Discrimination is choice made under the cost of information. Once you see it that way, the question stops being who is prejudiced and becomes who ends up paying for the sorting.
Source · Thomas Sowell, Discrimination and Disparities
— Reflections on Thomas Sowell, Discrimination and Disparities
Discrimination is the normal condition; there is no need for excessive sensitivity.
Discrimination is neutral, not inherently tendentious.
Discrimination is choice — the choice in which, being able to select only A, we are obliged to forgo B, C, D, and every other option.
We are always discriminating against the external world, and always responding to the external world’s discrimination against us.
This seems easy enough to grasp, yet in practice we are seldom aware of it.
The manner in which we discriminate is determined by the degree of information we command about the choice.
When we are able to bear the cost of seeking relevant information, the discrimination we exercise is a choice grounded in objective, clear information. Consider a multiple-choice examination: you make a judgment on the basis of the information you have accumulated and mastered, and simultaneously discriminate against the other options.
But in most real situations, we cannot bear the cost of seeking information. What then supports our choice is historical experience — or even prejudice.
By contrast with most of you, I arrive at my judgment on a multiple-choice examination by the roll of a die or by counting off — and simultaneously discriminates against the other options.
Sowell’s taxonomy in fact identifies three types of discrimination:
A small illustration. Suppose there is a large investment firm.
Ia. The firm’s aging chairman intends to select a successor from among three senior executives (the CFO, the COO, and the CTO). To secure the company’s future development, the chairman will in all likelihood conduct a thorough, detailed examination of all three candidates, seeking the most comprehensive possible information bearing on the company’s prospects, and only then make the choice he judges beneficial — discriminating against the other two candidates.
In the end, the chairman selects the CFO: outstanding in professional capability, somewhat less accomplished in social nuance.
Ib. During the recruiting season following the handover, the firm plans to select three hires from three hundred applicants. But to ensure a timely generational transition and maintain normal operations, the HR department must complete screening and onboarding quickly. HR will therefore, in all likelihood, screen the information in applicants’ résumés rapidly against some standard distilled from experience — not with the meticulous scrutiny the chairman applied to his successor in case Ia.
II. The new chairman, though outstanding professionally, is rather rigid and traditional in outlook. On taking office he amends the firm’s recruiting policy: eliminate all female applicants of marriageable age.
By the end of the recruiting season, the firm has hired three male new graduates, all holding undergraduate degrees from “985” universities — the tier of elite Chinese institutions designated under Project 985 — and master’s degrees from top-twenty American universities.
It is not difficult to see that all three cases — and the final hiring outcome — betray the sort of “discrimination” we take entirely for granted and yet find perpetually unsettling.
Discrimination is not merely choice; behind it is cost.
That HR is content to select, “without thinking,” graduates of elite institutions with rich social experience requires no stipulation as a premise; worldwide, this is in all probability simply the fact.
So why is the firm willing to select them?
Or rather — why does it not select us?
Time is finite, energy is finite, cost is finite. A firm can hardly know every applicant’s individual circumstances in full; sometimes it does not even see our résumé, because some firms announce a minimum educational threshold in advance.
To expect a corporate recruiter to understand our strengths, character, and talents the way our parents or friends do is fantasy — unless the firm belongs to our parents or friends.
From the foregoing we may conclude: the firm cannot bear the cost of collecting information on every applicant.
So in most cases the firm can only discriminate on the basis of historical experience. And what is historical experience?
Historical experience is this: good schools have a high probability of producing good students; good students have a high probability of becoming good employees; good employees have a high probability of producing good results; good results have a high probability of yielding good compensation; good compensation has a high probability of indicating a good firm…
This has nothing to do with the preceding pieces on success — though high social fitness does, in fact…
And what of the poor student?
As a long-serving “poor student” myself, I consider myself qualified to speak.
