避免人為非理性判斷錯誤，運用統計科學方法建立財務預警模型，有助於企業財務危機發生前即可統計出公司發生財務危機之機率高低，讓投資人與債權人做出理性的投資判斷。財務報表是企業營運活動的紀錄，而財報數字更是經營成果的表現，同時也是各界評估企業的首要條件。
本研究選擇臺灣地區上市公司為分析對象，試圖以相關財務比率做為研究變數，樣本選取期間為1998年至2005年間發生財務危機企業，並依同一產業別中資產規模相近之正常企業配對分析，共選出30家符合本研究定義之財務危機公司及30家配對之正常公司，並以Logit迴歸分析，建立財務危警模式。實證結果發現：
1. 財務危機公司與正常公司兩群體財務比率變數大部份具有顯著性差異。
2. 財務危機發生前一年，長期資金適足率、速動比率、存貨週轉率、固定資產週轉率、現金流量允當比率，為顯著變數。財務危機發生前二年，現金再投資比率、營業毛利率、股東權益報酬率、每股盈餘，為顯著變數。財務危機發生前三年，總資產週轉率、稅前淨利率、借款依存度，為顯著變數。
3. 整體而言，Logit迴歸分析模型預測能力均為顯著，具有實用性及預測效果。整體預測正確度對危機企業預測正確率皆有90%以上，危機發生前一年為96.3%，危機發生前二年為92.6%，危機發生前三年為90.6%。
To avoid none-rational judgment of human being, it will be helpful to figure out possibility of financial crisis of a company by using financial warming model through statistic method; to allow investors and creditors come out a reasonable investment decision. Financial report is the record of business activities; numbers inside the report represent the performance of management and they are also major factor to evaluate the enterprise.
My study is targeting at listed companies in Taiwan stock market, attempt to take financial ratio as variable for studying. Sampling and comparing 30 companies in distressed and normal financial status each from same industry category with similar scale of asset from 1995 to 2005. Through logit regression analysis and financial warming model set up, the analysis result indicates;
There is a significant difference in the variable of financial ratio between enterprises that with and without financial problem.
One year before financial distress happens; (L-T Liab.+SE)/FA %, Acid Test, Inventory Turnover, Fixed Asset Turnover, C/F Adequacy Ratio all are obvious variables. Two years prior to financial distress, the explanatory variables are Cash Reinvest %, Gross Margin %, ROE, EPS. Three years prior than financial crisis, the explanatory and significant variables are Total Asserts Turnover, Pre-Tax Income %, Debt/Eqity%.
Generally speaking, the Logit regression analysis got great functionality to predict financial distress in early stage. It is practical and precisely predictable overall. The result of the study shows that the abilities of correct prediction are all above 90% one year before distress is 96.3%, two year before distress is 92.6%, three years before distress is 90.6%.
