This paper is aimed at studying the determinants of currency crises suffered by Argentina from 1885 to 2003, on one hand, and at characterizing each particular currency crisis, on the other hand. Firstly, we identify crises episodes throughout the Argentine history. We apply the Eichengreen, Rose and Wyplosz (1994) methodology to sort crises from non-crises periods, and we distinguish among deep crises (crashes), mild crises and minor turbulences. Secondly, we look for regularities and common factors throughout history. We report the two- sample Kolmogorv-Smirnov test of equality of distributions and the Kruskal-Wallis test of equality of population. We complemented it by estimating a logit model including a set of variables chosen from the prescriptions of the existing currency crises theories. Thirdly, following Kaminsky (2003) we perform regression tree analysis to classify crises and crashes into different varieties proposed by the theories at stake. We use fifteen financial and macroeconomic variables suggested by the empirical literature. It is found that fiscal imbalances were always present, which is consistent with the predictions of the first generation speculative attack models. All three methods used to characterize currency crises in Argentina show the importance of the fiscal side. Adverse foreign factors had also a key role in explaining crises. Finally, in most of the crises, regularities in the behavior of macroeconomic variables can be detected.
