| The global economy as a whole presents great uncertainty in the post epidemic era,and China’s capital market is gradually becoming an important ways for financing the real economy in a market environment where economic development is stable and the economic scale is constantly increasing.In 2022,there were 425 domestic A-share IPOs,with steady development in all major listed sectors.The implementation of the comprehensive registration system in 2023 puts forward higher requirements for the capital market.However,compared with developed foreign capital markets,China’s stock market at this stage shows a strong synchronization between traders,a large degree of fluctuations in earnings,irrational behavior accumulation,and stock price crashes from time to time.In addition,the message conveyed by the report of the "20th National Congress" is that the country will pay greater attention and emphasis on growth and security(in all dimensions),and in the long run,the trend of China’s capital market to have sustained investment value and maintain stable operation will not change.This provides the conditions and necessity for us to study the correlation between the fundamental information value of securities in China’s capital market and the occurrence of stock price crashes.This article selects A-share listed companies in Shanghai and Shenzhen Stock Exchanges from 2005 to 2020 as the research object,and selects characteristic variables from the aspects of financial indicators,corporate governance,investor response,and macroeconomic to establish a stock price collapse risk prediction model,namely,the accounting economy investor response model.At the same time,using the processing method of mixed frequency data,using four weights to convert non annual frequency public information of listed companies into annual data to improve the utilization rate of information,using different machine learning methods such as random forest,boosting,and logistic regression analysis methods to prove that the accounting economy investor response model proposed in this article is effective in predicting stock price crashes.Due to the fact that the stock price collapse is an unbalanced sample,random oversampling,random undersampling,and machine learning threshold parameter adjustment methods are also used to alleviate the problem of result bias caused by sample imbalance,resulting in a full sample accuracy rate of up to 71.86%.In addition,this article comprehensively considers evaluation indicators such as Recall,Precision,and F1 value,improving the objectivity of model evaluation.On this basis,we further examine the sensitivity of listed companies in different industries,state-owned and non-state enterprises,and in different enterprise life cycles to the importance of the characteristic variables selected in this article.The results show that even in different sub sample situations,business performance and financial cash flow are still the most important factors.The research results of this article provide a new perspective for research in the field related to stock price crashes.Starting from the fundamental information of individual stocks,using machine learning methods,the effectiveness of the accounting economy investor response model proposed in this article has been verified.At the same time,it is also attempted to use mixed frequency data in the prediction of non macroeconomic issues,thereby improving the practicality of the model.At the theoretical level,starting from information asymmetry,principal-agent theory,and behavioral asset pricing theory,this paper verifies the importance of previous studies on the factors affecting stock price crashes in predicting stock price crashes.Overall,this study can provide retail investors with a certain investment cushion,and can guide investors to refer to specific indicators for more reasonable stock selection,providing some help for the stability of China’s capital market and the investment safety of retail investors. |
PDF Full Text Request