Department of Education's Budget Estimates Cannot Be Trusted

Today, the Government Accountability Office released a new report detailing how the Department of Education has completely mismanaged the budgeting process for Income-Driven Repayment plans. The report shows that the Department failed to factor in basic assumptions and adhere to standard rules for budgeting for the loan program. The Department was so derelict and incompetent in their duties that we can no longer trust their budget estimates when it comes to the loan program, and Congress should hold hearings on whether the Executive Branch should transfer budgeting responsibilities to another agency.

Income-Based Repayment and Public Service Loan Forgiveness were first passed in 2007, with the Obama administration and Democratic Congress making the plan much more generous in 2010, and again in 2011 and 2015 through executive action. As these expensive and generous changes were being made, Jason Delisle and I, along with other policy experts, began sounding the alarms on the cost of these programs. We ran models that showed that graduate students would receive large windfall benefits. In response, advocates of the programs pointed to official estimates from the government to show that our fears were unfounded. We now know that those estimates were not just wrong, but that those making the estimates in the Department of Education were negligent in their calculations.

What ED Got Wrong

Here are some things that the Department of Education (ED) failed to account for in their modeling:

1. ED assumed there would be no growth in enrollment. They assumed this even as growth in the program was skyrocketing and as the administration made a massive push to enroll 2 million new borrowers.

2. ED did not produce different cost estimates for the different types of IDR plans, despite substantial difference. That’s apparently because: 

According to Education officials, the student loan model, which it uses to generate official estimates of total Direct Loan costs, was created when only one IDR plan was available and cannot produce separate estimates for each IDR plan.

3. Ed can’t even fully explain how their model works. It’s worth quoting the GAO in full on this:

While a broad narrative summary of the model is available, agency officials confirmed that other technical documentation recommended in federal guidance for estimating subsidy costs does not exist. For instance, Education does not have a flow chart or other similar documentation specifying how elements of the estimation process—which is implemented by nearly 50 computer programs—are sequenced and interact with each other. Additionally, the numerous mathematical formulas embedded in these programs are not separately documented, and there is no data dictionary to decode the variable names and values.

In other words, ED doesn’t even have documentation necessary to perform a proper audit on their methods.

4. Until 2015, ED didn’t even count Grad PLUS loans, one of the biggest drivers of the growth in IDR. According to the GAO:

Education officials said that they had to make a model adjustment in order to include Grad PLUS loans in IDR estimates. Prior to this adjustment they assumed all Grad PLUS loans would be repaid in other repayment plans.

Every policy expert studying IDR would have told them this was a faulty assumption. It turned out to be at least a $3 billion mistake.

5. ED did not account for the fact that different types of loans carry different interest rates, which affects the amount that will be forgiven. That means that ED’s estimate of subsidy costs are based entirely on upfront fees and interest accrued in school. This is a big deal because it likely massively understates the subsidy cost of Grad PLUS loans, making them seem less costly to the government than they really are. But even worse is ED’s reasoning: 

According to Education officials, they could have separately estimated repayment patterns for each loan type, but did not believe that it was important to do so.

That is suspicious because it would have been obvious to most analysts that separating the loans was important (because of basic math and the way that interest accrues). In not separating them, ED allowed advocates to continue to claim that Grad PLUS loans make money for the government, when in fact, after accounting for IDR and using fair-value accounting, that may no longer be true.

6. ED failed to test a variety of models or account for potential error in income models. They also failed to adjust for inflation, which is standard practice, and a potential $17 billion error. They also failed to factor in that some people don’t re-certify, even as evidence mounted that this was a common occurrence.They also estimated that every person eligible for Public Service Loan Forgiveness will apply for the program, an assumption that Shahien Nasiripour of Bloomberg points out is unlikely to be true.

7. ED ran a sensitivity analysis on only one assumption-borrower income-and that was only after the Office of Management and Budget prodded them to do so. Sensitivity analysis is used to determine which assumptions have the biggest effect on costs by changing the assumptions by a fixed percentage. GAO points out that federal guidance states they should be conducting sensitivity analysis on all of their main assumptions. In one of the snarkiest instances I've ever seen in the usually dry language of GAO reports, the authors call out ED on its sorry excuse:

Education officials told us they only conducted sensitivity analysis when asked by others, preferring instead to focus their resources on developing a single set of assumptions they believed were best. Developing a sound set of assumptions is, of course, important. Sensitivity analysis supports, rather than detracts from, this effort. (emphasis added)

In other words, ED's excuse makes no sense and is misguided. But what stands out is that the only reason they ran any sensitivity analysis was because OMB "asked". What's remarkable is that even after OMB pointed out ED should be running sensitivity testing, ED appears to have dug in and only done so when forced. Reading between the lines, it looks like a turf war, with OMB begging ED to use normal budgeting practices, and ED, for whatever reason, refusing (except for in one instance).

8. ED management did not “review or approve” the new models, nor did they hire an independent firm to check their work, which the GAO identifies as a common and good practice. Basically, no one checked anyone’s work. And because ED is notoriously stingy about releasing their loan data, policy experts had little ability to check ED’s work, other than to just say it seemed wrong.

A Call for Reform

Congress and the American people have come to expect honest and straightforward budget estimates from the administration. The Department of Education and the Obama administration failed to appropriately estimate the budget impact of the Income-Driven Repayment programs. They failed to account for basic, common-sense assumptions, like enrollment growth, even as they pursued those policy outcomes. 

This is particularly concerning because, all together, ED’s mismanagement likely underestimated the cost of the Income-Driven Repayment programs during critical years when the program could have been fixed. Instead, costs skyrocketed. This conveniently hid the fact that the Obama administration’s changes were far more expensive and regressive than they had claimed (at one point an administration official even claimed the changes would save money) when the policy was proposed.

Moving forward, we cannot fully trust ED’s budget estimates for the IDR program, or any aspect of the loan program. Because nearly 50 percent of the outstanding Direct Loan portfolio is in an income-driven plan and because ED was not differentiating between loan types, the cost to each type of loan was also incorrect in previous budgets. Policymakers should view ED’s estimated costs for the Grad PLUS loan program with extreme skepticism.

This also further vindicates experts’ increased skepticism of ED’s stated default recovery rates. The Department states that they recover nearly all of defaulted loans on a present value basis, while evidence mounts that this is probably not true. After the GAO’s report, no one should take those numbers seriously either.

While some may call to beef up the Department of Education’s budgeting capabilities, it’s not clear it should be trusted with estimating the loan program at all. Even though ED has agreed to address GAO's critiques, the damage has been done and there's little reason to believe they will make things right. This is, after all, an agency that only did a sensitivity analysis on one assumption and only after being prodded by the Office of Management and Budget to do so.

As costs continue to increase from new policies, we need either a lot more oversight, or to simply move budget estimating for the loan program to some other part of the Executive Branch. The Department of Education does a good enough job at administering the loan program, but they clearly can’t determine its costs. Congress should further investigate what went wrong and call for reforming the loan budgeting process.