CMStatistics 2017: Start Registration
View Submission - CFE
Title: Maximizing discriminatory power of bankruptcy prediction models: Empirical evidence over short and long-term horizons Authors:  Christakis Charalambous - University of Cyprus (Cyprus) [presenting]
Spiros Martzoukos - University of Cyprus (Cyprus)
Zenon Taoushianis - University of Cyprus (Cyprus)
Abstract: Acknowledging the economic benefits associated with the development of powerful bankruptcy prediction models, a methodology is presented for maximizing their discriminatory power over short and long-term horizons. For our analysis, we use accounting and market-related information for a sample of U.S. public bankrupt and healthy firms between 1990 and 2015. Results show an improvement in the discriminatory power when we implement our approach as compared with traditional approaches, such as logistic regression models, using short (one and two years) and long-term (five years) forecasting horizons. Most importantly, this improvement in model performance is evident not only in-sample but also when employing three out-of-sample approaches and in several cases is substantial, even when making longer-term forecasts.