企業財務危機預警模型—EBITDA利潤率與營業利益率之分析

Abstract

本研究旨在比較扣除所得稅、利息、折舊及攤銷前利潤率（EBITDA利潤率）及營業利益率於預測危機企業之表現並判斷EBITDA利潤率是否為較佳之財務危機預測因子。研究樣本取自台灣經濟新報(TEJ)資料庫，並以2008年至2022年間之上市櫃企業作為研究對象，使用Logit迴歸分析及Probit迴歸分析來檢測EBITDA利潤率及營業利益率於財務危機發生前三年間之反應能力。研究結果指出，EBITDA利潤率及營業利益率於兩種統計方法下之預測能力並無顯著差異，雖然EBITDA利潤率最終模型及營業利益率最終模型於危機前三年內之整體預測準確率均達90%以上之水準，型一誤差(%)卻表明預測危機企業之能力仍不理想，顯示EBITDA利潤率無法作為危機判別之重要預測指標。另外，相關實證結果亦表明Logit迴歸分析於危機前二年內之預測表現略優於Probit迴歸分析，而隨著預測時點之接近，二最終模型之預測錯誤成本均呈現逐漸上升之跡象，顯示模型早期之預警效果較佳。

This study aims to compare the effectiveness of the EBITDA margin and the operating profit margin in predicting financially distressed companies and to determine if the EBITDA margin is a better predictor of financial crises. The research sample is taken from the Taiwan Economic Journal (TEJ) database and focuses on listed companies from 2008 to 2022. Logistic regression analysis and Probit regression analysis are used to examine the responsiveness of the EBITDA margin and the operating profit margin during the three years preceding a financial crisis. The results indicate that there is no significant difference in predictive ability between the EBITDA margin and the operating profit margin under both statistical methods. Although both the final EBITDA margin model and the final operating profit margin model achieved an overall prediction accuracy of over 90% within the three years preceding the crisis, Type I error (%) suggests that the ability to predict distressed companies is still not ideal. This implies that the EBITDA margin cannot be considered an important predictive indicator of crises. Additionally, empirical results show that Logistic regression analysis performs slightly better than Probit regression analysis in the two years before the crisis. As the prediction point approaches, the prediction error costs of both final models tend to increase, indicating that the early warning effect of the models is better.