Bankruptcy prediction by survival models based on current and lagged values of time-varying financial data

Quantitative methods to assess the performance of firms and to predict the bankruptcy event based on balance sheet indicators are widely used in the credit risk context. Logistic regression and survival analysis techniques based on hazard models are among the methods often employed. The risk of failure of Small Business Enterprises in Umbria, Italy, during the period of economic crisis (2008-2013) is investigated in a large data set of 11248 businesses. Training and holdout samples were used to develop and test survival models incorporating time-varying covariates, their lagged values at one and two years and weighted macroeconomic variables. ROC curves were used to compare models and obtain global performance measures. Lagged covariates improved performance. The same data sets were used to build logistic regression models for comparison. Logistic and survival analysis models produced similar results.