Modelli di scoring per il rischio paese

Country risk and sovereign risk are two of the most important topics in risk management. The first part of this work introduces these concepts and shows the differences between them. The following chapters fit linear and ordinal regression models to a data-set with more than 100 countries, where the response variable is an appropriate measure of their creditworthiness. The main purposes are to identify the most relevant explanatory variables and to make predictions for those countries whose response variable is not available. For the second aim it is important to verify that records with missing values are not systematically different from the complete ones: a Little test for the MCAR hypothesis is implemented. About model selection, ad hoc algorithms are used and the theory of reduction, proposed by David Hendry, is also briefly described.