This paper empirically analyzes the origins of currency crises for a group of OECD economies from 1970 through 1998. We apply duration analysis to examine how the probability of a currency crisis depends on the length of non‐crisis periods, contagion channels, and macroeconomic fundamentals. Our findings confirm the negative duration dependence of a currency crisis—the likelihood of speculative attack sharply increases at the beginning of non‐crisis periods and then declines over time until it abruptly rises again. The results also indicate the hazard of a crisis increase with high values of the volatility of unemployment rates, inflation rates, contagion factors—which mostly work through trade channels, unemployment rates, real effective exchange rate, trade openness, and size of economy. To address concerns regarding validity of the identified crisis episodes, we exploit crisis episodes that are identified by a more objective approach based on the extreme value theory. Our results are robust under various specifications including two different crisis event sets that are identified on monthly and quarterly basis.