|
This research is viewed from the standpoint of the credit of the bank to analyze the evidential risk of financial institute loan to the small-medium business who received the funds through the bank, and expects to find the risk factors of loans during the business pursued the business achievement. Moreover, we try our best to eliminate those risk factors at the credit process and construct a precursor model for the risk management to reduce probabilities of delinquent loans.
The small-medium business is about 98 percent of total businesses in our country and is also the main power of economic development in Taiwan. In the early period of domestic economic development, the small-medium business elaborated the strong tenacity and nice accommodation, creating the miracle of economy in Taiwan to be the focus of world attention. However, restricted to the connatural model, the most significant frailty of the small-medium business in administration is the weakly financial structure and the difficulty in finance. Therefore, the government fell over herself on the finance and financial management of the small-medium business.
This research is based on a small-medium business customers’ credit of a local bank, collocating the credit assurance fund mechanism of the small-medium business and according to the delinquent loans cases provided by institute of creditor’s right management of this bank, and applies the logistic regression analysis, which is one of the statistical methods, to find the main affecting factors of the delinquent loan based on the nine branches of the bank in the middle area of Taiwan. The empirical results reveals that the five factors, including weather the manager has capital asserts or not, a turnover, the turnover’s growing trend with two years, the credit period and interest rate, mainly influence weather the small-medium business will loan delinquently or not. Of the fitted logistic regression model in this study, the classified probabilities of the abnormal business and the normal business are 69.93% and 85.63%, respectively; the type I and type II error probabilities are 30.07% and 14.37%, respectively; the overall accurate classified probability of the model reaches 79.13%.
|