Stats Project.docx

Task:

Use the Data Set “Loan Pricing Data Set.xlsx”

You recently joined a boutique consultancy that specializes in providing data-driven insights and decision support to the financial sector.

For your first assignment, Isabella Euler, one of the firm’s thought leaders, assigns you to a client who wants to improve the procedures of its Auto Financing Group (AFG). The client’s newly appointed VP believes that the existing procedures for pricing car loans are outdated and adversely affect profitability. This state of affairs is the result of organizational challenges, as the division has been cobbled together from multiple groups over the years. The ultimate ambition is to introduce sophisticated price optimization models. However, for now the VP would like you to study a sample of recently approved car loan applications to provide clarity on the AFG’s current approaches and results.

Specifically, you are to provide guidance on the following questions:

1) What is the relationship between the annual percentage rate (APR) that is offered to a prospective customer whose loan application has been approved, and the characteristics of that customer’s loan application? The AFG makes each approved customer a take-it-or leave-it APR offer, but the process for determining the quoted APR is currently not standardized. It is therefore not clear exactly how the quoted APRs are determined.

2) What factors are related to a customer’s loan offer acceptance? The VP of the AFG would both like to have a way to predict how a given applicant pool will respond to offers under the current APR quotation approach, and also understand which factors are related to these responses.

3) Relatedly, which factors are relevant for predicting the profitability of a given applicant pool under the current APR quotation approach?

Overall, record your approach for each step; why you did what you did. And summarize/infer the data results for each question.

Concepts: Confidence Intervals, Hypothesis Testing, Multiple regression, Regression for prediction, Regression for prescription, Linear regression assumptions, Collinearity, Binary Dependent Variables: Linear & logistic regression

Rough Steps/Guideline:

1)

-0.867 adjusted R squared (expected answer) -Independent variables used: credit score, amount, term, competitor, prime, internal_risk = high, internal_risk = very high, internal_risk = very low, loan_type = refinance -If internal risk is very high it’s 1 and the other 2 are 0 and vise versa. -Use Interaction terms? -If you don’t use interaction terms, run the correlation matrix in Stattools (Summary statistic, correlation and covariance) to see how highly correlated they are. Then do a linearity test and then run the regression to see which gives best sample fit, who has the highest adjusted R squared, then check linearity again -Suspected Answer: If internal risk is high or very high than the interest rate is high -Explanatory not prediction for Q1 -Do not use just holdout sample, run regression on entire data set. Holdout is “1” and “0” is main sample data.

2) -Do a predictive? analysis to see whether they will accept the offer rate or not -Holdout sample column added using 15%. If it is “1” then it is holdout sample, if it is “0” then it is main sample data.

3)

-Profitability we will have to run it with an interaction variable. -Profitability is assumed to be acceptance rate – prime which gives us the spread. Create a column for this as the dependent variable? -Internal risk: make dummy variables (4 for each type). Regression take 3, leave one out -Refinance column: put 1 and 0 and see how the coefficient looks like -What interaction terms should we create? -Holdout sample column added using 15%. If it is “1” then it is holdout sample, if it is “0” then it is main sample data. -Refinance column: put 1 and 0 and see how the coefficient looks like