Predicting Financial Distress Of Chinese Listed Manufacturing Companies By CUSUM Model

With economic globalization,deepening reform of China market economic and rapid development of capital market,Enterprise not only get giant opportunities,but also are faced with unpredictable risk.Nowadays,it is common that a great number of listed companies are under special treatment because of financial distress.Several contributing factors will result in financial stress.Be concerned with company itself,low operating efficiency,poor profitablity,dishonest integrity have a considerable adverse impact.And in large scale,different industry risk,economic statute and finance risk can also contribute to corporate failure.A gigantic number of academic researchers from all over the world have been developing financial distress models,based on various modeling techniques.The majority of models are based on static cross-section data,neglecting the time-series behavior of financial variables.In fact,financial distress is a gradual process.Traditional models can not reflect dynamic behavior of corporate financial condition.Besides,traditional models uses single period data,which is hard to assess past influence.Hence,it seems that classical statistical failure model are not suited for corporate failure prediction.Thus this paper use the dynamic model CUSUM model which can "accumulate" past financial condition.This paper use the companies which is from manufacturing industry.The failing sample companies are under ST from 2013 to 2016.When the numbers of failing companies and non-failing companies are equal,it may result in over-sampling of failing companies,leading to an overstatement of accuracy of the model.For avoiding that problem,the ratio between failing companies and non-failing companies is 1 to 3.As CUSUM model is a dynamic model,the panel data consist of last 20 seasons before ST.After Mann-Whieney test and correlation test,this paper chooses Current ratio,Rate of Return equity,total asset growth rate,velocity of working capital as the best predictors.After testing stability of data,then use panel vector autoregressive model and build the CUSUM model.For assessing predictive power of CUSUM model,Logistic model is used as a comparable model.According to result of model in season T-1,T-3,T-5,T-7,T-9,T-11,the empirical research shows result of the CUSUM model is satisfactory with high accuracy.Logistic model behaves better in the short term,and CUSUM model has higher accuracy.