The Difference

Two plates of this volume have been about why the ceiling differs from worker to worker. The plate on the margin set the ceiling at the value of what a worker produces; the plate on skill explained why that value is higher for some — more training, more talent, scarcer abilities. Both locate the difference in the worker. But there is a third reason wages differ that has nothing to do with the worker at all, and everything to do with the job.

Consider two workers of identical skill, equally productive, who could do each other's work. One cleans offices at night; the other guts carcasses on a slaughterhouse line. One files papers in a warm room; the other roofs houses in January. If the wage were worth alone, and the workers are equally worth, they should be paid the same. Often they are not — and the theory of this plate says they should not be. The unpleasant, dangerous, or precarious job must pay more, not because the work is worth more, but to persuade anyone to do it at all. The difference in the wage compensates for the difference in the job. This is the compensating wage differential, and it is the oldest idea in the economics of wages.¹

Adam Smith set it out in 1776, in a chapter on why wages differ across employments, and he did it so thoroughly that little has been added since. The “whole of the advantages and disadvantages of the different employments of labour,” he wrote, tend in a free market to equality — not the wages, but the whole bundle, money and everything else together. Where a job carries some disadvantage, the money wage must rise to offset it; where it carries some hidden advantage, the money wage may fall. He named five such circumstances.²

The first circumstance is the agreeableness of the work — whether it is clean or filthy, safe or dangerous, honourable or despised. The dirty and dangerous trades must pay more; so, Smith noted, must the dishonourable ones, which is why the public executioner was, for the little he did, better paid than almost any common trade. The second is the cost of learning the work — the long, expensive training skilled trades demand, which the wage must repay; this is the human-capital story the plate on skill has already told, and Smith had it first. The third is the constancy of the employment — steady work or broken. A mason, idle through every frost and downpour, must earn more on the days he works to carry him through the days he cannot, so that irregular work pays a higher rate for the same skill. The fourth is the trust reposed in the worker — we pay the goldsmith, the physician, the lawyer partly for the responsibility we place in their hands and the harm they could do us. And the fifth is the probability of success — the lottery trades, law and the arts and the like, where a few succeed grandly and many fail, must dangle high rewards before the winners to draw entrants at all. Though here, Smith shrewdly observed, the compensation is usually too small, because people wildly overrate their chances; the “overweening conceit which the greater part of men have of their own abilities” leads them to crowd the long-odds professions and bid the reward down.³

Modern economics took Smith's insight and gave it sharper form. A job is a bundle — a wage, yes, but also a set of conditions: its safety, its hours, its pleasantness, its security, its standing. Workers care about all of these, and will trade between them: give up some wage for a safer or pleasanter job, or demand more wage for a worse one. Firms, for their part, can often supply better conditions at a cost — guard rails and ventilation and shorter shifts are not free. The labour market, putting the two together, ends up pricing each condition: there emerges, implicitly, a going rate at which wage can be exchanged for safety, for comfort, for security. The wage attached to any particular job is then the sum of two things — what the worker's product is worth, and a premium or discount for the bundle of conditions the job carries.⁴

This is the hedonic theory of wages, and its power is that it makes the implicit explicit. We never see a price tag on “safety” or “pleasantness.” But by comparing the wages of otherwise-similar workers in jobs that differ in some condition, we can read off the price the market places on that condition — the wage discount workers accept for a safer job, the premium they demand for a riskier one. The market has been quietly pricing the unpriceable all along, and the wage differential is the price.⁵

There is one condition whose pricing is so consequential, and so unsettling, that it deserves a section of its own: the risk of death. Some jobs are likelier to kill you than others — logging, fishing, roofing, mining. By the logic just laid out, such jobs must pay a premium, the extra wage that persuades a worker to accept the extra risk. And that premium, once measured, lets economists do something startling: put a price on a human life.

The arithmetic is simple and chilling. Suppose workers must be paid an extra few hundred dollars a year to accept a job carrying one additional chance in ten thousand of dying that year. Then ten thousand such workers, taken together, are accepting a few hundred dollars each — a few million in total — in exchange for one expected death among them. That total is what the market reveals these workers, by their own choices, implicitly charge for one statistical life. It is not the value of any particular person, who would not sell their life for any sum; it is the rate at which people trade small risks of death for money, scaled up to a whole life. Economists call it the Value of a Statistical Life, and have measured it, from wage-risk tradeoffs and other choices, at something around thirteen million dollars in the United States today — a figure that has drifted upward over the years as the country has grown richer.⁶

