Reconstructing Market Failure as Institutional Injustice — Epoche C2
The claim, and the theorem it has to get past The standard catalogue of market failures — externalities, public goods, asymmetric information, natural monopoly — is taught as a list of exceptions to an otherwise well-functioning mechanism, and this essay argues that it is better read as a list of places where the law has already made a distributive choice and declined to describe it as one. The compressed version of this argument asserted that conclusion without confronting the results that make the orthodox framing seem compulsory. It also opened with a misattribution that needs correcting at once, because the correction is the hinge of the whole case: it presented Amartya Sen's Collective Choice and Social Welfare (1970) as a seminal statement of the paradigm that ties market failure to Paretian efficiency. That book contains, in its starred sixth chapter, the theorem which does more than any other to show that the Pareto criterion cannot serve as a neutral baseline. Sen is not the paradigm here. He is the objection to it. What licenses the orthodox framing is a pair of theorems established in the axiomatic form given by Debreu in 1959. The first states that if consumers' preferences are locally non-satiated — for any bundle there is an arbitrarily near bundle strictly preferred, which rules out thick indifference bands and satiation — and if a complete set of competitive markets exists so that every commodity relevant to anyone has a price, then any allocation supported by a price system at which all markets clear is Pareto optimal: there is no alternative feasible allocation that every agent weakly prefers and some agent strictly prefers. Marginal-cost pricing, the condition $P = MC$ that price equals the additional cost of producing one more unit, is simply the first-order condition of this equilibrium for a producer, which is why departures from it are the traditional signature of failure. The first theorem is silent about who gets what. An allocation in which one household holds the entire endowment and everyone else starves is Pareto optimal, since no reallocation can improve anyone's position without worsening that household's. This is not a criticism of the theorem; it is what the theorem says. What conceals the silence is the second theorem, and this is where the trouble lies. What the second welfare theorem demands, and why it cannot be supplied The second theorem states that if preferences and production sets are additionally convex — averages are at least as good as extremes, and there are no increasing returns — then every Pareto optimum can be realised as a competitive equilibrium, provided the initial endowments are first redistributed by lump-sum transfers. This is the theorem that licenses the working division of labour in policy economics: let the market handle efficiency, let the tax system handle distribution, and treat the two as separable problems. The licence depends entirely on the phrase "lump-sum". A lump-sum transfer is by definition one whose magnitude depends on nothing the recipient chooses; that is what makes it non-distorting, since no behaviour alters it and so no behaviour is altered by it. But to redistribute towards those with low earning capacity a government must condition transfers on earning capacity, and earning capacity is not observable. What is observable is earnings, which are a choice; a transfer conditioned on earnings is an income tax, and an income tax changes the choice. The instrument the theorem requires therefore does not exist, and it fails to exist for a structural rather than an administrative reason. So the separation of efficiency from distribution is not a theorem but a hope. Every actual redistributive instrument operates through the same margins as the allocation itself, so every choice of instrument is simultaneously a choice about allocation, and the description of an externality as a technical deviation to be corrected while leaving the underlying arrangement untouched has lost its warrant. There is no untouched underlying arrangement. Why the catalogue is not a list of exceptions That is an argument about hypotheses. There is also a result showing that the hypotheses fail almost everywhere, and it is stronger than the textbook treatment suggests. Greenwald and Stiglitz proved in 1986 that in economies with imperfect information or an incomplete set of markets, competitive equilibria are generically constrained Pareto inefficient. Two words carry the weight. "Constrained" means the comparison is not with an omniscient planner but with an intervention that faces exactly the same informational limits as the market participants: a set of small taxes and subsidies, using no information the market does not have, can make everyone better off. This blocks the standard reply that the inefficiency is unavoidable because the information simply is not there. "Generically" means the property holds for all economies in the relevant parameter space except a set of measure zero — so efficiency is the knife-edge case, not the rule. The mechanism is pecuniary. In a complete-market world, price changes caused by one agent's action are pure transfers and cancel in the aggregate; where markets are incomplete or information is asymmetric, prices also carry information and determine who transacts at all, so a price change alters real opportunities and no longer cancels. Since no actual economy has a complete set of contingent markets — there is no market in which one may today buy wheat conditional on a specific harvest in 2040 — the theorem's antecedent is always satisfied. The catalogue of four failures thus has the topology backwards. Externality, public good, information asymmetry and natural monopoly are not four holes in an otherwise sound floor; they are four especially legible instances of a condition that obtains everywhere. Coase: the externality is a choice of rule The previous section shows the orthodox framing to be empirically mis