Research On Financial Crisis Early-Warning Of Listed SMEs In China

Small and medium-sized enterprises(SMEs)start to play a more positive role in our economic market with the rapid growth of China's securities market and the situation that the domestic market is no longer monopolized by the state owned enterprises.However,SMEs are faced with increasing pressure and more intricate financial risk for the reason of Economic Globalization?rapid development of economic market and cut-throat market competition.The financial situation is very important to the development of SMEs.Based on this,preventing and defusing financial crisis by establishing an effective financial crisis early-warning model becomes urgent.In this paper,200 SMEs are obtained from the SME board and start up board as research sample,and 30 financial indexes are selected as the initial index system.Lasso?Adaptive Lasso and stepwise regression are used to select the desired variables.Finally,we build a financial crisis early-warning model by using Cox proportional hazards model,and test the goodness of fit and the prediction accuracy of the model.Based on the features of sample data,Cox proportional hazards model in the Survival Analysis method is used in this paper.This model has the following advantages.First,due to the inability to observe the time of financial crisis in all samples,this kind of sample data is incomplete,the Cox model can deal with this kind of data.Second,the Cox model doesn't require specific distribution of variables,it matches realistic statistical situation better.Lastly,sample pairing is required when people use traditional statistical methods to select sample.It will influence the reliability and stability of the model if the standard of sample pairing is not same.However,the Survival Analysis method doesn't require sample pairing.It not only reduces the workload of collecting samples in the actual work,but also improves the prediction accuracy and stability of the model.The results of empirical analysis show that the prediction accuracy of Lasso-based model is 88.3%,Adaptive Lasso-based model is 90%and stepwise regression-based model is 85%.It clearly indicates that all three methods could select significant variable effectively.All three models get acceptable prediction effect.It demonstrates that Cox proportional hazards model has significant reference value on the construction of financial crisis early-warning model.Adaptive Lasso-based Cox proportional hazards model gets the best result based on the sample data in this paper.