This number is no academic curiosity. Governments use it, by the billions, to decide how much to spend on safety — whether a guard rail, a regulation, a recall is “worth it,” by weighing its cost against the statistical lives it saves at thirteen million each. The practice provokes understandable revulsion: it feels monstrous to price a life, and monstrous to refuse a safety measure because it costs too much per life saved. But the alternative — spending without limit to drive every risk to zero — is impossible, and pretending we do not trade safety against cost is a lie we tell with every road we build and every regulation we decline. The compensating differential, the humble premium for a dangerous job, turns out to be the place where a society reads its own price on death.⁷

So the theory is elegant, the evidence for it real, and its boldest application — pricing a life — genuinely in use. And yet anyone who looks at the actual world of work feels a deep unease, because the theory makes a clear prediction that the world routinely violates. The compensating differential says bad jobs should pay more. Look around: bad jobs very often pay less.

The jobs that are dirtiest, most dangerous, most precarious, most despised — the slaughterhouse line, the field labour, the care work, the gig delivery in all weathers — are disproportionately the worst-paid jobs in the economy, not the best. The roofer may earn more than the file clerk; but the migrant farmworker, bent in the heat for piece rates, earns less than the office worker in the cool, though the field is incomparably harder and more dangerous. The correlation the theory predicts — worse job, higher pay — is, across great swaths of the labour market, simply inverted. Far from being compensated for the badness of their jobs, the workers in the worst jobs are, on the whole, paid the least.⁸

Three things break the theory, and they nest inside one another. The first is the simplest: “all else equal” almost never holds. The compensating differential operates between jobs alike in everything but the condition in question. But the worst jobs differ from the best in more than their badness — above all, in the skill they require. The dangerous, dirty, precarious jobs are disproportionately low-skill jobs, and so the downward pull of low skill, the fourth plate's story, swamps the upward push of the compensating premium. The premium is still there, hidden inside the wage: the dangerous low-skill job may pay a little more than a safe low-skill one. But compared across the whole market, where skill varies enormously, the skill effect dominates, and the bad jobs come out at the bottom. The differential is real, but partial, and masked.⁹

The second is information. A worker demands a premium for a danger only if they know the danger is there. Many of the worst risks of work are invisible — the slow poison of a chemical, the long damage of a repetitive strain, the cancer that surfaces decades later. Where the risk is hidden, no premium is demanded, and the worker bears it unpaid. This is part of why we do not leave safety wholly to the wage: the market can price only the dangers workers can see, and much of the harm of bad work is precisely the harm they cannot.¹⁰

The third reason is the deepest, and it is the volume's. A compensating differential is a premium paid to induce a worker to accept a disadvantage they would otherwise refuse. It exists only because the worker can refuse — can take the safer, cleaner, steadier job instead, forcing the bad job to bid for them. But strip the worker of that power — of alternatives, of savings, of the safety net the last plate described — and the premium evaporates. A worker who must take this job or go hungry does not need to be compensated for its dangers; they will take it, dangers and all, at whatever it pays, because the alternative is worse. The compensating differential is a luxury of the worker who has somewhere else to go. For the worker who does not, the bad job is not a bargain to be sweetened but a necessity to be endured — and its badness, far from lifting the wage, sits perfectly comfortably beside a low one.¹¹

Gather the threads. The compensating differential began as a worth-side idea — a third reason, beside the marginal product and skill, that wages differ. But it turns out to sit uneasily on the worth side, for two reasons that bear on the whole volume.

The first is that the compensating differential is the clearest case in which the wage is not about the worker's product at all. The premium for danger, the discount for pleasantness — these track no contribution the worker makes; they price the qualities of the job. Even on the side of the ledger the volume has called worth — the side where the wage is supposed to reflect the value of the work — a large part of the variation in pay reflects not what the worker is worth but what the worker must endure. The wage is, even here, only loosely tethered to product.

The second is the one this plate has driven toward. The compensating differential, the most worth-like of mechanisms, works only when the worker has power — the power to refuse the bad job and take a better one. Where that power is present, the theory holds and the bad job pays its premium. Where it is absent — where the worker has no alternative, no cushion, no floor beneath them — the premium vanishes and the bad job pays badly. So the pattern we observe, in which the worst jobs so often pay the least, is not a refutation of economics but a map of power: the premium appears exactly where workers can command it, and disappears exactly where they cannot. The dirty, dangerous, ill-paid job is the signature of a worker who could not say no.¹²

This is why the differential belongs in a volume about power, though it wears the dress of worth. The bargaining range is still the frame. The job's badness, like the worker's skill, helps set the ceiling — what the match is worth, all its conditions reckoned in. But whether the worker actually captures the premium that badness should command depends, as everything in this volume depends, on where in the range the wage finally falls — and that is decided by power. Compensation is what worth would pay; what the worker receives is what power can take.¹³