Thomas Sowell
Thomas Sowell; drawing by David Levine

The 1980 elections marked the end of an era in American race relations. Between 1964 and 1980 federal officials had argued about the moral legitimacy and practical benefits of particular strategies for helping blacks catch up with whites economically, but none had questioned the principle that the government should actively promote this goal. This consensus about long-term goals had led to three kinds of federal activity. First, a lot of money was spent on education and job training, and spent in such a way as to give blacks a substantial fraction of the benefits. Second, the federal government used its power to get private employers to assign blacks to jobs traditionally reserved for whites. Third, the government increased both cash and other benefits for families that were unable to take advantage of the new job opportunities, especially families without a male breadwinner.

Some parts of this implicit social contract between blacks and whites were unpopular from the very start. Even in the early 1960s many whites felt that Aid to Families with Dependent Children (AFDC), which was the major source of income for many families without a male breadwinner, encouraged both promiscuity and idleness. As more mothers entered the labor force over the next two decades, the idea that the government should pay women to stay home and take care of their children became less and less popular. The fact that an increasing percentage of these women were black did nothing to enhance the program’s popularity.

Other elements of the implicit social contract were initially popular but lost support as time went on. Early federal pressure on private employers to eliminate discrimination against blacks was widely accepted. Later, pressure to discriminate against whites in order to offset the effects of past discrimination against blacks was extremely unpopular. Education and job training also started out with broad support, but much of this support was predicated on unrealistic expectations about what such efforts could achieve, especially in slack labor markets. As this became obvious, enthusiasm waned.

The Reagan administration has abandoned the notion that the federal government should try to make black incomes equal to white incomes and has won congressional support for cutting back all the major federal programs designed to accomplish this goal. Spending on education and job training has been cut, along with many other government services. Since government has been a major source of new middle-class jobs for blacks over the past two decades, these cutbacks have reduced job opportunities for blacks more than for whites. Federal pressure on private employers to hire more blacks has also dropped sharply. Eligibility for AFDC, food stamps, and Medicaid has been narrowed, making life much harder for families without breadwinners. Perhaps most important of all, the goal of full employment, which has always been crucial for narrowing the distance between “haves” and “have-nots” in America, has been assigned its lowest priority since 1932.

Even if the Democrats regain control of the White House in 1984, they will not have a mandate to revive affirmative action, expand government services, or raise AFDC benefits high enough to lift recipients out of poverty. Some Democratic presidential aspirants still use traditional liberal rhetoric to rally their supporters, but none is likely to risk his political career by pursuing the policies this rhetoric used to imply. The implicit social contract between blacks and whites that prevailed from 1964 to 1980 is therefore dead. The question today is whether anything will take its place.

Thomas Sowell, a black economist trained by Milton Friedman and his colleagues at the University of Chicago, has been campaigning against special treatment for blacks since the early 1970s. In Ethnic America and Markets and Minorities he tries to lay the intellectual foundations for a more general argument that ethnic minorities do better in laissez-faire economic systems than in systems subject to government regulation. Ethnic America describes how the Irish, Germans, Jews, Italians, Chinese, Japanese, blacks, Puerto Ricans, and Mexicans came to America, how they dealt with the discrimination they all encountered, and how they progressed economically despite the absence of government help and sometimes in the face of government opposition. Markets and Minorities is a textbook for undergraduate courses on minorities; it uses conventional economic logic first to analyze the effects of discrimination and then to analyze the effects of government programs aimed at eliminating either discrimination or its presumed consequences, such as low wages, unemployment, and squalid housing.

Both books are, in effect, briefs for Sowell’s view that governments should pursue colorblind policies and let different ethnic groups look out for themselves. Both make three general claims in support of this view.

  1. While racial and ethnic discrimination has been common throughout American history, the victims of such discrimination are often more affluent today than their former oppressors.
  2. In a competitive economy, discrimination against racial and ethnic minorities is expensive to those who engage in it. Discrimination therefore tends to disappear once governments stop enforcing it. By 1969, Sowell claims, American employers were paying their black and Hispanic employees the same wages they paid whites with comparable skills.
  3. Government efforts to eliminate both private discrimination and its presumed consequences do more harm than good, mainly serving the interests of white liberals and middle-class blacks, not the interests of the poor.

These three claims directly contradict conventional liberal assumptions about discrimination, namely:

  1. Racial discrimination has been the most important—perhaps the only—cause of economic inequality between blacks and whites in America.
  2. The costs of discrimination fall largely on the victims. As a result, nobody but the victims gains much from eliminating discrimination. In the absence of government action, discrimination may therefore persist more or less indefinitely.


