Three essays on currency crises.

Abstract

Chapter three empirically tests the effectiveness of interest rate defenses against speculative attacks. It is shown that previous empirical studies on this issue suffer a classic sample selection problem by restricting their sample observations to crisis periods only. To correct this selectivity bias, I employ the full information maximum likelihood method to a large unbalanced panel dataset that covers both crisis periods and peaceful periods. I also develop a rare-events-corrected probit model with sample selection that can be used to correct a second bias that is created by the rare events data used to estimate the selection equation. My empirical results show there is no significant statistical relationship between the interest rate and the outcome of a speculative attack. (Abstract shortened by UMI.)

This dissertation studies the roles of interest rates in currency crises. Existing studies in the literature suggest that high interest rates can play two different roles. The first role is an attack deterrent: a monetary authority can potentially deter a speculative attack at the first place by raising interest rates ex ante to reveal its willingness and ability to defend the currency. The second role is a weapon of defense: once an attack has already been launched, a monetary authority can possibly defend the currency by raising interest rates to increase speculators' costs.

The first chapter of my dissertation studies the attack deterrence effect of high interest rates. In this chapter, I present a signaling model of raising the interest rate as a deterrent to speculative attacks and then test the model using a dataset that covers 54 countries from March 1964 through December 2000. I am able to show that unsuccessful signaling can be equilibrium behavior in either a pooling equilibrium or a semi-separating equilibrium. Although it is still possible for a weak monetary authority to hide his type in a pooling or semi-separating equilibrium, in both cases, the weak monetary authority faces a higher probability of an attack. In the empirical part of this paper, I find evidence that raising interest rates in advance has significantly different impacts in different country groups. It significantly reduces the probability of attacks in countries that have a de facto hard peg but increases it in de facto soft-pegging countries. This finding is robust to alternative measure of monetary policy and to different specifications and samples. These results clearly support the predictions of the theoretical model.

The second chapter presents a theoretical model of interest rate defenses against speculative attacks. I argue that the battle between speculators and the government can be well modeled as a war of attrition game under asymmetric information. I then solve for a pure strategy weak perfect Bayesian equilibrium in which each party's time until concession depends on their benefits from winning and their costs of fighting. Using this model, I am able to produce results that are novel for the speculative attack literature. First, the model shows that failed defenses (attacks) can be ex-ante rational for governments (speculators). Second, the model predicts systematic variations in the durations of interest rate defenses and I find some support for these predictions in the context of attacks in 1997 and 1998. Finally, the model suggests that the relation between interest rates and the outcome of a defense is likely to be nonlinear, which is consistent with existing empirical evidence.