重新檢視中國大陸證券市場企業危機之研究

中國大陸上市公司一旦因財務狀況異常而被列為特別處理公司時，不但對企業是嚴重警訊，對廣大投資人及股東權益將造成巨大損失，因此公司發生危機所引起的衝擊，將對整體經濟社會產生極大的傷害。如何精確地評估公司可能發生危機的機率，以及危機發生後公司何時能安然度過或因而下市，進而提出公司危機預警模型，則將是本文關注的議題。經由重新檢視過去文獻中研究方法之優劣，以尋求更具效率的分析工具，經多重實證比較得知瀑布羅吉斯函數區別危機公司與正常公司之正確率高達92.6%，倒傳遞類神經網路則相對缺乏估計穩定性，再經由多項分類基準投入吸收馬可夫鏈鎖，可有效估計危機企業恢復上市或下市之時程，期望對中國上市公司風險評估提供有效量化依據。

關鍵字

中國大陸證券市場；企業危機；瀑布羅吉斯函數；倒傳遞類神經網路；吸收馬可夫鏈鎖

並列摘要

China's listed companies once been marked ”special treatment” (ST) companies in case of financial standing abnormality, it not only raised a serious alarm to the enterprises, but also caused the investors and shareholders to huge losses. The follow-up impacts of enterprises' operational crises could severely harm the whole economy society. This research concerns how to evaluate the distress probabilities and duration accurately from beginning to end, whether the end is safety or plunge, and try to design financial distress warning model. This article reexamines benefit and weakness of methodologies by literatures review, trying to find out the most effective analysis tools. The discriminate ratios estimated by cascaded logistic function could be higher than 92.6% from the comparative empirical evidences. The estimation results are lack of stability by back-propagation neural network. Furthermore, the enterprises distress duration estimated by absorbing Markov chain on the categorical critical values base. It could provide distress companies exactly approximations about relisted or failure state. Conclusively, we hope provide valid quantitative recommendation about risk assessment of China's listed companies.