The Predictive Validity To The Listed Company’s Financial Situation Of The Binary Classification Logistic Regressication Analysis Model

In the document’Opinions on reform perfection and implement strictness on delisting system of the listed companies’ (2014) proposed by China Securities Regulatory Commission (CSRC), it clearly stated that companies shall be delisted compulsively when certain financial signals alarm. Under this policy background, there exists delisting risks in listed companies when their financial health are deteriorating. In modern business management practice, financial health deterioration is a process of repetition and accumulation, which contains the deterioration of sorts of financial signals. Therefore, we make efforts to construct models based on series of financial or non-financial ratios and the application of binary classification Logistic regression analysis, in order to predict financial status tendency thus preventing delisting caused by persistent downward. This is where the significance and implication of this research lies.This paper apply the methodology of both qualitative and quantitative, both normative and empirical analysis and comparative analysis. It is built on the theory basis of MM theory, disequilibrium theory and agency theory. The research will be conducted following how the financial and non-financial indicators formulate the system, how the sample and data should be selected, how the binary Logistic regression model is constructed, and how to measure the predictive validity quantifiably.Specifically, this paper will select 19 financial indicators in terms of the companies’ solvency, profitability, operational ability, growth ability and liquidity; 8 non-financial indicators are selected from ownership structure, corporate governance, internal control, external guarantee, industry prospective and local economic development in the perspective of microcosmic-middle-macroscopic; furthermore,200 securities listed on Shanghai and Shenzhen A-share markets are selected sample data, which includes 146 sound financial health companies and 54 companies have delisting potential. On the basis of binary Logistic regression model, there are model I for financial indicators and model II for non-financial indicators. The model can predict the companies’financial situation of two year from now on by applying the data from year t-2 to year t and calculate the probability of financial failure.The model I based on financial indicators shows the predictive validity of 0.845 from training sample and 0.830 from testing sample, and 0.939 from training sample and 0.911 from testing sample when non-financial indicators are introduced in model II. Obviously, model II has better performance which leads to the conclusion that non-financial indicators can improve the predictive validity of the model.This research is innovative in two senses. Compared with other studies, this paper firstly introduce the industry prospective and macroscopic economic development index, thus reshape the financial diagnosis model. In addition, the paper defies the company is under financial situation of distress, sub-health and sound health, by settling the critical threshold of companies’ financial health.