Rescission of Borrower Defense Partial Relief Methodology (EA ID: GENERAL-21-51)

This electronic announcement provides additional information regarding the policy and legal rationale for the March 18, 2021 announcement by the U.S. Department of Education (Department) that stated it will adopt a streamlined approach for providing full relief to borrowers with approved borrower defense to repayment claims to date, in accordance with applicable regulations. This includes borrowers with previously approved claims who received only partial loan forgiveness.

This announcement also provides further detail on the Department’s rescission of the partial relief methodology for approved borrower defense to repayment claims, originally announced on December 10, 2019 and updated in August 2020 (“the 2019 methodology”). This methodology replaced a different partial relief methodology announced by the Department in December 2017. The Department’s use of the 2017 methodology was enjoined by the U.S. District Court for the Northern District of California in Calvillo Manriquez v. DeVos, 345 F.Supp. 1077 (2018). This announcement also officially rescinds the 2017 methodology. These methodologies were applied to claims filed by borrowers who attended certain locations and programs offered by Corinthian Colleges, Inc. and ITT Technical Institute.

Under the Department’s regulations in 34 CFR §§ 685.206 and 685.222, if a borrower’s defense against repayment is successful, the Secretary determines if the borrower will be relieved of the obligation to repay all or part of the loan. The Secretary also determines if the borrower should receive refunds of payments previously made. Prior to implementation of the 2017 methodology, the Department granted full relief to borrowers with successful borrower defense claims without requiring specific evidence from the borrower of the extent of the harm caused by the school. Based on that history, borrowers may not have been on notice of the need to present evidence with their application of the extent of the harm caused by the school’s misconduct.

In light of this as well as for other reasons explained below, the Department decided that as of March 18, 2021, approved claims will be assessed using a rebuttable presumption of full relief. This approach aligns with the applicable regulations in that a borrower may ultimately be granted full, partial, or no relief. Specifically, 34 CFR 685.206 and 685.222 do not specify whether the Secretary’s determination of relief begins with an assumption of 0% relief (increased as appropriate) or 100% relief (decreased as appropriate). The rebuttable presumption fills in this gap consistent with the Administrative Procedure Act because it does not provide a binding rule. Rather, it gives the Department discretion to determine the appropriate relief for borrowers in a case-specific manner consistent with applicable regulations. While the Department understands there could be cases that lend themselves to partial or some other measure of relief, we are not proposing a new, formula-based partial relief methodology at this time. Instead, the Department will use a presumption of full relief as the starting point, and will reduce the amount of relief provided in a manner consistent with the applicable regulations if warranted by evidence provided by the school, the borrower, or other sources.

The 2019 methodology made relief determinations by comparing the earnings of graduates from a program attended by borrowers with approved defense to repayment claims (program earnings) versus the typical earnings of graduates from all similar programs for which the Department of Education possesses data (comparison earnings). Where possible, similar programs were defined as those at the same 4-digit classification of instructional program (CIP) codes as well as the level of the credential.

A borrower’s relief was then tied to how close the average earnings of the attended program were to those of the comparison program. The cut points for different amounts of relief were tied to the standard deviation calculated when computing the median comparison earnings amount. The cut points for relief amount were as follows:

(Note: Regardless of these terms, all Corinthian borrowers were guaranteed at least 10% relief.)

The Department reviewed this methodology and rescinded it in the March 2021 announcement after finding several significant flaws, which are described below. The overall effect of these flaws is that the methodology unfairly determined relief for students based upon data that may not have included them, and then incorrectly used statistical concepts that were not suited for the task at hand. This led to unachievable cut points for full relief in many cases by requiring program earnings to be negative, and many other instances in which 75 percent relief would be unavailable to borrowers who attended programs with earnings below the amount earned at full-time annual work at the minimum wage.

The Department has concluded that the 2019 methodology for partial relief had the following significant flaws:

As applied to claims to date, the 2019 methodology used data obtained as part of the Department’s implementation of the gainful employment regulations. The gainful employment data, however, may not provide a reasonable approximation of earnings for borrowers with approved borrower defense claims. For one, the gainful employment data only include earnings of graduates who received Federal student financial aid, but many borrower defense claimants did not complete their programs. Given that, in general, graduates tend to have better outcomes than those who did not complete their programs, looking at earnings of graduates could present a distorted picture of the outcomes of some borrowers with approved claims.

Similarly, the gainful employment data only present data from a set time—mostly those who graduated between July 1, 2010 and June 30, 2012. But borrowers with approved borrower defense claims may have been enrolled after this time. The result is that the gainful employment data used to calculate program earnings may not actually include the specific borrowers whose claims are being reviewed for relief or may not be a fair comparison group.

The privacy-protected nature in which the Department receives earnings data makes them additionally ill-suited to judging whether a borrower was sufficiently harmed for borrower defense purposes. None of the sources of data available to the Department provide the agency with individual-level earnings data. They are all at the mean, median, or other percentile cut points. Half of the students in the earnings data will fall below the median point. Some of them may fall far below it. The limitations in the data thus make it impossible to judge the extent to which any given individual has been harmed.

