The estimate of default probability in internal rating system

Il progetto che il dottor Americo Todisco ha presentato al SAS Forum International riguarda un aspetto dell'accordo Basilea II, ossia la stima della probabilità di inadempienza nel sistema di rating interno.

Abstract
***********
In January 2001 the Basel Committee on Banking Supervision issued a consultative paper with a proposal (usually referred to as Basel II) of reform of the capital adequacy discipline of banks. The main change in the new framework is related to the strong incentive for banks to develop internal rating system in order to reduce their needs of capital for credit exposure less risky.

This research project is focused on the core of the construction of an Internal Rating System: the estimate of default probability (PD).

The dataset used contains the balance sheet of about 40.000 small and medium enterprise (3% of them in default). For each enterprise were used three years of accounting data. Using a random procedure we split the whole dataset in 30 different samples.
Using a comprehensive scheme of ratio analysis we compute 63 ratios to use as predictive variables in explaining the event of default. We use the logistic regression in order to build a model of default prediction for each of 30 samples. The model with the best classification performance was selected as the one to be used in the subsequent step of the analysis.

The fine tuning of the model is realized:

• trough the modification of the probability that result from the model for the average rate of default in Italy;
• comparing, trough the cluster analysis, three different solution about the number and the widht of the rating scale.