Predicting multilateral trade credit risks: comparisons of Logit and Fuzzy Logic models using ROC curve analysis

Abstract

Employing pooled data of 3344 listed firms from seven Asia-Pacific countries, this is the first empirical study to classify and predict trade credit risks in the international trade context. In addition, this paper extends previous work by applying receiver operating characteristic (ROC) curve analysis to compare the model performance of Logit to that of Fuzzy Logic (FL). We are unaware of any other paper that has discussed the application of ROC curve analysis in the business and finance literature. The results show that FL exceeds Logit in terms of overall classification accuracy and prediction accuracy. However, by incorporating measurement in the form of ROC curves, Logit is proven to outperform FL in classifying non-default firms. This suggests that though FL is superior in overall accuracy and in classifying default firms, Logit is preferable in situations where higher accuracy in classifying non-default firms is preferred. The stability of the models is also demonstrated.