Student Loans Cost Taxpayers $645 Billion More Than We Were Told

Among the debates over federal student loans, two of the most important are: 1) should student loans be used to subsidize college? and 2) are student loans subsidizing college?

Should Student Loans be Used to Subsidize College?

Regarding the first debate, scholars have long pointed out that there is a role for government facilitation of student loans, but that this involvement should not include subsidization. As Eric Hanushek wrote back in 1989:

Because of the difficulty of borrowing for higher education, it is also appropriate for the federal government to facilitate the market for loans. But again there is no strong argument for subsidizing these loans.

Susan Dynarski and Daniel Kreisman went one step further and laid down the gauntlet that no serious scholar or analyst has successfully challenged, writing:

The government should seek neither to make nor to lose money from student loans. Student loans correct a capital market failure … Federal student loans therefore solve a liquidity problem, not a pricing problem. Student loans are appropriate neither for raising revenue nor for subsidizing college.

I elaborated on their reasoning in a recent report, writing that even if you think we should increase subsidies for college,

Student loans are a terrible method of subsidizing higher education. Providing subsidies through loans … is poorly targeted, distributing the subsidy only to those who borrow. Yet around half of all community college graduates do not take out loans. Using loans to subsidize higher education also gives the biggest subsidies to those who borrow the most (relatively wealthy graduate students). Subsidies via loans also distribute the subsidy at the wrong time. Rather than lowering upfront costs when enrollment decisions are made, it offers a benefit to students far in the future, after they’ve already graduated.

Some advocates who would like to see free college will support anything that further subsidizes it, including zero interest rates on loans, loan forgiveness, or indefinite repayment pauses. But they have talking points, not logical arguments. If you ask whether the government should subsidize the rich more than the poor, they say no, but then proceed to push for policies like loan forgiveness or repayment pauses that shower the rich with more benefits than the poor. At best, these advocates simply don’t understand the effects of the policies they support. At worst, they are just another special interest seeking to plunder the taxpayer with whatever justification sounds good in a focus group.

The bottom line is that regardless of whether you think subsidies for college should be increased, decreased, or eliminated, student loans should be budget-neutral for the federal government.

Are Student Loans Subsidizing College?

This brings us to the second debate: does the government currently subsidize student loans? The answer to this question hinges on the subsidy rate. A subsidy rate of 10% means that for every $1 the government lends, it loses $0.10. Subsidy rates can be negative if the government is making money off student loans (e.g., a subsidy rate of -5% would mean the government makes a profit of $0.05 on every dollar it lends).

To determine the subsidy rate, you need estimates of future repayments as well as a discount rate to convert future payments into their present value, since $100 ten years from now isn’t worth as much as $100 today.

[Related: “Republicans are Pushing for the Wrong Student Loan Reform”]

Historically, most of this debate has centered on the discount rate, which boils down to using FCRA (Federal Credit Reform Act) or fair-value discount rates. FCRA uses the interest rate the government pays when it borrows as the discount rate, while fair-value uses an estimate of the interest rate the market would use. For example, suppose the FCRA discount rate is 2% and the fair-value rate is 5%. Under those discount rates, a payment of $100 ten years in the future would be worth $82 today under FCRA but only $61 under fair-value.

While fair-value is the appropriate method to use because it provides a more accurate assessment of the cost of extending a loan, under current law, FCRA is used to estimate subsidy rates. The Congressional Budget Office currently estimates a difference of 16 percentage points between the FCRA and the fair-value subsidy rate, meaning that every $1 in lending will cost the government $0.16 more under fair-value than FCRA. If this difference in subsidy rates is applied to the last 25 years of student lending, student loans cost taxpayers $334 billion more than originally estimated.

But the other side of estimating the subsidy rate, forecasting future repayments, has been all but neglected until now. Fortunately, a new bombshell report from the Government Accountability Office (GAO) analyzes how accurate estimates have been and finds that repayments have been massively overestimated for a quarter century. In particular, GAO found that the Department of Education (ED) originally estimated that the past 25 fiscal years of student lending (1997-2021) would generate $114 billion in profit for the federal government. The latest estimates are that those loans will cost the government $197 billion instead. In other words, the last quarter century of student lending is $311 billion more expensive than we were told it would be.

