| 摘要 | The most striking feature of U.S. racial relations today is how much we spend to regulate the expression of various kinds of racial reasoning in our behavior. The purpose of this spending is to promote racial justice. But in no small measure, the conspicuous social programs we have brought into being in the attempt to realize this purpose have themselves contributed to racial strife and mistrust, while doing little to ease the problems of racial inequality. Why?
There is general agreement that minorities, at least in the past, have suffered injustices at the hands of the majority. But agreement is not as readily reached on what the remedies should be. Much of the difficulty stems from the fact that racial problems are discussed mostly in moral terms and in value-laden language. Since there is no thoroughly accepted standard for testing moral statements, there is not always a rational way to settle these arguments. Another part of the difficulty stems from the imprecise and ambiguous use of language by scholars and laymen alike. In analytical usage, it is necessary not only to separate the connotative from the denotative (literal) content of words, but also to make precise and operationally useful distinctions. My modest aim here is to suggest more precise ways of analyzing racial problems and to show how these ways will help us to understand the problems better and hence promote more effective policy.
Working Definitions
The three words I am seeking to demystify—to restore to their strict (denotative) meanings—are discrimination, preference, and prejudice. I will also consider briefly the matter of real differences.
Racial Discrimination. Discrimination—as distinct from those things that produce it (preferences, prejudice, and real differences)—may be defined as an act of choice based upon utility maximization, and racial discrimination as an act of choice wherein racial attributes provide the criteria for choice. In this view, racial discrimination does not differ in any fundamental sense from other kinds of discrimination. All selection necessarily and simultaneously requires nonselection—that is, choice requires discrimination. When we preface the word “discrimination” with the word “racial,” all we do is to state the attributes selected as the criteria of choice.
Our lives are largely spent discriminating for and against selected activities, objects, and people. For example, many of us discriminate against those who have criminal records, who bathe infrequently, and who use vulgar speech. Some employers discriminate against applicants who speak with a foreign accent, or have low intelligence, or cannot read, or went to the “wrong” college. There is also evidence of discrimination in politics—not many short men have been elected to the U.S. presidency. Furthermore, personal discrimination is not consistent. Sometimes people discriminate against theatre in favor of parties, against women in favor of men; and at other times the same people do the reverse.
When a choice is made on the basis of race, that choice may reflect the preferences of the chooser for a particular race, but it also may not. It is impossible for an observer to say whether choices based on a particular physical feature reflect the indulgence of preference (tastes) or the attempt to minimize information costs (prejudice) or the recognition of real differences. Let me spend some time on the distinction among the three.
Racial Preference. In everyday language as well as in economic analysis, an individual is said to prefer A to B if he places a higher value on A than on B. In economic theory we postulate that each individual has a consistent set of preferences—that is, tastes—and chooses the combination he most prefers from the available alternatives. When we are speaking as economists, there are no objective criteria by which we can judge whether one set of preferences is “better” than another. We cannot prove, for example, whether it is better or more righteous for a person to prefer the wines of Bordeaux to those of Burgundy or to prefer blue cars to red cars. The most we can ever objectively say is that, given his preference pattern and income and price constraints, the chooser is—or is not—doing the best he can.
It is impossible for an observer to say whether choices based on a particular physical feature reflect the indulgence of preference (tastes) or the attempt to minimize information costs (prejudice) or the recognition of real differences.
This holds true as well when we come to individual preferences for physical attributes such as height, weight, “richly” endowed body, hair color, and so forth: these are solely matters of individual taste. And given that there are individual preferences for or against physical attributes in general, we expect people to exhibit preferences for or against racial attributes as well. Indeed, so far as our analysis is concerned, there are no conceptual distinctions between racial and other preferences.
