Financial Distress Prediction Of Listed Companies Based On Panel Data

In the market economy, the production and operations is full of risks.Competition not only brings enterprises opportunities, more of challenge. In the intense market competition,some enterprises will fall into financial distress inevitably,even a crisis of survival.Once enterprises fall into financial distress,which will impact on investors, creditors,enterprises and the government.If we can predict the likelihood of financial distress correctly,that will be practical significance for protecting the interests of investors and creditors, Operators’preventing the financial crisis, Government departments’ monitoring quality of listed companies and stock market risk.Based on the study of domestic and foreign theories and methods of financial distress prediction,the thesis selects chinese listed companies as objects of study and utilizes methods of normative research and empirical research’s combination.,then the thesis establishes the theoretical framework of financial early warning and analyzes the reasons for financial distress.Furthermore,it summarizes the distribution of chinese failed companies and proves the industry differences is an important factor that affects financial distress prediction.Lastly,the thesis builds indicators of financial distress prediction and constructs the financial distress predicting models on the basis of binary classification and multi-classification respectivly and tests the model prediction,then a new system of financial distress prediction forms The main research is as followed.1.The thesis analyzes the background of this topic and elaborates the theoretical and practical significance of the research on financial distress prediction. Then the common problems in the present study are summrized on the basis of reviewing a lot of foreign and Chinese literatures about financial distress prediction.2.The thesis builds a theoretical framework of financial distress prediction. Firstly, the defining principle is proposed based on the research of foreign and Chinese defining methods of financial distress.Then the reasons for financial distress are analyzed and the theory of financial warning is explained.Lastly,the methods of financial distress prediction are discussed from the qualitative and quantitative points. For each method the basic idea is introduced and the advantages and disadvantages are summrized. 3.The statistical distribution of the failed companies are gathered from the industries, regions, assets and survival time points of view respectively. Results show that the distribution of companies is different in each area.From the industry distribution, comprehensive industry, communication and cultural industry,information technology industry and agriculture, forestry, animal husbandry and fishery industry are high-risk.Menufacturing industry,real estate industry,wholesale and retail trade industry and social services industry are medium-risk.Mining industry, Electricity, gas and water production and supply industry,building industry and transportation and warehousing industry are low-risk.From the regional distribution, Hainan, Shanxi, Guangxi, Qinghai, Shaanxi are high-risk areas. Gansu, Chongqing, Heilongjiang, Ningxia, Hebei, Hubei, Liaoning, Tianjin, Jilin, Xinjiang, Tibet, and Hunan are medium-risk areas. Shenzhen, Sichuan, Shandong, Yunnan, Guangdong, Shanghai, Henan, Beijing, Inner Mongolia, Guizhou, Fujian, Jiangsu, Jiangxi, Zhejiang and Anhui are low-risk areas.From the assets distribution,the scale below200million is very high-risk.200million-1billion is a high-risk scale.1billion-2billion is a medium-risk scale and the scale more than2billion is low-risk. From the survival time distribution,1-2years after listing is a initial stabilization period.3-6yeears is a period of increased risk.7-10years is a high-risk period.10-15years is a period of risk reduction and more than15years is a stable period.4.The thesis analyzes the theoretical basis for the differences of financial indicators in different industries and makes the non-parametric Kruskal-Wallis H test for10financial indicators during2005-2009from10industries.Results show that all the indicators reject the null hypothesis except net profit growth rate and net assets growth rate’s accepting the null hypothesis in some years,which means there are significant differences in defferent industry.In order to test the stability of differences in industry,we made the Kendall W test. The results show that all the financial ratios reject the null hypothesis in the5%level except net profit growth rate and net assets growth rate’s accepting the nulll hypothesis in some time span.That means these financial ratios have considerable stability in the industry differences and the consistent test results are obtained from different time span.Lastly, the thesis selects the two biggest subordinate industries:machinery, equipmentandinstrument industry and oil,chemical, plastic industry in manufacturing for study.The Cox models are built to test industry difference is a factor that affects financial risk and a dummy variable on behalf of industry differences is used in the Cox model with mixed samples.. Empirical results show that the dummy variable coefficient is significant, which indicates that companies in different industries are facing different financial risk.In this research, the risk of financial distress in the oil,chemical, plastic industry is1.857times in the machinery, equipment and instrument industry.5. The thesis selects the oil,chemical, plastic industry in manufacturing for study,which is a bigger subordinate industry.Firstly, the Mann-Whitney U test is made for31primary indicators and6viables whose mean difference are not significant are excluded.Then factor analysis is made for the remaining25variables and we select the9factors from factor analysis as explanatory variables to build the Panel Logit model. Empirical results show that profitability factor, debt factor and growth factor are factors that can prevent the listed companies from falling into financial distress and the type of audit opinion is also an important factor.At the same time, the Panel Logit model can describe the relationship between the probability of financial distress and the factors. Financial distress is an asymptotic process and the factors that affect the enterprises’financial situation are constantly changing. So the static econometric models based on cross-sectional financial data can’t reflect thedynamic process,while the Panel Logit model can just make up the shortfall.6.The thesis devides the normal companies into two categories:financial healthy companies and financial sub-healthy companies,so the listed companies are classified into three categories:financial healthy companies, financial sub-healthy companies and financial distress companies.Then the SVM multi-classifier model is constructed with panel data. In variable selection, the method of Mean Impact Value is used. The empirical results indicate that the financial distress early warning model after variable selction has good forecast ability, which can give a higher classification accuracy by a fewer characteristics variables.On the one hand,it means that the MIV method is feasible and effective on variable selection.On the other hand,it means that SVM has high generalizing ability,which can get a high classification accuracy not only for the training set data but also the test set data.The thesis makes a more comprehensive research on the financial distress prediction of chinese listed companies with panel data and the methods of statistic and artificial intelligence and gets lots of meaningful conclusions.But because of my limited ability,this thesis still has its limitation and it still needs to be further researched in some places,which I will overcome and improve in my future research.