Plate iii in print

The Diploma

Spence and the signal: why a costly credential can be worth its price even when it teaches nothing, and what it means that we burn years to prove what we already are.

A costly mark can be believed precisely because it costs; if the diploma is dearer to win for the dullard than for the scholar, then to hold one is to prove oneself the scholar — though it taught one nothing.

— the principle of this plate, after Michael Spence, 1973

If hidden information can strangle a market, the first remedy is for the side that knows to find a way to prove what it knows — credibly, so that the proof cannot simply be claimed by anyone who likes. The difficulty is that words are cheap: every seller says his car is sound, every worker that he is able. What is needed is an action the good type can afford and the bad cannot, and the theory of such actions is the theory of signaling.

Michael Spence found the principle in the labor market. An employer cannot see how productive a job applicant will be; the applicant knows, but saying so proves nothing. Suppose, though, that there is some action — acquiring a credential, say — that is costly to everyone but more costly to the less able, who find study harder, slower, more painful. Then the able may choose to acquire the credential and the less able may rationally decline, since for them the cost exceeds the gain; and the employer, observing who holds it, can infer who is able, and pay accordingly. The credential has sorted the types — not by anything it taught, but by who was willing to bear its cost.1

1. Michael Spence, "Job Market Signaling" (1973): when a costly action (such as education) is less costly for high-productivity types, they alone find it worth acquiring; employers infer type from the signal and pay accordingly. The signal separates types by differential cost, not by anything it produces. Shared the 2001 Nobel.

The implication, which Spence stated without flinching, is unsettling. The credential can perform this sorting even if it teaches nothing useful whatever — even if the years of study add not one jot to the worker's actual productivity. All the signal requires is that it be costlier for the less able; its content is irrelevant to its function. To the extent that education works this way — and how large that extent is remains one of the genuinely contested questions of the field — society spends enormous sums, and individuals spend years of their lives, not to become more productive but merely to demonstrate a productivity they already possessed. The diploma, on this view, is not a transformation but a token: a costly proof of what one already was, burned into the record at great expense to separate the able from the rest.2

2. The signaling implication: a credential can sort types even if it adds nothing to productivity, since only differential cost matters, not content. How much of education is signaling versus genuine human-capital formation is empirically contested; insofar as it is signaling, the resources spent acquiring it produce no output, only sorting.
⁂

What makes this more than a curiosity is that the signaling can be an equilibrium no one can escape, and one that wastes resources in the aggregate. Once employers expect the able to hold the credential, an able worker who forgoes it is taken for unable and underpaid; so he must acquire it, though it teaches him nothing, merely to avoid being pooled with the worse. And once he does, the bar may rise: if everyone able holds a bachelor's degree, the truly able distinguish themselves with a master's, and credential inflation sets in, each tier forced to spend more to stay ahead of the one below. The private return to the signal is real — the holder is genuinely paid more — but a part of the social expenditure produces nothing, a deadweight cost incurred only because information is hidden and must be bought with a costly, contentless proof.3

3. Signaling can yield a wasteful equilibrium: able workers must signal to avoid being pooled with the unable, and credential inflation can raise the required signal over time. The private return is real, but resources spent purely to sort — rather than to build skill — are a social deadweight cost of asymmetry.

Once seen, the signal is everywhere. The firm that pays a costly dividend signals that its earnings are real, for a weak firm could not sustain the payout; the warranty signals quality, for a shoddy maker could not afford the claims; the lavish office and the tailored suit signal success, for the failing could not bankroll them; the peacock's tail and the gazelle's wasteful leap signal, to the biologist, a fitness the weaker animal could not squander energy to display. In every case the logic is identical: a signal is believed exactly to the degree that it would be too costly for the wrong type to send. The cost is not a regrettable side effect of the signal; the cost is the signal — without it, the message could be sent by anyone, and so would mean nothing.4

4. The signaling principle is general: costly dividends (Ross; Bhattacharya), warranties, conspicuous consumption (Veblen), and biological displays (Zahavi's handicap principle) all signal hidden quality. A signal is credible precisely because it is too costly for the wrong type to imitate; the cost is constitutive of the signal, not incidental to it.

Signaling does not always rescue the market, and its limits matter. The signal works only if costs truly differ by type; where the dishonest can fake it as cheaply as the honest can send it, it collapses back into mere talk and the lemons return. It can settle into the wrong equilibrium, with too much costly signaling or too little. And it shifts as well as solves: a signal that sorts the able from the less able may also, when the cheapest available proxy is membership in a group, harden into the statistical discrimination a later plate must confront. The signal is the informed side's great weapon against asymmetry — but a weapon that costs the wielder dearly, and society more — and the next plate's device approaches the same silence from the opposite side: not the informed proving what they are, but the uninformed building a sieve to make them show it.5

5. Signaling fails if the bad type can imitate it cheaply (it reverts to cheap talk); it may settle at inefficient levels; and where group membership is the cheapest proxy, it shades into statistical discrimination — see The Stereotype. The complementary device, in which the uninformed party induces revelation, is screening — The Screen.

The diploma is the signal made flesh: a costly mark, believed because it costs, by which the holder of hidden worth can prove it without a word — and which, in the proving, may teach nothing at all, so that we burn years and fortunes to demonstrate what we already are. It is the informed side's answer to the lemon's unravelling: where quality cannot be seen, let it be signaled, at a price the unworthy cannot pay. But there is another way to break the silence between the two sides, worked not by the one who knows but by the one who does not. The next plate takes up the screen: the menu, the test, the deductible — the sieve the uninformed party designs so that the informed, choosing freely among its options, sort themselves.