  3. Government efforts to eliminate private discrimination have had a major part in reducing economic inequality between blacks and whites since 1964. Further government effort will be required to achieve full equality, and perhaps even to maintain the gains made since 1964.

The truth is more complex than either of these approaches suggests. The effects of discrimination, for example, depend on its pervasiveness. Sporadic discrimination of the kind many European immigrants encountered in America does not seem to have harmed its victims and may even have helped them in the long run. Universal discrimination over many generations, which is what blacks have encountered in America, has had disastrous economic effects, many of which show no signs of disappearing even today, when many traditional forms of discrimination have been eliminated.

Eliminating discrimination is also more complicated than Sowell assumes. Some kinds of discrimination lower profits, so once a few firms abandon them, they disappear everywhere. Other kinds of discrimination raise profits and therefore persist unless the government stamps them out. As a result, government efforts to eliminate discrimination have played an important role in narrowing the wage gap between blacks and whites since 1964.

Nonetheless it would be a mistake to dismiss Sowell as a polemicist who plays fast and loose with the truth. He is a polemicist, and he does make many factual mistakes. Of the half dozen factual claims in these books that I checked, all but one were seriously misleading. Yet in some cases the actual facts supported Sowell’s theoretical arguments as strongly as the facts he cited, and in other cases they justified his rejection of traditional liberal views, even though they did not support his preferred alternative.

The remainder of this article looks more carefully at the nature and consequences of discrimination. A subsequent article will assess Sowell’s argument that federal efforts to eliminate discrimination are counterproductive.


It is hard to argue with Sowell’s claim that many ethnic minorities overcame discrimination and achieved extraordinary influence in America. Both Ethnic America and Markets and Minorities begin with a table showing the relative income of families from different American ethnic groups. The table covers only the largest groups of European origin, fails to distinguish affluent Irish Catholics from relatively impoverished Irish Protestants, and relies on a survey that underrepresented poor Jews. But correcting these errors strengthens Sowell’s argument instead of weakening it. My Table 1, for example, uses data collected from 1972 through 1980 by the National Opinion Research Center’s General Social Survey (GSS) to estimate the household income of people who said their ancestors came to America from Europe—people who, for convenience, I will simply call “Europeans.”


The table includes all sizable white ethnic groups except Hispanics, most of whom came to America via Latin America and most of whom are much poorer than those who came to America directly from Europe. The table supports Sowell’s claim that the victims of discrimination are often more affluent today than their former oppressors.

While Jews do not look as prosperous in my table as in Sowell’s, they are still far better off than any other American ethnic group. Irish Catholics are second. Americans of British origin—the “WASPs” who were once said to run the country—have only 1 percent more income than the average American of European origin. Northern Europeans are for the most part worse off than Southern or Eastern Europeans. Contrary to what one might suspect, these differences persist even when one looks exclusively at families living in the urban North.

There is plenty of room for controversy about which European groups encountered the most discrimination in America, but few would argue that the Dutch or the Scandinavians had a harder time than the Jews; that the country was run by a French establishment that kept immigrants of British origin in their place; or that Irish Protestants encountered more discrimination than Irish Catholics. Indeed, Irish Protestants (often known as “Scotch Irish” to distinguish them from Irish Catholics) blended so easily with their British cousins that few people are even aware that they constitute one of the largest ethnic groups in America—larger, for example, than Irish Catholics. Because Irish Protestants never established their own churches, political machines, or voluntary associations, they never entered American consciousness the way Irish Catholics did. The absence of organizations for mutual aid may, in turn, help explain why Irish Protestants are now worse off economically than Irish Catholics, though many other factors are also involved.

Far from suggesting that discrimination kept its European victims impoverished, Table 1 suggests that it spurred them to greater success. This is certainly consistent with Sowell’s argument. But Table 1 does not offer much support for another of Sowell’s recurrent arguments, namely that an ethnic group’s success in America depended to a great extent on the values, skills, and traditions it brought from the “old country.”


If cultural legacies were of critical importance, one would expect groups that prospered in Europe to have done the same in America. This may be true of the Jews, who may have been more affluent than their gentile neighbors in Europe as well as in America. But among gentiles the European economic order seems to have been almost completely reversed in America. Germans and Scandinavians are richer than Italians and Czechs in Europe, but poorer in America. Irish Protestants are richer than Irish Catholics in Ulster, but poorer in America.

Indeed, the whole tradition of Protestant affluence and Catholic poverty that inspired Max Weber’s reflections on the economic impact of religious ideas has been stood on its head here. Catholics from virtually every European country are today better off in America than Protestants from the same country, though the differences are seldom as large as those shown in Table 1 for the Irish. None of this means that Sowell is wrong when he suggests that a group’s initial economic position was heavily influenced by its values and traditions. But ethnic traditions that encouraged affluence in Europe did not always have the same effect in America, and in many cases a group’s skills and values changed rapidly once it arrived in the New World.

