The previous plate showed that wages may differ between equally able workers for reasons the worker can be said to earn — the danger, the dirt, the insecurity he is paid to bear. This plate takes up the harder and the uglier case: the differences a worker does not earn, the gaps in pay that track not the work or its burdens but the worker's race, or sex, or some other mark of the group he was born into, which has nothing whatever to do with what his hour produces. Discrimination is the name for the wage that follows identity instead of worth; and it is the starkest case in all of this volume of the price coming clean apart from the productivity it is supposed to measure. The plate sets out the two great theories of how it happens — one rooted in prejudice, the other in cold calculation — and presses the question that has long divided the field: whether the market, left to itself, dissolves discrimination, or quietly keeps it alive.
Begin with Gary Becker, who in 1957, in a doctoral thesis that startled his discipline, did a thing no one had quite thought to do: he turned the cold apparatus of price theory upon prejudice itself. Discrimination, Becker proposed, could be treated as a taste — a preference, like any other, for which those who hold it will pay. The employer with a taste for discrimination suffers, as he feels it, a genuine cost from employing a worker of the group he disdains; he behaves as though her wage carried a surcharge only he can perceive; and so he will take her on only at a discount deep enough to offset his distaste, or will pay a worker of the favoured group a premium for the comfort of his company. The wage gap, in this account, is the cash value of the employer's prejudice — the disfavoured worker paid below her product by precisely the measure of the aversion she must overcome to be hired at all. It was a cold way to describe an ancient cruelty, and its very coldness was the point: it made prejudice a thing the market could weigh.
And from this cold beginning Becker drew a conclusion that has echoed, and disturbed, ever since: in a competitive market, prejudice does not pay, and the market ought to punish it. For the employer who indulges his taste — who turns away cheap and equally productive workers because he cannot abide them — sacrifices profit to his feeling; and his unprejudiced rival, who hires those same workers gladly and cheaply, enjoys the lower costs and undersells him. In time the bigot is driven to the wall, or made to swallow his distaste in order to survive; the least prejudiced employers, hiring the most of the disfavoured group, come to set the going wage; and the gap is competed away. It is a striking result, and a genuinely double-edged one. It means the market is, in this one respect, a friend to the persecuted — which is exactly why those most determined to discriminate have so often had to reach past the market to the law, enforcing by statute the segregation that competition would otherwise have eroded, from the colour bar to apartheid: the prejudiced knew, better than anyone, that the free market would undercut them. But the result cuts the other way as well, for it leaves a hard puzzle behind it: if competition truly erodes discrimination, why has so much of it survived so long?
The market is, in this one respect, a friend to the persecuted — which is why those determined to discriminate have so often had to reach past the market to the law.
The puzzle has answers, and they mark out the limits of Becker's hopeful theorem. Competition erodes the employer's taste, and only where competition is genuine. It does nothing against the prejudice of customers: if buyers prefer to be served by the favoured group, the unprejudiced employer loses nothing — gains, rather — by indulging their bias, and the market rewards the discrimination instead of punishing it. It does nothing against the prejudice of coworkers, who may exact a premium to labour beside the disfavoured, raising their effective cost to any employer. And where competition itself is wanting — where the employer wields the monopsony power of the eighth plate, with no rival near enough to undersell him — he may indulge his taste at leisure, the disfavoured worker having nowhere else to take her labour. Becker's market dissolves discrimination only under conditions that very often do not hold; and just where they fail, prejudice pays after all. The cure is real, but it is partial, and it is slow.
But the deeper and the more disquieting theory came later, and it needs no prejudice at all. Edmund Phelps and Kenneth Arrow, in the early nineteen-seventies, asked what a perfectly unprejudiced and perfectly profit-seeking employer would do, confronted with the one inescapable fact of hiring: that he cannot see, in the applicant before him, the productivity he is paying for. He must guess at it, from whatever signs he has — and if the group a worker belongs to carries information, or is merely believed to, he will use it, exactly as he uses her schooling or her years of experience. If he believes the average member of one group the more productive, or only that the marks of productivity are easier to read in one group than in another, he will favour that group — not from animus but from inference, the very statistical inference that the signalling of the fourth plate made the whole of the employer's predicament. This is statistical discrimination, and it is far more troubling than Becker's taste, for three reasons together. It is, in the narrow sense, rational — profit-maximising, given the employer's beliefs — so that competition, which punishes the costly indulgence of a taste, does not punish it in the least. It wrongs the individual with a peculiar cruelty: the highly productive member of the disfavoured group, judged by her group's average rather than her own worth, is paid for what others like her are presumed to be, and not for what she is. And it may rest on beliefs that are simply false — mere stereotype mistaken for statistics — which the practice itself conveniently shields from ever being tested.
