Human capital contracts continue to be all the rage in higher education funding. Beth Akers of Brookings writes that they can tackle “the growing risk associated with investing in higher education.” They are also playing a role in the presidential debate over higher education. Greg Mankiw, discussing higher education costs, is excited that Marco Rubio “wants to establish a legal framework in which private investors help pay for a student’s education in exchange for a share of the student’s earnings after college. In essence, the student would finance college less with debt and more with equity.”
One thing never mentioned in these discussions is the way these kinds of financial instruments would exacerbate inequality. As we’ll see, even a preliminary financial model of these instruments shows that, when it comes to the percentage of income, women would pay 8–22 percent more relative to men, and a poor woman of color would easily pay 40 percent more relative to a rich white male, in order to attend college.
As normal, it’s tough to model an imaginary market that won’t exist at scale without extensive government intervention because of profound adverse selection problems. But let’s give it our financial engineering best. I’m following the format of “income-share agreements” (ISA) funded by private, profit-seeking markets, where tuition is paid upfront in exchange for a percentage of future earnings.
One of the most important parts of private ISAs to their advocates is that the percentage of future earnings you have to pay isn’t fixed, but instead is set depending on your school and predicted earnings. Many proponents say that this will drive people to better schools with higher graduation rates as well as in-demand majors. Why? Because, since students will end up making more money this way, the private ISA lender can charge them a lower rate.
It’s not clear if the consequences would be what proponents expect. A quick model I ran shows that there’s no reason to believe it would lead students to schools with higher graduation rates, because at reasonably high discount rates this instrument would prefer the smaller payments upfront that one would get from a dropout. More generally, it’s tough to model small changes in future payments from things you could discern at the age of 18. But there are three things you know at 18 that are correlated with future income: gender, race, and parental income.
Let’s take the AAUW report “Graduating to a Pay Gap,” which allows us to focus on immediate post-college workers and is consistent with other numbers. It finds that one year after graduation, “women working full time earned only 82 percent of what their male colleagues earned.” Controlling for major, hours worked, and employment sector, “women still earned just 93 percent of what men earned.” So a wage gap of between 82 and 93 cents on the dollar.
How do we model this? Two important things to remember. First, we can just abstract away from the debate over what causes the wage gap. Many reader are no doubt already saying that this gap is driven by women’s choices, such as work effort, type of firm, family choices, and so forth. Of course, lol at uncritical notions of “choice” under a system of patriarchy, but none of that even matters for the financial engineering. The wage gap could be caused by unicorn farts for all the financial instrument cares. All we care about is the predicted cash flow, and the gap will exist no matter what the causes turn out to be.
The second is we have to model what we’d know when the contract starts. Again, imaginary market, but it’s not clear to me how the instrument should handle the full wage gap (82 cents) versus the gap controlling for major and industry (93 cents). Remember we need the person to sign the contract at 18 because they need the money to start school. Yet majors can change until graduation and are only weakly tied to industry. Think of an Ivy League basket-weaving major who goes off to Wall Street, or a STEM graduate who goes and teaches high school or works for a nonprofit. This means you couldn’t control for industry and perhaps even major while determining the initial rate, so you’d probably have to model the full 82 cent wage gap.
So what does this look like? Let’s make some additional assumptions. It’s a 15-year contract, and during the first four years in schools you make no income. You go on to make 11 years of payments. There’s a discount rate of 12 percent.  Initial tuition payment of $29,000, the average debt for students who take out loans. $43,000 starting male salary. These are all reasonable, and the results are robust to different numbers. Note that these are the guts of how you do it; all the difficult modeling would be in predicting the cash flows, which we are taking for granted.  (Some R code for this too.) What are those results?
To get $29,000 for tuition, a man needs to pledge 15.96 percent of 10 years of future income. A woman facing an 82 cent pay gap would need to pledge 19.5 percent, 22 percent more relative to the male cost of funding education. Even at the 93 cent pay gap, a woman would have to pay 17 percent, or 7.5 percent more relative to men.
Also All That Other Structural, Intersectional Inequality Too
I think you know where this is going. We know that your parents are a major determinant of your income. Chetty et al. have your parents’ rank in the distribution of income determining 30 percent of your rank. There also exists a major wage gap for people of color. The Center for American Progress estimates women of color making 64 cents on the dollar compared to white men, and there are many other comparable numbers.
There’s a lot of cross-correlation in these numbers, so let’s take a stab at a low estimate. Let’s imagine that between a white man from an upper-income family and a woman of color from a poor family there’s a 30 cent gap. For every dollar the latter makes, the former makes 70 cents. Given the strong parental income link, this doesn’t seem far-fetched, though I’m happy to consider different numbers.
If the poorer woman of color makes 70 cents on the dollar in this scenario, she’d have to pledge 22.8 percent as compared to the male 16 percent—a relative difference of 42 percent.
But But But
I imagine many people would say that anti-discrimination laws should prevent this from happening. But there are two problems with that.
First, it’s not clear what laws would be in play, because one of the first arguments proponents of ISAs make is that they aren’t loans. An AEI and New America report by Kevin J. James and Alexander Holt on safeguards puts it upfront, saying we shouldn’t apply “existing consumer protection laws used for loans“ to ISAs, because “ISAs are not loans.” A sample funding terms document from Upstart has the third sentence, bolded and underlined, say “This agreement is not a loan.”
If it’s not a loan but some sort of business partnership, it’s not clear to me that the Equal Credit Opportunity Act, the main defense against anti-discrimination in lending, would apply. It’s not even clear to me that any anti-discrimination laws would apply. (The AEI/New America report doesn’t mention bringing over anti-discrimination from loans to the not-loans of an ISA.) Calling an ISA not-a-loan is good for all kinds of fun regulatory arbitrage but would also exempt it from the infrastructure designed to combat this unfairness.
Even if anti-discrimination was extended, ISAs would invite redlining on a mass scale. Lenders are known to redline when all they have is downside risk. Here lenders would also have upside benefit and every incentive to seek out all avenues under which people earn high incomes, fair or not. Outside of slavery and indenture, we’ve not had many experiments in which a third party has a claim on every dollar a person earns, and in those systems we’ve seen huge amounts of control over every detail of a person. The incentives to discriminate would be an order of magnitude higher here.
You can’t say that ISAs would tease out the minor pay differences that would come from marginal higher education choices without also noting that they would focus like a laser on the pay differences that come from structural discrimination. Efforts to tackle this through anti-discrimination would be much more difficult than with normal loans, even without a well-coordinated effort to exempt them from loan regulations.
When even the most clever of alternatives results in a cruel, regressive instrument that exacerbates income inequality and further breaks the link between equal opportunity and education, maybe there’s something to the centuries old idea of free higher education.
 I have no idea if that’s a high or low discount rate, because this is an imaginary market that doesn’t exist because of profound adverse selection problems. Remember it must be higher than student loans since an explicit point of ISAs is to transfer risks from debtors to lenders, which will need to be compensated.
 I’m tempted to do some Markov Chain Monte Carlo modeling and other financial engineering bedazzling to make you think this is all rocket science, but it wouldn’t add any value. Note how easy that temptation to complicate is, though.