inaccurate” and “worse than flawed.” The more accurate payroll data, he said, demonstrated that employment fell in New Jersey as a result of the minimum-wage increase. To support his position, Berman cited a March 1995 research paper by economists David Neumark of Michigan State University and William Wascher of the Federal Reserve Board, to whom the Em ployment Policies Institute had made the data available for analysis.

Berman’s allegations kicked off a public op-ed and private whispering campaign to undermine the credibility of the study and the reputations of Card and Krueger. The Washington Post, the New York Times, and other publications repeated, generally as fact, Berman’s claims about his own data as well as his critique of Card and Krueger. Business Week‘s conservative commentator Paul Craig Roberts hit the low point, with a column devoted almost entirely to savaging David Card, a mild-mannered Canadian who assiduously avoids making policy pronouncements on the minimum wage. Roberts questioned “the quality of the review process at The American Economic Review [where the study was first published] and the rigor of the selection process for the awarding of the John Bates Clark Medal.” He charged that “political correctness seems to have crept into the inner sanctum of the AEA, discrediting its scholarly journal and debasing its top prize.”

Of course, Berman’s op-ed piece saw the light of day only because the standards for publishing “research” on the editorial page of the Wall Street Journal are considerably less stringent than at the American Economic Review. While Card and Krueger explain in excruciating detail how they identified and selected restaurants for inclusion in their sample—the heart of proper survey design—Berman to this day has not revealed how the Employment Policies Institute assembled its own, much smaller sample.

Correctly drawn samples accurately describe the broader population and follow patterns well studied by statisticians. Improper or biased sampling design, however, leads to distorted results. For example, if Berman surveyed only firms that did poorly after the minimum-wage increase, or if only firms hurt by the wage hike took the time to respond, his results would be atypical and biased in favor of finding job losses.

Much about Berman’s piece was pure polemic. Even Berman’s statement that telephone surveys are less reliable than payroll records is not as plausible as it seems. The opposite would be true if workers were being paid under the table. And why should telephone respondents systematically overstate employment growth in New Jersey, while consistently understating it in Pennsylvania?


Neumark and Wascher, the economists on whom Berman and the fast food industry relied, became uneasy. A careful reading of their paper revealed that the payroll data did not, in fact, differ much from Card and Krueger’s. In one of two statistical tests, the Employment Policies Institute data showed no statistically significant loss of employment. In a second test, the job losses were only weakly significant by standard statistical criteria.

Neumark and Wascher clearly recognized that the fuzzy results obtained with the Employment Policies Institute’s small sample of restaurants would not convince most economists. They also felt obliged to answer concerns that the Employment Policies Institute, which they acknowledged in a later version of their study had “a stake in the outcome of the minimum wage debate,” may have provided them with data that were “falsified so as to undermine [Card and Krueger’s] results.” They set out to verify the Employment Policies Institute data and, separately, to gather their own payroll data from a larger group of restaurants.

Neumark and Wascher contacted those restaurants that supplied data to the Employ ment Policies Institute and checked the responses. They also conducted their own survey of restaurant payroll records in the same areas originally surveyed by Card and Krueger. In the end, Neumark and Wascher had a sample of 230 restaurants, 80 supplied by the Employment Policies Institute and 150 that they had gathered themselves. (Card and Krueger attempted to interview 473 restaurants and obtained responses from 410.)

When Neumark and Wascher analyzed the Employment Poli cies Institute sample and their own sample separately, a funny thing happened. The slightly expanded Employment Policies Institute sample indicated that the minimum wage did have a significant negative impact on employment in New Jersey (at least in one of the two statistical tests). But Neumark and Wascher’s own data found no statistical difference in employment growth in the two states. Quite unintentionally, Neumark and Wascher had vindicated Card and Krueger.

Neumark and Wascher scrambled to explain their results. They maintained that the combined data provided the best basis for determining the employment effects. But their position was undermined by major differences in the two samples. The Employment Policies Institute restaurants showed much more uniform employment changes than those in the Neumark and Wascher sample. In fact, basic statistical tests demonstrated convincingly that the Neumark and Wascher sample showed so much more variation in employment changes across restaurants that it was highly unlikely that the two samples were chosen randomly from the same population of restaurants. Since the samples were either not random or not from the same population, Neumark and Wascher would be wrong to lump them together.

