Finding discrimination in the data
The observation of racial disparities is not, by itself, evidence of racial discrimination in any meaningful sense of that term. I have discussed a number of likely causes of the incidence poverty that disproportionately affects black communities in the US.
More sophisticated investigations try to account for the various causes of any observed racial disparity. Consider, for example, a difference in pay. Clearly, a comparison of average earnings without adjusting for factors like educational qualifications, training, and experience are inappropriate. If one wants to isolate the effect of racial discrimination one needs to compare apples to apples. Some studies have attempted to make such adjustments, using available, though imperfect, measures of labor-relevant characteristics, like education, qualifications and experience. And some have found that the adjustments do not reduce the gap between black and white earnings for equally qualified and able workers to zero. This “residual” is then attributed to discrimination.
As another example, a similar study corrected for buyer characteristics other than race, finds a persistent disparites is housing prices paid by blacks and whites, a premium that blacks had to pay for the same/similar houses under similar conditions – and concludes that this must be because of discrimination.
Yet, another example, involves the observation that, despite having applicant pools of equally qualified people, men vastly outnumber women in certain lucrative, prestigious professions, like engineering and computer science. [Notably these observations do not control for the number of offers made and accepted.]
The details of these studies are not important for this discussion. What is important are the following two questions.
1. Is it legitimate or sensible to attribute the “unexplained” disparities to discrimination (prompting the corollary question: what does that mean)?
2. What follows from that; what do these remaining gaps imply for policy?
With regard to question 1 much depends who we imagine has the burden of proof. When we observe a disparity that cannot be explained by differences in the background data are we justified in concluding that discriminatory behavior must have occurred, or will we require that such a conclusion be proven by further evidence of actual discriminatory behavior? This will turn out to be very important when we consider the way the criminal justice system works in deciding accusations of discrimination.
Question 2. Is the more important one. What should be done about racial disparities that can with greater or lesser confidence be attributed to discrimination? To be more specific the possible scenarios include disparities in pay, hiring, firing, promotion, training and working conditions. To explore this we need to examine the economics of discrimination.
immoral, but should it be illegal?
In a free society people will be free to do some immoral things. Clearly freedom has been an indispensable part of the growth of our society and the achievement of prosperity for many. But, many will argue, if left unchecked it will produce gross inequality and injustice. For example, the argument goes, people with power to hire and fire, to pay more or less, to promote or not to promote, who have racial biases should not be free to exercise those biases with impunity. We need laws and regulations to check that behavior. Such is the justification for the extensive regulatory apparatus designed to counter any suspected bias that might lead to labor market discrimination.
The extent of the anti-discrimination, diversity promoting, bureaucracy and regulatory framework is formidable. In economic terms it constitutes a significant proportion of our GDP. Every company above a minimum size must devote considerable resource to diversity monitoring, ensuring that minorities are not being disadvantaged in any way. Besides the various and varied state agencies tasked with advising and monitoring such matters, there also exist powerful federal agencies for the purpose, notably the EEOC (Equal Employment Opportunity Commission) and the OFCCP (Office of Federal Contract Compliance Programs) – which is the watchdog for the implementation of affirmative action, that is preferential treatment for “equally qualified” minority workers and job applicants). This regulatory bureaucracy, besides its operating cost, imposes significant indirect costs on businesses as they try to ensure that no complaint is filed with the EEOC against them or that no deficiency is noted in their affirmative action plan (required by all federal government “contractors” – those who directly or indirectly do business with the federal government, which may be the majority of businesses in the economy. It is hard to overstate the reach of the anti-discrimination, diversity-promoting, micro-managing bureaucracy.
But, what is it all about? Is it simply a matter of trying to counter the consequences of labor market bias? And if so, does this make any sense? Like all regulation, labor market regulation faces serious knowledge and incentive problems. There are always unintended consequences and costs.
We may examine this by using the example of pay. Presumably the objective of the regulations is to ensure that racial bias does not result in lower pay for equally qualified workers. To construct a regulation to ensure this one has to provide a way of deciding who is “equally qualified”. More specifically, when a complaint of pay discrimination is lodged with the EEOC, the agency must decide whether the complainant is equally qualified in comparison to those who are receiving higher pay. Or, more simply, are doing the “same job”. The agency will have to insert in place of the judgment of the employer its own judgement to decide whether the pay gap is justified. But, clearly the agency cannot know as much about the business as the employer. How can the agency know how much of the gap is due to the legitimate judgment of the employer about differences in the productivity of the worker and how much is due to bias?
To complicate matters, should the issue come to a court of law, the procedure followed will be, as described above, to use “experts” to control for productivity factors as much as possible, and to observe if any pay gap still remains. If, after this exercise, a gap remains, it is common in such proceedings that such finding will be regard as sufficient for the plaintiff to have established a prima facia case of discrimination, and then the burden of proof will shift to the employer to “prove” that he has not discriminated. The presumption of innocence is now removed.
Needless to say, since it is very difficult to prove the negative of no discrimination, many of these cases are resolved with costly settlements. Even more likely, employers will work hard and spend significant resources to avoid such lawsuits by ensuring that minorities are paid sufficiently to deter such complaints even if it means paying them more than they are worth in terms of unbiased business judgments. In other words, this kind of regulation, in trying to prevent pay discrimination causes discrimination against non-minorities The same logic applies to hiring in an affirmative action environment where less qualified minority workers are likely to get hired in order to avoid any presumption of discriminatory bias. Affirmative action is actually affirmative discrimination. If discrimination is immoral, one wonders how it can be justified in this context.
