Economics: Minimal Evidence
Flawed studies drive the call for an increased minimum wage.
As the old saying goes, economists may not be able to agree on the time of day. But for many years, almost all economists have agreed on one issue: Raising the minimum wage leads to greater unemployment among workers with low skills. Both basic theory and many empirical studies support that conclusion. Textbooks preach it, and even liberal economists who chafe at the notion teach it to their students.
Or so it was until recently. But as nature abhors a vacuum, economists prefer controversy to consensus—especially when the consensus puts their professional judgment at odds with their political instincts. Accordingly, the decades-long agreement on the employment effects of the minimum wage now has been challenged by several scholarly papers examining survey data on employment in fast-food restaurants and other low-wage sectors. Four widely publicized studies argue that raising the minimum wage has not led to greater unemployment among low-skilled workers.
Unsurprisingly, this new research has quickly become politicized. Labor leaders, who like high minimum wages because they make union pay scales more competitive, have seized on the new research to disparage the traditional view. So have Labor Secretary Robert Reich and President Clinton. "Now, I've studied the arguments and the evidence for and against a minimum wage increase," Clinton claimed in his State of the Union Address. "I believe the weight of the evidence is that a modest increase does not cost jobs, and may even lure people back into the job market."
Well, not so fast, please. While the new studies are serious works, done by serious scholars, they are not particularly believable. Indeed, they have several problems in common, as well as various individual flaws. And the (seemingly) most persuasive study turns out to be based on abysmally bad data.
To understand the studies and their problems, it is important to understand the traditional argument that raising the minimum wage reduces employment among low-skilled workers. The reason is simple. The demand for labor—the number of manhours (not workers) employers want to buy at different wage rates—is determined by the expected productivity of the given workers. So if wages rise by decree, rather than because workers can produce more per hour, employers will hire fewer hours of labor.
Moreover, not all low-skilled workers are created equal. Some have more skills than others. Some have a few or all of such advantages as good looks, physical strength, a flexible schedule, greater intelligence, more or better schooling, stronger communication skills, better health, greater perceived honesty and character, a lower perceived likelihood of resorting to the litigation system, ad infinitum. In general, they have higher expected productivity than other workers with low skills.
Other people are not so lucky. For any number of reasons they may appear "riskier" to prospective employers. They may have physical or mental handicaps, weaker references, or small children interfering with work schedules. They may be from neighborhoods or ethnic groups that lead prospective employers to expect (rightly or wrongly) a greater-than-average chance of various problems. They may have transportation problems, annoying personalities, poor personal hygiene, and so forth.
Given that employers are in business to make money—to get the biggest labor productivity bang for their wage buck—all of the anti-discrimination regulation in the world cannot overcome the simple reality that an increase in the minimum wage leads employers to hire relatively more-productive workers rather than relatively less-productive ones, who otherwise might be cheaper. The minimum wage thus makes it harder for the low-skilled—and for the lowest-skilled among them—to compete for employment. Elementary economic analysis predicts unambiguously that increases in the minimum wage, other things being equal, will reduce the employment of low-skilled workers.
Or so we all thought. The revisionist studies claim that, at least in a few instances, this story has a different ending. But a careful examination of the studies suggests that the traditional view is far from discredited.
? The Card/Krueger New Jersey-Pennsylvania Study. The most frequently cited, and seemingly most convincing, new study takes advantage of a "natural experiment" created when New Jersey raised its minimum wage from $4.25 an hour to $5.05 in April 1992. David Card and Alan Krueger of Princeton reasoned that since economic conditions ought not vary greatly between southern New Jersey and eastern Pennsylvania, which are essentially a single economy, looking at employment trends in the two states ought to reveal the effects of the minimum wage.
Card and Krueger conducted telephone surveys of about 400 fast-food restaurants in February–March 1992, and then again in November–December 1992. They asked questions about full- and part-time workers, wages, benefits, and prices. From their statistical analysis of those survey data, Card and Krueger not only "find no evidence that the rise in New Jersey's minimum wage reduced employment at fast-food restaurants in the state," but "find that the increase in the minimum wage increased employment." Indeed, the Card/Krueger statistical analysis suggests that the 18.8-percent increase in the New Jersey minimum wage yielded a 20.8-percent increase in employment relative to the Pennsylvania sample.
One immediate problem is that the authors looked only at major fast-food chains: Elementary economic analysis does not say that if you increase the minimum wage, employment will go down in every business—or in any particular business. The higher minimum wage might have differing impacts across firms. Indeed, it is possible that the major fast-food chains might emerge better off if the increased minimum wage raises costs at such smaller competitors as mom-and-pop fast-food stands.