In any event, the fact that many European ethnic groups overcame discrimination and prospered in America has little direct bearing on the situation of non-Europeans. As we shall see, discrimination only poses a serious barrier to economic success if it is almost universal, or if its existence transforms the victims’ behavior in self-defeating ways. Discrimination against European minorities was common but never anything like universal. The descendants of European immigrants have almost all had the option of shedding their ethnic identity and “passing” as just plain “Americans.” Physical differences, combined with extreme social sensitivity to the significance of these differences, made this much more difficult for most blacks and somewhat more difficult for Asians, American Indians, and even Latin Americans.1 Perhaps as a result, discrimination seems to have had more effect on black culture and behavior than it had on the culture and behavior of European minorities. It has also had more effect on black incomes than on European incomes.

Table 2 gives Sowell’s estimates of relative income for non-European families in America.


The Japanese are doing much better than the average European, the Chinese somewhat better, and the Filipinos only marginally worse. Hispanics in general are doing worse than Europeans, and Puerto Ricans are doing much worse. American Indians and blacks are also doing much worse than Europeans, though West Indian blacks are only a little worse off than Europeans.2

How are we to explain the persistent poverty of American blacks and Hispanics? Sowell does not address this question directly. Instead, he obfuscates it.

Markets and Minorities suggests, for example, that “ethnic groups differ greatly in median age, and age in turn influences income.” Sowell reports that American blacks, Indians, and Hispanics have median ages between eighteen and twenty-two, while European groups have median ages between thirty-six and forty-six. He then reports that in 1974 families whose heads were between forty-five and fifty-four had incomes 93 percent higher than families whose heads were under twenty-five, leaving the reader with the impression that if the median black, Indian, or Hispanic were as old as the median European, he would be almost as affluent.

This is nonsense. The reason for the age differential between rich and poor ethnic groups is largely that poor groups have more children and die younger. These facts have almost no impact on family incomes, which depend on the age distribution of family heads. Such distributions vary little from one ethnic group to another and, contrary to Sowell, account for almost none of the income variation between groups. For all practical purposes, therefore, age is irrelevant to ethnic income differences.

Sowell also suggests that part of the cause of ethnic differences in family income is that different ethnic groups live in different parts of the country. What he fails to mention is that most non-European groups live in relatively rich areas, not poor ones. The Japanese mostly live in California, for example, and the West Indians mostly live in New York City. Incomes in both places averaged about 15 percent higher than those in the nation as a whole during the year covered by Sowell’s data (1969). Comparing the Japanese or West Indians to their neighbors would thus have reduced their apparent affluence, weakening Sowell’s claim that nonwhites with the right traditions and values have been able to overcome the effects of discrimination.

Other Asians and Hispanics are also concentrated in big cities where both incomes and living costs are above the national average, so Table 2 overstates their relative affluence. American blacks are also disproportionately urban, but because they tend to live in southern rather than northern cities, their white neighbors’ incomes are just about at the national average. Taking account of blacks’ geographic location would therefore leave the picture in Table 2 almost unchanged. It is true that many of today’s northern blacks grew up in the rural South, or had parents who grew up there, and it often takes migrants several generations to assimilate the attitudes and skills needed for success in an urban economy. Northern blacks who grew up in the rural South, for example, typically learned less while they were in school and quit school younger than those who grew up in the urban North. Yet even those blacks who were born in the North are far poorer than second-generation European immigrants. Like age, therefore, geography explains only a tiny fraction of the overall difference between black and white family incomes. Geography obscures the problems of Hispanics and some Asians.

While Sowell exaggerates the importance of age and geography, he is still right that a lot of factors besides discrimination contribute to ethnic differences in family income. Three are especially important. First, what constitutes a “family” differs dramatically from one ethnic group to another. Second, different ethnic groups have different attitudes toward work, especially toward women working for pay, and these differences have sizable effects on family income. Third, education and other measures of skill vary from group to group, so that even if employers paid no attention whatever to ethnicity, different groups would have different wage distributions.

The Census Bureau defines a family as any group of two or more related persons living together. If related persons do not live together they do not constitute a family. Living arrangements therefore have a major effect on Census statistics regarding family income. Asians, for example, are more likely than Europeans to live in extended families, which means that Asian families typically have more potential earners. Since Asian women are also more likely than European women to work for pay, Asian family incomes are high even when their wages are low. Blacks, in contrast, are more likely than Europeans to live in single-parent families composed of a mother and her children. Such families usually have only one potential earner. As a result, black family incomes are far lower than one would expect if one simply contrasted black and European hourly wages.

