This post looks at what factors go into predicting the outstanding principal and interest balances of student aid–both for direct loans as a whole and the sub-category of income driven repayment loans. The direct loan program is the federal student loan program under which eligible students and parents borrow directly from the U.S. Department of Education at participating schools. Income driven loans include Direct Loan and Federal Family Education Loan borrowers enrolled in an income-driven repayment plan.
I run a regression with outstanding principle and interest balances for student aid in both the direct loan scenario and also the income driven loan scenario as dependent variables. The data for the outstanding balances for both direct loans and income driven loans come from studentaid.gov and are current as of December 31, 2019. They are available by state.
The independent variables in the regression include percentage of men and women in each state in 2018, the average credit score by state for 2017, the median age in each state for 2018, the number of people in each race category (e.g. Hispanic, White, Black, Asian, American Indian) in each state for 2017, educational attainment (e.g. percent high-school or higher, percent bachelor’s degree or higher, and percentage advanced degree or higher) by state for 2017, the marriage rate by state for 2018, amount of disability as measured by the number of Supplemental Security Income (SSI) recipients by state in 2018, and the number of veterans by state for 2016.
Turning first to the regression of direct loan balances on the independent variables, Figure 1, below, shows that race and ethnicity are the only variables that are statistically significant. In other words, a one Hispanic person increase in a state leads to $2,042 increase in outstanding direct loans. A one White person increase in a state leads to a $4,487 increase in outstanding debt. A one Black person increase in a state leads to a $8,855 increase in outstanding debt. Finally, a one Asian person increase in a state leads to a $3,917 increase in outstanding debt. It turns out that Blacks have the highest outstanding debt of any race category.
| Figure 1. Regression Results of Outstanding Principal and Interest balances of Direct Loans on Demographic Variables | |||
| Independent Variables | Coefficient | P-Value | Signficiant? |
| Percent Male | 109,000,000,000 | 0.163 | No |
| Average Credit Score | -27,400,000 | 0.622 | No |
| Median Age | 241,000,000 | 0.184 | No |
| Number of Hispanics | 2,042 | 0.030 | Yes |
| Number of Whites | 4,487 | 0.000 | Yes |
| Number of Blacks | 8,855 | 0.000 | Yes |
| Number of Asians | 3,917 | 0.041 | Yes |
| Number of American Indians | 7,219 | 0.280 | No |
| Percent High School or Higher | -1,790,000,000 | 0.945 | No |
| Percent Bachelor’s Degree or Higher | 9,630,000,000 | 0.806 | No |
| Advanced Degree | 18,300,000,000 | 0.722 | No |
| Marriage Rate | 11,700,000 | 0.913 | No |
| Number of SSI Recipients | -10,084 | 0.424 | No |
| Number of Veterans | -2,874 | 0.567 | No |
As for the regression of outstanding balances of income driven loans on the list of independent variables, I find similar results with the direct loan regression. In other words, only race and ethnicity variables are significant. But, unlike the direct loan scenario, the number of Asians is no longer statistically significant and, in fact, is negatively correlated with income driven loans. Nevertheless, a one Hispanic person increase in a state leads to $1,125 increase in outstanding income driven loans. A one White person increase in a state leads to a $1,1,95 increase in outstanding income driven debt. A one Black person increase in a state leads to a $3,225 increase in outstanding debt. Blacks, again, have the highest outstanding debt balances of any race category.
| Figure 2. Regression Results of Outstanding Principal and Interest balances of Income Driven Loans on Demographic Variables | |||
| Independent Variables | Coefficient | P-Value | Signficiant? |
| Percent Male | 18,600,000,000 | 0.634 | No |
| Average Credit Score | 4,576,992 | 0.871 | No |
| Median Age | -16,400,000 | 0.857 | No |
| Number of Hispanics | 1,125 | 0.019 | Yes |
| Number of Whites | 1,195 | 0.000 | Yes |
| Number of Blacks | 3,225 | 0.000 | Yes |
| Number of Asians | -747 | 0.429 | No |
| Number of American Indians | 2,204 | 0.512 | No |
| Percent High School or Higher | 8,700,000,000 | 0.507 | No |
| Percent Bachelor’s Degree or Higher | -166,000,000 | 0.993 | No |
| Advanced Degree | 9,860,000,000 | 0.705 | No |
| Marriage Rate | 49,900,000 | 0.362 | No |
| Number of SSI Recipients | 5,279 | 0.409 | No |
| Number of Veterans | 3,240 | 0.206 | No |
Looking at the predictors of outstanding student debt, it turns out that race and ethnicity are the most significant factors. This is true for both the direct loan scenario and for only the income driven loans scenario. In both scenarios, blacks had the highest level of outstanding debt–followed by Whites and then Hispanics. The only difference is that, in the income driven scenario, Asians had the lowest levels of debt.
Sources: https://studentaid.gov/sites/default/files/fsawg/datacenter/library/IDRPortfolio-by-Location.xls, https://studentaid.gov/sites/default/files/fsawg/datacenter/library/DLPortfolio-by-Location.xls, https://www.thebalance.com/the-average-credit-score-by-state-4161310, https://www.cdc.gov/nchs/data/dvs/state-marriage-rates-90-95-99-18.pdf, https://en.m.wikipedia.org/wiki/List_of_U.S._states_and_territories_by_median_age, https://www.governing.com/gov-data/census/state-minority-population-data-estimates.html, https://en.m.wikipedia.org/wiki/List_of_U.S._states_and_territories_by_educational_attainment, https://www.ssa.gov/policy/docs/statcomps/ssi_sc/2018/ssi_sc18.pdf, https://www.va.gov/vetdata/veteran_population.asp