The usefulness of missing information on personal loan applications in differentiating approved from denied loans and late-paying from timely paying loans

This dissertation used the information from personal loan applications from a bank in the South and a logistic regression model based on a loan application to test the usefulness of dummy variables representing missing information to differentiate approved from denied loan applications and late paying from timely paying loans. In the test to differentiate approved from denied loans dummy variables representing the officer's identity, the type of collateral, the risk rate assigned to the loan application, the presence of a checking and or savings account, a co-signor or co-applicant, the name of a close relative, the applicant's years at present address, completion of the related debt section of the application, and the variable for the applicant's date of birth were all found to be useful. In the test to differentiate late paying loans from timely paying loans only dummy variables representing collateral and the presence of checking and or savings accounts were useful along with the actual loan amount and age of the applicant.;The logistic regression to differentiate approved from denied loans identifies one of the lending officers as a significant factor, and further tests indicate that this officer approved nearly 75% of his/her loan applications. None of the loan officer variables was a significant factor in identifying late paying loans. If the loan officer had been approving loans with greater risk than other loan officers are willing to accept, then the rate of late paying loans should have been higher and the loan officer variable would be a significant factor in this regression. Unless some intervening variable is present, it appears that the officer identified as O21 Possesses superior skills.