Recurrent financial crisis in LDCs
Doctor of Philosophy thesis
When estimating probabilities of repayment problems in Less Developed Countries (LDCs), the standard approach has been the application of logit models. Since there are few cases of LDCs, panel data has been used to increase the number of observations. The problem with this method is that when there exists state dependence the observations are not independent. In this case the logit models have to be adapted to deal with serial correlation and, possibly, with heterogeneity. Serial correlation is present because when a rescheduling occurs creditors will probably restrict credit in the future to the country having repayment difficulties. If the borrowing country has been using new loans to service the old debt the restriction of credit will probably create financial difficulties. Heterogeneity is a problem because different country-specific or time-specific covariates are unobserved. In this dissertation we use duration models to estimate probabilities of repayment problems, in particular probabilities of rescheduling. We consider a rescheduling as exit from the state of financial "health". Since a country can face more than one rescheduling, we have the case of multiple-spells multiple-states models. When using duration models to estimate probabilities of rescheduling we take into account the past. In this case, to estimate the probability of repayment problems at some particular period of time, we need to know the number of previous repayments problems and the number of periods the borrowing country has been "healthy". The importance of indentifying state dependence is that we can differentiate the first rescheduling from recurrent reschedulings. This differentiation helps to characterize the current debt crisis, when most of the debtor countries have been facing debt restructuring for the second or third time, and where there is a general withdrawal of creditors from the financial market for LDCs. In this case the probability of occurrence of financial problems will be different because these problems are recurrent. Economic policy has to differentiate simple from recurrent financial problems because they behave in a different way.