| Sustaining a controlled exchange regime requires credible and compatible policy. By extending and solving a basic discrete-time model of a collapsing fixed exchange rate, limits to monetary policy are determined. Insights are gained by studying the crises of eight countries and by applying the extended collapse model to Mexico and Argentina.; Purchasing-power-parity and interest rate parity, assumed in the basic discrete-time collapse model, are relaxed, allowing systematic over- and under-valuation of currencies, and providing for imperfectly substitutable domestic and foreign assets. Uncertainty, previously introduced only through nondeterministic domestic credit policy, also enters through randomized relative prices. In addition, real income and currency substitution are permitted to influence real balance demand.; Formulas for one-period-ahead exchange regime collapse probabilities, conditioned on assumed reserve floors, and shadow exchange rates are derived. It is shown that if a country has constrained access to external credit, the collapse probability can be eliminated only when the domestic credit growth rule is known and nonrandom.; Circumstances surrounding major devaluations and regime collapses are analyzed in the crises of Mexico, Argentina, Jamaica, Greece, Zaire, Portugal, Pakistan, and the Philippines. Quantitative restrictions and capital controls aimed at staving off exchange rate collapses are documented. The cross-country study did not uncover any patterns in reserve floor determination. Asset and liability positions of foreign banks vis-a-vis crisis countries provided evidence that some form of external credit rationing and capital flight preceded most collapses.; Optimal central bank holdings of reserves were studied using a 'demand for reserves' approach. Although some valuable insights were gained, reserve demand was too closely linked with external capital availability to warrant this single-equation approach.; To evaluate its performance, the extended collapse model was applied to Mexico (January 1977 to December 1983) and to Argentina (March 1977 to January 1983). Since all of the extensions to the basic model proved statistically significant and valid, the results improve upon those of earlier studies that used more restrictive assumptions. The extended model performed consistently well, generating both high collapse probabilities preceding crises and realistic lower bounds for the new post-collapse devalued rates.; The goal of providing a framework suitable for empirical application has been achieved. |
