The Second Theorem reconsidered

As the assumptions of the model have been altered, the Second Theorem may so far have appeared quite robust. It survives the introduction of extended preferences if they are assumed to satisfy the non-paternalist condition. Introduction of effort-aversion certainly destroys the optimality of the simple Rule, but not by automatic implication the Theorem: there is no obvious reason why it should not survive in a properly constructed model (which I do not provide). Introduction of risk is another matter; we lose not only the Rule but, in the absence of complete and costless information, the possibility of attaining First Best at all. It remains to consider increasing returns to scale, which, in the consideration of Lerner's Problem (ch. 5), I was careful to postpone. A sufficient reason was that a partial-equilibrium framework is inadequate to the treatment of this problem. The important result, which requires general-equilibrium analysis, is that, in the presence of non-convexities, the “divorce,” seemingly justified by the Second Theorem, between considerations of efficiency and of distribution, or equity, cannot be made. There is, of course, quite another reason for thinking this divorce impossible (which will not be further explored here). If the simple notion of the representative consumer cannot be employed (see again Blackorby, Davidson, and Schworm, 1991), then any Criterion Function employed to judge any change must aggregate preferences in a fashion that must depend on value judgments.