Moreover, the Card/Krueger study turns out to have a major flaw: The survey data upon which it depends are lousy.
Suspicious of the Card/Krueger data and findings, the Employment Policies Institute gathered the actual payroll records from the Burger King franchises in the Card/Krueger zip codes and compared them to franchises surveyed in those zip codes. The survey data were wildly inconsistent with the payroll records. (The payroll sample also includes some restaurants that Card and Krueger missed.)
Independently, David Neumark of Michigan State and William Wascher of the Federal Reserve noticed that the variation in employment changes across the surveyed restaurants in the Card/Krueger sample seemed implausibly large—some restaurants had supposedly added huge numbers of employees while others had supposedly cut large numbers. In relatively small businesses, this sort of fluctuation seemed odd.
So Neumark and Wascher reviewed the payroll employment data gathered by EPI. When they applied the payroll data to the same econometric model used by Card and Krueger, they got completely different results. The variation in employment changes declined markedly, and analysis of the new data yields an estimated 4.8-percent decline in New Jersey employment relative to the Pennsylvania sample as a result of the higher minimum wage. Where payroll data could be compared with survey data for specific restaurants, Neumark and Wascher also found numerous errors in the Card/Krueger data.
Looking just at Burger King restaurants, for instance, the Card/Krueger survey data show employment declines in two of three Pennsylvania zip codes, while the payroll data show employment increases in all three zip codes. Neumark and Wascher conclude that the questions used by Card and Krueger were too vague to generate precise information. For example, the survey asked how many "full-time" and "part-time" employees a restaurant had. But it didn't define either those terms (40 hours a week? 30?) or the relevant time period (within the last week? month? year?), leaving different restaurant managers to define the question differently. In short, using the actual payroll data instead of the survey "guesstimates" effectively refutes the Card/Krueger findings yielded by the New Jersey/Pennsylvania "natural experiment."
? The Card State-Group Study. In another study, Card uses the April 1990 increase in the federal minimum wage (from $3.35 to $3.80 per hour) to produce another "natural experiment." He reasons that the increase ought to have affected various states differently. The new minimum would make a big difference in states where relatively few low-skilled workers earned $3.80 an hour before the increase; it would not matter as much in states where many low-skilled workers were already earning at least $3.80.
Card notes that in 1989 the proportion of teenage workers earning between $3.35 and $3.79 an hour varied from less than 10 percent in the New England states and California to more than 50 percent in many southern states. Accordingly, he divides the states into "high-wage," "low-wage," and "medium-wage" groupings.
The central issue is what happened to employment across the state groupings. In the crude group comparisons, Card finds a larger fall in teenage employment in the high-wage states than in the low-wage states, an outcome inconsistent with the traditional view of the minimum wage. Indeed, he finds an increase in teenage employment in the low-wage states, with no effect in the medium-wage group.
But of course, the minimum wage isn't the only factor affecting employment. The state's general economic climate and growth rate are also important. Card recognizes that differences in economic conditions and other factors might account for his findings. After controlling for them econometrically, he concludes that such differences in labor market conditions might in fact explain all of the variation in teenage employment growth. But he says also that "there is no indication of an adverse employment effect [caused by the increased minimum wage] in the low-wage states…."
That latter finding is quite weak: It would be one thing to find that an increase in the minimum wage yielded an increase in low-skilled employment, other things being equal. But to say that no negative effect can be found in the data means next to nothing. It says little more than the data are so imprecise or there is so much measurement error that the predicted effect is difficult to discern. The effect of the minimum wage gets lost in the noise—a weak basis indeed for fundamental change in the traditional view of the minimum wage.
And employers don't necessarily wait for the minimum wage to rise to cut jobs. They may have sufficient advance notice to make gradual adjustments accordingly. The Card paper ignores this. Neither does it look at reductions in fringe benefits or, even more important, changes in manhours hired—the more relevant parameter—as opposed to numbers of teenagers working.
Most important, the Card analysis examines employment changes over a one-year period; but it is very easy to believe that the demand for low-skilled labor over so short an adjustment period is highly inflexible ("inelastic"). Looking over a longer adjustment period might very well show stronger employment effects. After all, wages on average rose by only 6 percent in the low-wage states; if businesses adjust gradually to wage increases—for instance, by reducing employment through attrition—the resulting employment effect might be real but too small to discern in the data over a short period.
