This past week, the Trump administration rolled back an order that the Obama White House put into place to collect information about various disparities in labor markets. Last year, the Obama administration issued a “Fact Sheet” grandly titled “New Steps to Advance Equal Pay,” which contained a proposal “to annually collect summary pay data by gender, race, and ethnicity from businesses with 100 or more employees.” In total, the proposal would cover some 63 million people. The purpose of collecting the data was to provide “better insight into discriminatory pay practices across industries and occupations,” which would in turn give “women additional tools to fight pay discrimination.”
The data do indeed reveal persistent pay differentials between men and women, which the Obama White House attributed largely to illegal discrimination. Under its view, progress in this area requires removing or reducing these pay differentials by coercive action if necessary.
One measure of the size of this supposed problem is that across all industries, white female wages are on average 82 percent of those of white male wages, according to Pew Research Center. Asian women earn about 87 percent of what white men earn, black women 65 percent, and Hispanic women 58 percent. Asian men, however, earn 117 percent of what white men earn. The common, but by no means universal, conclusion is that some large chunk of these differentials is the product of employer discrimination.
The recent decision by the Trump administration to challenge that consensus makes good economic sense given the serious gaps in reasoning and evidence behind the costly and counterproductive Obama initiative. Let’s start with the basic question of whether discrimination on the basis of sex and race can persist in competitive markets. In 1957, the economist Gary Becker first published his book The Economics of Discrimination, which argued correctly that discrimination could survive in markets in which firms possessed monopolistic power, but not in competitive markets where free entry and exit are possible. The logic here is that a monopolist is in a position to engage in discrimination with respect to prices or terms of service because its customers have few if any other options. Hence, from the earliest days, the law imposed duties on common carriers and public utilities to supply services on a fair, reasonable, and nondiscriminatory basis because of their market power.
There are, however, virtually no monopolies in labor markets, so here the possibility of entry and exit makes it difficult, if not impossible, for any firm to discriminate against workers who have alternative options. Today, labor markets are more diverse than ever on the employer as well as employee side. It is virtually impossible to imagine how any major firm could survive by engaging in discriminatory practices. New firms would enter the market to target the highly skilled workers that were overlooked or underpaid by existing firms. Current firms are aware of this risk and must adjust their practices to forestall massive exit. Hence, the observed wage differentials could not survive in competitive markets unless they were cost-justified by productivity differences.
In fact, any firm that paid less productive workers the same wages as more productive workers would be subsidizing inefficiency. As a theoretical matter, it is wholly implausible to think that in transparent and competitive markets, pervasive discrimination could survive. It does not take all firms to transform labor markets. It just takes enough firms at the margin to bid up the wages of underutilized workers.
In an effort to deal with these theoretical points, the defenders of the Obama directives point to empirical evidence that shows the various wage differentials are indeed the product of some unconscious form of discrimination. But here, the efforts to show market imperfections unravel. Virtually all of the studies that find discrimination do so by inference, not by producing direct evidence of discrimination. These studies try to control for some variables—such as education levels and working conditions, which tend to reduce observed wage gaps—but the research is imperfect and the list of variables is incomplete. Some significant portion of those residual gaps is then attributed to discrimination. Unfortunately, this common methodology systematically overestimates the extent and durability of discrimination, and may in fact overlook the distinct possibility that the figures can be explained by innocuous factors such as the preferences of women and minorities in the workforce.
The first objection here is technical. It is very risky to treat the topic of concern as a residual variable teased out from studies that control for only a fraction of the observable variables. As the economist James Heckman wrote in 1998, “there is a staggering gap between the list of productivity characteristics available to economic analysts in standard data sources and what is available to personnel departments of firms.” Think only of the vital information about how people interact in the workplace, their attitudes toward risk, their flexibility, their willingness to do certain kinds of jobs, the hours they are prepared to work, their willingness to travel away from home, and so forth. Adding these variables back into the mix would necessarily reduce the importance of the residual earnings gap and hence would undermine many charges of discrimination. But aggregate industry-level data does not give real information about firm-level activity.
The second objection is that these studies typically ignore all considerations of supply and demand on job selection and wages. Thus on the supply side for labor, does it really take a rigorous empirical study to establish that more women choose jobs that allow them to have greater flexibility to raise their families—jobs that have fewer hours, less travel and less stress? Women are also likely to remain out of the workforce for some period of time when their children are young.
These women make perfectly sensible choices to turn down some high paying jobs that would be inconsistent with their preferred work-life balance. In the short run this choice limits the kinds of jobs that women are prepared to take, and in the long run, absences from the workforce make it more difficult to keep up with the latest developments in a given field or to acquire the practical experience necessary to advance to higher positions. Women who want to raise children face higher opportunity costs by remaining in the work force, and hence tend to reduce hours over time. The sorting effects in markets thus reflect worker priorities. One of the vices of the standard models for discrimination is that they assume that worker preferences are the same for men and women, when in fact they are not. Wages and occupational choices are not solely driven by employers.
In addition, it is all too easy for these studies to ignore differences in demand. Thus it is common for researchers to sloppily compare studies across fields. Bureau of Labor Statistics data show that in the male-dominated profession of information technology, wages are on average 27 percent higher than in human resource management, a female-dominated profession. But it is wrong to conclude that these data offer evidence of discrimination. First, how can these fields involve “similar skills” when their training and job requirements are vastly different? Second, the data do measure differences in average wages between men and women in the same professions. Third, the higher demand for technical skills could easily explain the salary differentials.
This endless quest to find hidden forms of discrimination against women ignores the explicit and persistent discrimination in their favor. Many institutions, public and private, loudly trumpet their insatiable desire for diversity and inclusion, which translates into special preferential programs for recruitment, training, and development of “underrepresented” groups. To be sure, this groundswell ignores the color- and sex-blind provisions of Title VII of the Civil Rights Act of 1964, which, emphasis added, reads: “to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin…”—the provision ends with an explicit prohibition against preferential treatment in favor of any group.
It is therefore a matter of profound irony that Google, which runs massive diversity and inclusion programs to advance female employees, finds itself being sued by the Trump Administration’s Department of Labor for its “systemic compensation disparities against women pretty much across the entire workforce.” Evidently, the rush to find discrimination against women is bipartisan. Let’s hope that pulling back the Obama disclosure order is the first step in terminating the endless procession of misguided gender discrimination suits, Google’s included.