Modeling Default Prediction with Earnings Management

This study explores whether taking into account variables for real earnings management improves specification of the default prediction model based on the Z-score methodology for Chinese listed companies. We demonstrate that the Z-score model proposed by Altman (1968) overestimates the survival probability for firms engaging in earnings management. In contrast, our inclusion of the indicator variable for real earnings management considerably enhances the prediction power of Z-score methodology. With respect to the ability to predict out-of-sample default, our findings suggest that the accounting-based credit scoring model, adjusted for real earnings management, especially via accelerating sales, yields a greater prediction accuracy rate and a lower false loan rejection rate than the unadjusted scoring model for financially non-distressed firms