The first generation models of currency crises have often been criticized because they predict that, in the absence of very large triggering shocks, currency attacks should be predictable and lead to small devaluations. This paper shows that these features of first generation models are not robust to the inclusion of private information. In particular, this paper analyzes a generalization of the Krugman–Flood–Garber (KFG) model, which relaxes the assumption that all consumers are perfectly informed about the level of fundamentals. In this environment, the KFG equilibrium of zero devaluation is only one of many possible equilibria. In all the other equilibria, the lack of perfect information delays the attack on the currency past the point at which the shadow exchange rate equals the peg, giving rise to unpredictable and discrete devaluations.
Currency crises are characterized by two seemingly contradictory features. On the one hand, currency crises are usually “large,” in that they involve massive asset reallocations, wild swings in asset prices, and heavy output losses. On the other hand, currency crises are often triggered by shocks that seem too small to account for these effects. Although these characteristics of crises might suggest some form of irrationality, the literature has provided two types of models that account for some of these features in an environment in which agents are rational.1
First generation models of currency crises, starting with Krugman (1979) and Flood and Garber (1984a), view crises as arising from inconsistent policies; in particular, monetization of fiscal deficits together with fixed exchange rates. In these models, the drop in demand for real balances at the time of the crisis leads to a discrete drop in reserves at the Central Bank, even in the absence of a corresponding discrete deterioration in fundamentals. Although these models account for large “attacks,” they have the counterfactual implication that these attacks should be predictable and, consequently, lead to no discrete changes in the exchange rate or other asset prices.2
Second generation models of currency crises, starting with Obstfeld (1986), account for the unpredictability of crises by assuming the existence of multiple equilibria.3 In these models, consumers’ expectations can be self-fulfilling because they affect the Central Bank's decision of whether to devalue. One drawback of these models is that they have little to say about the dynamics and timing of crises, as these depend on unmodeled expectational dynamics. Furthermore, Morris and Shin (1998) show that the existence of multiple equilibria in second generation models may not be robust to the inclusion of private information. Morris and Shin present a canonical second generation model and show that, when consumers have private information about the level of fundamentals, the model has a single equilibrium. As a result, in their environment the indeterminacy on which second generation models depend to account for unpredictable crises and discrete drops in asset prices disappears.4
This paper proposes a first generation model of currency crises that accounts not only for the attack on Central Bank reserves, but also for the suddenness, unpredictability, and discrete devaluations associated with currency crises. The only difference between the model proposed in this paper and previous first generation models is that I assume that not all consumers are perfectly informed about the level of fundamentals. In particular, I analyze a generalization of the Krugman–Flood–Garber (KFG) model, in which only a subset of informed consumers knows the level of Central Bank reserves at which the peg will be abandoned, while other, uninformed, consumers have imperfect knowledge about this threshold level of reserves. In this environment, the timing of the crisis is determined by the interplay between the learning process by uninformed consumers in the run up to the crisis, their resulting portfolio reallocation, and their effect on the timing of the attack by informed consumers. As in previous first generation models, when all consumers are informed the attack is predictable and its timing is simply determined by the condition that the shadow exchange rate be equal to the peg. 5 When some consumers are uninformed, however, the value of the shadow exchange rate is no longer public information and the crisis becomes unpredictable even though its timing is still determined by fundamentals. 6 When the fraction of informed consumers is sufficiently low, I show that the unpredictability of the crisis increases the range of outcomes that can be sustained in equilibrium. In particular, discrete devaluations can no longer be ruled out, as the associated capital losses are suffered only by uninformed consumers. 7, 8
Fig. 1 illustrates the main results of the paper. The top panel presents the set of equilibrium times at which the crisis can take place, as a function of the fraction of informed consumers α∈(0,1]α∈(0,1]. It is the bifurcation diagram of the game for a particular realization of the threshold level of reserves. The bottom panel depicts the deterioration of fundamentals and the corresponding depreciation of the shadow exchange rate for each crisis time. When the amount of private information is low (αα high), there is a single equilibrium. In this equilibrium, which corresponds to the equilibrium of previous first generation models, the crisis takes place at a time TKFGTKFG such that the shadow exchange rate equals the peg and, thus, the size of the devaluation is zero. When the amount of private information is high (αα low) there are multiple equilibria, which differ on informed consumers’ propensity to attack the currency. At one extreme, there exists an equilibrium in which the crisis takes place at time TKFGTKFG and the size of the devaluation is zero. However, in all other equilibria the crisis takes place later and is associated with discrete devaluations. There exist equilibria such that the crisis can take place at any time between TKFGTKFG and some latest time TMAXTMAX. I present an equilibrium refinement that suggests that the equilibrium likely to be played is one in which the crisis takes place at a time TREF>TKFGTREF>TKFG. Thus, the presence of private information delays the crisis and leads to discrete devaluations.
