The wage pays for an hour, but the hour comes attached to a life. There is the hour spent in daylight and the hour spent while the household sleeps; the hour in which the work is safe and the hour in which an injury is possible; the reliable week and the week whose last three days may disappear. Two workers can bring the same skill to different bundles of conditions, and require different wages before they will agree to them. Adam Smith's theory of compensating differences begins with this ordinary recognition and gives it an extraordinary consequence: where workers can choose, an unpleasant employment must offer something for its unpleasantness. The money wage can differ precisely because the whole advantage is being equalized. Yet the worker enduring miserable conditions for little pay remains before us. Understanding the theory requires asking which alternative would have made that employer offer more, and whether the worker ever had it.
Smith named five circumstances that could make the money rewards of employments unequal. The work could be agreeable or disagreeable, easy or costly to learn, steady or intermittent, entrusted with little or much responsibility, likely or unlikely to yield a successful career. The list is broader than a premium for danger, and its parts should not be mistaken for five disadvantages faced by otherwise identical workers. The cost of learning a trade belongs also to the preceding plate on skill. The mason's higher daily rate has to be read against the days weather prevents him working. The standing of an occupation can attract people at a lower money reward, while a despised one may need to pay them to enter. Smith's decisive enlargement is to put the rest of the employment beside the money. A high rate for the day worked may contain the cost of the day that will not be worked; a smaller payment may accompany an advantage the worker values enough to purchase.
The freedom required by the argument is more demanding than the freedom to put one's name on an application. Smith's own qualifications include employments sufficiently familiar and established, in their ordinary condition, and serving as the worker's principal livelihood. He also sees how the hope of distinction can crowd people into careers whose expected money reward is poor, and how youthful confidence can discount a danger. The worker choosing the bundle chooses what he believes it contains. A promise of glory can draw labor just as surely as a payment, even when most entrants will never collect the promised distinction. The theory therefore starts with conditions on knowledge and choice already inside it. To read Smith as promising that each burden will always receive an adequate addition to the wage would remove the very passages in which he asks how the chooser comes to value the burden.
The modern theory gives this bundle a price schedule of its own. A worker weighs more money against a less desirable condition; an employer weighs the expense of improving that condition against the wages needed to recruit without the improvement. Safety may require a guard, a different process, a slower rate of production. If providing it cheaply saves enough on recruitment, the employer has a reason to provide it. Workers differ in how much money they require to bear a risk, and firms differ in what it costs to remove one. The market brings these differences together in the jobs it offers. At an interior choice, the extra compensation for a small increase in risk matches the choosing worker's valuation of that increment and the corresponding saving per worker to the choosing firm. The resulting wage is a joint settlement of the package. It need not be constructed by taking an independently fixed productivity wage and adding an equally independent charge for danger.
A premium within that market can coexist with a low position in the wage table. Take four invented offers. For workers able to enter one set of jobs, a safe post pays thirty thousand dollars a year and a more dangerous one thirty-two thousand. In a different set, requiring qualifications the first group does not possess, the corresponding offers are sixty thousand and sixty-two thousand. Within either set, the dangerous job pays two thousand more. Across the sets, the dangerous job paying thirty-two thousand still pays twenty-eight thousand less than the safe job paying sixty thousand. Nothing in these numbers contradicts a compensating differential. Nothing establishes that the two thousand is adequate, either. The example separates the comparison the theory proposes from the comparison an occupational table may happen to make. The burden can have raised a wage that remains low for other reasons; to infer the absence of compensation from the low total alone is to lose the comparison that would identify it.
Differences in the workers' preferences complicate even a comparison that holds qualifications steady. Someone who is especially averse to a danger may choose the safer job; someone who minds it less may take the premium. The observed wage schedule records both choices. Its slope at the second worker's chosen job can identify the compensation that worker requires for a small additional risk, under the model's conditions. It does not tell us what the first worker would demand to move there, or what either would demand for a much larger change. The line through the market is assembled from people with different valuations and employers with different technologies. Treating it as one person's indifference curve would silently give every point the same chooser. An elegant market relation can conceal a great deal of human difference without having eliminated any of it.
