An Access Advertising EconBrief:
When Fighting Fire With Fire Just Makes a Bigger Blaze
Fans of the classic television series Get Smart will recall the snappy comeback of secret agent Maxwell Smart to a malefactor indignant at the prospect of detention: “You’re not going to arrest me on this flimsy evidence, are you?” “No,” Smart replied confidently, “I’ve got some more flimsy evidence.”
The quality of empirical debate over public policy has deteriorated to this level. Just as politicians are now compelled to act virtually any time something goes wrong, no matter what it is or how slim the likelihood of successful intervention, no exchange of opposing views is complete without quantitative citation. As soon as one side unveils its numbers, the other side must respond with numbers of its own – no matter how far-fetched or badly compiled. It is a Newtonian law of equal and opposite polemical reaction.
As a result, public discourse is now debased to the point of decadence. The long-running debate over the minimum wage has plumbed these depths of intellectual degradation. In the August 21 Wall Street Journal op-ed, “Do Higher Minimum Wages Create More Jobs?” authors Liya Palagashvili and Rachel Mace probe for the bottom. It is as if they have rewritten Mel Brooks’ script: “You don’t expect me to believe this flimsy evidence, do you?” “Well, my flimsy evidence is a lot better than your flimsy evidence!”
The Left Wing’s Flimsy Evidence
Op-ed authors Palagashvili and Mace (hereinafter, P&M) correctly relate that the left-wing Center for Economic and Policy Research (CEPR) released a report purporting to demonstrate the success of state-level minimum-wage increases in increasing relative employment growth among states. The report was released in June, 2014, and used data compiled by the federal Bureau of Labor Statistics. It examined 13 states that increased their individual minimum wage (as distinct from the federal minimum wage) that month and compared them to the other 37 states whose minimum wage did not rise. The report claimed that the average overall employment growth among the 13 states exceeded that of the 37 states for the five-month comparison period.
The Obama administration appropriated these conclusions with the alacrity of a police department confiscating drug-dealer assets. As P&M note, there was the little matter of “why [the] firms [would] hire more workers when the government raises the cost of hiring workers?” The straight-faced answer was that “hiking the minimum wage raises the incomes of poor workers, causing them to spend more. This additional spending, in turn, is so great that firms hire even more workers.” No less a personage than Barack Obama himself got into this act. “That [worker spending] gets churned back into the economy. And the whole economy does better, including the businesses.”
A priori, this “theory” of economic development is so ludicrous that it would qualify for an evening comedy skit at an American Economic Association convention. “Ludicrous” means ludicrous a priori; its theoretical underpinnings are so completely lacking that nobody would take it seriously enough to investigate. Well, nobody should – these days, no premise is too ridiculous if it can backstop a political point. Our Economist-in-Chief in the White House needs to bolster his standing with the public and shore up two key constituencies. One of those is obvious – the poor, downtrodden low-skilled workers who allegedly benefit from the minimum wage. The other is hidden – the higher-skilled workers, particularly union members, who substitute for the low-skilled workers laid off after the minimum-wage increase.
The “spending rescue” thesis is the culmination of two decades’ worth of left-wing attempts to promote the minimum wage as the salvation of the poor. This crusade began in the early 1990s, when economists David Card and Alan Krueger published a now-legendary study purporting to show that imposition of a minimum wage in New Jersey increased employment there relative to Pennsylvania. The defects of this study have since become almost as legendary as its conclusions. It utilized phone surveys to gather data – a technique heretofore shunned within the profession but thereupon praised as innovative and groundbreaking. But when other economists attempted to confirm the results using payroll data, this change instead reversed the results of Card and Krueger. The study’s econometrics has been panned by expert econometricians. Card and Krueger themselves were unable to supply a theoretical rationale for their result. Ordinarily, this would have been a fatal defect, but the policy implications of the study’s results were so delicious to the left wing that Card and Krueger were lionized and have gone on to professional fame and fortune. The only valid theory that would support their result does not comport with the reality of labor markets.
Why is the left so desperate to validate such a worthless policy measure? Their anxiety derives from the unique qualities of the minimum wage: it hides the benefits to their treasured constituency (unions), masquerades as a godsend to the poor while actually screwing them, and visibly appears to screw the rich (business owners, all of whom are assumed “rich” by definition) while actually doing so only in the short run. What a deal! The “optics” of the minimum wage are ideal for the left; that is, its visible or apparent effects are politically beneficial to them. Of course, its actual effects are harmful to everybody except the special-interest monopolists who comprise the left wing’s leading constituency these days, but that is jake with the left. Their ultimate goal is power – increasing real incomes for special interests are only a means to that end.
The Traditional Economic View of the Minimum Wage
Until Card and Krueger came along, the minimum wage vied with tariffs and quotas on foreign goods for the title of “most unpopular policy measure” among professional economists. Nearly a half-century of empirical examination reaffirmed the verdict of a priori theory: minimum wages redistribute jobs and real income from some poor and low-skilled workers to other poor and low-skilled workers by reducing employment, closing some businesses and temporarily reducing profits earned by businesses utilizing low-skilled labor.
