The permit begins with a number the market has not supplied. So many tons may enter the air; the authority issues claims sufficient to cover them, and each emitter must surrender a claim for each counted ton. Then the claims may change hands. The plant that finds a cheap way to clean its exhaust can sell what it no longer needs to the plant whose next reduction would be dear. A common limit becomes a field of separate choices. The earlier Permit found its elegance here: a market deliberately made in an activity whose consequences the ordinary market had left outside its reckoning. Yet the elegance belongs to the allocation beneath the limit. It does not relieve anyone of deciding what that limit should be.
The invention joins two decisions that are often confused. One fixes the environmental requirement; the other finds a way to meet it. In W. David Montgomery's 1972 analysis, the requirement was given, and the problem was to achieve it at minimum joint cost. Competitive trading could do that under his assumptions, with licences designed to respect the relation between emissions and pollution at particular places. The argument developed the proposal associated with John H. Dales. It was not a theorem that any printed quantity of permits would make the world better, nor that a ton emitted anywhere must be interchangeable with a ton emitted elsewhere.
Imagine two plants, each initially emitting fifty tons, with a combined cap of sixty. Forty tons must therefore be removed. In this invented case the first plant's marginal cost of abatement equals the number of tons it has already cut; the second's is three times that number. Ordering each to cut twenty makes the first spend two hundred and the second six hundred. The last ton costs twenty at one plant and sixty at the other. That difference is an opportunity: shift a little abatement from the second plant to the first, and the same air, under the example's assumption that the emissions are interchangeable, is obtained with fewer resources. Equality of percentage cuts has no special claim to equality of economic sacrifice.
Trading carries the cuts towards thirty at the first plant and ten at the second. Both then face a marginal abatement cost of thirty. Total cost is four hundred and fifty plus one hundred and fifty: six hundred, two hundred less than under equal cuts. A permit price of thirty supports this allocation. Each plant compares the cost of removing another ton with the price of the allowance it would otherwise need, or could otherwise sell. If each starts with thirty allowances, the first emits twenty and sells ten; the second emits forty and buys ten. The three hundred paid between them divides the financial burden. It does not add three hundred to the resources spent cleaning the air.
A freely allocated allowance still has a cost when it is used: the sale that is forgone. That is why giving a plant its initial permits need not remove the incentive to abate. In the example, a ton avoided saves or releases a claim worth thirty regardless of whether the government sold or granted the original allocation. What the initial gift changes is wealth. An auction can bring the value of the claims into the public account; a grant leaves it with the recipients. This separation is sharp only within the stated setting. Financial constraints, market power, entry, or allocations conditional on future output can reconnect the distribution of permits with the decision to emit. A fixed initial grant in a competitive model should not be mistaken for every scheme that calls itself grandfathering.
The market can discover the cost of the cap. It cannot discover, from that price alone, whether the cap is right.
The price that emerges measures the scarcity of permission under the chosen rule. In the two-plant case it is the cost of the marginal required reduction. It is not automatically the damage another ton would do to the people breathing the air. For those numbers to coincide, the authority would have to choose the cap with that damage and the benefits of the activity properly in view. A permissive cap can leave the market efficient at achieving very little; a severe one can require costly reductions whose benefits do not justify them. The market supplies a method for allocating the obligation. The difficult judgment about the obligation itself remains where it began.
A tax of thirty on each emitted ton would produce the same thirty/ten abatement in this fully known example. The distinction appears when costs change after the policy has been set. With a fixed tax, a plant whose next cut becomes unexpectedly expensive can emit more and pay; with a binding cap and effective compliance, the total allowed quantity holds while the price of scarce allowances can rise. Weitzman's inquiry was which error would cost more. In his local, smooth model, with independent uncertainty in benefits and costs, the comparison depends on the relative slopes of marginal benefit from abatement and marginal abatement cost. A rapidly changing benefit can make departures from the planned quantity costly; a comparatively flat benefit can favour letting the quantity respond to cost. The rule compares two slopes, under specified uncertainty. Calling a pollutant a stock, or uttering the word catastrophe, does not by itself perform that comparison.
The ton itself also needs scrutiny. Suppose one plant's emissions contribute twice as much to a particular neighbourhood's concentration as the other's. Exchanging permission ton for ton can then worsen that neighbourhood's air while leaving the aggregate cap untouched. Montgomery's analysis gave this difficulty a place inside the market's design, distinguishing licences tied to pollution at receiving locations from unrestricted permissions to emit. The general lesson is physical: define what must be kept within a limit, then ask which trades preserve it. A market in undifferentiated tons is appropriate only to the extent that those tons are equivalent for the constraint being enforced. A ceiling on emissions is not a ceiling on every illness, every exposure or every local injustice they may cause.
The American acid-rain programme made such institutional design concrete. EPA's 1993 account of the rules described transferable sulfur-dioxide allowances, each covering a ton, and a requirement that utilities hold enough to cover their annual emissions. It also described banking unused allowances and stressed that an allowance did not excuse violation of federal ambient air-quality standards. The instrument was therefore a measured permission within a larger set of rules. Banking added time to the bargain: a claim saved from one year could support later compliance. Even the phrase the annual cap needs to be read with the provisions governing which years' claims may be used.
The choice between fixing a price and fixing a quantity is thus a choice about which adjustment the rule will permit, within a system able to count and enforce it. A cap does not make measurement unnecessary, and a tax does not make environmental judgment unnecessary. Either can share work with direct local limits; either can be revised as knowledge changes, at the cost of altering the expectations built upon it. The permit's achievement is to make one person's cheap reduction available to another without asking a regulator to name every plant's cheapest technique. The cap still has an author, the market still has rules, and the people exposed still have claims that the price of a traded certificate cannot settle for them.