? The Card California Study. California raised its minimum wage from $3.35 an hour to $4.25 an hour in July 1988; the federal minimum wage remained unchanged at $3.35. In this third paper, Card tries to test the effect of the minimum wage by comparing changes in low-skilled employment in California with changes in a group of "comparison states." Looking at the 1987–89 period, he concludes that the data suggest "a gain in [California] employment following the rise in the minimum wage." He also argues that "groups with a higher fraction of low-wage workers do not appear to have suffered any relative losses in employment" as contrasted with trends in the comparison states.
Card argues in this paper that his control areas—Arizona, Florida, Georgia, New Mexico, and the Dallas-Fort Worth area—were "a legitimate control group" for California over the 1987–1989 period in terms of economic growth and other important parameters. But that premise is highly debatable, as the data in Table 1 show. Real growth in California was substantially greater than the weighted average for the comparison group, with the exception of 1988; that anomaly is due largely to strong growth in Texas (included as a proxy for growth in the Dallas-Fort Worth area) that year. It is hardly surprising, then, that California produced more jobs for low-skilled workers.
More fundamentally, Lowell Taylor of Carnegie-Mellon has examined employment growth (or losses) across California counties and across California retail sectors when the minimum wage went up. In counties and retail sectors in which the increased minimum had the greatest wage impact, Taylor finds the greatest adverse employment effects as well, a refutation of the Card findings.
? The Katz/Krueger Study of Texas. Princeton's Krueger and Harvard's Lawrence Katz surveyed well over 100 fast-food restaurants in metropolitan areas of Texas (from the Burger King, Wendy's, and Kentucky Fried Chicken chains) in December 1990 and in July and August 1991. These surveys followed the increases in the federal minimum wage in April 1990 and April 1991. With complete data for 100 restaurants, the authors' statistical analysis finds greater employment growth in the restaurants most affected by the increase in the minimum wage. In other words, the increases in the minimum wage yielded increases in fast-food employment, or, at a minimum, no effects upon that employment.
First, the Katz/Krueger sample is limited to restaurants operating both before and after the increases in the minimum wage. If the increases forced some out of business, or reduced the rate at which new restaurants were opened, the analysis, as Katz and Krueger recognize, would fail to pick up that change. It would therefore underestimate the adverse employment effect of the higher minimum wage.
More important, the Katz/Krueger analysis emphasizes wage differences across fast-food restaurants in Texas. Logically, the increase in the minimum wage ought to have affected low-wage establishments the most, but Katz and Krueger find greater employment growth in those restaurants. This is less meaningful than it may appear.
At the time, low-wage regions of the United States were growing faster overall than high-wage areas. The same might have been true within Texas, but the Katz/Krueger analysis does not control for different growth rates. And if low-wage areas were growing faster, employment might still have jumped despite the increase in the minimum wage. In short, the Katz/Krueger paper provides little basis to reject the traditional view of the employment effect of the minimum wage.
Thus does the Law of Demand still stand. New research by Donald Deere and Finis Welch of Texas A&M and Kevin M. Murphy of the University of Chicago finds substantially greater employment losses after the 1990 and 1991 increases in the federal minimum wage for population groups with larger proportions of low-wage workers, high school dropouts, and minority dropouts.
Moreover, people who advocate increasing the minimum wage seem not to have examined its income distribution implications carefully; instead, they make the usual glib argument that an increase in the minimum wage will help "the poor" or even, in President Clinton's words, "the underclass." But the minimum wage tends to shift employment within the class of low-skilled workers, from those with relatively less productivity to those with relatively more—squeezing "the underclass" out of jobs in favor of less-risky, more-productive employees.
And, of course, many people working at the minimum wage are not "poor." They may be middle- or upper-class teenagers working part-time, second earners in a family, seniors working to keep busy, etc. Michael Horrigan of the Bureau of Labor Statistics and Ronald Mincy of the Urban Institute report data for 1987, dividing into income quintiles all "minimum wage families"—families in which at least one person works at the minimum wage. Table 2 presents their findings. Far from what advocates of the minimum wage would have us believe, most minimum wage families are not poor; indeed, about 60 percent are in the middle of the income distribution or higher.
The effect of the minimum wage is an important subject for scholarly exploration and debate. But it is one thing for scholars to write papers challenging conventional wisdom, which then can be subjected to debate and renewed analysis. It is quite another for politicians to latch onto new findings uncritically in their pursuit of interest group advantage. A willingness to argue that water flows uphill is precisely the present stance of the Clinton administration with respect to the minimum wage, and it is unlikely to yield salutary outcomes.
Benjamin Zycher is vice president for research at the Milken Institute for Job and Capital Formation in Santa Monica.