Efficiency Through Rawls's Distributive Lens — Epoche C1
When an economist tells a court or a legislature that a legal rule is "efficient", the word carries one of two precise technical meanings, and neither of them says anything about who ends up with what. This essay takes those two meanings apart, locates where distribution drops out of them, and asks what John Rawls's A Theory of Justice (1971) puts in its place. The answer is not, as the phrase "distributive efficiency" suggests, a rival conception of efficiency. It is something more useful and more awkward: a rule for choosing among the arrangements that the economists' criteria certify as equally good, together with an argument that the choice cannot be postponed. The two standard criteria, stated exactly Begin with the criteria themselves, because the whole argument turns on their precise form. The first is Pareto efficiency , the notion Vilfredo Pareto set out in his Manuel d'économie politique (1909): a change from one allocation of resources to another is a Pareto improvement if at least one person is made better off — by their own ranking of the outcomes, not an observer's — and nobody is made worse off. An allocation is Pareto-efficient when no such improvement remains available. The criterion's appeal is that it asks for no comparison of one person's gain against another's loss, and so avoids interpersonal comparisons of welfare, which economists in the 1930s had come to regard as scientifically ungrounded. Its practical weakness follows from its form. Because the test requires unanimity — literally nobody worse off — a single loser disqualifies a change, and essentially every real reform has losers: the incumbent whose licence is withdrawn, the worker whose plant closes. A criterion that cannot rank most real reforms is not much of a criterion. The second criterion was devised to repair precisely this. Nicholas Kaldor (1939) proposed that a change counts as an improvement if the winners could compensate the losers out of their gains and still be better off; John Hicks (1939), in the same journal that year, proposed the mirror-image test, asking whether the losers could profitably bribe the winners to forgo the change. The pair is now known as the Kaldor-Hicks criterion, or a potential Pareto improvement. The crucial word is "potential". Kaldor was explicit that the compensation need not be paid: the economist's task, on his view, is to report that a change would enlarge the total, and whether the gain should be shared is a political question outside the economist's competence. That division of labour is the origin of the ethical oversight this essay is about, and it was a considered methodological choice rather than a lapse. In applied work the criterion becomes cost-benefit analysis. Each affected person's gain or loss is expressed as a compensating variation — the sum of money which, taken from a winner or given to a loser, would leave that person exactly as well off as before the change — and the change is approved if these sum to a positive number, $\sum_{i=1}^{n} \mathrm{CV}_i > 0$. Distribution has now been compressed into a single scalar, and the persons behind the terms have vanished from it. Why "efficiency first, distribution later" looked respectable The separation Kaldor recommended was not mere convenience; it had a theoretical warrant, and that warrant is the strongest thing that can be said for the mainstream position. Two results of general-equilibrium theory, standardly called the fundamental theorems of welfare economics, frame it. The first states that under a demanding set of conditions — every good is traded on a market, all participants take prices as given, there are no external effects unpriced by those markets, and preferences are locally non-satiated — a competitive equilibrium is Pareto-efficient. The second runs the other way: given convex preferences and production possibilities, any Pareto-efficient allocation whatever can be reached as a competitive equilibrium, provided the initial endowments are first redistributed by lump-sum transfers, that is, transfers whose size does not depend on anything the recipient can alter. The second theorem licenses the division of labour: if every efficient allocation is attainable by markets once endowments are set, the choice of which one society reaches is a separate decision taken with a separate instrument — the tax and transfer system — and market rules may be designed for efficiency alone without prejudging it. That is the picture in which efficiency and equity are independent dials, and the lump-sum condition is where it fails. A transfer is lump-sum only if its size cannot be altered by the taxpayer's behaviour, which means levying it on unalterable characteristics such as innate earning capacity; governments cannot observe earning capacity, only income, a joint product of capacity and effort. Every real tax therefore falls partly on effort and alters the labour supply it taxes, so the separation of efficiency from distribution is exact in the theorem and approximate at best in practice. Why the compensation test is not distribution-neutral even on its own terms One further point must be made before turning to Rawls: the mainstream criterion is not merely silent about distribution but quietly dependent on it. Compensating variations are sums of money, and how much money a person will pay to secure or avoid an outcome depends on the prices they face and on how much they already have. Tibor Scitovsky (1941) showed that this can make the test approve a move from state $x$ to state $y$ and, evaluated afresh in $y$, also approve the move back from $y$ to $x$: the criterion is not even consistent as a ranking, and his repair — require the Kaldor test to pass in one direction and fail in the other — is itself evidence that the underlying measure moves with the distribution. The ethical consequence is sharper. Willingness to pay is bounded by ability to pay, so a wealthy household's compensating variation for clea