Corporate Financial Crisis Prediction Incorporating ESG Indicator System

With the highly globalized and informationized world economy and the rapid development of China’s financial market,listed companies are facing increasing competition,and the probability of financial crises is also increasing.The occurrence of corporate financial crises can affect investors,creditors,and the macroeconomy,among others.Therefore,conducting a reasonable analysis of corporate financial conditions and building a scientific and efficient corporate financial crisis prediction model is of great significance for risk control in the entire financial market.In existing financial crisis prediction research,the selection of indicators in the models is mostly based on financial indicators.However,information about corporate risks is not fully contained in financial information alone.Environmental,Social,and Governance(ESG)indicators,as measures of a company’s long-term value,can provide more comprehensive information on corporate sustainability and risk conditions.They have higher transparency and are among the most important non-financial indicators.Relevant studies have also shown a correlation between ESG indicators and corporate financial risks.Based on this,this study creatively introduces ESG indicators to explore their performance in financial crisis prediction.This study focuses on A-share listed companies excluding the financial industry.Companies that have received a Special Treatment(ST)designation are considered to have experienced financial crises,and 214 ST companies from 2015 to 2021 are selected.Based on the criteria of industry similarity and comparable size,an equal number of 214 non-ST companies are chosen.In terms of indicator selection,this study incorporates 14 thematic indicators from the first-level ESG indicators according to the Huazheng ESG rating system.Additionally,33 financial indicators reflecting the abilities of debt repayment,growth,operation,profitability,and cash flow are selected.To discuss the possibility of predicting corporate.financial crises over a longer time period and the improvement of model performance by ESG indicators at different times,financial crisis prediction models for T-2,T-3,and T-4 years are constructed.Through normality tests,significance tests,and correlation tests,indicators are screened,and based on the selected indicators,four individual machine learning models—Random Forest,Gradient Boosting Decision Tree(GBDT),XGBoost,and LightGBM—are constructed.Furthermore,a Stacking model is developed using the four individual machine learning models as base learners and logistic regression as the metalearner.By comparing the classification results of each model for T-2,T-3,and T-4 years,it is found that the closer the indicator data is to the year in which corporate financial crises occur,the better the predictive performance of the model.By comparing the classification results of the models before and after the inclusion of ESG indicators,it is found that the inclusion of ESG indicators can effectively improve the predictive performance of the model.Additionally,it is discovered that the degree of performance improvement by ESG indicators varies for different years,with the smallest gain for T-2 year and larger gains for T-3 and T-4 years.This demonstrates that the inclusion of ESG indicators enhances the short-term predictive ability of the model and significantly improves its long-term predictive ability.Furthermore,by calculating feature importance values,it is observed that ESG indicators account for 20%30%of the top ten indicators in terms of importance for T-2,T-3,and T-4 years,and their rankings are relatively high.This further confirms the contribution of ESG indicators in enhancing model performance and increasing interpretability.Finally,by comparing the individual machine learning models and the Stacking model,it is found that the Stacking model outperforms the individual machine learning models in terms of predictive ability.