Econometric analysis of currency crises in the developing countries

This study develops early warning system model to predict currency crises for 16 developing countries covering the period 1991-2008, based on the unbalanced panel data. First, the binary dependent crisis variable was obtained with the help of exchange rate pressure index proposed by Kaminsky et al. (1998). Here, the crisis variable was established both normal distribution assumption and in the light of extreme value theory. Then, the leading indicators (explanatory variables) in early warning system model were determined by using panel causality test based on meta-analysis proposed by Emirmahmutoglu and Kose (2010). According to panel causality test results, overvalued real exchange rate, lending and deposit rate spread, the ratio of M2 to foreign exchange reserves, the ratio of external debt stock to reserves and real gross domestic product growth were determined as the leading indicators of currency crisis.Finally, using the significant leading indicators as explanatory variables in the binary panel logistic model, early warning system models were estimated by maximum likelihood method. The objective of these models is not to predict the exact timing of a crisis, but to predict whether a crisis occurs within a specific time horizon. Thus, crisis dependent variable in early warning systems was converted binary forward looking variable for 3 different time horizon and predicted probability of currency crisis from six early warning system models was estimated. According to results of estimates, overvalued real exchange rate, the ratio of the current account to GDP, real gross domestic product growth, M2 growth and regional contagion are statistically significant and these leading indicators have more marginal effects on the probability of currency crisis.