Discussion #7-Dummy-Variable-Regression

ANALYTICS REPORT TO: TAIWANESE CREDIT CARD COMPANY FROM: KAEDEN CALDWELL SUBJECT: DUMMY VARIABLE DATE: NOVEMBER 13 TH , 2023 Introduction This document was created to accurately predict a person’s credit limit and chances of default on next months payments. To find this, we ran two regression analysis test and a several different variables within each test. We used Age, Education, Martial Status, Gender, Average Bill Amount, and Average Payment Amount as part of our variables when we ran the test. Data Analysis Credit Limit: ^ Credit Limit = 3353.72 + 363.21 Feamle ( d ) − 1140.98 High School ( d ) + 2319.56 GradSchool ( d ) − 2974.09 Single ( d ) + 66. We are 11.86% of the way toward perfectly predicting credit limit using this model. Predictions of credit limit using this model are off by an average of $4029.06 Females have a credit limit of $363.22 higher than males, on average and all else constant. Those with a high school degree have a credit limit of $1140.98 lower than those with a university degree, on average and all else constant. Those who went to Graduate School have a credit limit of $2319.56 higher than those with a normal degree, on average and all else constant At zero years old, single people would have a credit limit 2974.02 lower than married people, on average and all else constant For married people, as age increases by 1 year, credit limit increases by $51.10, on average and all else constant. For single people, as age increases by 1 year, credit limit increases by $118.02, on average and all else constant. ^ Credit Limit = 3353.72 + 363.21 ( 1 ) − 1140.98 ( 1 ) + 2319.56 ( 0 ) − 2974.09 ( 0 ) + 66.92 ( 0 ∗ 45 ) + 51.10 ( 45 ) = 4875.45 Default Payments: P ¿ Page 1 of 4

We are 15.31% of the way toward perfectly predicting chance of defaulting using this model. Predictions of chance of defaulting are off by an average of 41 percentage points. As average bill amount increases by $10, chance of defaulting increases by 0.0067 percentage points, on average and all else constant As average payment increases by $100, chance of defaulting decreases by 0.016 percentage points, on average and all else constant. As age increases by 1 year, chance of defaulting increases by 0.00086 percentage points, on average and all else constant. Females have a 2.98 percentage point lower chance of defaulting than males, on average and all else constant. Prediction: P ¿ or 18.84 Recommendation/Insight: Throughout the data, we saw that the men have higher bill amounts meaning the men spend more per month on average when compared to women. There is a gender bias, however it is in the opposite direction. While the women on average have higher credit limits than men, men tend to spend more money even with a lower limit. One recommendation I would have to make our model more accurate, would be to include customers salary/earnings, or maybe include someone debt (if any) in the data to get a better understanding of people’s spending. Lastly, we could also include customers living situations, for example we know the credit limit may be different for someone who lives in a house when compared to someone living in a apartment. Conclusion A key finding, we saw throughout the data was that there was a gender bias, however it was in the opposite direction than we thought. Women have a higher credit limit compared to men, but men still tend to spend more money per month than women. With these finding we concluded that our models were not accurate enough to use, but it is important that we add other variables into the analysis (Living situation or Income). If you have any questions on our model or need further clarification, please reach out to me at KaedeCaldwell@YLC.com Page 2 of 4

SUMMARY OUTPUT Regression Statistics Multiple R 0.123012675 R Square 0.015132118 Adjusted R Square 0.014969973 Standard Error 0.411797253 Observations 24301 ANOVA df SS MS F Significance F Regression 4 63.30287335 15.82571834 93.32468669 6.64348E-79 Residual 24296 4120.042256 0.169576978 Total 24300 4183.34513 Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Intercept 0.227591111 0.01147998 19.82504461 8.79358E-87 0.205089643 0.250092579 Avg Bill 6.66509E-06 1.3433E-06 4.961719827 7.03436E-07 4.03213E-06 9.29805E-06 Avg Payment -0.00016774 9.23222E-06 -18.16893243 2.7819E-73 -0.000185835 -0.000149644 AGE 0.000865721 0.000290412 2.98101523 0.002875791 0.000296497 0.001434946 Female (d) -0.029772152 0.00543564 -5.477211953 4.36362E-08 -0.040426341 -0.019117963 Kaeden Caldwell Figure 2– Default Regression Analysis Page 4 of 4