| Summary: | 碩士 === 國立高雄第一科技大學 === 風險管理與保險所 === 92 === A Study on slack cash-cards for Banks System
Student: Chen Chin-Chuan Advisor: Dr. Hsien-Chueh Peter Yang
Dr. Min-Sun Horng
Department of Risk management and Insurance
National Kaohsiung First University of Science and Technology
ABSTRACT
Consumer banks provide the main stream of profits for most banks. Especially, the sales from cash cards make up the largest portion of the revenues of consumer banking. This financial product has the advantages of high spreads, high efficiency and high returns. Cash cards has already become the most important product of generating profits for banks along with the increasing of the number of holders and the using rates year by year. The purpose of this study is to provide a whole completed set of lending evaluation model in which the main features of cash cards are incorporated and lending risks are diversified. Among the main features are: providing the smallest operation cost and the fastest approving speed, reducing the loss of slack cards and increasing operating profits.
In the study we would like to find the characteristics of slack card holders. The sample data are taken from one financial institution locating in the southern part of Taiwan. There are eleven variables including the type of branch offices, gander, source of applicants, income levels, occupations, family members, the type of housing, education levels, the number of children, working experiences and age. After all, there are four factors found to be associated with the probability of slack cards. They are gender, source of applicants, the type of housing, and education levels.
The final logistic regression model of slack cards is
logit(π) = -3.0024 + 0.6783X3 + 0.5269X2 + 0.4728X8 - 0.3614X7.
The final probit model of slack cards is
Φ-1(π) = -1.6964 + 0.3572X3 + 0.2878X2 + 0.2539X8 - 0.1936X7.
X2 :gender ，X3 :source of applicants ， X7 :the types of housing
X8 :education levels ，
The correct rate of slack card through logistic regression model is 69.90%.
The correct rate of slack card through the probit model is 69.39%.
Both the logistic regression model and probit model show that gender, the source of applicants, the type of housing and education levels affect the probability of slack cards. We also find that a female, issued card through marketing agents, living in her own property and with higher education level has a higher probability to become a slack cardholder. The findings are consistent with our expectation.
|
|---|
