Course Project Part II: Bertram Proprietary Credit Scoring Model

Jordan Spence

Student ID: D01326483

In partial fulfillment of the requirements for

MATH 533 – Applied Managerial Statistics

Keller Graduate School of Management

Dr. Gerard L. Kiely

June 24, 2015

PROJECT PART II: Bertram Proprietary Credit Scoring Model
The preliminary analysis carried out in Part I of our project has shown that the data is consistent and reliable, with no missing values. The next step is to construct preliminary and final models.

A. Create indicator (dummy) variables for the qualitative variables Own/Rent and Location using Minitab.

First, label columns for the new indicator variables: Own, Rent, Urban, Suburban, Rural.

Pull down the Calc menu in Minitab and select “Make Indicator Variables”. In the box labeled “Indicator Variables for … “ put the variable for which indicators are desired. Minitab will automatically code the values of Own/Rent and create two new variables named “Own” and “Rent”. Repeat this process for the variable Location. See screenshot below for new indicator variables:
[pic]

B. Develop a preliminary model and display its output in your paper. Describe its statistical characteristics and state your conclusions. Identify which variables you will keep and those you will drop. Be sure to explain why you made your choices. Be specific.
[pic]

Demonstrated above are variables for potential credit card owners for the proposed system. The current variables are: Credit Score, Income, Household Size, Own/Rent and Location. Some of the variables might be unnecessary for this particular system. For example, household size does not determine whether or not someone has a good credit score. Based on previous statistical information in this project, owning versus renting might also be unnecessary. Previously, it was noted that almost everyone in the suburban areas, whether they own or rent homes, have very high credit scores. Consequently, I will remove both Household Size and Own/Rent variables. All the other variables I will keep such as: Credit Score, Income and Location. These variables will be indicators of whether or not the applicants are worthy of receiving Bertram’s credit card. Interesting qualitative findings are displayed below.

[pic]
The graph above shows less people in the suburban areas rent than those who own their homes. Orange indicates the amount of people who rent, while green represents those who own.
[pic]
The graph above shows that more people in the urban areas rent than those who own their homes. Orange indicates the amount of people who rent, while green represents those who own.

[pic]
The graph above shows that almost everyone living in rural areas rent rather than own their homes. Orange indicates the amount of people who rent, while green represents those who own. C. Run a final model with only the variables you decide to keep. Again, present its output in your paper.

Explain why you think this is a good model. Discuss its statistical characteristics in a way that senior management will understand.

Displayed below are the actual variables that will be used for the Bertram’s Credit Card System. As shown, the three variables are: Credit Score, Income and Location. This is a good model because these variables are the most critical information needed to filter out applicants who are not likely to become delinquent in credit card debt.

[pic]

As shown in the graph below, people living in Suburban areas have the highest credit score ratings out of all three locations. Rural areas are ranked the lowest in terms of credit score and my understanding of this is because country areas operate differently than urban and suburban. Urban areas act as the median/mean of all three areas as shown in the graph.

[pic]

D. Discuss how this model will benefit Bertram House & Hardware. Make the business case.

The aim of this project is to create a Credit Scoring Model for Bertram House & Hardware Company. This system will be part of the new Senior Vice President’s aim to bring all credit processing and risk management in-house at Bertram House & Hardware. This model is effective because it uses the key variables which will determine people who are delinquent from those who are dutiful and trustworthy to receive Bertram’s credit card. This model is also cost-effective because only a few vital variables will be used to determine applicants’ eligibility.

This system is cost-effective because with few strong variables, production time will be reduced and thus, Bertram will incur less associated fees for the system. As Bertram brings all credit processing and risk management in-house, the company will quickly see a Return on Investment (ROI), especially since the credit card will help to keep current customers and attract new ones.

.

NOTE: Part II of the course project will be due at midnight on Thursday, June 25, 2015. ___________________________________________________________________________