It may be rejoined that racial preferences are not comparable to other kinds of preferences in the consequences they have for society and for individuals. However, although the indulgence of racial preferences has specific effects that the indulgence of preferences for certain wines does not have, are the preferences basically different? If so, how do they differ? The preference for Bordeaux wines “harms” Burgundy producers by reducing the value of resources that are held for Burgundy production. If the consequences of preferences are generally to reduce the value of some resources and increase the value of others, then it can be said that preferences for physical attributes have effects similar to those of other preferences. The essential difference—by no means small—between preferences for racial features and those for wines is that the latter are not as specialized as the former. In other words, if Burgundy producers see that consumers prefer Bordeaux, they will try to shift their resources to Bordeaux production. On the other hand, for example, people who are black cannot become white.
But the fact that racial characteristics are unchangeable does not put them in a class by themselves. Persons with average IQs are preferred to those with below-average IQs, and persons who are not physically disabled are preferred to those who are. In each of these cases, the less-preferred characteristics are unchangeable and in each the less-preferred person suffers a competitive disadvantage. This disadvantage is to be expected. Disadvantage and advantage are inevitable consequences of differences in individual tastes, abilities, and traits, on the one hand, and of freedom of choice in a democratic society, on the other.
What I have said could be construed as trivialization of the problems of minorities and a defense of the status quo in racial matters—in which case it will necessarily offend the sensibilities of decent people. I do not consider the problems trivial, do not defend that status quo, and do personally find choices made solely on the basis of racial preferences obnoxious. However, in what follows, it will be shown that many racial outcomes are the results not of racial preferences per se but of political restraints on voluntary exchange. In a world where racial preferences are not the sole cause of these outcomes, policies that assume they are will fail.
Racial Prejudice. The term prejudice has the Latin root meaning “to judge before,” and a prejudiced act may thus be defined as a decision made on the basis of incomplete information. (Note that prejudice, like preference, is not the act but the cause of the act.) Making decisions without complete information is necessary in a world of scarcity and uncertainty, and in a complex world erroneous interpretation of the available evidence is by no means uncommon. Moreover, different individuals may arrive at different conclusions even if confronted with the same evidence, and the behavior of any given individual may sometimes not respond to changes in the evidence.
Consider a simple example of how decisions are made on the basis of incomplete (and perhaps misinterpreted) information. If a fully grown tiger suddenly appeared in the room, most of us would rush for the door, not because of what we knew about that particular tiger but because of our stock of information—and misinformation—about tigers as a class. The response is based on a stereotype—that is, the individual makes a prejudiced decision. He makes no attempt to seek additional information, but rather ascribes known or surmised group characteristics to the individual tiger. Examples of prejudiced behavior abound—not talking to strangers, running in response to rustling in the bushes, not buying bent cans of food, not recruiting employees from certain schools, and so on.
Decisions to prejudge are inextricably tied to individual judgments on what constitutes optimal information search. Information is not a free good; it is acquired by the expenditure of time, effort, money, and income foregone. As a result, wealth-maximizing individuals can be expected to economize on information costs. For any of us there will be a point at which the cost of acquiring one more unit of information is equal to the expected gain from that unit—which means that further information will not be acquired because the added cost would exceed the expected added gain.
A vast number of decisions must be made during our lives. Some of them, such as deciding to greet a passerby in the morning, require relatively small amounts of information. Others, such as selecting a marriage partner, require relatively larger amounts. A person is not prejudiced or unprejudiced. Rather, a person always exhibits prejudiced behavior to the extent that he substitutes general information (prejudgment or stereotypes)—which is less costly—for more costly specific information. What distinguishes among people are their comparative degrees of prejudiced behavior when facing similar situations.
In the literature on racial behavior the word “prejudiced” is most often used pejoratively to refer to those whose optimum quantity of information is deemed by the observer to be too small. Behavior based on racial or sexual stereotypes is commonly viewed as making use of too little information and thus viewed as opprobrious—and in many cases, of course, it is illegal. However, the quantity of information effectively collected before decisions are made (which is the reciprocal of prejudice) is up to the individual’s calculation—for there is no social standard or optimum amount of search that is applicable to all individuals in all cases. For example, for the prospective house buyer, there is no socially determined optimum number of houses to be looked at before making a decision. Instead, the amount of information collected by free individuals before acting is determined by and reflects, among other things, the efficiency of the individual in converting resources into information and the value of those resources as measured against the expected value of a “correct” decision.
