..

This paper extends the work by Morris and Shin (Am. Econom. Rev. 88 (1998) 587–597) where multiple equilibria in the self-fulfilling currency attack models can be reduced to a unique equilibrium when agents observe fundamentals privately with small errors. We find that under a more general specification with realistic parameters, noisy private observations are generally insufficient to prevent the multiplicity of equilibria. The pivotal role played by the transparency of fundamentals/policies in currency crisis is also examined. Surprisingly, transparency may trigger rather than eliminate currency crises when fundamentals are relatively healthy. Our results may be relevant to research in other coordination problems.

There are two genres of models on currency crisis in the literature. The first generation models, à la Krugman (1979), explain currency crisis from the inconsistencies between the exchange rates and domestic macroeconomic fundamentals. The second generation models (see Obstfeld 1986, Obstfeld 1994 and Obstfeld 1996) view currency crises as shifts between different equilibria generated from self-fulfilling speculative attacks. In a recent paper, Morris and Shin (1998) (MS, hereafter) suggest that the multiplicity of equilibria in this type of models is in fact “apparent” (p. 588) and is due to an unrealistic assumption that all agents observe the same signals without error. By using a more realistic assumption that agents privately observe fundamentals with small errors, they establish a very strong result that multiplicity of equilibria in a standard second generation model is eliminated. This result makes self-fulfilling expectations and their required belief coordination irrelevant. Based on their analysis, MS urge governments to adopt transparent policy. According to MS, transparent policy means policy that makes the fundamentals common knowledge. Since we plan to clarify the policy suggestion that MS advocate, we will adopt their notion of transparency. However, their notion may not be the only accepted one. For instance, Heinemann and Illing (1999) define transparency as a reduction in the dispersion of private signal noise.1 The reader is referred to King (1999) for a thorough discussion of the notion of transparency. It appears that, when making their policy suggestion, what MS have in mind is fundamentals that are relatively weak. However, when the fundamentals are relatively strong, the suggested policy restores self-fulfulling currency crisis as a possible equilibrium outcome. In this latter case, transparent policy is in fact inferior to policy that is not transparent. It prompts us to examine the robustness of the insight. To this end, we expand the parameter space of the original MS model to investigate equilibrium outcome in those previously unexplored scenarios. The present paper enlarges the analytical framework of MS by including various boundary conditions. We find that transparency alone in many cases cannot discourage currency attacks, contrary to MS's assertion. In fact, for some cases, the opposite is true. We also clarify that the MS's uniqueness result depends on the nature of the equilibrium at the boundaries, and it is the uniqueness of the boundary equilibrium that generates their results. If there are multiple equilibria at the boundaries, then the lack of common knowledge of the fundamentals may also lead to multiple rather than a single equilibrium as found by MS. In other words, private observational errors are generally not sufficient to eliminate multiplicity of equilibria. The ultimate equilibrium will still need to be determined by self-fulfilling expectations. It still leaves open the question how agents succeed in coordinating their beliefs. A recent paper by Sbracia and Zaghini (2001) shares our view that the lack of common knowledge may not be sufficient to rule out multiple equilibria. Their paper assumes that all speculators observe the same public information about the fundamentals. In our paper, however, we retain MS's assumptions that the signal of each speculator is privately acquired and this signal may differ from others. There has been a growing volume of work that employs MS's approach to study currency crisis and other coordination problems.2 We hope that our study can serve as a reminder to this growing literature that scenarios with multiple equilibria may still be plausible and interesting. The rest of the paper is organized as follows. Section 2 presents the model and the insight into why multiplicity of equilibria at the boundaries can drastically affect the results in the MS model. Taking the MS model as the norm, Section 3 examines the cases of comparatively weaker and stronger economies. Section 4 presents a broader picture of the framework and a proposition on the conditions under which non-transparency decreases the incidence of speculative attacks. Section 5 concludes.

The present paper has shown that under a broader variety of boundary conditions, noisy private observations of the fundamentals do not completely eliminate multiple equilibria. In particular, we have included into the MS framework certain “sound” economies where defending the peg is still viable at their worst possible fundamentals and certain “shaky” economies where attacking the peg at their best possible fundamentals can be beneficial for speculators. We have shown that in these extended scenarios, one equilibrium with successful attack and one without attack co-exist under certain conditions. Our multiple equilibrium result is intimately linked to the non-monotonicity of the u(k,Ik) function. The non-monotonicity of the function arises from the fact that an observed signal, when close to a boundary, is no longer a good estimator of the true state since the true state cannot go beyond the boundary. This suggests that we cannot eliminate the non-monotonicity by reducing the signal noise, and multiple equilibria always remain a viable possibility. Our study suggests that self-fulfilling expectations are generally unavoidable in analyzing currency crises. Another advantage of the self-fulfilling expectation paradigm is that it gives more flexibility in explaining field events than paradigms based on market fundamentals. It remains an empirical question whether speculative crises occurring in reasonably viable economies are attributable to MS's unique equilibrium mechanism or to the more general mechanism of self-fulfilling attacks. The present paper has also found that, in general, when the economic fundamentals are strong, the lack of common knowledge of the fundamentals among agents will generate less incidence of speculative attacks than the presence of common knowledge, while the opposite is true when economic fundamentals are weak. Hence, the claim on the superiority of transparent policy in alleviating the incidence of currency attacks has to be qualified in a more general framework. The present result raises doubts on the contention that the non-transparency of the financial sectors alone can explain the East Asian currency crisis.13 Moreover, the contention itself is also difficult to sort out because of the ill-defined multi-faceted notion of transparency. Finally, a few comments related to Obstfeld's (1996) classification of fundamentals is in order. For one thing, it appears that MS have provided a very general conclusion, with the space of fundamentals comprising all of the three regions that Obstfeld (1996) classifies. Nonetheless, on close scrutiny, when the fundamentals are not common knowledge, an element of randomness is introduced and the state space should contain all possible distributions of fundamentals. The restriction to distributions that have a non-zero probability in each of the region classified by Obstfeld – as what MS assume – is actually a special case. Secondly, certain cases, where the distributions of fundamentals put a zero probability in some of these regions, in fact correspond better to the crisis economies observed in the field. For these cases, it turns out that the uniqueness of equilibrium is no longer guaranteed. Thirdly, the co-existence of the three regions in an economy is not an indispensable feature of a standard second generation model. The insight of self-fulfilling crises as a consequence of the time-inconsistency does not depend on such an assumption. Under common knowledge of fundamentals, the second generation models produce identical conclusions, save minor details, for all of Cases A–D (see Obstfeld 1986 and Obstfeld 1996; Jeanne, 1997). Hence, under common knowledge, focusing on Case A alone seems adequate. Nonetheless, under uncommon knowledge, this modeling strategy misses other important dimensions and insights.