Research On Financial Distress Early-warning Models Of Listed Companies Based On Panel Data

List companies are the basis of the stock market, and their performance will directly influence its health development. So, to establish the models for predicting financial distress is a very important thing that the listed companies, investors, creditors and others concern.Now, the domestic research mainly regarded "ST" as the definition of financial distress. However it is only the skin-deep definition based on the listed company’s surface without analysising the financial substance which determined whether the listed companies get into financial distress really. So this study, based on previous studies, proposed a new definition of financial distress combining with the current actual needs.In addition, most previous studies mainly based on data of a given year or single-section data to predict the probability of financial distress, without taking the continuity of the corresponding time into account, so the practical application of financial distress prediction is limited. This study established a Logit model based on panel data, comparised the prediction accuracy with standard Logit model, got the results of various financial indicators’influence on the probability of financial distress through the marginal analysis on Logit panel data model, and finally found that Logit panel data model has advantages of higher prediction accuracy, stronger stability and longer predictive time-span.The main contents of the article as follows:The first chapter "Introduction" introduced the background, purpose, significance of the research and structure of this article.The second chapter "Literature Review" first introduced the definition of financial distress of domestic and foreign studies then carry out a more comprehensive summary on the financial distress theory, early-warning models of domestic and foreign studies.The third chapter "related theories" detailed the panel data, panel data models, logit model and other theories, which laid the theoretical foundation for this article. The forth chapter "establishing logit panel data model" first discussed and defined the standard of financial distress followed by sample extraction and variable selection, and introduced the form of logit panel data model established by this study. This chapter and the fifth chapter is the core of this article.The fifith chapter "predicting and analysising the results" consisted of two main parts: First, prediction and result analysis of T-1-year model, including the test of model type, the estimation of model parameter, model interpretation, model accuracy comparison and marginal analysis; Second, prediction and result analysis of T-2-year model, including the test of model type, the estimation of model parameter, model interpretation, model accuracy comparison and marginal analysis. Finally a comparative analysis was carried out on warning models of T-1-year and T-2-year.The sixth chapter "conclutions and outlooks" mainly summed up the conclutions, the main contributions, the shortcomings of the study and problems in further study.