The National Bank of Fort Worth, Texas wants to examine methods for predicting sub-par payment performance on loans. They have data on unsecured consumer loans made over a 3-day period in October 2013 with a final maturity of 2 years. There are a total of 348 observations in the sample. The data, which have been transformed to provide confidentiality, include the following:
PAST DUE: Coded as 1 if the loan payment is past due and zero otherwise
CBSCORE: Score generated by the CSC Credit reporting agency from 400 to 839 with higher values indicating better credit rating
DEBT: Debt ratio calculated by taking required monthly payments on all debt and dividing it by gross monthly income of applicant and co-applicant. This ratio represents the amount of the applicant’s income that will go towards repayment of debt
GROSS INC: Gross monthly income of applicant and co-applicant
LOAN AMT: Loan Amount
You have been asked to examine the feasibility of predicting past-due loan payment. Report your results to the bank in a two-part report. The report should include an executive summary with a brief non-technical description of your results (less than 1-page) and an accompanying technical report with the details of your analysis. The data are in an excel file posted on eLearn.
For the report, you should consider the following: Use of logistic to analyze the data; appropriate variables which are useful in predicting performance; the hit-rate in the estimation sample and how it compares with appropriate benchmark criteria.
