Predicting Default and Non-Default Firms using Discriminant Analysis: Adaptation of KMV-Merton's Default Probabilities and Financial Ratios

Abstract

The KMV-Merton model provides conceptual determinants for predicting firms' default risk, but its accuracy was tested long ago, and it contains insufficient statistics for default prediction. Therefore, previous literature adapted the KMV-Merton model into a statistical model involving financial ratios to improve its predictive capabilities. Discriminant Analysis (DA) is a widely used statistical model for predicting financial distress. The objectives of this study are to identify financial ratios significant to KMV-Merton's default probabilities using DA, to predict default and non-default firms using the DA model obtained, and to compare the performance of the KMV-Merton and DA models in predicting default risk. The study uses 11 years of data from Malaysian publicly listed firms, applying the KMV-Merton model and stepwise DA in SPSS. DA identifies the significance of selected financial ratios to firm default, with KMV-Merton's default probabilities as the dependent variable, forming a discrimination function to predict default and non-default firms. Credit ratings and Type 1 and Type II errors are used to compare model performance. The DA using SPSS reveals a discriminant function with net profit margin and return on assets significantly related to KMV-Merton's default probabilities. The DA model is more biased in predicting non-default firms due to the need for more information on default firms, yet it slightly outperforms the KMV-Merton model. This study offers guidance on adapting KMV-Merton's default probability estimates with financial ratios in the DA model and highlights the significant financial ratios related to KMV-Merton's default probabilities.