Financial Distress Prediction Of Listed Companies Based On Mixture Cure Model

The prediction of corporate financial distress has been a hot topic in the field of risk management for a long time.The financial distress prediction model can be applied to many issues,including monitoring the solvency of listed companies,evaluating the risk of loan defaults,and pricing of bonds,credit derivatives,and other securities that face credit risk.At the same time,timely prediction of the financial distress risk of listed companies is closely related to the company's own development and order supervision of the capital market.Therefore,it is particularly important to establish a scientific and effective early warning system for corporate financial distress.Previous studies using the Cox regression model to characterize the risk of financial distress for enterprises assume that the survival function will approach zero over time.However,through empirical research,it is found that the survival curve of Chinese listed companies doesn't tend to zero,and doesn't satisfy the theoretical assumption that the survival function approaches zero in the classic Cox model.On the other hand,most of the previous financial distress prediction models based on Cox regression models only used financial indicators as covariates,which can only reflect the financial characteristics of listed companies,while leaving the characteristics of macroeconomic factors outside the dynamic prediction model of financial risk of listed companies.Mixture cure models are an extension to the standard survival model.The model consists of a weighted average of two parts,which respectively explain if and when financial distress occurs,and can not only consider the probability of financial distress of listed companies,but also reflect the time characteristics of the company's financial risk changes.The main contribution is reflected in two aspects.First,the macroeconomic indicators are added to the model to reflect the explanatory power and prediction performance of the model better;second,the use of the mixture cure model introduces the probability of financial distress of listed companies,which is in line with the fact that the survival function of Chinese listed companies doesn't tend to zero in practice,and solves the problem that the classic proportional Cox model is not applicable,and more effectively characterize the change law of the risk of listed companies' financial distress.The results of the empirical analysis show that:(1)the accuracy,AUC value and KS value of the prediction model are improved after adding macroeconomic indicators to themixture cure model and the Cox regression model with only financial indicators,which means that the introduction of macroeconomic indicators has a positive effect on the prediction of financial distress of listed companies;(2)the accuracy,AUC value and KS value of the mixture cure model with only financial indicators are higher than the Cox regression model with only financial indicators;(3)the accuracy,AUC value and KS value of the mixture cure model with financial indicators and macroeconomic indicators are higher than the Cox regression model with financial indicators and macroeconomic indicators,indicating that the mixture cure model is suitable for the prediction of financial distress and has a good effect.In summary,this paper may provide a new way to effectively predict the financial distress risk of Chinese listed companies.