A Study Of The Predictability Of Company Quality Under The ST Regime

The ST system is an important institutional arrangement for China’s stock market to bring into play the market mechanism for the elimination of winners and losers,and its key institutional arrangement under the full registration system is important for optimising the quality of listed companies,as well as giving listed companies the opportunity to develop during economic cycle fluctuations.This paper conducts a comprehensive study on 744 ST companies and 145 delisted companies in China’s Ashare market from 2000-2021,using logistic regression and random forest to construct ST and delisting warning models.The study predicts the probability of "cap","delisting" and delisting of listed companies and determines the predictive effect in different years to determine whether there are certain indicators that play an important or decisive role in the development of ST companies.The research results show that:(1)the smaller the earnings per share and earnings per share growth rate of a listed company,the greater the concentration of shares and the greater the proportion of shares held by outstanding shareholders,the greater the probability of being "hatted".(2)There are significant differences in the indicators of net sales margin,return on total assets,gearing ratio and cash ratio between "delisted" companies and delisted companies,and the stronger the profitability and solvency,the lower the probability of delisting.In addition,the audit opinion plays an effective early warning role for the delisting of listed companies;restructuring is also an effective means to successfully "remove the cap".(3)When predicting delisting,the factors of financial indicators that are significantly different in year T-1 are not significant in year T-2,indicating that the influence of financial indicators on delisting gradually strengthens as the time of delisting or "delisting" approaches.In addition,the accuracy of the model using T-1 data is higher than that of T-2,indicating that the model is more robust the closer the time is.The model’s ability to predict delisting risk gradually strengthens as business conditions deteriorate.(4)The prediction accuracy of the random forest model is higher than that of logistic regression.By tracking the company’s financial and non-financial data,it helps listed companies to be prepared for danger,assess their own risks and find the causes,and predict the risk of being ST in advance.Companies that are already in ST distress can determine the extent of their current distress and take targeted measures to avoid delisting.Investors can further balance the risks and benefits of investing in the company by focusing on certain indicators based on their own risk appetite.For the regulator,it can provide a reference for the regulator to formulate relevant rules and regulations for ST companies.