Andrew Atkeson | Macro Paper Warehouse

International Lending with Moral Hazard and Risk of Repudiation

This 1991 Econometrica paper by Andrew Atkeson examines optimal international lending when a sovereign borrower can both repudiate its debt and privately choose whether to invest or consume borrowed funds (moral hazard), and shows that recurrent debt crises – involuntary net transfers of capital from borrower to lender precisely when the borrower’s output is low – are a necessary feature of the constrained-optimal lending contract, not a market failure. Atkeson first establishes, following Worrall, that repudiation risk alone (with state-contingent debt contracts) is not a sufficient friction to generate recurring crises: when output is exogenous and fully observable, the optimal repudiation-proof contract achieves complete consumption smoothing for the borrower in finite time. Once the borrower’s investment decision is private information, however, lenders must condition future credit terms on realized output as a (noisy) signal of whether the borrower actually invested, and Atkeson shows – using a generalization of Abreu-Pearce-Stacchetti’s self-generation and factorization techniques to an environment with a persistent state variable – that the optimal incentive scheme punishes suspected underinvestment through large current debt-service payments rather than through a cutoff in new lending, which forces the borrower to make a net capital outflow to lenders whenever output is low. The source text is the November 1987 Federal Reserve Bank of Minneapolis working-paper draft that preceded the 1991 Econometrica publication; it contains the paper’s full theoretical apparatus (Sections 1-5 and the appendix) but its worked numerical example of an economy displaying debt crises is marked ‘Section 6 (Forthcoming)’ and was not yet written at the time this draft circulated, so this record’s grounding is confined to the paper’s proven propositions rather than its published numerical illustration.

Money and Exchange Rates in the Grossman-Weiss-Rotemberg Model

This 1997 Journal of Monetary Economics paper by Fernando Alvarez and Andrew Atkeson asks whether the segmented-markets model of money pioneered by Grossman and Weiss (1983) and Rotemberg (1984, 1985) – in which only infrequent, staggered visitors to the asset market absorb each monetary injection – can reproduce three salient features of post-Bretton-Woods exchange rates: that nominal exchange rates are far more volatile than inflation, money growth, or interest-rate differentials; that changes in nominal and real exchange rates are highly correlated and similarly sized; and that both are highly persistent. Using Lucas’s (1990) device of organizing agents into extended families to handle wealth effects, the authors solve the model in closed form under log utility and show analytically that a monetary injection has a disproportionate impact on the currency holdings, consumption, and asset prices of whichever small fraction of agents happens to be transacting in the asset market that period. A calibrated numerical example shows the model can match the data’s roughly 37% annualized volatility of real exchange-rate changes and near-unit correlation between nominal and real exchange-rate movements, but only when money growth is persistent and not highly correlated across countries and when the share of agents in the asset market at any time is calibrated to be very small (about 3%); the authors contrast this with related Lucas/Fuerst-style liquidity models (including their own earlier cash-in-advance work, Grilli and Roubini 1992), which cannot generate exchange-rate volatility much beyond that of interest-rate differentials.

Money, Interest Rates, and Exchange Rates with Endogenously Segmented Markets

Alvarez, Atkeson, and Kehoe build a cash-in-advance model in which households face a fixed real cost to transfer funds between a goods market and an asset market, so that in any period only a fraction of agents are “active” traders; the rest hold their real balances passively. Because a central bank’s open-market money injection is absorbed only by the currently active traders, it moves their consumption and marginal utility directly – a “segmentation effect” on real interest rates that is distinct from, and can outweigh, the usual expected-inflation (Fisherian) effect. With moderate amounts of this endogenous segmentation, the model generates, from pencil-and-paper analytics, three patterns that standard cash-in-advance models cannot: a negative relation between expected inflation and real interest rates, persistent (not just one-period) liquidity effects and a steepening-then-reverting yield curve after a money shock, and – in a two-country extension – exchange-rate behavior that differs with a country’s inflation rate much as in the data. Because higher steady-state inflation makes agents trade more often, the model also implies that segmentation (and hence these effects) weakens as average inflation rises, which the authors show matches the empirical pattern that real exchange rates are relatively more volatile in low-inflation countries and relatively much less volatile than nominal exchange rates in high-inflation countries.