Being a “poor student” means that the cost to a teacher of collecting future information about you is greater — so great, perhaps, that it is not worth bearing. The teacher therefore abandons the collection of information about you and can only judge you on historical experience.
I will not address the moral and ethical dimensions of the teacher–student relationship here; I confine myself to the economic.
Suppose I have repeatedly disrupted class discipline and provoked the teacher, but I believe — and the whole class agrees — that I sing very well. At a forthcoming class singing competition, the teacher will in all likelihood not choose me as lead vocalist, because the cost of collecting the information that “I sing very well” is simply too high: she risks provocation from me and risks my disrupting class discipline and delaying the syllabus. That cost is too large to bear. The teacher can therefore only act on experience, and I can only be discriminated against.
(The case is illustrative and not autobiographical — thank you for understanding. I sing badly.)
Thus: because firms cannot bear the cost of collecting information about graduates of “bad” schools, they can only evaluate those graduates on the historical experience that graduates of “bad” schools have generally been “bad,” and are therefore unwilling to hire them.
Over time, this experience-based discrimination gradually hardens into habit.
But because our understanding of the mechanics of discrimination is defective, we readily elevate discrimination grounded in the limits of information cost into the prejudice and personal distaste of others.
Our preference for Type Ib discrimination may derive from a biological tendency toward energy conservation — reducing thought conserves biological energy, ensuring we retain enough to meet future risk. In practice, this is a mismatch between instinct and an age of information saturation.
The book cites Professor Walter Williams’s point about economizing on information cost: we tend to select information that is physically cheap (skin color, age, sex), and cheap information sometimes substitutes for the expensive information correlated with it.
In reality, however, experience-based discrimination and prejudice-based discrimination are joined like the Huangpu River and Suzhou Creek. But unlike the line delivered by Uncle in the television series Blossoms Shanghai — the Huangpu and the Suzhou, one clear, one muddy, with a definite boundary between them — we cannot specify or fix where the boundary between these two kinds of discrimination lies, or on what scale.
The book observes that we cannot automatically attribute both men’s contributions in engineering and women’s in education to prejudice — men and women choosing different disciplines at university do not do so entirely out of prejudice.
True enough. This is like our inability to explain clearly how the self is formed: to what degree we are determined by genes, to what degree influenced by parents, and how much is produced or altered in the process of socialization.
We cannot simply and automatically attribute an outcome to one type of discrimination. Indeed, the world does not contain only those three types; in principle, the three above can be subdivided into countless further types blending them in varying proportions.
Statistical results are experiential, historical, and conceptual. Conclusions should therefore be understood as systemic outcomes, produced by the mutual influence of many factors within a complex system. And it deserves particular emphasis that the historical character of an outcome refers not only to the process of its formation: the outcome may also evolve in the future, with different results corresponding to different points in time.
Data results do not merely represent themselves; data are historical. Behind the data lie historical, complex, and varied factors.
The distinction between numbers and words is not merely one of objectivity versus subjectivity — and for anyone harboring a particular purpose, such distinctions are trivial.
To achieve an end, omitting data and substituting concepts — methods that appear crude but remain effective over the long run — are persistently difficult to detect.
(1) Errors of omission.
Errors of omission are not solely a product of subjective bias; sometimes they are a reluctant necessity imposed by the limits of our own capacity, since, as noted above, we can rarely bear the cost of collecting all the information.
Beyond the simple and effective expedient of ignoring data, ignoring the statistical range of the sample, the comparability of samples, and the concept of time are harder still to detect.
(2) The scope of the statistic.
I. The narrower the statistical scope, the lower the likelihood of finding equal group representation — and the higher the probability of finding inequality.
We may find more men clustered at the top of the academic rankings in university engineering programs, but this does not affect the possibility that the women ranked below them — and indeed other men — may outperform those men in music, chess, calligraphy, painting, or athletics. Under the current educational system, in fact, the higher a student’s academic ranking, the more likely they are to lag in other areas, notwithstanding a small population of genuine all-rounders.
