The smoke has acquired a place in the account. Now the account asks how much to write beside it. The question is easy when the consequence is a bill for damaged cloth, harder when it is an illness that may occur, a landscape that may disappear or a loss falling a century hence. None of these consequences becomes unreal because it lacks a price. Yet the instrument proposed in the opening plate needs a number: a charge capable of making the private decision answer to the harm outside it. The shadow price attempts to supply the missing term. Its difficulty is that several judgments must pass through the same narrow opening before the consequence emerges as money.
The first judgment concerns what would actually change. A cleaner chimney can alter concentrations, exposure, the probability of illness and the conditions under which people work or rest. These are connected consequences, not interchangeable descriptions. An estimate of emissions avoided does not by itself establish the health gain; an association between a neighborhood and poor health does not by itself identify what closing one plant would prevent. The comparison needs a baseline, a causal account and a time horizon. Valuation begins after enough of that comparison has been specified to say what is being valued. Refining a monetary coefficient cannot repair a physical effect attributed to the wrong cause.
The troubling phrase value of a statistical life concerns small changes in mortality risk, not the price of an identified person. Suppose, in an invented example, one hundred thousand people would each pay eighty dollars for a reduction of one in one hundred thousand in their probability of dying over a specified period. Their payments sum to eight million dollars; the expected number of deaths avoided sums to one. Eight million is the implied value per expected fatality avoided in that comparison. It does not follow that any participant would accept eight million to die, or that the group knows whose life will be spared. The probability change and the population are essential to the meaning of the number.
The distinction makes the calculation intelligible without making every estimate sound. A trade observed in a wage or an answer elicited by a survey must be interpreted in light of the risk understood, the alternatives available and the people represented. Moving an estimate from one setting to another adds another judgment. DOT's 2021 guidance makes a separate policy choice: it assigns a single nationwide value in a given year rather than varying it with the affected population's income, age or other characteristics. The estimate draws on evidence; its application also embodies a rule about how people are to count.
Money also carries the circumstances of the person offering it. The same sacrifice can purchase less safety for someone whose income is already committed to necessities. The Willing brought this difficulty into the surplus account, and it survives the move from a market good to cleaner air. A technically careful measure of willingness to pay remains a money measure under particular resources and entitlements. Asking what someone would pay to obtain an improvement and asking what compensation would make its loss acceptable need not yield the same answer. Neither answer, merely by being measured accurately, decides the social weight to be given to that person's claim.
The social cost of carbon carries the calculation through time as well as across people. Its object is the additional harm associated with an additional emission, compared with an otherwise specified path. The analyst must follow the increment through the carbon cycle and climate, into consequences for people and production, and then into a present monetary value. An integrated assessment model makes those links explicit. That is its service and the source of its vulnerability: the final number inherits the assumptions and omissions of the links. Climate response, future exposure, adaptation, economic conditions, damage relationships and discounting all matter. A disagreement about the result cannot be assigned to ethics alone, nor can an ethical disagreement be dissolved by improving the physical model.
Discounting changes the weight of a distant consequence even when its physical description is held still. Suppose a certain loss of one thousand units of real consumption occurs exactly one hundred years from now, with no other consequence. At a constant annual discount rate of one percent, its present value is about 369.71; at three percent, about 52.03. These are invented rates and amounts, calculated with annual compounding. The loss has not become smaller in the second future. Its weight in today's comparison has become smaller. That makes the choice consequential, but does not mean that every component of the rate is simply impatience with the unborn. Expected growth, the value of consumption in different circumstances, opportunities forgone and the risk of the claim also enter the argument.
The distant future makes uncertainty about the rate consequential too. Averaging present values across possible paths need not give the present value obtained by inserting an average rate. Low-rate paths can retain much more weight at long horizons than high-rate paths. Arrow and his coauthors explain why persistent uncertainty can produce a declining certainty-equivalent discount rate; they also distinguish that case from uncertainty that is independently renewed each year. The conclusion depends on the process, not merely on attaching an error bar to a number. The useful discipline is to expose the assumptions, compare defensible alternatives and identify whether the decision changes with them. Sensitivity analysis shows where judgment matters. It does not select the judgment on the reader's behalf.
Some consequences resist credible monetization even after this work has been done. A habitat may be described with considerable ecological precision while the available valuation evidence remains too weak to assign its loss a defensible sum. Entering zero would convert the weakness of the evidence into a conclusion about the habitat. The account can instead retain physical and qualitative descriptions, distinguish reversible changes from losses that cannot readily be repaired, and show how large the missing benefit would have to be to change a ranking. Such a threshold is useful because it identifies the unresolved issue. It supplies neither the missing estimate nor permission to pretend the issue has been settled.
Choosing a protected outcome first changes the role of the price. A community might establish a limit because it judges a particular exposure unacceptable, then ask which means can meet the limit at least cost. The resulting shadow value records the cost of tightening that constraint; it need not equal a monetary estimate of every harm the constraint prevents. Conversely, a benefit-cost calculation may help compare policies without deciding that all rights and protections should be exchanged whenever the sum is favorable. These are different uses of economic information. Keeping them distinct permits the calculation to inform a decision whose legitimate purposes may be wider than maximizing the reported money total.
A number can expose a judgment without making the judgment disappear.
Who, then, is fit to set the numbers? The question divides into work that evidence can discipline and choices for which institutions must take responsibility. Analysts must make the causal chain, the population, the time horizon and the uncertainties inspectable. Public decision makers must explain the weights, protections and purposes under which the comparison will be used. People exposed to the harm must be able to contest both the description of their circumstances and the treatment of their claims. No single coefficient can discharge all those responsibilities. The shadow price is most useful when it illuminates the decision that produced it—and remains open to correction by the people and consequences it was meant to bring into view.