In credit markets where borrower types are observable, is it welfare-maximizing for a rate-setting institution to offer different rates to different borrower types, or to pool them at a common rate? Moral hazard favours separation; deadweight default costs favour pooling, since compressing rates reduces aggregate defaults through hazard rate heterogeneity. We derive the condition determining which force dominates: the ratio of default cost intensity to moral hazard intensity. We show that there is a single threshold value of this ratio such that separation strictly dominates below it, complete pooling strictly dominates above it, and the two are welfare-equivalent exactly at it, for every possible value of the ratio. The result does not depend on information being scarce: even though borrower types are observable throughout, pooling can still dominate separation when default costs are sufficiently large, provided the maintained compression and effort conditions we state precisely continue to hold; we conjecture, but do not formally establish, that the same logic extends to imperfect information if the threshold, recomputed for that weaker information structure, continues to be exceeded. The welfare criterion is utilitarian surplus; the efficiency claim is surplus maximisation, not Pareto improvement.

This paper shows that under plausible assumptions, the inability of lenders to discover all of the relevant characteristics of borrowers results in investment in excess of the socially efficient level. Raising the rate of interest above the free market level will restore optimality. This conflicts with generally held views and is contrasted with the Stiglitz-Weiss model. It is shown that the assumptions which yield overinvestment support debt as the equilibrium method of finance. However, under the Stiglitz-Weiss assumptions, used to derive an underinvestment result, equity is shown to be the equilibrium method of finance.