The State of Higher Education

While there are many areas of concern, there is little the federal government can or should do about some of them. For example, while there are valid concerns about curricula and grade inflation, federal involvement would not fix the problems and would likely make things worse.

But there are several areas where the federal government could have a positive impact, largely by addressing undesirable and unintended consequences of existing government policies and practices. In particular, there is a strong case for federal policy changes to address affordability, accountability, and innovation.

The federal government provides around $140 billion in financial aid each year, with a goal of promoting equality of opportunity and increasing college affordability. Most of the aid takes the form of grants, loans, or tax benefits. The effectiveness of these programs varies widely. For instance, Pell grants are well designed and implemented, and go far toward achieving their intended purpose. At the other end of the spectrum are the tax benefits, which are

Because they are so badly designed and ineffective, tax benefits should be eliminated.

Student loans fall somewhere in the middle. Loans are necessary for many students, but our current loan system is badly designed. One of the main problems is that a substantial portion of this aid is harvested by colleges without making college more affordable. There is considerable evidence that colleges raise their prices and reduce institutional aid when federal financial aid is available. This phenomenon is referred to as the Bennett Hypothesis. The scholarly literature finds solid evidence that the Bennett Hypothesis is real:

• Professors Cellini and Goldin found “large and significant differences between the tuition charged by T4 and NT4 institutions… The magnitudes are comparable to average per-student federal grant aid awards, suggesting that T4 institutions may indeed raise tuition to capture the maximum grant aid available.”²
• An old paper of mine explores how the competitive pressure will lead almost all colleges to raise their price when aid is available.³
• Professors Gordon and Hedlund find that increases in student loan borrowing are the most important factor in explaining rising college costs, accounting for 40% of the increase.⁴
• Analysts at the New York Federal Reserve Bank found that each additional dollar in loans led to 40–60 cents in higher tuition and a reduction of 20 cents of institutional aid.⁵
• Professors Black, Turner, and Denning find that “the creation of Grad PLUS led to significantly higher program prices… sticker prices went up approximately dollar for dollar with increases in federal loans.”⁶

Given that the Bennett Hypothesis is sabotaging student loans, this Committee should seek to reduce or eliminate the threat.

There are two policies that would limit the damage from the Bennett Hypothesis. First, aid could be targeted only to low-income students who would otherwise not be able to afford to attend college. Colleges cannot raise tuition as much in response to such aid because that would price these students out of the market again. Second, aid programs should be capped. Pell grants and undergraduate loans are already capped. Grad PLUS and Parent PLUS loans are not — there is no aggregate cap and the annual cap is entirely up to the college. While I would argue for eliminating PLUS loans entirely, at the very least, PLUS loans need to be capped.

There is also a way to neuter the Bennett Hypothesis. The way aid eligibility is determined right now, an increase in tuition will automatically lead to an increase in aid. So if a college raises tuition by $1, their students get $1 more in loans. This link between higher prices and more aid is the key driver of the Bennett Hypothesis. So sever the link. The best way to sever the link is to use the median cost of attendance among all colleges to determine aid eligibility instead of each colleges’ own cost of attendance.⁷ When using the median, an increase in tuition at a particular college has no effect on how much aid a student can receive to attend that college. This vastly reduces each college’s incentive to raise tuition when aid is available, which in turn neuters the Bennett Hypothesis. Notably, the House’s reconciliation includes such a provision. The Senate should support this change too.

We desperately need more accountability in higher education. While most colleges provide a good education at a reasonable cost, there are substantial portions of higher education that are too low quality, too overpriced, or too oversupplied to justify an investment. But right now, colleges still benefit from enrolling students in such programs, because the college gets to keep all the money students paid even if the student fails to repay their loans. In other words, colleges can win even when students and the government lose. An accountability system should change this so that colleges only win when students and the government win too.