And there is a further turn, the darkest the theory has to offer. Statistical discrimination can manufacture the very group differences it claims only to observe. Suppose that employers, rightly or wrongly, expect less of some group, and so offer its members the worse jobs and the lower wages. Then the members of that group face a poorer return on their own effort and their own investment — why acquire, at cost, a skill that will not be rewarded? — and they rationally acquire less of it; and their lesser investment, in the fullness of time, confirms the employers' original expectation, whether or not it had any truth in it to begin with. The prophecy fulfils itself. Discrimination, on this account, is not merely a response to differences between groups but a cause of them, sealed in a circle that no single party has any incentive to break, and that can persist indefinitely without a particle of prejudice anywhere within it. It is the bleakest finding in the whole economics of the subject: that an injustice can be self-sustaining, self-justifying, and entirely “rational,” all at the same time — and that the wage which follows identity can, in the end, make identity follow the wage.
Discrimination is not merely a response to differences between groups but a cause of them — the wage that follows identity making identity follow the wage.
How much discrimination, then, is there — and of which kind? The question is hard, for the same reason the return to skill was hard: what one cannot measure, one cannot easily hold fixed. The raw gaps in pay between the races and the sexes are wide, but a great part of each is explained, on inspection, by differences in things one can measure — schooling, experience, occupation, hours of work — and the residue that remains unexplained is suggestive of discrimination yet proves nothing, for it may as easily be unmeasured productivity as prejudice. The cleanest evidence comes instead from experiments contrived to hold productivity fixed by design: the matched pairs of résumés, identical down to the comma but for a name that signals the applicant's race, of which the versions bearing the favoured name draw reliably the more invitations; the orchestra auditions conducted behind a screen, which, on the most-cited account, raised the hiring of women once the judges could no longer see who played. Such studies leave little doubt that discrimination is real, and present, at least at the threshold where workers are let in or turned away. And yet the largest pay gap of all, between the wages of men and of women, has been traced, increasingly and persuasively, not to unequal pay for the very same work but to the structure of the work itself — to the steep premium that modern jobs place upon long and unbroken and inflexible hours, colliding with the unequal burden of the care of children, so that the gap yawns widest at parenthood and falls hardest upon mothers. The wage that fails to reward equal contribution equally, it turns out, does so through several different mechanisms at once, of which raw prejudice is only the crudest and, very possibly, no longer the largest.
Return, then, a last time before the volume's end, to its question. Discrimination is the plainest violation in all these pages of the principle that the wage measures worth — the case where the price of an hour follows not what the hour produces but who it is that produces it. And the two great theories of it divide along the very fault this volume has traced from its first plate. The taste theory makes discrimination a matter of power and preference — the employer paying to indulge an aversion, the wage bent to his prejudice — and holds that competition, the great leveller, will in time bend it back. The statistical theory makes it a matter of information and inference — the individual sacrificed to the average, the worker paid for her group and not herself — and holds that competition will do no such thing, since the discrimination costs the employer nothing to keep up. Between them the two theories bracket a truth that holds some of each: there is prejudice the market punishes and prejudice it rewards; there is calculation that is accurate and calculation that is only stereotype in the dress of arithmetic; and there is the long self-fulfilling circle in which the two breed one another. What unites every case is the gap between what the worker is worth and what the worker is paid — the same gap this volume has found beneath every wage, here laid barest of all, because here the thing that overrides her worth is nothing she has done, or left undone, but only what she is. The market is, at its best, a force against this; at its worst, the very instrument of it; and which it will prove, in a given time and place, is not handed down by nature but settled, like everything in this volume, by the contest of worth and power. That contest, in this its ugliest form, remains — as the table is bound to say — open.