Prediction of Credit Risk in the Peruvian Financial System

This study intuitively constructs a credit risk measure that captures the deterioration of a loan portfolio in normal condition over a 12-month window. In this way, we address the issues and biases associated with using the delinquency ratio as an approximation of the probability of default for financial system borrowers. Additionally, using various univariate and multivariate prediction techniques, we forecast the future trajectory (6 months and beyond) of the probability of default. Time-varying parameter models demonstrate better performance in out-of-sample predictions compared to predictions derived from time-invariant parameter models for longer-term horizons. However, for shorter-term horizons, predictions from time-invariant parameter models exhibit quite good performance