- 學位論文
上櫃公司全額交割股之信用違約風險探討：羅吉斯迴歸之應用
The study on the full-delivery stocks of credit default risk for OTC companies： Application of logistic regression.
摘要
台灣的經濟發展一直是個以中小企業為主體的國家，這使得台灣產業具有高度的企業家精神、對環境變化的快速反應、高度彈性、完整的產業網路特性，面對金融海嘯所伴隨而來景氣衝擊，能夠快速反應調整，對於穩定我國就業市場以及經濟發展具有舉足輕重的地位。而營運週轉資金的充裕與否，常常是中小企業能否永續經營的關鍵要素之一。本研究採用羅吉斯迴歸分析，從量化實證的角度來探討中小企業之財務變數，建構一套適合中小企業之信用風險預警模型，找出關鍵具影響的變數，期望作為金融機構建立信評機制時重要參數之參考，降低中小企業中資訊不對稱所帶來的違約風險，提高金融機構對中小企業的放款意願。本研究之模型建構流程乃針對所蒐集之中小企業財務變數，區分為經營能力、獲利能力、現金流量、成長率和償債能力五大類。參考企業財務危機衡量變數為基礎選擇41個變數，再以t檢定和羅吉斯迴歸篩選出9項最有預測力之財務變數，以建立違約機率預測模型。事件發生前一年在訓練模型上所得到之結果與測試模型所得之結果相近，第二年及第三年結果的比較上有較大的差異，顯示隨著危機發生時點接近，預測準確度會有上升效果，無法窗飾危機即將發生。 經過實證研究發現，當採用0.5的機率做為分割點時，整體預測準確率以危機發生前一年最高，接著遞減，若採用0.3的機率做為分割點時，平均整體正確率不如以0.5的機率做為分割點，無論是在測試樣本或訓練樣本都是同樣的結果。 但是若是預測型一誤差，採用0.5的機率做為分割點時，預測準確度不如以0.3來的好，因此我們可以採用前一年的數據做參考指標，但是要區分違約公司是否被歸類為正常公司，要用0.3分割點作為判斷依據。 關鍵字：中小企業、信用風險、羅吉斯迴歸、全額交割股、上櫃公司
並列摘要
As the mainstay nation, Taiwan's economic development has always been a small and medium enterprises. This made Taiwanese industries have a high degree of entrepreneurship with rapid response to environmental change; it’s highly flexible, and has complete industrial network properties. From the impact of a financial tsunami, it is capable of rapid response adjustments to stabilize the employment market as well as the economic development for our country, which is in a pivotal position. Whether operating working capital is sufficient or not, sustainable development is one of the key elements in SMEs. This study uses logistic regression analysis from the perspective of small and medium enterprises in financing of quantitative empirical variables constructed a credit risk early-warning model suitable for SMEs; identify key influential variables, expected as financial institutions credit rating mechanism reference of key parameters, reduce the SMEs defaults caused by information asymmetry risk, and improve financial institutions ' willingness to loan for SMEs. A model construction processes of this study aims to gather small business financial variables; distinguished for viability, growth rates, profitability, cash flow, and solvency for five categories. References of enterprise financial crisis is based on the measured variables, there is 41 variables selected; the t-test and logistic regression identified 9 of the most predictive power of financial variables, to establish a default rate forecasting model. Events in training a year earlier with the results of tests on a model is similar to results of the model, the comparison of the results for the second year and third year shows a considerable amount of difference, it displays the points close to the crisis, the predicted accuracy can be increased by the effect, and prevent the tracery crisis from happening. Researchers found that when using a 0.5 chance of doing a split point, the overall forecast accuracy for a year before the crisis the Supreme, is decreased. With a 0.3 chance of doing a split point, the average overall probability of accuracy is better than 0.5 as a division point, they are both in the test samples(s) with the same result. If we want to predict type I error, using a 0.5 chance of split point, which is not recommended compared to 0.3 chance of split point; we can use the previous year’s data as reference pointers to identify whether the defaulting company is classified as a normal company, using the 0.3 chance of split point is the first step when judging. Keywords: SMEs; Credit risk; Logistic regression; Full delivery unit; OTC companies.