Title: Structural models for firm bankruptcy prediction
Authors: Ludovico Rossi - Colegio Universitario De Estudios Financieros - CUNEF (Spain) [presenting]
Abstract: The role of the Merton distance-to-default model in forecasting corporate bankruptcies is investigated. Given the high number of bankruptcy predictors proposed in the literature, we use Logistic Lasso regressions to perform variable selection and Post Lasso Logit regressions to estimate conditional hazard models. In contrast with the existing literature, the Merton distance-to-default model consistently predicts bankruptcies in all time periods and industries. This result is robust to different model specifications. Moreover, we show that the Merton distance-to-default model is one of the most accurate variables to predict bankruptcies. Out-of-sample forecasts confirm that Post Lasso Logit, which includes Merton distance-to-default, produce more accurate bankruptcy predictions than previous models proposed in the literature.