While the psychological effects of growing up in households headed by females have often been exaggerated, the economic effects are clearly disastrous. Such households not only have fewer potential earners than the average family, but these potential earners are women, and women’s wages are still less than three-fifths those of men. In theory, families with children and an absent father should also receive child-support payments from the absent father, but in practice only 43 percent of white mothers and 14 percent of black mothers in this situation receive anything whatever from the father; and the payments that are made average only $150 a month among whites and $100 a month among blacks.3 Such families are also eligible for AFDC benefits, but these benefits are so low in most states that recipients end up below the poverty line, and benefits are sharply reduced if the mother tries to supplement them by working. The result is that families headed by females typically have incomes about half those of married couples. The ratio is even less favorable for families that are headed by women with young children.

The proportion of black families headed by women rose dramatically during the 1970s, reaching 40 percent in 1980, compared to only 12 percent among whites. Had black parents been as likely as white parents to live together, the improvement in black earnings shown in Table 3 would have led to a substantial improvement in black family incomes.


But because fewer and fewer black parents were married, black family incomes rose less rapidly than those of white families during the 1970s. Indeed, after adjustment for inflation black family incomes actually fell during the 1970s. The increase in families headed by women had a particularly unfortunate effect on black children’s standard of living. By 1979 black adults had incomes that averaged 68 percent of white adults’ incomes. Black children typically lived in families with incomes only 55 percent of those of white children’s families. The discrepancy between black adults and black children arose largely because 43 percent of black children got no support from their fathers, compared to only 8 percent of white children.4

A second major reason why non-Europeans’ family incomes deviate from the national average is that their participation in the labor force deviates. Asian men and women of all ages are more likely to have paid jobs than Europeans in similar family situations. This inevitably boosts Asians’ family income even when their wages are low. Historically, black women have also been more likely to hold paid jobs than their white counterparts, but this difference has almost disappeared as more and more white women take jobs. Black men, in contrast, have been less likely to hold jobs than their white counterparts since World War II, and this gap is widening. By 1980, 15 percent of all black men between the ages of twenty-five and sixty-four were telling the Census Bureau that they had earned nothing whatever during 1979. Among whites and Hispanics this figure was only 7 percent. A few of these men said they were looking for work; but most said they were not.

The high percentage of black men who are neither working nor looking for work seems to reflect three facts. First, black men have more trouble finding work than white men do. Second, black men earn less when they do find work. Third, black men are much more likely than white men to suffer from chronic illnesses. Because their earnings are low, blacks with chronic illnesses have more incentive than whites to quit working altogether and apply for federal disability benefits.5

Because ethnic groups have such different living arrangements and laborforce participation patterns, it is almost impossible to say anything sensible about the effect of discrimination on family incomes. If we want to assess the economic impact of discrimination, we must look at individuals, not families. More particularly, if we want to assess the impact of labor market discrimination, which is Sowell’s primary concern, we must compare the earnings of workers with similar skills and similar work habits. When we do this, the picture that emerges is very different from the one that emerges when we compare family incomes. The reader can verify this by comparing the data on individuals in Tables 2 and 3 to the data on families. It is to labor-market discrimination of this kind that I now turn.


Perhaps the most astonishing single assertion Sowell makes is his claim in Markets and Minorities that “the data do not show current employer discrimination in pay among black or Hispanic male, full-time workers….” Sowell cites an article by Eric Hanushek to support this claim, but Hanushek’s article does not even address the question of employer discrimination in pay. What Hanushek shows is that after making a somewhat tenuous adjustment for variation in what students learn in a given year of school, an extra year of elementary or secondary schooling increased both black and white men’s earnings by about 5 percent in 1969, while an extra year of higher education increased both groups’ earnings by about 9 percent. But this does not mean that blacks and whites with comparable skills end up with comparable earnings, as Sowell implies. Since blacks with very little schooling earn less than whites with equally little schooling, increasing both groups’ earnings by the same percentage each time they get an extra year of schooling ensures that blacks will earn less than whites at every educational level. How Sowell managed to miss this point I do not know.

Table 2 summarizes Sowell’s own estimates of the 1969 incomes of non-European college graduates. Looking first at men, which almost everyone concerned with ethnic inequality habitually does, we see that blacks with BAs earned only 61 percent of what whites with BAs earned in 1969. The ratio was no more favorable for West Indians or for American Indians. It was somewhat more favorable for Asians and Hispanics, but did not approach parity even for the Japanese. These same patterns hold for men who did not attend college, and also for those who attended graduate school.