The irony is sweet. Neumark and Wascher, the two experts used by the fast food industry to impeach Card and Krueger, effectively ratified them. Neumark and Wascher’s own original work also leaves one wondering about the Employment Policies Institute’s data.

There are two ways to resolve that issue. The first is through a full account of how the Employment Policies Institute gathered its data. But neither the Institute nor Neumark and Wascher have so far provided such an account. The second is a direct analysis of the Employment Policies Institute and Neumark and Wascher samples by all parties in the debate. But the most outrageous feature of the whole attack on the New Jersey study is that it has been waged entirely with a “secret data set.” While Card and Krueger have shared their data with Neumark and Wascher and any other interested parties (it is posted on the Internet), the Employment Policies Institute and Neumark and Wascher have turned down all requests to make their data available to the public or even to Card and Krueger, despite Richard Berman’s assurance in a Washington Times op-ed last summer that the data would be made public in July 1995.


The fast food industry and their conservative allies have cooked up a whopper. But when the statistics settle, where does the debate on the employment effects of the minimum wage stand? Despite the attacks on Card and Krueger, or perhaps because of them, the economics profession has shown some signs of accepting the notion that moderate increases in the minimum wage may have little or no effect on employment. In October, 101 economists, including three Nobel laureates and four other past presidents of the American Economics Asso ciation, signed a letter calling for a 90 cent increase in the minimum wage. They stated their belief that a moderate increase would not “significantly jeopardiz[e] employment opportunities.”

A panel discussion on the minimum wage at the most recent American Economics Association annual conference also showed progress. Even critics contended that a 10 percent minimum-wage increase would reduce employment only by, at most, about 3 percent—a more modest claim than in previous years. That might still seem like a lot, but think of it this way: If the minimum wage rose by 20 percent, 94 out of 100 minimum-wage workers would get a 20 percent pay increase, while 6 out of 100 would “lose their jobs.” Since most minimum-wage jobs have a high turnover, it is unlikely that 6 out of 100 workers would be fired. More likely, employers would fail to fill a certain number of vacancies when the current employees left. With fewer vacancies, those looking for minimum-wage work would have to wait a little longer to find work—on average, 6 percent longer. But, once they found a minimum-wage job, it would pay 20 percent more. On an annual-income basis, minimum-wage workers would still come out 14 percent ahead.

Further, given that minimum-wage jobs also tend to be more flexible than most, employers could also compensate for higher wages by cutting hours worked by 6 percent. If this were the case, minimum-wage earners would work fewer hours (about an hour less on a 20-hour week) but make 20 percent more for each hour worked. They would work fewer hours, but their weekly pay would go up by 14 percent. (The most recent version of Neumark and Wascher’s New Jersey study provides some evidence that firms behave in just this way.) This is the arithmetic that explains the overwhelming support that low-wage workers voice for regular and reasonable increases in the minimum wage.

A recent op-ed in the Wall Street Journal by conservative economist Robert Barro suggests a tactical retreat and a new set of arguments from an odd quarter. Barro concedes that “[w]hile the net negative effect on employment is small,” minimum wages lead to “disturbing compositional changes.” Citing other research by Neumark and Wascher, he argues that a higher minimum wage would induce “more advantaged,” “nonblack/non-Hispanic” teenagers to drop out of high school in order to work full time at the new, higher wage. These dropouts would presumably displace “more disadvantaged, notably black and Hispanic” workers who now hold those jobs.

Note the sublime irony of this argument. Barro, who made his name in economics by asserting the pure rationality of individuals, is now claiming that advantaged white teenagers will volunteer to be branded “high school dropouts” for the rest of their working lives, in exchange for an extra 90 cents an hour. That hardly sounds rational.

So we have made some progress after all. Opponents of the minimum wage are having to resort to ever more far-fetched arguments. The honor of Professors Card and Krueger has been restored. Their attackers are on the defensive as statistical short-order cooks. And more than a few mainstream economists seem newly willing to entertain the concept that it is efficient for work to pay a living wage.



Support Cepr


If you value CEPR's work, support us by making a financial contribution.

Si valora el trabajo de CEPR, apóyenos haciendo una contribución financiera.

Donate Apóyanos

Keep up with our latest news