We know that accountability mechanisms can work. Of colleges sanctioned for having a Cohort Default Rate that was too high, 95% lost access to aid.⁸ And of the 38 programs at Vatterott College that failed the Obama administration’s gainful employment test, all were closed several years later.⁹ But while we technically have three accountability systems operating right now, none of them are effective. The first, Cohort Default Rates, is obsolete now that we use income driven repayment plans since such plans allow for $0 payments that don’t count as defaulting. The second, Gainful Employment regulations, are routinely implemented by Democratic administrations and are just as routinely scrapped by Republican administrations. The third, accreditation, has probably never been used. I don’t know of any college that has lost accreditation for being too expensive for students or losing too much taxpayer money.

Without any functioning accountability system, we are in dire need of a new one. A well-designed accountability system would:

1. Focus on programs rather than institutions

Historically, accountability has been applied at the institutional (meaning the entire university) level, but program level (meaning a credential and field of study combination, such as a bachelor’s in nursing) is much better because it avoids punishing good programs at bad schools will also ensuring that bad programs at good schools don’t escape accountability.

2. Utilize labor market outcomes

Accountability systems should also utilize labor market outcomes. Not only do around 90% of students attend college to enhance their careers, but because colleges can’t control these outcomes, these metrics are harder for colleges to manipulate and game.

3. Be applied universally rather than selectively

An accountability system should apply equally to all of higher education. There has been an unfortunate tendency with the Gainful Employment regulations to target only certain segments of higher education, notably the for-profit sector, while giving the vast majority of public and nonprofit higher education a free pass. But this is a fatal flaw in an accountability system because most failing programs were not located at for-profits. For example, applying the Obama administration’s gainful employment test to all of higher education revealed that only targeting for-profits would have missed “89% of failing programs and 73% of students graduating from a failing program.”¹⁰

4. Use both carrots and sticks

Historically, higher education accountability has only used sticks to punish poor performers. It should therefore come as no surprise that higher education is reflexively hostile to accountability systems – they can only hurt. But it doesn’t have to be that way. Accountability systems could use carrots too. In particular, high performing programs could earn performance bonuses or regulatory relief, including waivers of requirements to obtain accreditation and state authorization.

5. Use relative performance cutoffs instead of numerical cutoffs

Most accountability systems have used numerical cutoffs. For example, the Cohort Default Rate (CDR) has a cutoff of 30%, so a college with a CDR of 29.9% for three years has unlimited access to federal aid programs, whereas one with a CDR of 30% for three years loses all access to all aid programs. But it is difficult to determine a reasonable threshold, and historically, we’ve been too lenient. A CDR of 29% is still much too high to escape accountability.

A better approach would set thresholds of relative performance among programs (and combine this with the use of carrots and sticks). Relative performance cutoffs are then determined based on a program’s CDR relative to all other programs. Programs with the lowest CDRs could receive carrots, programs with typical CDRs would receive neither carrots or sticks, and programs with the worst CDRs would receive sticks. Tiers of three (each tier accounting for 33.33%), four (each tier accounting for 25%) or five (each tier accounting for 20%) are simple, easy to understand, and provide opportunities to apply and scale carrots and sticks.

The relative performance approach has several advantages over the numerical approach. To begin with, it avoids the problem of choosing thresholds that are too stringent or too lenient. It also automatically adjusts to common shocks. For example, when the economy enters a recession, CDRs might rise for all programs, even if the quality remains unchanged. A numerical threshold would require Congressional action to avoid becoming more stringent then intended, whereas the relative performance approach would adjust to the recessionary environment automatically. Relative performance also encourages continuous competition among programs – if a program’s peers improve, the program must improve too to maintain its relative position.

So what are the some feasible options for an accountability system?

• Earnings floors

Earnings floors would terminate aid eligibility for programs where students don’t earn enough. Floors are easy to understand and would eliminate some of the most problematic underperforming programs. But earnings floors ignore debt. Programs that just barely pass the floor but load students with excessive student loan debt could have low or even negative returns while still passing an earnings floor test. Earnings floors are certainly a good start, but they can’t do the job alone.

