Two weeks ago, Twitter scolds looking for excuses to criticize a group of popular federal student loan repayment programs had a field day with a dense new study from the U.S. Government Accountability Office, or GAO. In 100 pages, the GAO report lays out a number of methodological and data challenges that the U.S. Department of Education faces in properly estimating the cost of income-driven repayment, or IDR plans, noting that these issues could raise or lower the programs’ price tags by billions of dollars.
Reactions to the GAO report offered a disappointing case study in confirmation bias. Voices that were already critics of these programs only focused on the possibility that program costs might be higher than anticipated. These critics trumpeted GAO figures suggesting that the federal government might end up forgiving about $108 billion in loan principal. The Wall Street Journal editorial board also used the report for the latest in its endless set of fact-free attacks on 2010 changes to the federal loan programs that saved taxpayers tens of billions of dollars.
These critics ignored the nuance in the report that is crucial for understanding what is really causing cost uncertainty. Here’s what the overblown criticisms of the programs missed. Hopefully policymakers will not ignore methodological subtleties for the sake of a catchy one-liner.
Forgiven loan principal does not equal program cost
The most basic interpretation error in a lot of the report’s coverage is a failure to understand that the amount of forgiven loan principal does not equal a program’s cost. The $108 billion figure is the amount of forgiven principal. It represents the portion of balances originally borrowed that are expected to be forgiven. But because of interest accumulation and a longer time in repayment that comes with income-driven plans, it is entirely possible that some borrowers will have principal forgiven but actually pay more in total than initially expected under a 10-year payment plan. In cases like this, not only has the government not lost money on the loan, it may have even turned a profit depending on how much extra in interest was paid back. Even if a borrower does not cover all the cost of loan forgiveness, additional interest payments will offset some of those expenses. That’s why factoring in these interest payments along with other cash flows from borrowers drops the program cost 31 percent lower than the level of forgiven principal, according to the GAO.
Costs grew due to increased participation
What complaints about the rise in cost estimates miss is that income-driven repayment programs are not necessarily getting more expensive because the expense per student is rising. Rather, income-driven repayment costs are growing primarily because more people are using income-driven repayment.
The GAO report notes, for example, that the expected costs of income-driven payment plans doubled from 2009 to 2016. And what happened from 2013 to 2016? The number of borrowers using these programs doubled. As GAO puts it, the costs of these plans have doubled “primarily due to the growing volume of loans expected to be repaid in IDR plans.”
The fact that the Department of Education did not properly estimate how many people might need these programs should not be a shock. After all, the more generous plans only started in 2012 or later as a result of congressional legislation and executive actions by the Obama administration, meaning their newness makes participation hard to project.
Cost estimates reflect 20 years or more of loans
Time matters when understanding cost estimates. Multiplied across enough years, any expense becomes a very large number. That is exactly what GAO’s report does. The cost figures it reports represent expectations for what will happen to loans issued over the course of 23 years. For instance, GAO estimates the cost of the program at $74 billion for debts issued between 1995 and 2017—the course of an entire generation.
GAO’s data representation deprives the reader of this context. In contrast, the program’s average cost on an annual basis is $3.2 billion—though this too masks a range of costs, which will vary based upon loan volume.
Fixing some cost estimation issues would reduce expenses
The key point that a lot of the coverage of the GAO report missed is that correcting several flaws in the department’s methodology would reduce costs.
One example highlights this challenge. The GAO raises a concern that the Department of Education used income data on past student loan borrowers to project future loan payments without adjusting earnings for inflation. That’s a fair critique. But adjusting a borrower’s projected income upward for inflation would result in a larger payment. Larger payments, in turn, mean lower forgiveness amounts. Such methodological adjustments would drive overall costs down by about $17 billion, according to the GAO. Not adjusting for inflation in effect acts as a more conservative, higher estimate of cost.
Estimates are not actuals
Many media outlets incorrectly portrayed the GAO’s forgiveness estimate as the minimum amount that the federal government will forgive. That’s wrong. Payments are tied to a borrower’s income, which means that cash flows on a loan will change annually. The 20-year time horizon for repayment also means there is a lot of time for borrower’s income and payments to jump, reducing and possibly eliminating any forgiveness. That means costs could go up or down, and it is very difficult to predict how costs will all play out.
As the GAO puts it, “estimates will continue to change over time, and actual subsidy costs of a loan cohort will not be known until all loans in the cohort have been repaid, which may take 40 years.”
The federal government has so far forgiven nothing through these programs
The hand-wringing over loan forgiveness is particularly noteworthy because the federal government hasn’t actually forgiven a single dollar of loans through these programs yet. That’s even true for the Income-Contingent Repayment Program, which was created in 1994. The first sets of loan forgiveness will not occur until next October, when borrowers who have been in public service for 10 years could seek forgiveness. Department of Education data suggest that approximately 139 people are in range of earning that eligibility next year.*
The lack of any forgiveness history matters because again it shows all the commentary on this issue is just prognostication with no actual results to bolster these claims.
IDR borrowers are not rich
Implicit in many of the critiques about income-driven repayment is the idea that the borrowers on these programs are well-off people who don’t need government help with their loans. Data from the Department of Education suggest otherwise. Below are average incomes of borrowers by repayment program reported by the Department of Education at a recent presentation:
- Income contingent (455,000 borrowers): $42,450
- Income-based with a financial hardship (2.1 million): $35,071
- Income-based without financial hardship (743,000): $32,675
- Pay As You Earn with financial hardship (643,000): $30,617
- Pay As You Earn without financial hardship (180,000): $23,787
- Revised Pay As You Earn (787,000): $32,769
These data suggest that many borrowers are people struggling to get by, not well-heeled professionals gaming the system.
The Department of Education is already improving its model
Several of the issues raised by the GAO are fair concerns. For example, the Department of Education model did not include Grad PLUS loans when they were not part of a larger consolidation loan. And the GAO found the department needed to better justify assumptions about the usage of Public Service Loan Forgiveness. These are concerning issues and clearly the cost estimation model needs to be improved. But as the GAO report and the department’s response to it note, the agency already started doing just that in spring 2015, with input from the U.S. Department of the Treasury and the Office of Management and Budget.
The details matter
It’s very easy to cherry-pick statistics from a report to write breathless hot takes that confirm one’s opinions. But a closer reading of the GAO report, as well as the facts it covers, reveals a more complex story.
If anything, the underappreciated discussion in the GAO report is just how difficult it is to properly calculate the cost of programs that tie payments to borrower incomes for two decades. Historical data will only be so good, since the profile of a borrower in 1995 looks different from that of a borrower graduating in 2015. Data privacy rules create further complications. For example, a 14-page appendix to the report goes into great detail about the difficulties of income data imputation and how they make it hard to properly model payments. Policymakers and budget experts also have no way of knowing how changes in the economy, graduate school attendance, and other factors, will affect long-term earnings. Add in the fact that most of these repayment options were created in 2008, and the most generous ones were not created until 2012 or later. The result is a complex system with no track record to work from in estimated its costs where even enrollment levels have yet to stabilize.
The good news is that estimates should get better over time as more data become available and actual repayment behavior becomes easier to observe. Then, perhaps, it may be possible to have a meaningful discussion about costs.
* Author’s note: These are individuals who submitted a voluntary form saying they are seeking public service loan forgiveness and have between 97 and 120 months of qualifying payments.
Ben Miller is the Senior Director for Postsecondary Education at American Progress.