A Reverse Engineering Approach to Price Credit Spreads in the Qualitative Rating Process

The paper presents a model to estimate qualitative variables to estimate credit spreads. The main purpose of our study was to price loans and verify whether interest rates depend on credit portfolio weights, applying a reverse engineering process. In particular, loans may be priced from credit portfolio composition applying the Sharpe model (1974) originally devoted to expected returns of asset classes. Our overall percentage of default forecast is higher for the unbalanced sample (around 6% higher) even though we believe the balanced sample is more robust. Using a regression tree we showed how to estimate the probability of default for each rating notch. Monotonicity property of PDs appears to be confirmed for both the samples (unbalanced and balanced) we tested. The final output will be the equilibrium portfolio, which depend upon the bank risk aversion