The situation today is only a little different from that in 1969. Table 4 shows that the earnings of black males with BAs rose much faster than those of comparable whites during the 1970s.


By 1979 young black men with BAs were earning 84 percent of what their white counterparts earned, whereas young black high-school graduates were still earning only 74 percent of what whites earned. Attending college thus raised a black man’s earnings more than it raised a white man’s earnings—not because black BAs earned more than their white counterparts but because black high-school graduates earned so much less.

Graduate education is also worth more to black BAs than it is to white BAs. The reason is that blacks with graduate degrees now earn about the same amount as their white counterparts, whereas black BAs still earn somewhat less. The Association of MBA Executives reports, for example, that Masters of Business Administration hired in 1980 typically started out with salaries of $24,259. Members of minority groups with MBAs started out at $24,145—a difference of less than 1 percent. (Executive recruiters say that demand for black executives has decreased since the Reagan administration took office, but salary statistics for 1981 and 1982 are not available.) Sowell reports that black PhDs earned as much as whites in the same field during the 1970s. Fragmentary evidence suggests that the same is probably true among recent law-school and medical-school graduates.

Among men without graduate degrees, however, blacks still earn less than whites. Sowell makes no systematic effort to explain this fact, though others have tried to do so. One explanation that appeals to scholars is that blacks learn less in school about the subjects that schools try to teach. While this gap narrowed during the 1970s, blacks at all educational levels still do worse than whites on virtually every known achievement test. Yet for better or worse, mastery of what schools try to teach has only a modest relationship to subsequent earnings for either blacks or whites. As a result, differences in test performance explain only a small part of the earnings gap between blacks and whites with the same amount of schooling.

Evidence of this kind has traditionally persuaded liberal and radical scholars (including myself) that employers paid blacks less than equally competent whites. Conservatives like Sowell have rightly argued that such evidence is not conclusive. Employers are as likely to complain about their black employees’ work habits and motivation as about their technical competence. If these complaints were well founded, blacks would earn less than whites with comparable skills even in a colorblind world.

Measuring work habits and motivation is almost impossible. I have been struck in my own research, for example, by the fact that blacks say they are less satisfied with their jobs than whites in comparable occupations who make the same amount of money. This remains true even when one takes account of differences in fringe benefits, job security, hours, unionization, and the like.6 If dissatisfaction is linked to job performance, as countless organization theorists claim, the fact that blacks are dissatisfied could mean that they perform poorly. If so, even unprejudiced employers would end up paying blacks less than whites with similar skills and credentials. This would remain true even if, as one can readily imagine, blacks had good reasons for being dissatisfied. If, through no fault of their own, blacks had worse relations with their supervisors or fellow workers, and if this led to poor performance, it would make economic sense for employers to pay blacks less than whites with similar credentials and skills—unless employers could somehow make white supervisors and workers treat blacks better without incurring significant costs.

Still, one must be somewhat skeptical about such explanations. Asian men, whom Sowell describes as model workers, also earn less than Europeans with the same amount of schooling. This is not because they attend worse schools, score lower on standardized tests, or live in the wrong parts of the country. If it isn’t because of discrimination, what is the explanation?

But if discrimination is really crucial, as liberal doctrine claims it is, how are we to explain the other striking fact in Table 2, that black women with college degrees earn more than their white counterparts? This pattern is not confined to college graduates. Black women at all educational levels earn about as much as white women with the same amount of formal schooling. They do this despite having lower test scores and higher levels of dissatisfaction. One might argue that black women have benefited more than black men from affirmative action, since employers know that hiring black women allows them to meet two quotas instead of just one. But as Table 2 shows, black women were making as much as white women with comparable schooling even in 1969, before affirmative action programs paid much attention to women. Furthermore, since white women’s wages rose more slowly than white men’s during the 1970s, it is hard to believe that affirmative action with respect to women ever had much effect.

Another tempting explanation for the fact that black women have caught up with their white counterparts, while black men have not, is that employers have eliminated discrimination in most middle-level jobs but not in most top jobs. Yet if this were the case black male high-school graduates should also have caught up with their white counterparts, while black male college graduates should have lagged further behind. As Table 4 shows, the opposite has happened. While young black BAs are still having trouble landing top jobs, young black high-school graduates are having even more trouble getting the steady blue-collar jobs that are the economic mainstay of white high-school graduates.

