The Study On Financial Distress Prediction Of Listed Manufacture Companies Based On Non-paired Samples

With the increase of the financial distress company year by year and constant strengthen of the Chinese relative law; it is particularly a great theoretical and practical significance focus on how to prevent the happen of distress. Nowadays, most of relative research assumed the assets scale would influence the early warming model, so they adopted one to one paired samples. However, after research it can be found that the assumption doesn't have any basis and the proportion of the real two paired companies is unconformity seriously. Moreover, the foresee effect of assets scale is not obvious,the former research have overestimated prediction ability of models. So, the paper adopts non-paired samples because that it accords with the fact better.This paper chooses the companies listed before January 1,1999, so the samples include the new 31 ST companies of Year 2005 and 2006 and 238 non-ST companies. It adopts non-paired samples and selects the pre-two and pre-three years' data of listed manufacture companies and chooses thirty-three variables including traditional financial ratios, cash flow ratios and non-financial ratios. Through test of normality and notability, it compares the two Logistic regression models by selecting variables using stepwise and factor analyze. The results show that the predicting effect of the latter is better; especially the forecasting accuracy of distress companies achieves 100% in the two models, and the accuracy of the health are 98.3% and 93.8% respectively, and the total accuracy are 98.3% and 94.7%. After test of the models, it can be found that the effect is not bad. Besides that, the paper puts forward the corresponding measures for ridding alarm and the early-warning countermeasures according to the causes of distress.