These results are the outgrowth of the impact felt by business upon imposition of the minimum wage. Formally, it acts like a tax on the employment of low-skilled labor, which is the kind of labor directly affected by the minimum wage. That tax has three kinds of impact: a substitution effect, an output effect and a profit effect. (The first two of these are analogous to the substitution and income effect of a price change in consumer demand theory.) The substitution effect causes firms to employ less low-skilled labor and more of other inputs, including the higher-skilled labor previously mentioned as well as machinery that substitutes for labor. The output effect causes businesses employing low-skilled labor to produce less output, thereby employing fewer inputs of all kinds including labor. The profit effect reduces the profits earned by firms employing low-skilled labor. This third effect is only temporary, because the exit of some firms from the industry due to insolvency or better opportunities elsewhere will eventually raise the rate of return back to its previous, competitive level. That is why so-called rich business owners are adversely affected only transitorily by the minimum wage. The “permanent” gains go to workers who retain their jobs at the higher minimum wage. The “permanent” losses are suffered by workers who lose their jobs, some of whom may leave the labor force altogether. This phenomenon of exit from the labor force is by now well-known to most Americans; it has reached its highest level in over thirty years.
This is a formidable a priori case against the minimum wage. Economists never doubted that the minimum wage adversely affected employment of poor and low-skilled workers; they only doubted the degree to which this was true. Empirical studies of this issue began in the late 1940s, conducted by luminaries like future Nobel laureate George Stigler. Over the succeeding decades, economists used formal statistics to enforce the conditions necessary for a valid empirical examination of the issue.
One common defense of the minimum wage made by newspaper editorialists and readers over the years is that “the minimum wage went up but the U.S. unemployment rate did not go up; in fact, it went down, which proves that the minimum wage does not adversely affect employment.” This argument is invalid for several reasons. First, the minimum wage only affects employment within firms and industries that hire low-skilled labor. That does not begin to comprise the entire U.S. economy. Second, even within those industries directly affected by the minimum wage, the overall effects on employment of labor are equivocal. The substitution effect causes employment of less low-skilled labor but more higher-skilled labor, while the output and profit effects cause less employment of all inputs. It is not unusual at all to find that a liberal administration increases both the minimum wage and the money supply, with the latter causing temporary gains in income and employment that can swamp job losses associated with the minimum wage. This is not only ironic – since it harms the very people purportedly highest among the concerns of the left – but fully compatible with a condition in which the minimum wage causes job losses while the overall unemployment rate falls.
To avoid being fooled by effects outside the scope of the minimum wage, economists confined their studies to low-skilled workers and corrected their statistical methods to correct for trends and outside influences. That has been the traditional focus of econometrics, to compensate for the ways in which social sciences differ from the laboratory experiments common to the physical sciences.
Now, though, traditional econometrics has taken a back seat to raw political desire. And this corrupting influence has infected both sides of the political spectrum.
The Right Wing Retaliates With Its Own Flimsy Evidence
P&M disdain virtually all of the history and a priori theory cited above. They have their own flimsy evidence to present against the minimum wage. Their case is purely quantitative; clearly they believe in fighting fire with fire. They begin by finding the portion of the labor force comprised of low-skilled labor, which is roughly 2%, insufficient to generate the high-powered spending necessary to outweigh the minimum wage’s disincentives.
While no doubt true, this leaves room for counterargument by the left. Minimum-wage proponents will respond by accusing P&M of “overlooking” the greater propensity to spend by the poorest families. This is a feeble rebuttal, but the average person won’t know the difference and will probably rule the point a draw at best.
P&M then make a stronger point – that the logic of proponents’ case should mean that bigger minimum-wage boosts should have bigger effects on employment. In fact, the opposite was the case in January-May, 2014. The three substantial minimum-wage increases took place in Connecticut, New Jersey and New York, the three falling between 5% and 14%. Yet these three states had the worst job growth of the 13 increase-states, an average of 0.3% compared to the 1.28% average increase in the other 10 states. “Indeed, job growth was worse in each of these three states than it was, on average, in the 37 states that did not raise their minimum wage at all,” P&M report. And “in New Jersey, the state that hiked [the] minimum wage the most – to $8.25 an hour from $7.25 – employment actually fell by about 0.56%.” In the state with the largest job growth, WashingtonState’s 2.1%, the minimum wage went up by a whopping 13 cents per hour, or almost $24 per month for a full-time employee.
If P&M had rested content with this demonstration, they could have escaped criticism. Up to this point, they were merely using the left’s own evidence against it without accepting its methods. They were showing that the left’s argument wasn’t consistent even in its own terms, albeit without demonstrating how hopelessly confused those terms really were.