The fact that black-white wage differences are greatest among poorly educated black males suggests that employers may be reacting more to ghetto culture than to skin color per se. Poorly educated black men behave in all sorts of ways that employers dislike. Black men, for example, are five to ten times more likely than whites to be arrested and convicted of serious crimes. This difference is partly due to the fact that black men are less well educated than whites and more likely to be between the ages of fifteen and thirty, which is the “criminal age.” Police, prosecutors, and judges may also be harder on blacks than on whites. Nevertheless, a prudent employer would certainly have to assume that young black high-school graduates are more likely to end up in trouble with the police than young white high-school graduates.7

Black men are also more likely to father illegitimate children whom they don’t support, and more likely to abandon their wives and children after they marry. This is of no direct concern to employers, but if young black men approach their work in the same way that they approach contraception, or when that fails, parenthood, employers would have good reason to avoid hiring them for responsible jobs.

The puzzle is why so many young black men still act in ways that not only make employers reluctant to hire them but make many employers want to fire those they have hired. The rewards for “going straight” are now almost as high among blacks as among whites, and the costs of refusing to do so remain far higher among blacks. In light of the physical, economic, and psychological price America exacts from black men who refuse to conform to employers’ expectations, one might expect virtually all young blacks to do whatever they had to do in order to get and keep a steady job. But this has not yet happened. Sowell argues for patience, pointing out that it took European immigrants several generations to adapt to the new conditions they found in American cities, and that we must expect American blacks to take at least as long. He may be right. Some economic indicators certainly show the kind of progress that Sowell’s theory of assimilation would lead us to expect. But despite the improvement in economic opportunities for black women and college-educated black men, statistics on youth unemployment, crime, living arrangements, and child support suggest that for many blacks the situation has gotten worse in recent years, not better.


The possibility that employers may have sound economic reasons for discriminating against black workers suggests that we should distinguish between varieties of discrimination. By law, the term subsumes almost all instances in which an employer considers a worker’s race when making decisions about hiring, promotion, or pay, regardless of the employer’s motive. As I will argue in the second of these articles, any other definition is fraught with dangers. But if we want to analyze the causes and consequences of discrimination, employers’ motives for engaging in it are critically important. Sowell makes some helpful distinctions among motives, but his analysis is too simple. Employers have at least four possible reasons for discriminating: malice, myopia, statistics, and consumer preferences.

“Malicious discrimination” involves conscious hostility toward the ethnic group of which the victim is a member, though not necessarily toward the victim himself. If an employer refuses to promote a black worker he knows is better qualified than any available white, on the grounds that putting blacks in positions of power over whites will undermine white supremacy in society as a whole, that is malicious discrimination. Sowell calls this “pure” discrimination.

“Myopic discrimination” is often rooted in malice, but its immediate motive is an erroneous ethnic stereotype that convinces an employer it is in his interest to hire or promote members of one ethnic group rather than another. An example is refusing to promote qualified blacks because you consistently underestimate their performance in their present job. Sowell calls this “perceptual” discrimination.

“Statistical discrimination” is a term invented by Lester Thurow. It differs from myopic discrimination in that it is based on accurate rather than inaccurate ethnic stereotypes. Refusing to promote black workers because those promoted in the past have performed worse than whites with the same qualifications is statistical discrimination. Sowell treats this as a special case of “perceptual” discrimination.

“Consumer-directed discrimination” is akin to malicious discrimination in that the employer recognizes that the victim could perform a given job satisfactorily. In this case, however, the employer refuses to hire the victim not because of his own malice but because of malice or myopia among his customers. Owners of major-league baseball teams refused to hire blacks for many years, for example, because they felt that the fans preferred an all-white sport. If we assume the owners were right, the fans were engaged in malicious discrimination. In other cases employers suspect their customers of myopic rather than malicious discrimination. If a significant fraction of the public still believes that blacks cannot master complex technical skills, for example, people may avoid airlines with black pilots. Airlines may then avoid hiring black pilots, even if the airline itself is completely satisfied with their competence. Sowell ignores this kind of discrimination.

Both malicious and myopic discrimination raise firms’ costs, since they cause firms to hire inferior workers without reducing their wage bill. If all competing firms engage in the same amount of discrimination, they can pass on these costs to their customers through higher prices or inferior products. But if some employers stop discriminating, competition will ensure that the public no longer has to bear the cost of discrimination. Instead, employers who continue to discriminate will have to absorb the cost.

So long as no major-league baseball team hired blacks, for example, the costs of discrimination fell on promising black players, who couldn’t get work, and on fans, who saw worse games than they would have seen in the absence of discrimination. But once a few clubs hired blacks, those that refused to do so had a harder time assembling winning teams and hence attracting fans. Owners who persisted in hiring only white players therefore had to pay for their prejudice through lost revenue. Black players no longer had to pay, since if even a few clubs were prepared to hire them on the same basis as whites that was sufficient to ensure them competitive salaries.

