The Skill — The Wage, Plate iv

The marginal principle says a worker is paid what his labor adds; but workers add wildly different amounts, and the differences are vast, systematic, and persistent. The doctor earns a multiple of the orderly, the engineer of the laborer, the graduate of the school-leaver, and these gaps hold across countries and across decades. What explains them? The mid-twentieth century gave an answer of great reach and considerable charm: workers differ in their skill, skill is a kind of capital, and the wage is the return on that capital — the yield on an investment that happens to be lodged in a person rather than in a machine.

The idea, pressed by Theodore Schultz, Gary Becker, and Jacob Mincer, was to treat schooling and training as investment in the strict economic sense. Time and money spent learning are costs borne now — tuition paid, wages forgone while studying — in exchange for higher productivity, and so higher earnings, later. A worker, or his family, or the state, trades present income for a larger future stream, exactly as a firm trades cash now for a machine that pays out over years. On this account the difference between a graduate's wage and a school-leaver's is not a mystery and not an injustice; it is the return on the years and money the graduate sank into building the human capital the job now rents.¹

1. The human-capital school: T. W. Schultz, Investment in Human Capital (1961); Gary Becker, Human Capital (1964); Jacob Mincer (1958, 1974). The term met early resistance for seeming to treat people as machinery, and Becker reports being uneasy with the title; it became, nonetheless, the organizing concept of modern labor economics.

The framework's great practical gift was to make the wage structure measurable. Mincer's earnings function — log earnings written as a straight-line function of years of schooling plus a curve in years of experience — became the most-run regression in the discipline, and it yielded a number that policy could chew on: a rate of return to a year of schooling, generally estimated somewhere between six and ten percent, comparable to the return on other investments. Education could now be appraised the way a bond is appraised, and the appraisal could be set beside its costs.²

2. Mincer's specification (1974) — log earnings as linear in schooling and quadratic in experience — and the "Mincerian return" of roughly 6–10% per year of schooling. The equation's ubiquity is itself a fact about the field: it turned the wage structure into something estimable from a household survey.

Becker added a distinction that did much work. Human capital comes in two kinds. General skill is portable — useful to any employer — and so, Becker argued, the worker must pay for it himself, accepting lower wages during training, since a firm that paid to build a portable skill would simply watch the trained worker leave for a rival. Specific skill is useful only to the present employer, and so its cost and its return are shared between firm and worker, which explains a tangle of otherwise puzzling facts: why wages rise with tenure, why firms hold on to workers through downturns rather than rehiring later, why both sides find a parting costly. A single idea — skill as investment, portable or not — organized the lot.³

3. Becker's general/specific distinction: general human capital is financed by the worker (through depressed training-period wages) because the firm cannot capture its return; specific human capital is a joint investment, generating the upward wage-tenure profile, labor hoarding, and the mutual cost of separation.

⁂

And yet a single, devastating question hangs over the whole apparatus, and it was put with great clarity by Michael Spence. Suppose schooling raised no productivity whatever — taught nothing a job uses — but was simply easier, less costly in effort, for the more able to endure. Then able workers would still acquire it, not to learn anything but to distinguish themselves from the less able; and employers, unable to see ability directly, would still pay more to the schooled, reading the diploma as a reliable sign of an ability that was there all along. In this world wages would rise with schooling exactly as the human-capital theory predicts — the same regression, the same return — and yet the schooling would have built nothing. It would be sorting workers, not improving them; a signal, not an investment.⁴

4. Michael Spence, Job Market Signaling (1973). In the separating equilibrium, education credentials convey pre-existing ability rather than create productivity. The unsettling result is that human capital and signaling can be observationally equivalent in wage data, while implying opposite answers to whether education makes workers more productive — and so opposite social, as against private, returns.

The signaling possibility is not merely a logician's worry, because human capital and signaling are, in the cross-section of wages, nearly impossible to tell apart. Both predict that the schooled earn more; the data alone cannot say whether the schooling caused the earning or merely revealed the worker who would have earned it anyway. And this is sharpened by a second, more mundane problem: the people who get more schooling differ, on average, in ability, family resources, and drive — traits that raise earnings on their own — so that a naïve comparison of the schooled and unschooled credits to education a premium that selection had already built in. Decades of ingenious work, using twins, and using natural experiments such as changes in compulsory-schooling laws, have tried to isolate the true causal return; they find it positive and real, but smaller and less certain than the raw correlation, and they cannot fully dissolve the signaling ambiguity.⁵

5. The ability-bias and identification problem: simple return-to-schooling estimates conflate the causal effect of schooling with the selection of who gets it. Twin studies and instrumental-variables strategies exploiting compulsory-schooling laws (notably David Card's) recover a positive causal return of contested magnitude. None settles whether the return reflects productivity built or ability signaled.

For the purposes of this volume, the human-capital theory has two limits that matter, and they point in the same direction. The first is the one just drawn: even granting that skill is real and its return real, the theory cannot, from wage data, separate worth created from credential displayed. The second is quieter and more important. Human capital explains the differences among workers — why this one earns more than that — while saying nothing about the level the whole structure sits at, or about who captures the surplus a skilled worker produces. A worker of given skill, with a given marginal product, may still be paid well below that product if the employer holds the upper hand; human capital sets what a worker brings to the table, not what he carries away from it. The flagship plate takes up exactly that gap, between what the worker is worth and what he is paid — and finds in it not skill but power.