- 學位論文
大陸企業違約預測之探討─複合加權模型
The Study of Assessing the China Business in Predicting Default Risk: Weighted Hybrid Model
摘要
本研究針對預測企業違約之(1)市場型違約預測模型─選擇權評價法(簡稱KMV Model)以及(2)會計型違約預測模型─Altman Z-score Model 與 Logit Model設計建構限制導致部分實證的效果不彰課題，但本研究兩個模型整合設計的突破點在於模型之非固定權重係數─模型權重依照企業特性變數改變，亦即探討何時會計模型的預測情形優於市場模型及何時優劣反置，本研究應用複合模型(Revised Hybrid Model：RHM)基於預測結果及型一(誤授)錯誤、型二(誤拒)錯誤，也將參考樣本分類的差異進入模型的實證比較，篩出最佳樣本模式及較適合的違約複合模型。 本研究的樣本含隱性違約公司及僅有真實違約公司，在上述分析比對後選出影響違約預測模型優劣參數設為複合模型權重變數。實證發現，本研究創新─複合模型(Revised Hybrid Model：RHM)比起KMV Model & Logit Model具有更好的違約預測能力，KMV Model 以及Logit Model 在同樣誤拒水準下達到更低的誤授。
並列摘要
Aiming at better estimating the probability of corporate defaults, which has typically been the key element in risk management, this study explores the limitations of KMV Model and Altman Z-score as well as Logit accounting-based Model. Specifically, it incorporates both market and accounting-based models and examines the suitable weights on the two types of models of based on characteristic variables including earnings-price ratio, and liquidity. The new model, Revised Hybrid Model, (RHM), outperforms in type I and type II errors the two individual models. This study include both firms that actually defaulted and the ones that should be defined as with stealth default; Using the weighing variable based on previous literatures, I document that the refined-Revised Hybrid Model (RHM) performs better on predicting default rates than KMV Model and Logit models. It also has lower Type I error when Type II error has been controlled than the original models.