As this example indicates, competition tends to eliminate both malicious and myopic discrimination. These forms of discrimination persist only when competing firms can collude with one another to ensure that they all discriminate equally, and hence that their customers will pay the costs. Such collusion usually requires sanctions against employers who refuse to discriminate. Without sanctions there will sooner or later be an entrepreneur who puts short-term profits ahead of noneconomic principles like white supremacy, or who figures out that he can improve the quality of his labor force by ignoring some traditional ethnic stereotype.

Unlike malicious and myopic discrimination, statistical or consumer-directed discrimination serves an employer’s immediate economic interest in a competitive market. Indeed, they may be essential to his survival in a competitive market. Suppose, for example, that a large firm has found that its black cashiers are slightly more likely than its white cashiers to be caught with their hands in the till. If the firm knows that there is a lot of undetected theft, and if the amounts of money involved are substantial, it may conclude that it would be better off hiring only white cashiers.

Whether such statistical discrimination makes economic sense depends on the alternatives. If more careful screening of applicants’ previous employment records could produce an applicant pool in which blacks were no more likely to steal than whites, this would clearly make more sense than using skin color to predict probity. Likewise, if a new accounting system could ensure that thefts were detected more easily, so that the amounts lost were small, the firm would probably be better off hiring on a colorblind basis and then firing cashiers who stole. But in situations where there are genuine statistical differences between ethnic groups, where these differences persist even after employers have screened applicants by using all other demonstrably relative criteria, and where the cost of hiring the wrong person is high, failure to engage in statistical discrimination can cost an employer a lot of money.

Confronted with arguments of this sort, liberals usually challenge the factual premise that blacks make less satisfactory employees than whites. Statistics can lie. In the case of cashiers, for example, blacks might get caught stealing more often than whites not because they stole more but because white (or even black) supervisors watched blacks more carefully than whites. In a society pervaded by racist stereotypes, skepticism about alleged racial differences certainly makes sense. But to assume that careful inquiry will always prove that blacks are indistinguishable from whites is folly. Two centuries of slavery and a century of Jim Crow have left scars on many blacks that will take a long time to eradicate, especially if statistical discrimination remains widespread.

It is important to remember that statistical discrimination can be based on considerations that have nothing to do with the skills or work habits employees bring to their work. Economists like Sowell habitually talk about work as if it were an isolated activity, in which performance depended only on the amount of “human capital” workers had accumulated. But most work is done by groups, not isolated individuals, so most workers’ value to their employer depends partly on how well they get on with other members of their group. If most of the workers in a group are white, and if some of these whites have trouble getting along with blacks, black workers will be less valuable as a result. Blaming this on blacks rather than on whites is obviously unjust. But from the employer’s viewpoint justice is irrelevant. He needs a work force capable of collaborative effort. If this means that all workers must be the same color, or that only blacks with an unusual talent for getting along with whites can fit in, he must live with that fact or lose money. The same is true of consumer-directed discrimination.

While competitive pressure encourages employers to engage in both statistical and consumer-directed discrimination, regardless of their own preferences, these forms of discrimination need not have a significant effect on the typical black worker’s earnings. Discrimination only has a major impact on a group’s earnings if it is almost universal. If discrimination affects only a limited number of jobs, its effects are largely psychological, not economic.

Consider the case of a Jew who graduated with a good record from a top law school in the 1930s. If he sought a job with a leading New York firm he soon discovered—if he did not already know—that most of these firms hired only gentiles. But he also discovered some exceptions. The likely result of such a lawyer’s job search was therefore twofold. First, he would have concluded that there was a great deal of discrimination against Jews and would probably have become a supporter of both “fair employment” legislation and the Anti-Defamation League. Second, he would have gotten a job that allowed him to contribute generously to these causes. If a social scientist had surveyed his law-school class thirty years later and had compared the earnings of Jews and gentiles, he would probably have found no evidence that Jews suffered economically from discrimination. This would not have meant that discrimination did not exist.

If our hypothetical lawyer had been black rather than Jewish, the story would have been quite different. No leading New York firm hired blacks in the 1930s. Even liberal firms assumed that their clients would never accept a black attorney. Blacks therefore looked elsewhere for work, sometimes entering government, sometimes joining a civil rights organization, sometimes establishing their own practices dealing with black clients. As a result, even a leading black lawyer was likely to earn far less than his white classmates.

