Abstract (Translated):
In the present decade, in emerging economies such as those in Latin-America, mixed logistic models have been started applying to predict the financial failure of companies. However, there are limitations for the methodology linked to the feasibility of predicting the state of new companies that have not been part of the training sample which was used to estimate the model. In the literature, several methods have been proposed for predicting random effects in the mixed models such as, for example, the nearest neighbor. This method is applied in a second step, after estimating a model that explains the financial situation (in crisis or healthy) of companies by considering the behavior of its financial ratios. In this study, companies from Argentina, Chile and Peru were considered, estimating the random effects that were significant in the estimation of the mixed model. Thus, we conclude that the application of these methods allow for identifying companies with financial problems with a correct classification rate of over 80%, which becomes important in modeling and predicting this risk.
