The Costs Nobody Was Asked About

A factory makes a product, sells it at a price that covers its wages, materials and machines, and turns a profit. Downwind, a laundry finds its washing dirty, and spends money it would not otherwise spend. The laundry was not consulted, is not compensated, and does not appear anywhere in the factory's accounts or in the price the customer pays.

That is an externality: a cost or benefit falling on someone outside the transaction. The defining feature is not the harm (plenty of legal activity harms people) but the absence of the affected party from the deal. Because they are absent, their interests never enter the price, and the price is the signal everybody is acting on.

The consequence is a predictable error in quantity. Where the uncounted effects are costs, the price is too low and society gets too much of the activity. Where they are benefits, the price is too low for the producer to capture and society gets too little. Vaccination protects people who never receive it, an attractive building is enjoyed by everyone who passes, and basic research produces knowledge whose value its funder cannot possibly appropriate, which is the standard economic argument for public funding of research and public health, and it is an argument about uncaptured value rather than about generosity.

Arthur Pigou's answer in 1920 was direct: tax the activity by the amount of the damage. The purpose is not to stop it and not to raise revenue, both of which are side effects. It is to make the price tell the truth, after which each decision-maker facing the full cost will do the arithmetic themselves. Those who can abate cheaply will abate; those for whom the activity is genuinely worth more than the harm will continue and pay for it; and nobody has to decide centrally which is which.

The difficulty is empirical and it is severe. Setting a Pigouvian tax requires knowing the marginal damage, and for carbon that means a number which depends on climate sensitivity, on how much weight to give people who are not born yet, and on events a century away. Published estimates of the social cost of carbon span an order of magnitude, and most of the spread comes from the discount rate: a judgement about the future rather than a measurement of anything.

Forty years after Pigou, Ronald Coase reframed the problem in a way that is still argued about. Suppose the factory and the laundry could negotiate costlessly. Then, he showed, they will reach the efficient outcome whoever holds the legal right: if abatement is cheaper than the damage, the factory abates, paid by the laundry if the right to pollute is the factory's, at its own expense if it is not. The allocation of the right determines who is richer. It does not determine what happens.

Two things are usually made of this, and the first is a misreading. Coase is routinely cited as having shown that regulation is unnecessary. His paper argues close to the opposite: the bargaining result holds only under conditions that essentially never obtain (costless negotiation, perfectly specified rights, no strategic holdouts, a manageable number of parties) and since they do not obtain, the real question is which institutional arrangement copes best with the friction. Ten thousand households downwind of a power station will not be negotiating with it.

The second is genuinely illuminating and does survive. Coase insisted that harm is reciprocal. We instinctively describe the factory as harming the laundry, but the laundry's presence is equally what makes the smoke costly; shutting the factory harms the factory. Stated that way the question stops being who is the wrongdoer and becomes which arrangement produces the least total damage, which is a more tractable question even if it is a less satisfying one.

In practice the two live instruments are the tax and the tradable permit, and they are mirror images. A tax fixes the price of emitting and lets the quantity land wherever it lands. A cap fixes the quantity and lets the permit price move. Which you prefer depends on which uncertainty you would rather carry, and on politics, since a cap can be legislated as a target while a tax has to be defended as a tax. The characteristic failure of real schemes has not been theoretical at all: permits were over-allocated at the start, the price collapsed to near zero, and no abatement occurred.

For a case where all of this is visible at once, consider a road at rush hour. Every driver imposes a delay on everyone behind them, pays nothing for it, and experiences only the delay imposed by others. Road space at peak times is scarce and unpriced, which is the exact setup for over-use, and the intuitive remedy (build more) runs into induced demand: added capacity is substantially absorbed by trips that were not previously worth making, and congestion returns.

London and Stockholm both priced it. Traffic fell sharply in both. In both, public opinion was hostile beforehand and supportive afterwards. Stockholm ran the charge as a trial and then held a referendum, which passed. That sequence is common enough to be worth expecting whenever a price is put on something that used to be free: the cost is concrete and immediate, the benefit is diffuse until people have lived in it, and the argument is much easier to win after the fact than before.