The Gainful Employment NPRM is Here

The foundation is more robust and thought out. What you think of the house on top of it depends on where you sit. That's probably the easiest way to think about the Gainful Employment Notice of Proposed Rulemaking (Gainful Employment NPRM), which the U.S. Department of Education published this morning at 7 a.m. At first glance, the rule, which attempts to hold career-oriented programs accountable for leaving students with too much debt or overly high dropout rates, has roughly equivalent or slightly weaker accountability provisions and no borrower relief compared to the the last public version released in December. That said, the fact that measures are independent from each other, there are new certification requirements, and ineligibility can happen faster means that this version is almost certainly stronger than the one released in 2011. The regulation will now be out for comment for 60 days.

This thing is an 841-page behemoth, so I'll be reading and updating throughout the morning, but here's a current read on the regulation based upon just the draft text, not the explanatory materials.

Overall Structure: Similar Accountability, Lost Relief

At the 10,00-foot view, the overall structure looks the same as the last version we saw in December. The metrics, thresholds, and time period before ineligibility are all the same. That means programs would be separately judged on a set of debt-to-earnings ratios and a program cohort default rate. For the debt-to-earnings ratios, a program that failed both an annual and a discretionary measure (which allows for a higher ratio after subtracting out basic living costs from income) twice in three years would lose eligibility. In addition, a program would have to pass one of the debt-to-earnings measures at least once in a four-year period. That means having an annual debt-to-earnings rate at or below 8 percent, or a discretionary rate at or below 20 percent. The measures would be based only upon the earnings of program graduates who also received some form of Title IV aid, which includes Pell Grants and Stafford Loans.

The Program Cohort Default Rate, meanwhile, functions almost the same as the similar measure that's currently applied to colleges through the Higher Education Act. Under that measure, a program would lose eligibility if 30 percent or more of its former students who entered repayment defaulted on their loans by the end of the third fiscal year. This test would operate separately from the debt-to-earnings rates, meaning that passing one set of measures does not absolve failure on the other.

The next layer down reveals some differences from December to today. The biggest is the loss of provisions that would have required programs that lose eligibility to provide some measure of debt relief for enrolled students. That idea was not included in the 2011 final regulation, but would have provided a way to get relief to at least some students who are highly likely to struggle with their loan payments. In addition, the Department also removed a cap on the amount of federal financial aid that programs about to lose eligibility. This would have stopped programs from growing at the same time they very well might be losing access to aid dollars a few months later.

Beyond the regulatory text, it's clear that this draft is designed to be much more legally sound than prior drafts. The upfront language spends a good bit of time talking about legal authority, there's substantially more data, and even a severability clause that would allow parts of the rule to stand even if others are struck down. These elements all clearly seem to anticipate a legal challenge, which is not unreasonable given that a substantially watered down rule in 2011 still produced a legal challenge within days. It also indicates a rule with a much stronger foundation, regardless of what you think of the house built atop it.

Debt-to-Earnings Rates: Maybe Stronger for Some, Weaker for Others

While the thresholds on the debt-to-earnings rates are unchanged, the way they are calculated is different. The minimum program size to measure debt-to-earnings rates increased from 10 to 30. This does not necessarily remove accountability for smaller programs, since ones with fewer than 30 graduates will have data aggregated over four years, but it potentially delays their loss of eligibility. The assumed repayment time frame for debt calculations also changed from a flat 10 years for everyone to 10 years for all certificates and associate degrees, 15 years for bachelor's and masters degrees, or 20 years for doctoral or first professional degrees. A longer repayment time frame helps programs by lowering their average annual debt payments. Both these changes essentially would return the rule to where it was in 2011.

Two changes would make the debt-to-earnings rate slightly stronger, though whether this is more than balanced out by the change in repayment length for degrees at the bachelor's level and up is not clear. First, the interest rate used to calculate payments will now be different for undergraduate and graduate programs and be based on an average of the past six years, not the lowest as suggested in December's iteration. That change will raise the assumed interest rate, making the deb-to-earnings rate slightly tougher to pass. It probably matters the most for certificate programs, since they will not benefit from the longer repayment calculation. Second, loan debt will still be capped as a way of preventing anecdotal claims of over borrowing," but the cap will now include books, supplies, and equipment. This will increase the maximum level of debt that can be considered.

There is one new provision that somewhat gets at the question of what to do about programs with low borrowing rates, which community colleges have raised concerns about. Programs that are failing or in the zone on the debt-to-earnings rates could file a "mitigating circumstances" appeal to the Secretary by showing that the borrowing rate for all completing students (those who received federal aid and those who did not) is less than 50 percent. This would help any college that has a high borrowing rate of students receiving federal aid, but large numbers of unaided students who do not borrow.

Other notable elements: The Department would now only use median debt, not the lower of mean or median as suggested in the past. Presumably this is because median debt was always lower, though there's no way to know for sure. The types of students who could be excluded from a debt-to-earnings rate are the same, but the conditions around them are broader. For example, there is no minimum enrollment period or time in military status to be excluded from the cohort. The December draft had considered minimum time periods of 60 days. In addition, the Department would still exclude the highest remaining debt for each student who does not have earnings information pulled by the Social Security Administration (SSA). In other words, if 35 students went over to SSA and two did not match, the two highest debts would be excluded. The Department would also continue to offer a transition period of four years, in which programs could use the debt levels of more recent completers judged against the earnings of older graduates. [UPDATE: The Department would also not count years in which a program does not have rates toward ineligibility. This means that if a program failed, had no rates, passed, and failed again, it would fail by virtue of having failed for two years in a three-year period for which data were available.]

