本文以中國上海和深圳上市公司A股為研究對象,首先分別比較不同財務危機預警模型之區別正確率,選定以羅吉斯迴歸及馬可夫吸收鏈來建構財務危機預警模型並推估危機時程,預測特別處理公司由發生危機到下市或恢復正常交易所需的時程。經樣本期間選擇後發現,以2003年之解釋能力優於其他年度,本文應用逐步羅吉斯建構財務危機預警模型,實證結果發現羅吉斯迴歸模型有五項解釋變數皆顯著,且整體正確區別率達96.28%。再以2004及2005年特別處理公司進行驗證,其正確區別率各為84.38%與96.29%,顯示本文模型具有相當地穩定性;第二部份本文使用馬可夫吸收鏈,投入上述模型計算得到之危機機率,藉以探討特別處理公司在下市或撤銷特別處理之前的預期停留時程。本文發現驗證樣本的實際停留時間與馬可夫吸收鏈估算出的結果大致吻合,顯示應用馬可夫吸收鏈可以有效估算出特別處理公司撤銷特別處理與終止上市前之預期停留時程。
The research constructs prediction model with logistic regression and estimates enterprises distress duration with Markov absorbing chain, using the companies of the A-share in the stock market of Shanghai and Shenzhen as a sample. The empirical result finding that there are five explanatory variables significantly, and uses special treatment companies of 2004 and 2005 to test predictable ability of model. The first, the discriminate rates of 2004 and 2005 are 84.38% and 96.29% respectively by logistic regression. Secondly, the enterprises distress duration estimated by Markov absorbing chain could provide exactly approximations before to rescind special treatment or to drop out from the stock market .