II. Influences on the sample prior to its entry into the statistical scope are also folded into that scope — that is, the imperceptible influence of history on people is ignored.
Take cultural capital in sociology: the cultural capital people inherit from family in early life typically confers an advantage in the later accumulation of new cultural capital. Even among children of the same age, placing a child from a scholarly household and a child from a poor one in the same sample for statistical purposes does not seem especially wise.
(3) Household income.
Income points to the individual; a household is a collection of individuals.
Wealthier households generally have more children — which means that the gap relative to ordinary households exists not only in income difference, but is further widened by the number of children and the potential income they will generate in the future.
(4) Capital gains.
If income is measured as the sum of wages and capital gains, then the conversion, within the measurement period, of capital gains accrued before it will substantially affect the measured income.
Moreover, capital gains do not grow at simple interest the way wage income does; they compound. And ordinary households typically accumulate far less in capital gains than wealthy ones.
If annual income is the standard, then even where your wage income and mine are identical, if you have realized substantial capital gains within the period through experience accumulated in long-run capital operations, your annual income will obviously exceed mine. Extend the measurement period further and this gap will likely widen. The risk of capital gains coexists with the return — but the compiler of statistics may still, by various means, disregard and obscure data that challenges the prevailing vision.
(5) Revealed preference > stated preference — facts outweigh eloquence.
What people do expresses their values better than what they say.
Consider our differing definitions of “effort”: my effort is reading for five minutes a day; yours is reading for five hours a day. This substantive difference is hard to capture in answers obtained by verbal report — as with the study-habits questionnaires we filled out at school.
The most direct illustration comes from the free market. However great my enthusiasm for a new pair of sneakers, however besotted I may be, if for whatever reason I never buy them — if I am never willing to pay a price to obtain them — on what basis can my affection be demonstrated?
(6) Survivorship bias.
The information we can collect often originates with those who were not eliminated by competition; the eliminated can no longer speak. The data we obtain therefore cannot fully reflect the competitive process, because it is shaped by the survivors’ subjective accounts.
Consider successful founders and entrepreneurs summarizing their own careers: they frequently attribute their success to luck — even a pig can fly if it stands in the right wind. In reality, the difficulties and challenges they encountered are beyond our imagining.
(1) Ex ante and ex post concepts.
I. Privilege and achievement.
Privilege is a right held prior to the event; achievement is the reward conferred after the fact on the one who succeeded.
To treat another’s success as the exercise of privilege, without evidence, is to put the cart before the horse.
The book notes that the achievements of Irish, Jewish, Chinese, and Japanese immigrants in America are now verbally erased by some through a single word — “privilege.” Even today’s middle-class Black Americans are described by some as a “privileged class,” notwithstanding that their ancestors arrived in America as slaves.
II. Mobility rate and mobility.
Mobility is a right existing prior to movement; the mobility rate is an ex post evaluation of movement.
Mobility rate and mobility are in fact not directly connected. Mobility does not determine the mobility rate; the rate is properly measured against those with a willingness to move, not against everyone who possesses mobility — the latter being meaningless, since not everyone with mobility genuinely wishes to move. And we readily overlook that movement is not only upward. Although sociology classifies social mobility in many ways, our attention here is chiefly on vertical movement between social strata — upward and downward.
It is rather like counting toward the mobility rate those renunciants content in poverty and detached from worldly ambition: they possess mobility, but including them injects water into the final figure.
(2) Citation, fabrication, and insinuation.
I. Theories or statements whose provenance cannot be traced are treated as universally known convictions and cited repeatedly to advance or reinforce a position, simply because they circulate widely among the like-minded. In reality we cannot locate their source at all — or the original statement or theory has been altered.
II. On insinuation, the book’s formulation is apt — a response to the claim that a minority of the world’s people hold a majority of its wealth:
Collectivize the wealth individuals create, and then describe those individuals who created more wealth and therefore received more of it as having deprived others of their fair share.
