| The main theme of the dissertation is that the routine procedure of ruling out divergent paths in rational expectations models may not be convincing. It has been argued that explosive bubble paths should be excluded since they, if followed into the infinite future, will lead to something absurd. Crucial to this type of reasoning is the implicit assumption that the underlying dynamic system, especially the government policy regime, remains unchanged indefinitely despite an occurrence of an infinite explosion. I assume, instead, that policy rules will be changed in order to stop explosions once an explosion occurs.;The first paper, "Anticipated Stopping of Hyperinflations: Can We Really Rule Out Speculative Bubbles?", discusses the above problem using a simple hyperinflationary model, and compares the result with arguments in the literature against bubble paths. The second paper, "The Indeterminacy of Short-Run Exchange Rates in a Managed Float Regime", applies the same idea to models of exchange rates and interest rates. Indeterminacy of exchange rates is particularly interesting in view of the current lack of convincing models that explain exchange rate movements. The third paper, "A Note on the Large Speculator Assumption in a BOP Crisis Model", is an attempt to correct some misunderstandings in the literature concerning the proof of the key proposition in a BOP crisis model--the proposition that a speculative attack occurs when the shadow rate equals the pegged rate.;Under this assumption, the usual distinction between market fundamentals and bubbles disappears. Because no path is infinitely explosive, no explosive path may be ruled out according to the usual criterion. A variable whose current value is dependent upon its expected future value is indeterminate in this more dynamic model where policy regimes change in response to explosions. A deeper question that arises from this exercise is whether or not the government should refuse to stop explosions. The dissertation shows that the argument in the literature that rules are superior to discretion in a saddle path model remains to be proved. |
