We report an attempt to use financial market data to predict the onset of a variety of international crises. A quasi-experimental design for studying interrupted time series data is applied to the historical behavior of the Hong Kong and New York financial markets. The results are generally disappointing. These markets seem more likely to react to rather than anticipate crisis developments, and frequently issue "false alarms." Comparison of results for periods where crises took place with those for randomly selected periods where no crisis happened does not engender confidence in markets as crisis predictors. However, there are some notable exceptions to this pessimistic generalization. We conclude that financial indicators may be profitably used in some cases to supplement warning assessments derived from other methods and sources.