The smallest probability change can nevertheless produce one of the largest-looking numbers in economics. Suppose, purely for illustration, that six hundred dollars a year compensates a worker for an additional annual chance of death of one in five thousand. Across five thousand comparable workers, the additional pay totals three million dollars and the additional expected deaths total one. The ratio is a value of a statistical life in this example. No person has agreed to sell a life for three million dollars. The choice concerned a small change in the chance of survival; multiplying the probability across many people makes an expected count, not an identified victim. That distinction is the entire justification for the calculation and the limit on what it can mean. The number can help compare small safety improvements with their cost while saying nothing about a bargain for certain death.
The precision of the division should not conceal the circumstances of the people whose choices supplied its numerator. The money needed to induce a small risk change is drawn from an offer made at a particular income, within a particular set of opportunities, to someone with particular responsibilities and preferences. Moving that estimate into a public decision requires judgment about the people and risks to which it will be applied. Equal concern for two lives does not imply that two people with different resources will make identical monetary tradeoffs. Nor does a smaller observed payment establish that the person receiving it has a life of less worth. A price of risk can be useful without becoming a hierarchy of persons. The temptation to let the measured figure settle that further question is the same temptation this table keeps finding: to treat a market's answer to the choice available as a complete answer to what ought to be valued.
A danger can enter the employment without entering the worker's comparison at all. Imagine two jobs that appear equally safe, with one concealing a hazard its applicant has no reason to suspect. The worker's decision between them cannot reveal a valuation of that undisclosed difference. The hazard may later become known, at which point an opportunity to leave or negotiate could change the terms; it may remain unknown until after harm is done. Either way, the fact that someone took the job does not establish that the risk was knowingly priced. Information is part of the mechanism by which a burden becomes a demand for compensation. Where that mechanism is missing, an observer cannot restore it by interpreting a signature as proof of a tradeoff the signer never had occasion to make.
An alternative can likewise exist in the economy without existing for the worker. A safer job may require qualifications not yet acquired, a move the household cannot finance, or a wait the worker cannot afford. These restrictions matter to what an employer must offer to attract and retain labor. They can leave both the wage and the conditions worse than they would be with readily reachable rivals. They do not imply that every premium must vanish: a constrained worker can still prefer one reachable job to another and demand something for a burden between them. The conclusion must follow the actual alternatives. A low wage with bad conditions could reflect a positive premium on a very low starting point, a limited ability to bargain, a misunderstood risk, or several of these together. The observed combination calls for investigation; it does not identify its own cause.
The hardest question is therefore what it would mean to say that the worker has been compensated. One inquiry asks whether the burden raised the pay relative to a credible alternative with the burden removed. Another asks whether the worker understood the burden and had a meaningful opportunity to choose otherwise. A third asks whether the resulting terms meet the standard society is prepared to permit. These questions are connected, but a favorable answer to the first does not dispose of the others. Nor does an unfavorable answer emerge simply from comparing an arduous low-paid occupation with an easier high-paid one. The discipline of the differential is to keep the whole bundle in sight and the relevant comparison within reach. Without the first, one mistakes hardship pay for prosperity; without the second, one mistakes differences between people for the price of a difference between jobs.
The wage can contain a payment for hardship without containing enough to make the hardship bearable.
Smith's insight survives with all of its force once its conditions are restored. A worker who can choose between comparable employments gives a disagreeable job a reason to pay more, and gives an agreeable one something it can offer besides money. That force can operate within an economy whose worst work remains badly paid; it can also be obstructed by ignorance, exclusion, and the cost of reaching an alternative. The original question presses upon both possibilities. To answer it, the investigator must recover the work concealed by the wage and the choices concealed by the work. A price is a verdict on a hidden contest, mistaken for a fact of nature; the compensating difference is the part of that verdict concerned with what the worker bears, and the absence of an obvious premium is a beginning for inquiry rather than its end. The next plate asks why wages differ when the work and qualifications are held steady and the employer's reading of the person changes. The argument remains open.