The Relationship Between Student Debt and Earnings

Note: This piece first appeared on the Brookings Institution’s Brown Center Chalkboard blog.

Student loan debt in the United States is now over $1.25 trillion, nearly three times as much as just a decade ago. The typical student graduating with a bachelor’s degree with debt (about 70 percent of all students) now owes between $30,000 and $40,000 for their education, about twice as much as a decade ago. Although taking on modest amounts of debt in order to pay for college is generally a good bet in the long run, colleges with similar admissions standards and resource levels leave students with different amounts of debt.

College Scorecard data highlight the large amount of variation in what high-debt undergraduate students borrow across colleges with similar admissions criteria.1 The figure below shows the distribution of the 90th percentiles of debt burdens (in 2016 dollars) for students who left 1,156 four-year public and private nonprofit colleges in 2006 or 2007, broken down into three selectivity categories.2 Not surprisingly, the most selective colleges, which have the resources to offer more scholarships and fewer students with financial need, have lower debt burdens than somewhat selective or less selective colleges. These differences in borrowing by selectivity are larger than by type of college, as median debt at public colleges was only about $2,400 more than at private nonprofit colleges.

brookings_fig1_sep16Attending college and taking on $40,000 or even $50,000 in debt can be an outstanding investment in a student’s future—but only if students from that college actually end up getting good jobs. I then examined the relationship between 90th percentile debt burdens upon leaving college in 2006 and 2007 and the median earnings of students in 2011 and 2012 who began college in 2001 and 2002.3 The figure below shows that colleges that tend to have higher amounts of student debt also tend to have lower earnings in later years, which is in part due to student characteristics and their prior family income rather than the causal impact of the college. The correlation coefficient between debt and earnings is about -0.35 overall, but between -0.10 and -0.20 within each selectivity group. This suggests that colleges with higher debt burdens also have higher earnings, but much of the relationship between debt and earnings can be explained by selectivity.

brookings_fig2_sep16An old rule of thumb in paying for college is that students should not borrow more for a bachelor’s degree than they expect to earn one year after graduation. Although the presence of income-driven repayment programs allows students to repay their federal student loans even if they make less money, the debt-to-income ratio is still a useful way to judge colleges. The final figure shows the distribution of colleges’ debt-to-income ratios using the initial debt upon leaving college for high-debt students and annual earnings approximately five years later. A ratio over 1 at this point is a major concern, as earnings should grow considerably during a student’s first few years after college.

brookings_fig3_sep16Few high-debt students at the most selective colleges likely have issues making enough money to repay their loans, as just one of the 191 colleges in this category had a debt-to-earnings ratio above 1. Just under 15 percent of the somewhat selective colleges had ratios above 1, while about one-third of the least selective colleges had ratios above 1. This reflects the fact that financially-struggling students who attend less selective colleges (roughly 13% of the undergraduates in my sample, or about 800,000 students) take on more debt and earn less money than high-debt students at highly-selective colleges.

With student debt being a growing concern among Americans and playing a key role in the presidential campaign, students and their families are wise to consider the likely return on their investment in higher education. As the data show, some colleges do leave their former students with less debt than other similar colleges. But among less-selective colleges that enroll large numbers of lower-income or minority students, some amount of debt is almost unavoidable. Students should not seek to avoid all debt, but they should be mindful that even among broad-access institutions, colleges vary in how much debt their students take on.


1 Debt and earnings data from the College Scorecard combine students who graduate with those who dropped out.

2 Colleges in the Barron’s categories of special, noncompetitive, and less competitive are in my lowest selectivity tier (n=195), colleges in the competitive, competitive-plus, and very competitive categories are in the middle tier (n=770), and all others are in the highest tier (n=191).

3 The College Scorecard does not track debt burdens by when students started college (only when they left), so I estimated that students either graduated or left college in about five years.

Do Presidential Debates Increase Student Applications?

Tonight is the first presidential debate of the 2016 general election season, and this clash between Democrat Hillary Clinton and Republican Donald Trump could top 100 million viewers. (I won’t be one of them, as I’m teaching tonight.) The host site, Hofstra University, is actually the second choice—as Wright State University pulled out over the high price tag this summer. Hofstra is paying about $5 million to host the debate, with the costs generally covered through three donors.

Hosting a presidential debate is undoubtedly a great public relations opportunity for a university, similar to making a big run in the NCAA basketball tournament or making a big football bowl game. Some research has shown that big-time athletics success is associated with increased student applications in the following year, so the media circus following a presidential debate (Hofstra is trending on Twitter as I write this post) could have similar results.

