Higher Ed Data Central: Input-Adjusted Measures

The fifth capability of Education Sector’s new Higher Ed Data Central that we would like to highlight (see links for the first, second, third, and fourth capabilities) is the ability to develop input-adjusted measures. For example, the Department of Education recently released student loan default rates by college. A list of colleges with the highest default rates could be compiled, but would be pretty uninformative: There are several other factors, or inputs, that influence default rates. Accounting for these helps paint a much clearer picture of where and among whom defaults persist most, and least.

Take, for example, family income. Students from affluent families can often get financial help with their loans from their parents if they are in danger of defaulting, while this option is typically not available to students from low income families. Thus, we would expect default rates to be higher at colleges that enroll more low-income students.

While there is no publicly available data source that would allow us to adjust default rates for family income, IPEDS does contain the percent of Pell grant recipients at each college; and since Pell grants are typically restricted to low-income students, this is a good proxy variable for the number of high-risk students at each college. For instance, since only 7 percent of students at Washington University in St. Louis receive Pell grants, but 73 percent of students at Florida Agricultural and Mechanical University do, we would expect default rates to be higher at FAMU even if everything else about the schools were exactly the same.

The same logic holds for other risk factors, such as the percent of students attending part-time (a proxy for nontraditional students) and the average loan taken out by first-year students (a proxy for debt load). These inputs can be used to create an input-adjusted default rate, which is essentially a prediction of what we would expect the default rate to be based on the various inputs. This value can be compared to the colleges’ actual default rate, allowing us to answer the question, “Which colleges are performing above expectations given their student population?”—a much better question than “Which colleges have the highest default rates (ignoring differences in student population)?”

The chart below shows the predicted (based on percent of part-time and Pell students and average student loan) and actual default rates for four-year colleges. Those colleges below the line have a lower default rate than we would expect given the types of students they serve, and they are therefore doing a good job. Those colleges above the line have a higher default rate than we would expect and are therefore doing a poor job. On both the far left and far right, there are colleges with actual default rates around 10 percent. But even though they have the same default rates, the ones on the far right are doing a much better job than the ones on the far left because the colleges on the far right have a higher predicted default rate (meaning they have a more risky population of students).

We hope you’ve enjoyed this whirlwind tour of the capabilities of Education Sector’s Higher Ed Data Central. Now it’s your turn.  Is there something you want to know? A potential trend in higher ed you want to identify? Leave comments or email us about what topics or relationships you would like to see analyzed, and we’ll post a new chart each Tuesday. (We’ll also occasionally update prior charts based on reader suggestions, so send those our way too.)