Since all of us will seek to economize on information expense, we will substitute less costly forms of information collection for more costly forms. Physical attributes are easily observed and hence constitute a cheap form of information. If a particular physical attribute is highly correlated with some less easily observed attribute, then the physical attribute may be used as an estimator or proxy for the other. The cheaply observed fact that an individual is short, or an amputee, or a Negro, or a woman thus provides “sufficient” information for predicting the presence of some other unobserved attribute. Most of us, for example, if asked to identify individuals with advanced academic degrees only by observing race and sex, would assign a higher conditional probability that white males would have such degrees than black males or women. Such behavior is what decision theory expects where unobservable variables must be estimated from observable variables.
Real Differences. Discrimination need not, of course, be based on preferences as the word is generally understood (that is, tastes) or on prejudice (that is, judging before all the returns are in). It may be based on virtually complete knowledge of real differences and have nothing to do with aesthetic (or other) sensibilities. My reluctance to remain in the room with an uncaged tiger is, though reasonable, prejudice—since it is virtually certain that my sample of tigers previously experienced is statistically insignificant. My reluctance to ride in a bus may be a matter of preference—taste—if it is based on the fact that I find buses, though cheap, aesthetically unappealing. But there are other kinds of choice, and though one may object to their being included under the general rubric of discrimination, it will be seen later on that acts based on full knowledge and rational calculation have been considered “discriminatory” in the bad sense.
My reluctance to remain in the room with an uncaged tiger is, though reasonable, prejudice—since it is virtually certain that my sample of tigers previously experienced is statistically insignificant.
Prejudice in Action
Some may think my discussion of discrimination and prejudice renders the words meaningless, since it can be said that all human acts involve choice and all choices are based on incomplete information (as well as on tastes). But the discussion is useful because it permits us to avoid confusing one form of behavior with another. It enables us to see that certain kinds of choices—those made on the basis of racial, sexual, and other physical attributes—may be intelligent optimizing or may be the result of tastes. An example will clarify these points.
Suppose we are on a university campus where the racial and sexual composition of the student body is the same as that of the U.S. population, and suppose we play a game of trying to identify students who can find the integral for the mathematical expression ∫x2dx. Players are given zero information about the students’ mathematical proficiency and may not communicate with students except to ask, What is the integral? In other words, players can distinguish between students only by observable attributes such as race, sex, mannerisms, dress, and speech accent. The game is played for money, and it is assumed that the player’s objective is to maximize his winnings.
The player is in a situation where choices must be made on the basis of incomplete information. He is faced with identifying the observable attributes that will be the best indicators of student proficiency in calculus (the unknown and unobservable attribute). If he thinks that mathematical proficiency is equally distributed by physical attributes, his choice process will be essentially random. But if he thinks that mathematical proficiency is not so distributed, he must adopt a different (nonrandom) decision rule. In his first cut at such a rule, he may decide not to choose females because he knows that women are not well represented in the quantitative sciences. (Note that such a rule might not be as valuable in the Soviet Union, where a greater portion of women enter the quantitative sciences.) In successive cuts at a decision rule, the player may discriminate against — not choose — Negroes, Puerto Ricans, and American Indians, perhaps reflecting his awareness that math skills are related to the quality of pre-college schooling and that these particular minorities have historically received grossly inferior elementary and secondary education. In the end, the player may settle on a rule that confines his choices to males of Jewish or Oriental ancestry.
I am not saying that there is a genetic or causal relationship between race or sex and mathematical proficiency. What I am saying is that these variables are correlated in the real world (though I think it risky even to suggest possible explanations).
