中小企業案件逾期放款之預測

The Prediction of Overdue Loans for Small and Medium Enterprises

許光昇(Kuang-Shen Hsu)

指導教授 ： 陳忠榮 ； 陳禮潭

國立中央大學/管理學院/產業經濟研究所碩士在職專班/碩士(2008年)

摘要

依據最新2007年中小企業白皮書(White Paper On Small And Medium Enterprises In Taiwan)統計報告指出，中小企業家數佔全體企業家數比例97.77％，在台灣的經濟發展中佔有相當重要的角色。雖然銀行近年推展中小企業放款業務，但金融機構放款與中小企業本存在著嚴重的資訊不對稱，係因中小企業財務報表普遍失真的情形，而無法取得融資。
長期以來，銀行審查人員對中小企業授信，多是憑經驗累積的主觀判斷。本研究以銀行實際承作之中小企業正常戶及已發生違約的案件做研究分析。
本研究以Logit及Probit模型對中小企業的基本資料及財務報表的各項因子做實證分析，期盼能提高預測其違約之準確率。
實證結果為：
(1)在適度的修正後，Logit及Probit模型總分類預測準確率都已提升到74.78%~75.65%左右。
(2)統計顯著之解釋變數有：公司成立年數、負責人年齡、負責人婚姻狀況、負責人授信額度使用率、負責人近三個月聯徵查詢家數及財務報表是否經會計師簽證等，多以非財務性變數較財務因子還要顯著。
(3)本研究與其他文獻之實證研究相同，以Logit模型較Probit模型佳。

關鍵字

預測準確率 ； 聯徵(聯合徵信中心) ； Logit model ； Probit model

並列摘要

According to 2007 White Paper On Small And Medium Enterprises In Taiwan, small and medium enterprises is 97.77 percent of total enterprises in Taiwan. They played a very important role of economic development in Taiwan’s history. The banks have popularized the loans of small and medium enterprises in recent years. But it has serious asymmetric information between banks and enterprises. The most significant failing of the small and medium enterprises is that the financial statement is always untruthful and causing the difficulty in finance.
For a long time, the loans of small and medium enterprises in banks always depend on the credit censor’s judgment call that accumulated experience in those fields. This research is using the case of Taiwan’s small and medium enterprises both in normal and in default on loans.
This research is an empirical analysis by Logit and Probit models for fundamental data of small and medium enterprises and financial statements factors. We hope that we could improve the percentage of correctly predicting on overdue loans. The empirical results of this analysis are as follows,
(1)The moderate corrections elevated the overall classified correctly predicting probability of both the Logit and Probit models reaches about 74.78% to 75.65%.
(2)We got that the explanatory variables are statistically significant as follows, existent years of enterprise、ages of chairman、marital status of chairman、the chairman’s use ratio of credit、inquiry bank numbers of chairman within 3 months in JCIC and certification of finance by accountant. The non- financial factors are more significant than the financial factors.
(3)This study is the same with the other paper’s empirical results. Logit model is better than Probit model.

並列關鍵字

Probit model ； correctly predicting probability ； JCIC (Joint Credit Information Center) ； Logit model

參考文獻

Altman, E.I. (1968), Financial Ratios, Discriminate Analysis and the Prediction of Corporate Bankruptcy, Journal of Finance, 589-609.

Black, F. and M. Scholes, (1973), The Pricing of Options and Corporate Liabilities, Journal of Political Economy, 81, 3, 637-654.

Bongini, P., L. Laeven and G. Majnoni, (2002), How Good Is the Market at Assessing Bank Fragility? A Horse Race between Different Indicators, Journal of Banking and Finance, 26, 5, 1011-1028.

Boyes, W., D. Hoffman, and S. Low (1989), An Econometric Analysis of the Bank Credit Scoring Problem, Journal of Econometrics, 40, 3-14.

Crouhy, M., D. Galai and R. Mark, (2000), A Comparative Analysis of Current Credit Risk Models, Journal of Banking and Finance, 24, 1, 59-117.