Consider an idealized case: current authorities, in order to protect the existing structure of social power, use media, education, and cultural transmission to invest historically combative vocabulary with gentler meanings, and then socialize those words and meanings into present and future generations. Liberty, for instance, ceases to be resistance to violent rule and becomes a claim on social welfare. Plainly, the cost of granting people higher welfare is trivial compared with the cost of suppressing their resistance, still less a revolution.
Of course, the mechanism by which people refuse to admit error — and even act further in the direction of the error — may resemble what psychology calls confirmation bias: the tendency to seek evidence favorable to one’s existing view. This tendency toward consistency between prejudice and “fact” becomes still more severe when the prevailing views are entangled with politics, interest, and prestige.
A vision is necessary — for individuals, organizations, societies, sciences, and nations alike.
There is nothing wrong with having a vision.
But when we examine today’s social visions — “social justice,” “equality for all” — it is difficult not to notice how far they diverge from genuine social justice and genuine equality.
Here I find nothing more incisive than Sowell’s own formulations:
A social vision built upon a fallacy is enormously infectious.
Ask yourself: on first hearing that only 14% of white applicants were denied subprime loans, against 52% of Black applicants, did your heart not stir? But in reality: only 54% of Black applicants were qualified for a loan, and the 52% who actually received one constitute 96% of that qualified group; whereas 18% of white applicants were qualified, and those who received loans constitute only 77.7% of that qualified group — lower than the Black figure.
1. That the causes of differing outcomes can be determined by collecting statistics displaying unequal outcomes — that is, that data can adequately reveal the causes of outcomes.
As set out in “Numbers and Words” above, finding historical facts that negate this assumption is not difficult across the long river of human history. What is difficult is whether we are willing to look for them and use them to challenge the prevailing view.
2. That different groups are typically similar in goals and potential, so that differences in outcome mean they have been treated unfairly by others — that is, that prejudice produces inequality.
Sowell’s response to this assumption:
I take Sowell’s meaning here to be roughly this: even outcomes generated under conditions of equal potential, equal goals, and procedural equality typically run counter to the prevailing social vision — that is, they are unequal, undispersed, and non-random. Unhappily, we remain deeply bound by that vision.
3. That if the more fortunate are not entirely responsible for their good fortune, then government can produce better outcomes — or morally better outcomes — through intervention — that is, that third-party decision-makers can help those “they judge to have been treated unequally” achieve “equality.”
The essence of discrimination is a question of cost: who bears it, and how much. That is the subject of the final section.
The essential difference between Type Ia discrimination and the others lies in information, and being unable to bear the cost of collecting information is the normal condition.
Take the free-trade economy as the representative competitive market.
Free competition means that individuals must bear the costs of their conduct along with its corresponding benefits, and weigh their conduct on that basis. In plain terms: you lie in the bed you make, and you get what you work for.
People’s weighing of the costs and benefits that fall on themselves is the incentive and the constraint of the free market.
In an insulated venue, by contrast, the bearer of costs and the beneficiary are in most cases not the decision-maker.
The various preferential policies proposed in service of a social vision — well-intentioned but perennially counterproductive — leave their designers under no obligation to bear the harm the policy inflicts on those subject to it. The minimum wage is a case in point.
A policymaker with the “noble vision” of “social justice” proposes a minimum wage, thereby depriving labor not worth such a wage under free-market conditions of the right to work:
Suppose a dishwasher earns 2,000 yuan a month under free-market conditions. The reason is simple: he generates 2,000 yuan of value beneficial to society.
After the minimum wage is enacted, the government raises the floor to 3,000 yuan a month.
In one scenario, the restaurant owner cuts some dishwashers in order to protect existing returns. But it is not the systemic decision of the market and its consumers that produced the dishwashers’ unemployment; it is the discrimination the restaurant owner was compelled to exercise in order to stay in business.