Hofstra also hosted a presidential debate in 2012, so I looked at what happened to the number of applications they received before and after the debate compared to their defined group of peer institutions. The data are below:

Name 2011-12 2012-13 2013-14 Pct increase, 2012-13 to 2013-14
Adelphi University 8278 9184 8654 -5.8%
American University 18706 17039 17545 3.0%
Boston University 38275 41802 44006 5.3%
Drexel University 48450 40586 43945 8.3%
Fordham University 31792 34070 36189 6.2%
George Washington University 21433 21591 21756 0.8%
Hofstra University 18909 21376 22733 6.3%
Ithaca College 13436 13813 15658 13.4%
LIU Post 7369 7209 6001 -16.8%
Marist College 11399 11466 10351 -9.7%
New York University 41243 42807 45779 6.9%
Northeastern University 43255 44208 47364 7.1%
Pace University-New York 10623 11778 12885 9.4%
Quinnipiac University 18651 18825 20699 10.0%
Seton Hall University 6436 10180 10735 5.5%
St John’s University-New York 54871 52972 51634 -2.5%
Syracuse University 25884 25790 28269 9.6%


Hofstra did see a 6.3% increase in applications between 2012-13 and 2013-14, compared to a 4.6% increase across its peer institutions. But other peers, such as Ithaca, Quinnipiac, Syracuse, and Pace saw even larger increases. So it appears that the debate brought plenty of pride to Hofstra, but there was not an unusual jump in applications after the debate aired.

Students Shouldn’t Be Terrified of Borrowing for College

I wrote the below letter to the editor of the Star-Ledger, New Jersey’s largest newspaper, in response to a truly woeful editorial piece that they recently published on student loan debt. (Note: They eventually ran the letter, but here is a slightly revised version for your enjoyment.)


As a college professor who researches the implications of student loan debt, I was dismayed to read the Star-Ledger Editorial Board’s recent piece titled “Why we should all be terrified of student loans.” Yes, the $1.25 trillion in outstanding student loan debt is a concern, but the typical amount borrowed for a bachelor’s degree is more manageable—about $30,000 per student. Students who borrow from the federal government can also enroll in income-driven repayment programs that allow them to make small or no payments if their income is low.

The “terrified” headline has the potential to scare students and their families away from making a worthwhile investment in college. Research shows that low-income, first-generation, and minority students are particularly averse to borrowing for college, even when borrowing a reasonable amount of money would help them attend and graduate college. Students and their families should be careful about taking on too much debt, particularly from programs like New Jersey’s state student loan system that do not allow payments to be tied to the student’s income. But students should not be terrified of taking out modest loans from the federal government to make college a reality.

How Did ITT Tech’s Outcomes Compare to Other For-Profit Colleges?

Last week, ITT Technical Institute announced that it would close all of its colleges, affecting approximately 40,000 students and 8,000 employees. This closure was expected after the chain of for-profit colleges stopped enrolling new students in late August after the U.S. Department of Education cut off federal financial aid dollars for new students a few days earlier. This closure, which could cost taxpayers up to $400 million through forgiven loans, has generally been celebrated by those on the political left while conservatives and those in the for-profit sector are concerned that the federal government is trying to severely restrict the for-profit college industry.

Given that some of the concerns about ITT Tech were about poor student outcomes, I examined ITT Tech’s outcomes relative to other degree-granting for-profit colleges on three important metrics: median debt burdens of former students who took out loans, the percentage of students seven years after entering repayment, and median earnings of former students ten years after entering college.1 I restricted my analysis to the 415 degree-granting for-profit colleges that reported data on all three of the outcomes, combining branch campuses that reported the same outcomes as other colleges in the same system.2

Median debt

The median debt burden of all former ITT Tech borrowers was $12,473 (as indicated by the red line on the below chart), slightly above the median amount of $11,993. This suggests that among for-profit colleges granting associate and/or bachelor’s degrees, ITT Tech’s debt burden was fairly typical.


Loan repayment rates

Seven years after entering repayment, 58.2% of former ITT Tech students paid down at least $1 in principal on their federal student loans. This is slightly worse than the median rate of 61.3% across similar for-profit colleges.



On this metric, ITT Tech looks pretty good relative to other for-profit colleges. ITT Tech students who received federal financial aid had median earnings of $38,400 ten years after college entry, well above the median of $29,200 and close to the 90th percentile among similar institutions. However, these data are based on students who entered college in the early 2000s, when ITT Tech looked much different than it did in recent years.


Based on financial outcomes, ITT Tech’s former students (at least those who enrolled at least several years ago) did as well or better than other for-profit colleges. This does lend some credence to defenders of ITT Tech who were concerned about the Department of Education targeting the institution. However, others have noted that ITT Tech’s closure may have been self-inflicted through an ill-advised private loan program that led to fraud charges. In any case, other for-profit college chains are likely to face additional scrutiny in the future—from politicians and accreditors alike.


1 I did not examine graduation rates, as many for-profit colleges have very few students in the first-time, full-time cohort of students that are currently used to calculate graduation rates for the federal government.

2 ITT Tech had 143 branch campuses in the College Scorecard data, and 141 of them had the same reported outcomes. I analyzed those campuses as a single institution, dropping the two small campuses that had different reported outcomes.