Validation of a Rolling-Logit Model to Predict TSE Corporate Bankruptcy

Publication Year:
Usage 9
Downloads 5
Abstract Views 4
Repository URL:
Lin, Chia-Liang
thesis / dissertation description
For more than 50 years, predicting corporate bankruptcy has been a critical topic of global interest resulting in significant research devoted to the development and refinement of corporate bankruptcy prediction models. However, most studies overwhelmingly concentrated on using single period's information to identify bankruptcy risks and very few studies investigated the role of previous information in prediction models. Morris (1997) and Um (2001) used a rolling-logit model based on previous and present information to measure the risks of corporations, but presented an inconsistency about the predictive ability. This inconsistency suggested more research efforts to verify the predictive ability of the model. Furthermore, there was a lack of empirical studies that adopted this method to predict the bankruptcies of Taiwanese corporations. Therefore, this study attempted to validate a rolling-logit model in predicting TSE (Taiwan Security Exchange) corporate bankruptcy.Using non-probability purposive sampling, 52 TSE corporations were matched with 52 non-bankrupt corporations during the period 1999-2005. Based on the sample, a predictive and secondary research design was conducted to (a) describe the characteristics of the sample TSE corporations, (b) compare financial and auditing characteristics between bankrupt and non-bankrupt TSE corporations, and (c) validate the usefulness of the rolling-logit model in predicting TSE corporation bankruptcies.The results of the mean comparison tests in financial ratios supported that a failing TSE corporation usually experienced declining earnings and liquidity, while simultaneously having a higher leverage position. Furthermore, the results of the mean comparison tests in auditing indicators disclosed that an aggravated financial condition would be critical in determining whether an auditor issues an unclean opinion or a corporation changes its incumbent auditor. Finally, the results of the validation of the rolling-logit model demonstrated that the model, compared to the benchmark model, exhibited higher overall accuracy. The successful performance was attributed to a recall mechanism which allows the model to measure a corporation's risks based upon consistent information across time. However, this study had limitations due to the selection of the sample and variables. Future research could validate its applicability in other markets or select other explanatory variables in order to improve the predictive ability.