A particularly stark manifestation of this is the preferential treatment given to black student applicants to college. Black applicants are in many institutions admitted with test scores a few hundred points below those of their white and Asian counterparts. The result is that many underprepared black students are admitted to programs in which they are likely to fail and to attract attention for doing so. This program is thus doubly counterproductive in, first, failing to benefit the people it is designed to benefit, and, second, in increasing discrimination and race-consciousness by discriminating against whites and Asian students. In addition, it fails to address the fundamental reasons so few black students are unprepared for the colleges to which they apply. By targeting diversity of admissions (notably not of graduations) the focus is on the symptoms not the causes.
All in all, in similar ways, anti-discrimination regulation is a problematic bag of tricks with unpleasant surprises. Which prompts the question, is it even necessary? A brief look at the economics of discrimination suggests it is not, in fact, suggests that such regulation may actually retard the erosion of discrimination.
In any reasonably competitive labor market, competition works to erode discriminatory behavior (and also discriminatory attitudes as a result, because if people behave in a non-discriminatory way they interact with a greater diversity of individuals and often overcome their initial biases and prejudices). No system has ever been more successful in reducing the effects of bias in race, class and other such designations than competitive capitalism.
To understand this, consider again the example of pay discrimination. Imagine an employer who will hire black workers only if they are paid less than whites (or hired into lower paying jobs for which they are overqualified) In a competitive labor market such an employer pays a price, and it could be a heavy price. If blacks have to discount their wages in order to get employed in equally qualified jobs, non-discriminators will have a cost advantage if they hire blacks and have an incentive to do so in large numbers. Over time such cost advantages will translate into higher market shares. In the long run discriminators will have to adapt or go out of business.
The same applies to hiring, promotion, training, work conditions, etc. Those employers who are color-blind will make better use of the resources to economic advantage and outcompete discriminatory behavior. This applies as much to black employers as anyone.
This story is well known. Some find it unconvincing because they think, incorrectly, that it depends on the employer having “perfect information” about the qualifications and productivity potential of black and white workers. If this were the case, then the story would be just a nice theory, applicable to a world where knowledge was perfect and there was no uncertainty. In fact in such a world there would never by any discrimination at all, as long as there were sufficient non-discriminators, though there might be segregated work forces.
In the real world discrimination is a fact of life. The question is what system minimizes discriminatory behavior or minimizes the existence of racial bias. And the answer is a competitive unregulated labor market. To think that regulation, like the current affirmative action regulation, can improve on it is to ignore the very heavy regulatory costs I have enumerated above, and to assume that somehow the regulation results in a better match between qualifications and employment regardless of race. Establishing that a free market would not bring about a perfect outcome is emphatically not to establish the case for regulation, which far from being able to make things better, is very likely to make things worse.
In this four part essay I have tried to show why the current race-loaded narratives are not helpful but are, rather, destructive of understanding, empathy and progress. My recommendation is to talk about how we individuals of all races and creeds can work to improve lives in general, and the lives of the poor in particular, a significant proportion of whom are black.
My decision to write this extended blog was provoked by the precipitous deterioration of the already abysmally low level of our public discourse, particularly around race, inequality and public policy. Such discourse shows less and less evidence of knowledge about the workings of markets, of American history and of the workings of government. In addition, myths and distortions abound in service of an unprecedented radical public policy agenda, one that, sometimes consciously, targets the very foundation of the American free market system and the freedom and prosperity with which it is associated.
I am particularly sensitive to these matters because race and economics is an area of specialization for me as an economist. I was born and grew up in Johannesburg, South Africa. My father was an extraordinarily liberal and tolerant man and we experienced considerable dissonance living in the apartheid police state. But, though he was a very passive man in general, he was easily angered by any manifestation of racial bias or bigotry. I grew up with an appreciation of tolerance and of individual rights, in sprite of seeing manifestations of the very opposite everywhere around me.
Upon leaving South Africa, I did my doctoral dissertation on the Economics of Apartheid, under the supervision of the famous economist, Nobel laureate, Gary Becker at the University of Chicago. Becker pioneered the field of the economics of discrimination. But he did not consider situations in which the government, supporting a white minority, imposed racially discriminatory laws. This was the case in South Africa under apartheid and in the Jim Crow south in the US.
I realized that in such situations, governments serve to prevent, or retard, the process by which free markets erode discrimination. This leads to the prediction that, when such discriminatory laws are removed, allowing freedom of choice, including biased choice, discrimination and bias will gradually, if not immediately, disappear. This is exactly what has happened in South Africa and in the American south. The current discourse evidences no understanding of this or, indeed, of the negative role that government must play if it enters into the realm of what should be private decision making.
 It should be obvious that, in a perverse way, labor market regulation gets in the way of this erosion process, by focusing on race and encouraging non-optimal use of resources.
 For the sake of brevity I have omitted consideration of the egregious violations of individual liberty that these regulations entail.
 Supporting such regulation, since it is coercive in nature, should carry the burden of proving that it does.