Program Cohort Default Rates: Largely Unchanged

Because the program cohort default rate pegs off existing legislative provisions, it is largely unchanged from the December version. The Department still chose not to include the statutory provision that results in immediate eligibility loss for a program with a default rate over 40 percent in any one year (a change that absolves 238 programs). The regulatory text itself now includes all the various challenges and appeals that are also allowed under institutional cohort default rates, though tweaked to the program level. I'll continue checking these to see if they are radically different from the institutional regulatory text. These appeals include things like having a low borrowing percentage (known as a participation rate index), improper servicing (things like missing payments that were filed), inaccurate data, etc.

Certifications and Approvals: It Comes Down to Implementation

The biggest x-factor in the new rules is not the formula-based calculations, but certification requirements for programs. These are requirements around things like showing the need for training, showing necessary approvals and certifications, and showing that programs are not overly long. But a lot of these conditions are going to enter into grey areas that are hard to stipulate clearly through regulation, and for them to have any meaningful outcome will require diligent oversight and implementation.

The regulatory text requires the most senior official at an institution to certify by December of the year the rule goes into effect that all of its programs meet certification requirements. In addition to attesting that the program has accreditation or state recognition either on its own or through its institution, the certification also requires stating that programs have necessary programmatic accreditation if that is required by the federal government or other governmental entity in the state and metropolitan areas served, and that it satisfies licensure and certification requirements, as applicable, in its home state and any other state in its metropolitan area so that students can take any required exams. On the one hand, this is good language that gets at an important issue whereby colleges were offering programs that could get federal financial aid through their regional or national accreditor, but could not actually let graduates sit for licensing exams due to a lack of programmatic accreditation. But there are two main flaws. First, the requirement only applies to the home state of the institution and other states in its metropolitan area. This means a distance education student could be in a state where he or she not get licensed and still not fail the certification requirement. Second, there are some types of programs, such as medical assistant, in which the certification requirements are from employers, not a governmental entity. In those cases, a program that's less desirable to the people doing the actual hiring could still pass.

The draft text would also add in some language around colleges' program participation agreements (the documents they sign to be in the federal student aid programs). This includes establishing that there is an actual need for the training to obtain employment in the occupation they are being prepared for. In addition, there is a requirement to demonstrate that the length of training in terms of clock hours is no more than 50 percent greater than the length established by the state in which the institution is located or a federal agency. This could in theory tamp down overly long programs, though 50 percent is a fairly sizable cushion.

Finally, the regulations add a definition of a recognized occupation, which must link back to codes established by the Department of Labor or the Office of Management and Budget or be determined by the Secretary in consultation with the Department of Labor. In theory, this provision could provide some accountability against programs that have no recognized occupation attached to them, but would again come strongly down to implementation.

Policy Odds and Ends

• Because the measures are independent of each other, a program does not have to have both a debt-to-earning result and a program cohort default rate result to potentially lose eligibility. If results are only available for one of the measures, then that one measure is still used to hold the program accountable.
• The definition of a prospective student now includes those who are contacted first by the school, not just students who reach out. This means proactive recruitment efforts will have to come with disclosures too.
• Debt-to-earnings rates are calculated off of award years rather than fiscal years. Still looking to see if that has a practical effect or is just a technical difference.

Fun Facts and Figures From the NPRM

• The Department estimates that there are about 50,000 gainful employment programs enrolling about 4 million students. In terms of breakdown, 60 percent are at private for-profits, 30 percent are at publics, and 10 percent are at private nonprofits. [UPDATE: It's 60 percent at public colleges, 30 percent at private for-profits, and 10 percent at private nonprofit.] In total, it estimates that these programs got $9.7 billion in federal grants and $26 billion in federal loans.
• Of programs where the mean or median earnings are below those of a high school dropout, 24 percent of borrowers defaulted within three years.
• Programs that fall into the "zone" category, meaning they are below the failing standards but not good enough to pass, have mean or median annual earnings below $18,000.
• At a minimum program size of 30, the odds of a passing program losing eligibility by being inaccurately characterized as a zone program for four years is close to zero. Same thing for the odds of a passing program being accidentally labeled a failing program. Even at a minimum program size of 10, the odds of a passing program being labeled failing in a single year is only about 0.7 percent. (Page 108)
• Most borrowers are on the standard 10-year payment plan. This includes 80 to 90 percent of undergraduate borrowers from two- and four-year colleges and 63 percent of graduate students. (Page 146). Of borrowers who entered repayment between 1993 and 2002, 83 percent paid off their loans within 20 years, 74 percent within 15 years, and 65 percent within 12 years (Page 147).
• In real terms, average loan size has increased by 6 percent from 1999 to 2011. (Page 147)
• There are a lot of very small (either in aggregate or completer terms) gainful employment programs. After the 2011 final rule, the Department received information on 37,589 programs. Of those, just 5,539 had at least 30 completers. In other words, 85 percent of programs were too small to be evaluated. Whether those programs would hit the minimum size of 30 over four years is unclear. (Page 639)
• But the smaller programs do not include that many students--the Department estimates that 73 percent of students in gainful employment programs would be captured by either the debt-to-earnings ratios, program cohort default rate, or both (Page 640).
• Approximately 15 percent of students had their debts capped in the informational rates at no higher than reported tuition and fees (Page 644).
• Demography does not appear to be destiny. The percentage of Pell Grant recipients and percentage of students with minority status explained less than 2 percent of the variance in annual earnings data. Those same factors explained less than 20 percent of the variance in program cohort default rates (Page 648).
• It looks like some of the identified problems are concentrated in a few bad actors. According to the Department, a subset of 11 percent of for-profit institutions account for 90 percent of students in failing or zone programs at these types of colleges. (Page 718)

Correction: I had the breakdown of gainful employment programs reversed. It's now fixed.