By definition, currency crises are high-frequency events. They happen suddenly, within a matter of days or even hours. To date, most empirical studies of currency crises use low-frequency data, usually monthly or quarterly. For some questions, this timing mismatch is not important. However, for other questions it could be crucial. For example, crisis theories are typically divided into fundamentals-driven first-generation models and sunspot-driven second-generation models. Distinguishing these theories requires an answer to the following question: Do policies cause crises, or do crises cause policies? Clearly, it is impossible to answer this question using data sampled at a frequency coarser than the interval between government policy choices.

An analogous problem arises in distinguishing theories of “contagion” – that is, the transmission of crises across countries. Is contagion driven by fundamentals, like trade flows, or do crises spread through their effects on expectations, with trade flows being merely an ex post response to an exogenous exchange rate change? Once again, if we are to distinguish between “real” and “financial” theories of contagion, then it would seem to be essential to use high-frequency data. This is exactly what Kristin Forbes does in her analysis of the recent financial crises in Asia and Russia. To my knowledge, this is one of the first essays that is even capable of empirically distinguishing first- and second-generation theories of contagion.

Forbes employs a standard event-study methodology, using daily data on the stock returns of more than 10,000 firms in 46 countries.