Comparison of Parametric, Semi-Parametric and Non-Parametric Methods in Bankruptcy Prediction
24 Pages Posted: 2 Mar 2007
Date Written: March 1, 2007
This paper compares parametric, semi-parametric and non-parametric methods in prediction of bankruptcy. Special care is devoted to the effect of choice-based sampling. The choice of the sampling and estimation method lead to a similar trade off. Using choice-based sampling and logit model leads to minimization of risk exposure. Samples unbalanced across groups and Klein and Spady (1993) semi-parametric method allow for better overall prediction accuracy and thus profit maximization. Both the choice of sampling method and the choice of estimation method should be thus made conditional on an explicit objective function of the financial institution in assesing credit risk.
Keywords: bankruptcy prediction, semi-parametric methods, CART
JEL Classification: G32, G33, C14, C25
Suggested Citation: Suggested Citation