Failure prediction of Chinese A-share listed companies : comparisons using logistic regression model and neural network analysis : a thesis presented in partial fulfilment of the requirements for the degree of Master of Business Studies in Finance at Massey University, Palmerston North, New Zealand
This study compares the relative prediction accuracy of corporate failure between two prediction methods –logistic regression model and neural network analysis– based on a sample of 3598 observations and companies data obtained from the Chinese A- Share market during the period 1991 to 2002. Seven criteria have been set up to define failure according to attributes of Chinese listed companies. Using forty financial ratios and seven misclassification cost ratios of Type I and Type II error, two models achieve ranges of minimal misclassification cost at optimal cut-off points for two years prior to business failure; The logistic regression model is slightly superior to neural network analysis. Compared with random prediction, both models are efficient. In addition, the study points out that Total Asset Turnover (TATR), Cash Ratio (CASR), Earning per Share (EPS), Total Debt to Total Asset (TDTA), Return on Assets (ROA) and the natual log of Total Market Value (MVLN) could be significant financial indictors of corporate failure. Results of the study have important implications in credit evaluation, internal risk control and capital market investment guidelines.