• Repayment rates

Repayment rates are arguably the most natural choice for an accountability mechanism for student loans. The main problem with using repayment rates is that there is very little information about current repayment rates, which would make setting reasonable numerical thresholds difficult. This can be overcome by using relative repayment rates, with programs with repayment rates above the median being rewarded with various carrots, while programs with repayment rates below the median face sticks of increasing severity.

• Gainful employment for all

Another option would be to apply gainful employment like debt to earnings tests across all of higher education. The metrics are familiar, and we’ll soon have almost all the data needed to implement this due to the Financial Value Transparency regulations. But it could be argued that GE arbitrarily defines excessive debt. Are the 8% and 20% cutoffs in the most recent version the right numbers? If we implement GE for all, I recommend scrapping the current numerical cutoffs and implementing relative performance thresholds instead. I would also recommend introducing carrots for high performing programs.

• Risk sharing

Risk sharing or skin in the game systems require colleges to reimburse the government when students fail to repay their loans. The best feature of these systems is that they align the college’s incentives with those of the students and government. A college can longer profit from offering an education the leaves the student and government worse off.

The main problem with risk sharing proposals is that they tend to hit sympathetic colleges hard. For example, community colleges tend to face high risk sharing payments because they are open access and many students drop out before graduating, leading to repayment problems. This can be addressed by introducing safe harbors or compensating funding, but this tends to make the system more complex.

For example, the House recently introduced a risk sharing system where colleges would be required to reimburse the government for a portion of losses when students fail to repay their loans. The reimbursement share essentially creates a safe harbor for community colleges, and the Promise grants provide additional funding to compensate for their remaining risk sharing payments.

Some argue that the risk sharing system proposed by the House is too complicated. This concern has some validity but is largely overstated. The House’s risk sharing metrics are no more complicated than the debt to earning metrics in gainful employment. The main GE formula is debt / earnings. The main risk sharing formula is earnings / price paid. These are comparable in their level of complication.

Moreover, the House version could be amended to make is simpler. The risk sharing payments currently take government losses on a program’s loans (consisting of missed payments, waived interest, and forgiven loans) and then apply the reimbursement percentage based on earnings / price paid. But the reimbursement percentage could instead be based on relative performance of the government’s losses per student. One simple formula that would accomplish this is: reimbursement percentage = (program’s relative performance – 50%) * 2, with a cap at 0%. So the first 50% of programs that have the lowest government losses per student would have a 0% reimbursement percentage. The 51% program would have a reimbursement percentage of (51%-50%)*2 = 2%. The program with the highest government losses per student would have a reimbursement rate of (100%-50%)*2 = 100%. This would protect open access community colleges (since they tend to have low debt per student), while requiring risk sharing payments from the worst offenders.

While individual professors and even departments or entire colleges can be quite innovative, the industry as a whole is remarkably stagnant. One of the primary drivers of this stagnation are government policies that suppress innovation, with accreditation being the key impediment.

Accreditation erects enormous barriers to entry for new colleges, which prevents new innovative colleges from emerging. Over 20 years, the seven largest accreditors approved only 9 new public four-year colleges, less than one new college every two years.¹¹

Accreditation also largely requires all colleges to use the traditional recipe of inputs and processes of existing colleges.¹² If there are innovative models to deliver a higher quality and more affordable education, the accreditation system ensures that we won’t find and adopt it.

Unfortunately, the potential replacements for accreditation are likely even worse, so our best course of action is an overhaul of the accreditation system. The most important reforms are to increase competition among accreditors and to ensure that accreditation decisions are made based on outputs and outcomes rather than inputs and processes.¹³

Higher education has certainly seen better days. But with federal reforms that address issues with affordability, accountability, and innovation, we could turn the corner and unleash higher education to provide a better and more affordable education than we’ve ever seen before.

Thank you again for the opportunity to provide this testimony and I look forward to answering any questions you may have.