A multivariate study of Spanish bond ratings

This paper analyses the ratings given in 1993 to the main Spanish banks, both private and governmental. It uses 24 financial ratios obtained from the balance and the profit and loss accounts. Multidimensional scaling, a multivariate technique which is intuitive and robust to the data, forms the basis of the study. This is complemented with other multivariate statistical techniques such as cluster analysis, property fitting and discriminant analysis. The results identify the financial information that has been used by the rating agency. They also confirm the conjecture that other factors, such as the public or private character of the institution, have also been taken into account by the rating agents.