But P&M couldn’t stand prosperity. To a roll of drums, they unwrapped the crown jewel in their collection. “We conducted a statistical analysis of the Bureau of Labor Statistics’ data called a two-sample “t” test for comparing two means. We found, for this time period, no difference in the job-growth trend in the states that raised their minimum wages from states that did not. In other words, the correlation cited as debunking the economic case against the minimum wage is not statistically significant.”
Ta-daaaaaaa!!! Too bad there are no bows taken in print media; P&M would surely rate a round of applause in a run-of-the-mill graduate school economics seminar for their performance. It is surely no coincidence that “Ms. Mace studies economics at GeorgeMasonUniversity” while Ms. Palagashvili is a law-school fellow at NYU. Alas, they have displayed academia at its worst.
That is not to say that P&M flubbed their econometric dubs by conventional standards. We don’t know because we can’t see their results and have only their word as to their findings. But taking their comments at face value, it seems that they followed what have become standard econometric procedures. The t statistic is the standard one for small-sample tests of statistical significance. A comparison of sample means is a basic econometric procedure. Almost certainly, they assumed the standard “null hypothesis” of no difference between average job growth in the 13 states as compared to job growth in the 37 states. In this context, “no difference” does not mean that the two averages are exactly the same, which they obviously aren’t. It means that the degree of correspondence between the two is not sufficient as to enable us to be confident that the correspondence was not due to random chance. And just what does “confident” mean? The standard meaning for it is that we must be at least 90% certain. Lacking that degree of confidence, we enter a finding of “statistically insignificant” – which means that the minimum-wage increase did not “cause” the increases in job growth.
It is overwhelmingly likely that the readers of this op-ed – who undoubtedly make up a sample of Americans that is far more intelligent than any randomly chosen sample – fall into two categories: those who have no idea what P&M’s “statistical significance” paragraph meant and those who think they know but are wrong. Those who correctly understand it probably represent a statistically invisible sliver of its readership. And a majority of economists and statisticians are excluded from that sliver.
P&M thought that they were “one up” on the minimum-wage proponents at CEPR because they (P&M) were using the tool of statistical significance as it has been used for decades in academia and government. That statement would be correct only if the word “misusing” were substituted for “using” in two places. That is why they were fighting fire with fire – they were responding to CEPR’s misuse of numbers with their own misuse of statistical inference. Their mistakes were just fancier than CEPR’s, that’s all.
The Flaws of Statistical Significance
Various authors have expounded the flaws of statistical significance as developed by the late statistician Sir Ronald Fisher. The most comprehensive treatment is probably that of Deirdre McCloskey and Stephen Ziliak, The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice and Lives. For our purposes, it is sufficient to summarize how one of the two groups referred to above views the notion of statistical significance and compare it with the truth.
Ask readers of the Wall Street Journal op-ed to explain the meaning of P&M’s statistical significance paragraph in layman’s terms. Those who think they know the answer will probably say something like the following: “Well, it means that the effect of an increase in the minimum wage on overall job growth is insignificant, the opposite of significant. That means it is “too small to matter.” It’s so small we can’t be confident that something else might not be causing what we’re seeing in job growth.” That’s an intuitively appealing explication for at least two reasons. First, it incorporates the familiar meaning of the words “significant” and “insignificant.” Second, it incorporates the kind of answer we are looking for when we do empirical research on issues like this. Typically, we want “how big” or “how much” kinds of answers rather than “yes or no” types of answers.
Unfortunately, the concept of statistical significance is not what most people think it is. Its findings do not convey any quantitative sense of how big an effect is or how much influence one variable (such as an increase in the minimum wage) has on another (such as state-level growth in employment). Rather, it is a binary, “yes-no” type of concept. It registers the likelihood that the influence of one variable on another is random, as compared to systematic or non-random. Because the variables involved are invariably derived from sample data, it can be viewed as a verdict on the representativeness of a chosen sample.
This is useful information, to be sure. But it is not the most useful information we could wish to obtain. And that is a crying shame because the obsession with statistical significance has pretty much overshadowed everything else in empirical research in the social sciences and even in much of the physical sciences today. This has reached such epidemic proportions that McCloskey, a leading economic historian and econometrician, declares that most statistical work in economics done over the last thirty years is useless and must be done over. That is tantamount to saying that we might as well junk the leading academic journals published during that interval.
Fighting Fire With Fire
The proper reaction to P&M’s reaction to the CEPR study and the left-wing minimum-wage ballyhoo is a polite yawn and a “So what?” This should be followed by a trip to the woodshed and back to the drawing board for P&M, where they would be schooled in proper econometric practice. Alternatively, they can do what true free-market economists have done while their colleagues were practicing pretend-Science: spend the time honing their understanding of concepts like the time-structure of production and capital theory. That will better inform their grasp of reality than the most esoteric econometric model.
Fighting fire with fire can work in specialized cases like oil-well fires. But in today’s debates over economic theory and policy, fighting fire with fire does not extinguish the original fire. It does not even provide intellectual illumination. It merely makes the blaze bigger.