本研究是以Logit迴歸模型建立台灣2004年Q1至Q3某一本國銀行授信 行業別為電子科技產業、定義為大額授信戶之公開發行公司共46家，特性為同質性電子產業其中資產及營收達一定規模以上之樣本公司作為研究對象，並運用Logic模型導入變數，將此類財務危機樣本公司予以分析，探討發生財務危機是否在銀行借款比率、財務操作比率、或有負債比率等方面產生相關性之研究。 本研究實證結果顯示，在銀行借款變數方面，銀行借款金額佔營收之 比率與發生財務危機企業呈現正相關，顯示危機企業的銀行借款比率偏高，對金融機構舉債過高易週轉不靈; 在財務操作方面，衍生性金融商品部位佔營收比率與發生財務危機企業呈現正相關，顯示危機企業對衍生性金融商品之財務操作太過度，易發生財務危機; 在或有負債方面，對子公司之背書保證金額佔營收之比率與發生財務危機企業呈現正相關，顯示危機企業對子公司或有負債之過度保證，易發生財務危機。 由實證結果得知將這些變數納入模型，會增加模型的預測能力，因此可 以從這些變數歸納出一些端倪，達到事先預防的功能。

本研究以1998年至2003年曾發生財務危機之上市櫃公司為研究樣本，採配對樣本研究設計，依McKeown et al. (1991)之樣本分類方法，將樣本公司區分為四類：(1)失敗/財務危機公司(FS)、(2)非失敗/財務危機公司(NFS)、(3)失敗/非財務危機公司(FNS)及(4)非失敗/非財務危機公司(NFNS)，探討不同分類下之公司其盈餘操縱行為是否存在差異。其次，針對台灣證券交易所及櫃檯買賣中心近期修訂之危機預警指標，分析其偵測財務危機之能力。最後則綜合證交所及櫃買中心之預警指標，並參考相關文獻指標及實務作法，建立財務危機預測模型。 實證結果顯示，除FNS公司與NFNS公司僅部分變數有差異外，其他各組樣本之間在各類盈餘操縱變數上都具有顯著差異，代表財務危機公司在發生危機之前，確實有操縱盈餘之行為。對投資人而言，這些存在顯著差異的變數可能代表曾被危機公司所操縱而在事前沒有顯現異常之會計科目。 針對證交所及櫃買中心之危機預警指標，實證結果顯示相較於正常公司，危機公司在稅後損益及稅前損益之表現上惡化的程度較嚴重，有較高的長期負債到期還款疑慮，在預付款項上有較異常的變動情形。不過，在現金及約當現金佔資本額比率上，危機公司並無較正常公司惡化之情形，顯示多數的危機公司在發生財務危機前，通常不會列示較高的現金餘額。另外，本研究所發展之其他預警指標發現，危機公司在發生危機之前，有較正常公司顯著不足支應下期日常營運支出之情形；危機公司董監事持股質押比例，以及控制股東之董監席次控制權與盈餘分配權之偏離程度，顯著高於正常公司，代表危機公司的內部治理環境較差。 本研究所建立之實證模型顯示，綜合考量證交所及櫃買中心指標後，各指標惡化程度越高、董監事質押比例越高、董監席次控制權與盈餘分配權偏離程度越大之公司，發生財務危機的機率越高。分析結果亦顯示，危機發生前流動性及償債能力越弱、營運報酬能力越差之公司，發生財務危機之可能性較高。 / This thesis focuses on a group of firms listed in Taiwan Securities Exchange (TSE) and Over the Counter (OTC) experienced financial distress over the period of 1998 to 2003, by applied to a control sample design approach. Based on the pre/after criteria employed in McKeown et al. (1991), this thesis classifies the sample into four subgroups: (1)failure/distress (FS), (2)nonfailure/distress (NFS), (3)failure/nondistress (FNS), and (4)nonfailure/nondistress(NFNS). This thesis first examines the difference in earnings manipulation behavior between the four subgroups. The detecting ability of the precaution indices recently amended by the TSE and OTC is also investigated. Finally, an empirical model is proposed by this research by inclusion of the precaution indices, governance structures and variables employed by the practice communities. The empirical findings indicate that with exception of FNS and NFNS firms, significant differences in all variables are found, representing that the distress firms manipulate earnings before the crisis taken place. The results may imply that these variables might be manipulated by distressed firms to cover the financial failing signals ex ante. For the precaution indices proposed by the TSE and OTC, the empirical findings indicate that compared to healthy firms, the distress firms are found to have worse after-tax/pretax earnings, higher doubt of repaying long-term debts, and unusual fluctuation in prepaid items. The results show that distressed firms have less cash before the occurrence of crisis. As to the indices proposed in this thesis, the analysis shows that distressed firms usually cannot generate adequate cash flow to pay the operation expenses necessary for the next year. It is also found that the distressed firms have higher pledge ratio of share held by the board members and larger deviation in control right from cash flow right. The empirical results of the proposed model show that firms with worse situation indicated by the indices suggested by TSE and OTC, higher share pledge ratio by the board member, larger deviation between control right and cash flow right, have higher probability of experiencing financial crisis. Firms are weak in liquidity, cash imbursement ability and operating return would have relatively higher possibility to face financial distress.