次貸風暴時期影響中小企業授信違約之實證研究-以直接保證為例

An Empirical Study of Small & Medium Enterprise (SME) Credit Defaults During the Subprime Financial Crisis Using Evidence from Direct Credit Guarantee Programs of Taiwan

高又雅(Yu-ya Kao)

指導教授 ： 劉錦龍

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(2011年)

摘要

本研究為討論影響中小企業授信違約之因子，分別架構出影響企業停業、貸款展期及貸款逾期之迴歸模形，蒐集2008-2009年次貸風暴爆發及延續期間信保基金直接保證460筆貸款保證之原始資料為樣本，以2010年底為基礎統計停業、展期或逾期以進行實證模型分析，為深化模型預測可行性，變數除企業財務資料，並延伸至非財務因子及產業屬性探討，以Logit 迴歸模型架構影響企業停業重要變數，以Multinominal Logit迴歸模型架構影響企業展期及逾期之重要變數。
實證結果指出，營收、成立年限與停業及逾期之違約機率有顯著正向關係，資本額、毛利率、核准金額與停業及逾期之違約機率有顯著負向關係，具有會計師財簽、負責人為已婚與展期及逾期有顯著負向關係，另在產業屬性方面，實證結果發現電子業、營建工程產業與停業有顯著正向相關。在預測正確率方面，以Logit迴歸模形評估影響停業之關鍵因子，預測正確率達到90%以上，而Multinomial Logit迴歸模形評估影響貸款展期及逾期之情形，預測正確率達到70%以上。
本研究之實證結果期對於後續信用保證申請案之准駁因素及風險控管因子提供參考及建議。

關鍵字

邏輯斯迴歸 ； 信用保證 ； 違約 ； 中小企業授信

並列摘要

The primary purpose of this study is to investigate the determints of non-performance loan (NPL) of small and medium size enterprises (SMEs). The 460 lending datas from 2008 to 2009 of Director Credit Guarantee of Small and Medium Enterprises Guarantee Fund (SMEG) in Taiwan were collected for empirical analysis. The key determints we examine include not only the financial indicators, but also the non-financial indicator as well as industrial attributes. The Logitic regression model is employed in determining the key factors affecting closure of business. Moreover, the multinomiatl logit regression model is employed in estimating the factors of credit extension and credit default.
The major empirical findings revel that the indicators including revenue and duration are significantly positive related to closure and credit default. It also suggests that the indicators including capital, gross ratio and approved amount are siginificantly negative related to closure and credit default. Futhur, it shows that firms with finance certificated by CPA and marital status of the chairman are important indicators which are significantly negative related to credit extension and default. Finally, it appears that elecronic and construction indusy are significantly positive related to closure of business. The prediction accuracy rations of Logistic model and Multinomal Logit model of the study reach more than 90% and 70%, respectively. Results of the empirical study may serve as reference for credit approving and risk control policy.

並列關鍵字

Logitic model ； Multinomial logit model ； credit default ； SME loan ； credit guarantee

參考文獻

[2]陳錦村（1998），「競爭、往來關係與銀行授信行為之研究」，Journal of Financial Studies，561-90。

[5]陳家彬、賴怡涒（2001），「台灣地區銀行放款有無擔保之決定因素-Logit模型之實證分析」，管理評論，20（1），129-159。

[7]陳家彬、賴怡洵（2003），「銀行往來關係和授信行為之研究」，經濟論文，31（3）497-539。

[12]陳家彬、賴怡洵、許茲盈（2004），「商業銀行關係放款之實證研究」，財務金融學刊，12（2）43-86。

[13]鄭宗明（2005），「影響商業銀行授信決策因素之研究」，中興大學應用經濟系碩士論文。