| Summary: | 碩士 === 國立臺北大學 === 統計學系 === 93 === Recently, because of the economic recession, unstable political environment, and unclear cross-strait relationship, many enterprises faced financial distress and bankruptcy. The failure of an enterprise would impact not only individual but also the whole society. Therefore, the purpose of this study was to establish an effective prediction model to detect corporate financial distress in advance and lower the impact. A matched pair of 35 distressed firms and 35 normal firms was random sampling from listed companies in the period between Year 2001 and 2004. 18 financial ratio variables one year before the crisis occurred and 6 non-financial variables three years before the crisis occurred were collected and divided into 5 aspects. The method of this project was first use descriptive statistics to analyze basic characteristics and the relationship among variables. Two-pair population test was also used to select critical variables to distinguish distressed firms from normal firms. Next, the factor analysis was employed to reduce variables, lower the multicollinearity among variables, and extract representative factors for the original data character. Taking these representative factors into Logistic Regression, an effective model to distinguish distressed firms was created. The research outcome showed that the explained variance of the factor analysis was 69.09% when only considering financial variables and the distinction accuracy for modeling sample and model-examining sample were 93.3% and 90%, respective under Logistic Regression model. After adding non-financial variables, the explained variance of the factor analysis increased to 72% and the distinction accuracy for modeling sample and model-examining sample also increased to 98.3% and 100%, respective under Logistic Regression model.
|
|---|
