因素分析對羅吉斯迴歸預測力之影響－在財務歸類模式上之應用

Abstract

企業的財務狀況健全與否，對於銀行業者及證券投資者而言，都是決策時 考量的重要因素。本研究利用財務報表所提供的資訊，以財務比率為解釋 變數，建立一套有效的財務歸類模式，使得企業的財務狀況可以經由模式 來反映健全與否。先透過對財務比率分配型態的驗證，模擬財務健全與不 健全企業的財務比率，來研究因素分析運用於羅吉斯迴歸建立財務歸類模 式時對預測力的影響；並且利用實際的財務比率資料以作實證。所探討的 羅吉斯迴歸分析是以逐步篩選變數方法為前提。從實證結果可以看出，未 經因素分析萃取共同因素，以原有的變數進行逐步羅吉斯迴歸分析，其預 測力比用共同因素來得好。因此，若使用羅吉斯迴歸分析的目的是要得到 較佳預測力的模式，則以原有變數直接進行羅吉斯迴歸分析較好。由於實 際財務不健全公司的數目有限，藉助於模擬資料進行多次分析來比較兩種 分析過程。從模擬分析的結果來看，兩種分析過程的預測力在發生財務危 機前一年並無顯著差異。顯示在接近發生財務危機的年度，利用因素分析 萃取共同因素，再進行羅吉斯迴歸分析，仍具有相當的預測力。但愈遠離 財務危機發生的時間，兩種分析過程的預測力之差異愈顯著。因素分析雖 可縮減構面並增進解釋能力，但此舉也會付出使預測力降低的代價。 Financial ratios has been used to predict business failure or bankruptcy. They can also be used as explanation variables of financial classification models to identify financial character- istics. The purpose of this study is to explore the effect of factor analysis on the predictive ability of logistic regression which was applied to financial classification models. This study uses stepwise selection technique on the logistic regression. The study simulates the financial ratios of bankrupt and non- bankrupt firms by means of the validation results of financial ratios' probability density functions. The simulative and empi- rical data are used to study the effect of factor analysis. The empirical results show that the predictive ability of financial ratios is higher than of common factors. The simulat- ive results show that the difference between the predictive ability of financial ratios and of common factors are not signi- ficant in the first year before bankruptcy. The more the time is away from bankruptcy, the more the defference is significant. The results indicate that factor analysis technique can reduce dimensions and make it convenient to explain the meanings of variables, but it will pay by lowering predictive ability.