Some of this was not ED’s fault. For example, about a third of the difference ($102 billion) is attributable to the COVID-inspired student loan repayment pause initiated by President Trump and renewed by President Biden. This giveaway is costing taxpayers dearly, but it is not something that ED could have foreseen and accounted for when it was making loans 20 years ago.

But 61% ($189 billion) was due to flawed assumptions and predictions by ED.

Some of these mistakes are more forgivable than others. For example, we are virtually certain that recessions will happen in the future (there’s a very good chance we’re in one right now), and that recessions reduce student repayment (costing the government more). But recessions aren’t predictable enough for ED to easily account for them beforehand. I would argue that ED is still making a mistake by assuming that there won’t be any recessions over the life of any loan it makes, but my recommended solution (a weighted average of repayment scenarios accounting for the frequency and severity of recessions) is open to debate as well.

Less forgivable are ED’s flawed projections about what types of repayment plans students would choose and how much students would pay under those plans. The income-driven repayment plans cost the government more, and more students have been using these plans than ED predicted. Under these plans, payments are set as a certain percentage of income rather than a fixed dollar amount like a mortgage, and any remaining debt after 10 to 25 years of repayment is forgiven. While income-driven repayment plans established decades ago required 20% of income for 25 years, the most generous recent plan now only requires 10% of income for 10 years. ED apparently failed to realize that much more generous plans would be much more popular. This mistake cost taxpayers $70 billion, and the related overestimate of how much they would repay under the plans cost taxpayers another $68 billion. ED also underestimated the cost of defaults to the tune of $23 billion.

[Related: “Biden’s Student Loan Forgiveness Plan Would Be a Huge Mistake”]

GAO provides the annual and cumulative effect of these errors in both dollars (figure 3) and in terms of the subsidy rate (figure 6). Forecasting is difficult, but if your methods are sound, you should overestimate some of the time and underestimate some of the time. ED has underestimated the cost of student loans to taxpayers in 25 out of 25 years. This indicates that ED’s method is deeply flawed.

Over the past 25 years, ED missed the mark by $311 billion. In terms of the subsidy rate, ED originally estimated a subsidy rate of -6% (meaning the government would make a profit of $0.06 for every $1 it lent), but in reality, the subsidy rate was 8.88%, meaning the government lost $0.0888 for every $1 it lent.

What Should Be Done?

For a quarter century, the government claimed that student loans would generate substantial profits ($114 billion). But changes in repayment programs and flawed assumptions about student repayment underestimated costs by $311 billion. Using an inappropriate discount rate (FCRA instead of fair-value) underestimated costs by another $334 billion. Thus, instead of making $114 billion, the government likely lost $531 billion. This means that the last 25 years of student lending cost taxpayers $645 billion more than we were told they would. This is unacceptable.

Improving ED’s estimates of future repayment is clearly necessary. ED is working on a new model for forecasting future payments, but it won’t be ready until 2026, meaning we’ll have another four years of inaccurate estimates. And there is no guarantee their new model will be more accurate.

But more important is being realistic about the extent to which student loans are subsidized. For loans made in 2022, the subsidy rate is estimated at -1.9%, meaning that the government is expecting to make a profit of $0.019 for every $1 it lends. But that estimate relies on the same flawed assumptions highlighted in the GAO report and on the use of FCRA discount rates. If ED’s predictions for 2022 are off by the same amount as they have been for the last 25 years, the subsidy rate will be 15% higher. And using the more accurate fair-value discount rate adds another 16% to the subsidy rate. So rather than making a profit of $0.019 for every $1 lent, I would estimate the government will lose $0.29. Current projections are for around $89 billion in new loans during 2022, which the government claims will generate a profit of $1.7 billion. In reality, those loans will probably cost taxpayers $26 billion, approximately the same amount we spend on Pell grants.

Student loans should be budget-neutral (a 0% subsidy rate)—they are simply not an appropriate tool to subsidize college. On paper, we are close to this goal, with an official -1.9% subsidy rate. But persistently flawed assumptions and inappropriate discount rates have massively underestimated how subsidized student loans are (the real subsidy rate is around 29%). To provide an accurate assessment of the cost of student loans to taxpayers, ED needs to make more accurate repayment projections and Congress needs to mandate fair-value accounting.

Image: Siam, Adobe Stock