Extreme Value Theory and the Incidence of Currency Crises

A common feature of numerous studies on early warning systems (EWS) of currency crisis is the use of an index of exchange market pressure, defined as a weighted average of the rate of depreciation, the monthly percentage changes in international reserves, and sometimes the inclusion of the monthly change in the interest rate, in order to identify and proxy the occurrence of currency crisis. Crucial to this approach is the appropriate definition of a currency crisis, and the literature has usually defined currency crisis occurring when the measure of exchange market pressure exceeds a certain threshold. The main theme of the paper is that not only is the use of a threshold in defining currency crisis as arbitrary, but a much more careful examination of the basic statistical distribution of the measures of exchange market pressure will reveal that the conventional method of defining currency crisis is statistically flawed or inaccurate in capturing the 'true' dispersion of any given exchange market pressure series. This study applies an alternative statistical method known as Extreme Value Analysis (EVA) to three different weighting schemes popularly adopted in the literature in constructing exchange market pressure indexes, namely the Eichengreen-Rose-Wyplosz (ERW), the Sachs-Tornell-Velasco (STV) and the Kaminsky-Lizondo-Reinhart (KLR). The application of EVA leads to more incidences of currency crises being identified or 'captured' compared to the conventional method across a number of countries in East Asia and Latin America from 1985 to 2003