It seems to me that young, highly educated black men today confront a situation more like the one Jews confronted fifty years ago than like the one blacks confronted fifty years ago. As Sowell points out, studies of specialized fields usually suggest that black men now earn as much as whites with similar records. It is true that social scientists typically initiate such studies in fields where competence is easy for outsiders to evaluate using some relatively objective criterion, like the number of times a scholar’s publications have been cited. These are also the fields in which ethnic prejudice is least likely to affect an employer’s evaluation of a worker’s performance. Among corporate managers, whose performance is much harder for both insiders and outsiders to evaluate, discrimination may have more serious economic effects than among, say, engineers.

Yet even if discrimination remains widespread among corporate managers, it may not have much impact on black earnings once blacks figure out who discriminates and who doesn’t. If, for example, even one firm in four were a true “equal opportunity employer,” there would be enough jobs in these firms for all the available black managers, since blacks constitute only a tenth of the labor force. Black managers could then earn as much as equally competent whites simply by moving to colorblind firms. Discrimination in other firms would still be a moral and political problem, but it would have little economic impact. Of course, if only one firm in fifty dealt with managers on a colorblind basis, only a minority of black managers would be able to find jobs in such firms, and others would get paid less than whites with comparable qualifications. But if this became obvious, black undergraduates would presumably turn away from careers in corporate management, entering professional and technical fields where their performance would be evaluated more even-handedly.

Table 4 suggests that most young black men with college degrees are usually able to circumvent the economic effects of discrimination. Census data like that in Tables 2 and 3 also suggest that black women at all educational levels have been able to circumvent the economic consequences of racial discrimination, although they have been no more successful than white women in circumventing discrimination based on sex. But as Table 4 indicates, black men who have not attended college still lag far behind their white counterparts. It is hard to believe that customer prejudice accounts for this, since most lucrative jobs open to white high-school graduates involve very little contact with the public. It is also hard to believe that firms that pay black women as much as white women, and do the same for black male MBAs, pay black male high-school graduates less than their white classmates because of a malicious commitment to maintaining white supremacy.

It is equally hard to believe that firms that have been under continuing legal pressure to hire and promote more blacks, and that have succumbed to such pressure when hiring and promoting both black women and highly educated black men, have not even experimented with moving less well-educated black men into the kinds of well-paid jobs that comparable white men often hold. In light of the extraordinary progress of both black women and highly educated black men over the past decade, I think we must assume that firms are telling the truth when they say they have trouble finding black male high-school graduates who perform well in the skilled, responsible, fairly well-paid jobs traditionally reserved for white male high-school graduates.

This painful conclusion does not imply that discrimination has no impact on poorly educated black men’s earnings. My guess is that it has an important impact. Assume for the moment that black male high-school graduates are, in fact, somewhat less likely to perform well in the kinds of jobs that have allowed many white high-school graduates to join the middle class. If this were the case, three consequences would be likely to follow.

First, firms would avoid hiring or promoting black males who had not attended college whenever they thought mistakes in hiring were difficult to correct.

Second, having committed themselves to this sort of statistical discrimination, managers would often exaggerate the real differences between black and white male workers. As a result, many managers would avoid hiring blacks even when a rational economic analysis would lead them to do so, namely when the black-white difference was small and the cost of firing second-rate workers was relatively low.

Third, having convinced themselves that poorly educated black workers spelled trouble, plant managers and foremen would be more likely to detect and punish such workers’ mistakes. This would nourish the anger and resentment many black workers already feel toward their white employers, reducing the chances that they would do a good job. A modest initial difference in performance would thus create a heady brew of statistical and myopic discrimination that perpetuated the initial difference.

Such a conclusion obviously contradicts the rhetoric of many liberals and black activists, whose position—at least in public—is that wage differences between blacks and whites with comparable schooling have no relationship to performance and derive entirely from malicious, myopic, consumer-directed discrimination. My hesitant conclusion is equally at odds with Sowell’s conservative views, which imply that wage differences are caused by differences in job performance and that current discrimination in the labor market has no effect on black earnings.

Sowell is certainly right to emphasize both the ways in which minorities can circumvent the effects of discrimination and the possibility that wage differences between blacks and whites with comparable educational credentials need not be due to discrimination. But by almost ignoring the crucial distinction between universal and haphazard discrimination, by arguing that discrimination tends to disappear once markets become competitive, and by neglecting the impact of long-term discrimination on its victims’ behavior, Sowell seriously underestimates discrimination’s potential economic consequences. As a result, he sees few benefits from making discrimination illegal. All he can see are the costs. In the next issue I will discuss his criticism of efforts by government to eliminate both discrimination and its consequences and will try to show that here, too, Sowell is no more interested in how the world really works than are the liberals he criticizes.

(This is the first of two articles.)

This Issue

March 3, 1983