Mixed-Model Prediction Of Financial Distress

A company usually experiences a gradual worsening process before running into financial distress (FD). Stakeholders paid much attention on how to detect the symptoms of distress dynamically and give a timely and accurate prediction about its financial prospects. Although there were much new research findings about prediction of financial distress in the past few years in China, it's suspectable for the rationality in some aspects such as using special treatment (ST) as criterion event, sampling with the approach of single-year paired and so on.Based on the review of correlative researches home and broad, this paper discussed several issues such as how to decide criterion event, how to improve the technique of samples derivation and the approach of checking models. Firstly, we defined financial distress as a listed company loss in one year and earned small profits in the next year. Further more, we made the year when listed company earned small profits as the time point it ran into financial distress. Based on the financial data of manufacturing industry companies between 2001-2003, 56 ones were defined as FD companies. And under the approach of year-to-year paired sampling, 56 ones were defined as non-FD companies for each of the three years before financial distress, added up to 168 non-FD companies. Secondly, we constructed three models (model Z₁,Z₂,Z₃) from 32 traditional financial predictor variables for the previous 1-3 years, and the prediction accuracy rate were 87.5%,80% and 76.25%. It proved that the new criterion event was viable. Then, with five cash flow variables added to model Z₁,Z₂,Z₃, model Z₁,Z₂ and Z₃ were constructed and the prediction accuracy rate wasincreased, which proved cash flow variables could improve the ability of prediction. From the point of view of application of prediction model, this paper compared the efficiency of single model prediction and mixed model prediction, and came to a conclusion that mixed model prediction was better.