The single most reliable fact about wages is that the schooled earn more than the unschooled — steadily, almost everywhere, and by a margin that has mostly widened. The orthodox reading is the hopeful one: schooling and training are an investment, building into a person a stock of productive capacity, human capital in Gary Becker's phrase, on which the market afterward pays a return exactly as it pays a return on a machine. But there is a colder reading, owed to Michael Spence, in which much of what the diploma earns is owed not to anything it taught but to what it signals — that its holder was, all along, the sort of person able to obtain it. The two wages look identical. The distance between them is the distance between a thing made and a thing merely revealed, and it has defeated every attempt to measure it cleanly.
Take the hopeful reading first, and grant it its full strength, for it is no straw man. Its founding move was a genuine enlargement of the idea of capital. Capital had always meant the made means of production — the machine, the building, the stock of goods — things one invests in, that yield a return and wear out with use. The economists who built the theory of human capital around 1960 saw that a person's skills answer to exactly the same description. A year of schooling, a course of training, an apprenticeship: each is a cost borne now — in fees, in effort, in the wages forgone while learning — for the sake of a higher stream of earnings later. Skill is a thing one invests in and draws a return on, a stock of productive capacity lodged not in a factory but in a person. Set it down that way and the wage premium of the schooled needs no further explanation: it is the return on the capital they built, paid as the market pays the return on any other.
The theory proved unusually fertile, dividing and refining as good theories do. It distinguished the capital a worker builds that any employer could use from the kind useful only where it was learned, and drew from the distinction a whole account of who pays for training and who reaps it. It explained the shape of a working life — earnings rising steeply when young, when the investment is fresh and the years to recoup it many, then flattening with age. And it carried, like the marginal-productivity theory it serves, a moral hard not to find attractive: that the premium for skill is the cleanest desert there is. The skilled worker is paid more not because he cornered anything or out-waited anyone, but because he gave up something — years, ease, earnings — to make himself genuinely more useful, and the market merely returns what he laid out. If any wage is earned in the plainest sense, surely it is this one.
And then there is the colder reading, which grants every fact the hopeful one cites and explains them all another way. Suppose, said Michael Spence in 1973, that schooling taught nothing useful whatever — that it added not a particle to anyone's productive power. Might the schooled still be paid more? They might, and for a reason that survives the supposition intact. If schooling comes easier to the abler — less tedious, less likely to defeat them, cheaper in effort — then the able will acquire more of it than the rest; and an employer who cannot see ability directly, but can see the diploma, will rightly read the diploma as a sign of the ability behind it, and pay accordingly. The credential need teach nothing to be worth having. It need only be harder for the less able to obtain, and it sorts the workforce all the same. On this telling the premium is real, and rational to pursue, and earns the worker nothing he did not already possess: the schooling did not build the ability; it revealed it.
The unsettling part is how much of the evidence bends the cold way. If the wage paid for skill were truly a return on the capacity schooling builds, then years of schooling should pay smoothly, each adding its increment of capacity and its increment of wage. They do not. The years that end in a credential — the final year that earns the degree — pay far more than years that do not, a jump at the line of completion that the capacity stored in that one last year cannot begin to account for. Something is being paid for over and above the learning, and that something is the certificate itself, the sortable mark — exactly what the signaling story predicts and the human-capital story must scramble to explain away. The diploma, it seems, is worth more than the education it is supposed to stand for.
To settle which reading is right, one would need to know what the schooled would have earned had they not been schooled — and that can never be observed, for they were. This is the difficulty at the bottom of the whole subject, and it has a name: selection. The people who get more schooling are not drawn at random from the population; they are, on the average, the abler, the steadier, the better-raised and better-connected, and such people tend to earn more whatever their schooling. So the plain gap between the wages of the educated and the rest cannot be read as the work of education, for it folds together what the schooling did with who the schooled already were. The raw premium flatters schooling by crediting it with the prior advantages of those who pursue it.
Economists have gone to remarkable lengths to cut the tangle apart. They have compared identical twins who happened to get different amounts of schooling, holding the lottery of birth and upbringing fixed. They have seized on accidents that pushed some people through more schooling than others for reasons having nothing to do with their ability — a compulsory-attendance law, a wave of college-building, even the month of the year one was born. The labor of decades has yielded a fairly stable answer to one question: there is a real causal return to schooling, something like a tenth for each added year, large enough to refute the claim that education is purely a signal and builds nothing. But the same labor has failed at the harder task, which is to say how much of that return is the wage of a capacity genuinely created and how much the price of a credential that merely sorts. The honest verdict is the unsatisfying one: both are at work, the proportions varying by level and field and stubbornly resisting the knife.
There is, in any case, a prior confusion in speaking of what a skill is "worth," as though the figure were fixed in the skill itself. It is not; it is a price, and like every price it answers to scarcity. The premium the schooled command has not held steady — it narrowed through the middle of the last century, then widened sharply from around 1980, and neither move had anything to do with schooling suddenly building more or less capacity than before. What changed was the balance of supply and demand for skill. Technology, since the computer above all, has tended to lean on skilled labor and dispense with the unskilled, lifting the demand for skill year upon year; whether its price rises or falls turns on whether the supply of the schooled keeps up. When the colleges pour out graduates fast enough, the premium falls and the wage spread narrows; when they fall behind a quickening demand, as they have for decades, the premium climbs. The wage of skill is the prize in a race between education and technology — a scarcity rent that rises and falls with the relative numbers, and no fixed measure of what skill is worth.
And there is one more contest folded into the premium, the earliest and least visible of them, which decides who is in a position to earn it at all. The return to schooling can be collected only by those who reach the schooling, and the road to it is not equally open. It runs through the schools of one's childhood, which are good or bad by neighborhood; through a family's ability to feed and house a young adult who earns nothing for years; through the financing on offer, and the willingness to wager debt against an uncertain return; through a hundred advantages of birth that have nothing to do with the effort the premium is supposed to reward. So the skill premium, the most meritocratic-seeming of all the wage's parts, is also among the most efficient engines for turning the inequality of one generation into the inequality of the next: those who can afford to invest reap the return, and pass on the means to invest again.
Gather the readings and the premium for skill comes apart in the hand. Some of it is exactly what the hopeful theory says — a real return on a real investment in a capacity genuinely built, and the worker who made that investment did earn it. But some of it is a rent to an ability the schooling only certified, sorted out by a credential that taught little. Some of it is a scarcity rent, set by where the race between education and technology happens to stand, and worth more this decade than last for no change in the skill at all. And some of it is the quiet return to a starting advantage, collected by those who could afford to reach the schooling in the first place. The single number on the skilled worker's contract is all of these at once, and the theory that reads it as the pure wage of merit has simply declined to look at three of the four.
The wages of merit, on inspection, are partly the rent of scarcity, partly the price of a signal, and partly an inheritance.
Of all the components of the wage, the premium for skill wears the most respectable face. It is not the spoils of power, like the gap the single buyer keeps, nor the prize of combination, like the union's gain; it looks like the reward of foresight and effort, the wages of having made oneself useful — and in part it is precisely that, which is what makes it so much harder to see through. But what the skilled are paid is not the worth of their skill. It is a verdict, compounded of the productivity their schooling built, the ability it merely revealed, the scarcity of their kind that season, and the fortune of having been able to acquire it at all — a verdict mistaken, because it arrives looking like the just reward of merit, for the plainest desert there is. A price is a verdict on a hidden contest, mistaken for a fact of nature; the premium for skill is that verdict at its most flattering, the contest dressed as the wages of merit. The argument remains open.