Study On The Financial Distress Prediction Model Of Listed Companies In China

The financial distress is not only a common phenomenon in a market economy, but also a global problem. Since the 1960s, along with the increasingly serious problem of bankruptcy,the scholars have tried to predict ahead of bankruptcy by quantitative analysis. In the past 50 years, from the multiple linear discriminant analysis to a variety of non-parametric model represented by the neural network model, the relevant research results have been emerging. However, the research of financial distress prediction has just begun in China. The rapid development of China's securities market is not yet mature; the financial health of listed companies is in the interest of investors and creditors, so the study on financial position is extremely important. For enterprise managers, investors, creditors and other stakeholders who evaluation of the operational situation, investment value of the credit situation, the study on financial distress is great practical significance.In this context, this paper reviews the field of classical literature at home and abroad from the defining of financial distress and forecasting model. Forecast models are reviewed with emphasis on the development, compared the advantages and disadvantages of various models. The current study is the use of more pluralistic model of discriminant analysis, Logistic conditional probability models and artificial neural networks.On the basis of summing up the results of previous studies and research, in this paper, 75 listed companies of the 2001-2004 in Shenzhen and Shanghai stock market, which have been ST by"abnormal financial situation", were chosen for the study sample. And 225 normal companies, which is the same size same period and one industry, were chosen for matching samples. Selected indicators as predictor variables, which can reflect the overall financial position of the target, including 32 traditional financial ratios indicators and 6 cash flow indicators.At first, the result of the normal testing (K-S test) to predict variables, showed that, except for a few indicators, most of the predictor variables that do not meet the overall forecast variables normality assumptions. For the test of mean difference of predict variable, this paper uses a nonparametric test method: M-W test and K-S Z test, but also as a reference for the T test. The results show that: various financial indicators which are calculated in accordance with basic financial data, 70% indicators are still significant between normal enterprises and financial distress enterprises in the ST T-3 period. Z value with the approach or ST taken place in a marked increase, indicator of the average difference between the two groups that happened with the approach of expanded as a predictor variables selected financial indicators show the content and timeliness of information.Fisher discriminant analysis and logistic regression model were used in this study to build financial distress predict model in form ST T-1 to T-3 period. And used multiple cross-validation to test accuracy rates of the six models. After testing, the model accuracy rate is satisfactory. Fisher Linear Discriminant Analysis Model of the overall accuracy rate is 91.25%. 81.67% and 71.28%. Logistic Regression Model and the overall accuracy rate is 93.33%. 84.58% and 72.5%. One hand, we can see that the two models have a good predictive capability .Before three years ago, accuracy rate is more than 70%. The other hand, discriminant model from the corresponding annual rate of accuracy is obtained. Logistic regression model has an excellent forecasting capability than Fisher's linear discriminant analysis model.Finally, according to comparative analysis of the predictor variables, currency and capital ratio, debt-to-asset ratio into four separate models have a strong predictive capability, in addition to EPS, the proportion of fixed assets are also important role model in predict model. In addition to the traditional indicators of financial ratios, cash flow indicators into three models.It is noteworthy that, fixed assets ratio was negatively correlated with the probability of the enterprise into distress, and the liquidity ratio is positively correlated. Comprehensive data analysis enables us to realize that the root cause of the financial distress suffered substantial damage to its ability to continue operating. Of particular concern is that too much emphasis on the highly liquid assets will lead to a long-term shortage of capital expenditures, resulting in an average profit levels tended to decrease, the more balance of liquidity is one of the reasons caused financial distress.The innovation of this paper: First, using multiple cross-validation approach to testing the model's prediction accuracy rate. Average accuracy rate from of multiple cross-validations, accuracy as model, reduced the sample to the accuracy of the prediction model, and the accuracy of a true and reliable; 2. In this study, the sample has bigger size, the longer-time span, newer data, a stronger value. But also 1:3 proportion choice samples, more realistic and to avoid matching the accuracy of the model matched exaggerating drawbacks; 3. This paper covers the short-term solvency, long-term solvency, asset management, profitability, risk level, Growth of the six categories 32 financial ratios indicators, and 6 cash flow targets for the financial distress prediction model is a comprehensive indicator of financial information.