In this scenario, although it is the restaurant owner who cut the staff, can we really say the owner is the one who truly caused the discrimination?
There is of course a second scenario: out of a sense of responsibility, or of his own “social justice,” the owner keeps all the dishwashers. But if the business is to keep operating, he has only one option — to pass the cost on to consumers, whether by cutting corners in the kitchen or by raising menu prices.
Consumers have varying elasticities, but faced with costs forcibly passed through to them, and guided by the law of demand (price and quantity demanded are inversely related), the unfortunate restaurant will in all likelihood disappear into the long march of market competition.
And, following the earlier discussion, when we subsequently investigate the minimum wage’s effects on firms and workers, we will in all likelihood obtain only this: no substantive impact on firms, and a general improvement in employee welfare — the answer of the survivors.
Thus policymakers, who bear neither the cost falling on those made unemployed nor any public criticism for that unemployment (since they can readily ignore or conceal the data, and since it was not they who executed the layoffs), convert a policy full of “noble vision” into “noble outcomes” — ex post statistics heavy with survivorship bias — and convert those in turn into far more votes than expected.
Beyond the pass-through discussed above, the bearer of the cost of discrimination is not necessarily “the person who ought to be discriminated against.” It may be one’s own group — the consequences of my misconduct are borne not by me but by “our own people.”
When I was in middle school — my delinquent period — teachers routinely compared my friends and me to rotten fish and spoiled shrimp. But in the beginning we were all merely small fry; it was only through the misconduct of certain among us that we gradually turned rotten and spoiled.
My conduct affects more than myself. To put it charitably: every one of us has influence.
Sowell’s example is more worth pondering — the reason prices are generally higher in Black (poor) neighborhoods.
Theft and crime rates in such neighborhoods exceed those in ordinary neighborhoods. The consequence is that few business owners are willing to open a store in such a place. Set aside breaking even and turning a profit — keeping one’s life is fortune enough.
But the power of the market appears precisely here: the owner can raise prices to meet these latent risks. In other words, the elevated price is compensation to the owner for bearing so much risk.
So who bears this premium over ordinary market prices?
Plainly not the criminals — criminals do not need to purchase goods to meet their needs at all; they have what they take to be lower-cost methods.
The consumers who ultimately bear the risk premium can only be the law-abiding residents who live there and go about their lives. Beyond having chosen to live in a dangerous place, they have done nothing wrong — yet they must bear the cost others have created.
Lowering standards is undertaken in service of the social vision of “social justice” and “equality for all,” and is the politicized result of “compensating for historical inequality” — relaxing admissions requirements for minority students, certain national welfare systems (minimum-income guarantees, minimum-wage floors…).
In essence, lowering standards is itself discrimination — discrimination against the students and citizens who are held to the original standard. Yet among the latter we rarely see reporting or data on the discrimination they suffer at the hands of “those whose standards were lowered.”
The result of lowering standards is that it reduces the obligations those people ought to bear.
When a student needs only 44 points to pass, they need not expend the same effort as their classmates; those able to subsist on a national minimum subsidy need not bear their reciprocal obligations within the family.
The outcome is that the human-capital gap that minority students actually accumulate widens further over the course of their studies, while the moral sense and values of those relieved of obligation are steadily eroded.
Sowell, trained as an economist, analyzes inequality in historical and contemporary fact chiefly from the standpoint of the constrained vision.
The book’s emphases on human limitation, evolutionary outcomes, and systemic outcomes are the principal markers of that vision.
As a student with strong interests in economics and sociology, I find myself deeply in agreement with the constrained vision. But knowing so little of the political and philosophical material in A Conflict of Visions, I regret that I cannot share that book with you systematically.
I will certainly read it again in the future and share it — and I recommend it to you as well.
I only answer from what is written on this site.
Type a few words — I will show you where it says so.