Use of Recursive Partitioning in the Development of Credit Scoring Models

The farm financial crisis in the mid 1980s brought increased interest in credit evaluation models. Many agricultural lenders and financial advisors have adopted formal credit evaluation models to monitor and forecast financial performance. Various non-parametric and parametric methods have been utilized to estimate the models, such as: experience-based algorithms (Alcott, Splett et al.); mathematical programming (Hardy and Adrian, and Ziari et al.); logistic regression (Mortensen et al.); probit regression (Lufburrow et al. and Miller et al.); discriminate analysis (Hardy and Weed, Dunn and Frey, and Johnson and Hagan); and linear probability regression (Turvey). There is not unanimous agreement on the best method for estimating credit evaluation models and new methods continue to be researched.