Suppose we relax the implicit assumption of neutral racial preferences and assume, instead, that the player has a distaste for Jews and Orientals but still believes these groups to be disproportionately represented in the quantitative sciences. So long as we retain the assumption that the player seeks to maximize winnings—that is, economic payoff—his decision rule, his prejudice, will not be distinguishable from that of a player with no racial preferences.
This illustrates an important point that is lost in most discussions of racial issues: choices made on the basis of race (or sex) do not always permit us to put the preferences of the chooser in unambiguous categories. Moreover, the example raises a question whether anyone should care if the player in the game chose to indulge his preference and not select Jews or Orientals? In our scenario (assuming that Jews and Orientals are disproportionately represented in the class of individuals knowing calculus), the player who because of tastes discriminated against Jews and Orientals would win less than other players. Even the most fervent advocate of civil rights would have little reason to seek a social policy that required anti-Jewish or anti-Oriental players to give Jews and Orientals an equal opportunity to be selected. The racist (or, for that matter, any individual who permitted his choices to be determined by economically irrelevant “preferences” of whatever kind) would bear the full cost of such an action. He would lose money.
Racial Preference, Optimizing Prejudice, or Real Differences?
For some reason these points tend to get lost in discussions of racial discrimination. We overlook the fact that not every discriminatory action reflects preferences—a dislike of Negroes per se, for example: certain discrimination may come from the rational behavior of individuals minimizing search costs or confronting real differences in the market, whether that market is free or institutionally constrained. And we overlook the fact that, in a free market, economically irrational preferences will—as in our hypothetical game—impose costs on whoever indulges them. Institutional restraints may, of course, render that indulgence costless to the indulger. If they do, the answer is to lift the restraints and reimpose the costs—in other words, to free the market.
When we are formulating policy, we must be careful to distinguish among the three sources of “discrimination”—preference, prejudice, and real differences. If we assume that racial tastes cause the problem we are addressing, when in fact the problem is caused by something else, our policy will be at least ineffectual and quite possibly harmful to its intended beneficiaries. Let us look here at three areas where whites are generally charged with discrimination against blacks, and where the assumption is generally made that the discrimination is based on racial preference—that is, tastes. These areas are hiring, home mortgages, and the prices ghetto shoppers pay.
Hiring and Employment Discrimination. Many recruitment and hiring practices are said to reflect racial preference, but an alternative explanation can be drawn from our knowledge of hiring procedures. When a firm seeks labor, it must find out how productive those seeking jobs are likely to be and must train the persons it hires. Since this process costs money, the firm has reason to search for recruits that appear to have a high probability of success. If the firm believes there is an important relationship between a recruit’s high school performance (and the quality of his high school), on the one hand, and employee productivity, on the other, it can reduce some of its recruitment costs just by knowing the job candidate’s record (and high school). If a firm knows that blacks at grade twelve are typically three to five years behind whites in scholastic achievement, it can assign a higher probability to a white recruit’s having the desired productivity.
In the mind of the employer, skin color may be a first approximate indicator of expected worker productivity. To observe a process that selects in part by skin color and to attribute the selection to taste (in this case, to employer “racism”) would be misleading. It would be like concluding that auto insurance companies charge drivers under twenty-five years of age higher premiums and that life insurance companies charge women lower premiums because these companies dislike young people and men. In both cases, a physical attribute may act as a general proxy for some other attribute (in the case of drivers, the higher probability of an accident) that is individually more costly to ascertain.
Suppose an employer who has racially neutral preferences incorrectly perceives that, on the average, a Negro worker is less productive than a white worker. What kinds of laws would cause him to seek more information and perhaps revise his perception? Clearly, laws that require him to pay all workers identical wages or that make it very costly to fire an employee have the opposite effect because they reduce his incentive to experiment. That is, employers will be less willing to hire someone of uncertain productivity if the wage must be the same as that paid a proven worker and if they are unable (or if it is costly) to fire that employee should hiring him have been a mistake. |