After establishing the median program and comparison earnings, the amount of relief a borrower received under the 2019 methodology was tied to how close the former is to the latter expressed as portions of a standard deviation. The methodology, however, misapplied the use of standard deviation for properly calculating earnings distributions. The cut point for full relief was set at median comparison earnings minus two standard deviations. This idea was based on the statistical concept that in a normally distributed set of data, points that fall below or above two standard deviations from the median are considered outliers. If one thinks of a normal distribution of data as a bell curve, these are the far left and right tails. In a perfect normal distribution these tails will each represent 2.5 percent of the sample.

However, this approach is inappropriate for analyzing earnings. Earnings are not normally distributed. For one, they are lower bounded, because they cannot be negative and minimum wage laws also set some limits above zero, assuming that people are working. Consider the earnings distribution of the entire United States population, which would range from $0 for those who are not working, including children, to tens if not hundreds of millions of dollars for chief executive officers, professional athletes, and others. The resulting distribution would have a very long tail on the right while it captures the richest people and no leftward tail because it stops at zero.

Therefore, because the data are not normally distributed, the two standard deviation approach will not capture the bottom 2.5 percent of the distribution. Instead, applying the standard deviation test as if earnings are normally distributed produces mathematical impossibilities. In some cases, two standard deviations may be larger than the comparison earnings figure. The result is that comparison earnings minus two standard deviations can result in a negative number.

An example highlights this flaw. The business administration diploma at Corinthian Colleges had comparison earnings of $18,104. Two standard deviations was $20,654. That resulted in a threshold for 100 percent relief of -$2,550. There is no chance that the median earnings from a group of at least 10 borrowers would be negative, thereby making it impossible for former attendees of this program to receive full relief. In fact, program earnings were just $11,699. Despite earnings of less than the annualized minimum wage, claimants from that program only received 25 percent of their loans cancelled.

Many programs would not have a chance to receive 75 percent relief for similar reasons. Many of the borrower defense claims are associated with certificate programs where the comparison earnings are not that far from the annualized minimum wage. That means subtracting 1.33 standard deviations could create a cutpoint for 75 percent relief that falls below annualized earnings at the Federal minimum wage.

For example, the diploma in accounting technology/technician and bookkeeping at Corinthian Colleges had comparison earnings of $19,966. That was between $10 and $13 an hour depending on whether you assume someone works 30 or 40 hours a week for 50 weeks. One standard deviation is $8,783. The result was that an award of 100 percent relief would have required program earnings at or below $2,399, and 75 percent relief would have required program earnings at or below $8,284. To meet the 100 percent threshold, borrowers would only have annualized earnings of between $1.20 and $1.60 an hour, while the 75 percent threshold would be between $4.14 and $5.52 an hour. The actual program earnings were $11,371.50—between $5.69 and $7.52 an hour—an amount below or right at the minimum wage. Borrowers from this program with approved claims would have had only 50 percent of their loans cancelled.

Regardless of the quality of their postsecondary education, borrowers must work to afford food, housing, and other necessities. Denying 100 percent or 75 percent relief because most borrowers in the program appear to have held some sort of job—regardless of whether it was connected to the training—will not yield appropriate relief.

The Department understands that there may be some cases where partial relief could be appropriate. However, we have not yet identified a formula-based partial relief methodology that would result in consistent determinations and not run into flaws with data availability or other issues. Instead, the Department will assess all approved claims applying a rebuttable presumption of full relief as a starting point.

Under the Department’s regulations in 34 CFR §685.222(e), (g) and (h), a Department official or a hearing official may approve a borrower defense claim in whole or in part. The decision on the borrower’s claim will be based on the information provided to the Department official or the hearing official, including any information provided by the school or the borrower. In determining the amount of relief on the claim, the Department will apply a rebuttable presumption that full relief should be awarded to the borrower. That presumption may be rebutted or bolstered by the evidence received by the Department official or the hearing official from the school, borrower, or other sources. Here is an example under which partial relief would be based on a rebuttal of the presumption of full relief. A borrower files a borrower defense claim based on the school’s misrepresentation of the amount of the cost of a program for which the borrower had to pay $25,000, all of which the borrower financed with student loans. There is no evidence of any other misrepresentations in the enrollment process or of any deficiency in the value of the school’s education. If the rebuttable presumption applied, the borrower could receive full relief. However, if evidence received by the Department official demonstrated that the school told the student that the cost of the program was $20,000, the Department would determine that full relief was not appropriate and that, in this case, the borrower should receive partial relief of $5,000. The school provided the education that the student was seeking but misrepresented the price.

The Department will continue to look at this issue, solicit stakeholder input, and consider other avenues to better address the issue, such as through the development of new regulations.

Because of the concerns outlined above regarding the flaws in the actual methodology, the Department rescinded the partial relief methodologies released in 2017 and 2019 and updated in 2020. Going forward, the Department will assess approved claims based on the merits of the claims under the applicable regulations, applying a presumption of full relief as a starting point.