Crises: The Next Generation?

Paul Krugman March 2001

CRISES: THE NEXT GENERATION?*

When I first began working on the theory of currency crises in 1977, I imagined that it was a subject mainly of historical interest. The motivating events were the speculative attacks that brought down the Bretton Woods system in 1971 and the Smithsonian system in 1973. Given the end of fixed rates for major economies, it seemedunlikely that such events would recur. Of course, that’s not how it turned out. The fixed rates of Latin American nations offered a target for large speculative attacks in the runup to the debt crisis of the 1980s; the fixed rates of the European Monetary System offered targets for a wave of speculative attacks in 1992-3; and the more or less fixed rates of Asian and other developing nations offeredtargets for yet another round of attacks in 1997-8. Yet while the continuing relevance of the general idea of speculative attacks has justified the original theoretical interest in the subject, the actual models have not fared as well. When Eichengreen, Rose, and Wyplosz (1995) introduced the terminology of “first-generation” and “second-generation” crisis models, they also highlighted the somewhatdisheartening fact that each wave of crises seems to elicit a new style of model, one that makes sense of the crisis after the fact. And sure enough, the Asian crisis

*

Very rough draft prepared for Razin conference, Tel Aviv University, March 25-6, 2001 1

led to a proliferation of “third-generation” models, quite different from either the first or the second generation. (Likethird-generation mobile phone service, the third generation of crisis models has not yet quite lived up to its billing. Producers have not been able to agree on a common set of standards, and with the fading of the Asian financial crisis there is also question about whether we have a “killer ap.” But with recent financial news in Japan and the U.S., things may be looking up.) This paper represents a veryrough effort to get ahead of the curve, by asking what a “fourthgeneration” crisis model might look like. The main insight, if there is one, is that third-generation currency crisis models are actually not very specific to currency crises: the mechanisms for speculative attack and self-fulfilling pessimism that these models identify, while they do make room for an Asianstyle crisis in whichcapital flight leads to plunging currencies that validate the initial loss of confidence, also allow with small modification for other types of financial crisis. In particular, some third-generation crisis models are very close in spirit to the closed-economy “financial fragility” models of Bernanke and Gertler (1989). What this suggests is that a fourth-generation crisis model may not be a currencycrisis model at all; it may be a more general financial crisis model in which other asset prices play the starring role. Moreover, I will argue that even the open-economy aspects of third-generation models may not be all that crucial. It’s true that a simple story of financial collapse is easier to tell if one assumes that capital has someplace else to run to; otherwise, a loss of confidence leadsto a fall in the price of capital, which at first sight seems to rule out the kind of self-fulfilling loop that plays so central a role in many models. However, this need not always be the case. In particular, I will argue for a tie-in between the possibility of financial crisis and another one of my obsessions, the possibility of Japanese-style 2

liquidity traps. This paper is in fourparts. The first part briefly summarizes the evolution of currency crisis models, from first generation to third. The second part focuses on third-generation models, and in particular on what they say about policy during a crisis. The third part then offers a highly stylized open-economy fourth-generation model. The fourth part offers a loose translation of that model into closed-economy IS-LM-type...