a11ce.com/isa.html
An income-share agreement (ISA) is a contract between a student and an ISA provider. The provider pays some amount of money towards the student’s education in exchange for a share of the student’s future earnings after graduation.
ISAs are generally defined by five values, all effectively set by the provider before the contract is signed: the initial payment amount, the ISA percentage, the payment cap, the duration cap, and the income threshold.
Upon signing, the provider pays the student (or really pays the school on behalf of the student) the initial payment amount. When the student finds employment after graduation that pays above the income threshold, she pays the agreed-upon percentage of her income to the provider each month. These payments are made until either the total amount paid by the student exceeds the payment cap ot the total number of payments exceeds the duration cap.
For example, consider an ISA with an initial payment amount of $10,000, an ISA% of 5%, a maximum repayment of $20,000, a maximum duration of 5 years (60 monthly payments), and an income threshold of $40,000/year. This ISA is offered to three students: Alice, Bob, and Charlie. After graduation, Alice immediately finds a job earning $40,000/year and stays there. She will pay $166/month for 5 years and then hit the duration cap, paying $10,000 in total. Bob instead starts earning $100,000/year. He will pay $416/month for four years, then hit the payment cap and have no further obligations. Charlie gets a job under the income threshold for 3 years, then finds one earning $40,000 and holds it for one year, quits, then finds a new one after a year of searching earning the same amount. He will not make payments for the first 3 years (meaning that they don’t count towards the duration cap), pay $2,000 in his first year of employment, nothing in his next year of unemployment, then $2,000/year for 4 more years before hitting the duration cap 9 years after graduating.
Here you can see some of the unique features of ISAs. From a student’s perspective, ISAs offer a sort of un/underemployment insurance in that they are shielded from compounding interest while unemployed and from onerous payments when struggling to find good employment. From a provider’s perspective, ISAs have a capped downside (because of the income threshold), low default risk (basically only from intentional misrepresentation, not from borrower inability), a shorter duration than traditional student loans, and very high potential upside if the provider is able to make good predictions about its clients.
Ideally, ISAs could also lead to incentive alignment between students and providers. This is mostly underexplored but a few companies like Clasp are looking into similar mechanisms (they call it ‘investing in students’ which is not a bad way to think of ISAs in general). I would like to explore this idea further in a later article.
Let’s define a simplified model of a $10,000 ISA from the provider’s perspective in order to look at pricing. We’ll start by disregarding the time value of money. It doesn’t change the fundamental structure of ISAs (unlike loans in which it’s a central factor) and we can mostly say providers build it into their terms based on forecasts anyway. This lets us assume students immediately get a job above the income threshold upon graduating and hold it for the duration of the ISA (because we can just elide any time periods in which this is not the case) and that their income does not change (i.e. they always earn what would otherwise be their average income over the ISA duration; without TVM this is equivalent).
By the same logic, we can fix the duration cap. I’ll choose 5 years which seems to be a common duration. The vast majority of ISA providers offer the same duration across all provided ISAs. In theory, a provider would prefer a longer duration if they expected the student’s salary to grow slower than the provider’s discount rate. This is kind of neat but not relevant for our model.
We will also fix the repayment cap because it is mostly just bounded above by explicit regulation. It is potentially subject to market competition below regulatory caps, but providers seem to prefer not to cut off their high-upside tails and instead compete on ISA%. I’ll arbitrarily choose a cap of $30,000, which is on the high side but still reasonable. Since we’re looking at equilibrium pricing, the payment cap is not so important except when considering example cases. Also notice that (for the same expected return), a lower cap means higher offered ISA%s.
Now, the offered ISA percentage is the only variable in our pricing model. Lets see how we should set it and take a look at some of the consequences that make ISAs worthy of special regulatory consideration.
We’re going to look at ISA pricing in equilibrium. By pricing, we mean the offered ISA percentage, as in our model it’s the single lever we have to affect the price that the student will pay. By equilibrium, we mean the price we should set such that we expect to exactly make out money back. This is the theoretical outcome of market competition: any higher and we get undercut, any lower and we lose money. In practice, ISA providers obviously want to actually make money, but we should expect real-world ISA pricing in a free market to resemble our model with some sort of offset somewhere in the terms.
First, consider a single student. We want to try to make sure that the ISA% times the income of the student over 5 years after graduating is equal to the initial payment amount. The duration is constant and we get to set the ISA%, so the only real unknown is the student’s future income. Ideally, we would know the actual future income of the student and set the ISA% based on that knowledge. Unfortunately, we can’t see the future! Instead, we must make predictions given whatever information we have about the student.
A common case is that providers are only aware of (and price ISAs based on) the college and major (cohort) of a student. We’re going to focus on this case both because it is very common (even when schools themselves provide ISAs) and because it produces some of the counterintuitive example cases that seem to be red flags for regulators. Given cohort information, we as an ISA provider should try to predict the average income of a student in that cohort and set the ISA% based on that prediction. If our prediction is good, we will end up very close to the actual ideal equilibrium ISA%.
For example, say we know that graduates from major X at Y college average $40,000/year in their first 5 years after graduation. So, on average, we’re going to be paid ISA% times $200,000 in return for our payment of $10,000. Thus the equilibrium ISA% is 5% because that will get us $10,000 back. Graduates from college Z in the same major, however, average only $30,000/year. We calculate 10,000/(30,000 * 5) to find the equilibrium ISA% of 6.6%.
What does this mean for individual students? Consider students Dan, Erin, and Frank, all graduates from college Z majoring in X. Dan earns $30,000/year, so he ends up paying us back exactly. Erin makes less than we predicted, and ends up paying us less than we expected. This is good for her, and demonstrates the underemployment insurance mechanic of ISAs, a nice social benefit. Frank earns $35,000/year and ends up paying back about $11,666 in total. In effect, he is subsidizing his unluckier colleagues such as Erin.
Finally, Grace, a graduate from Y in major X, gets the same job as Frank. Even though they’re making the same amount, Grace will end up paying only $8,750 in total. At first glance, this can look like the ISA is favoring a student who went to a better school. In fact, this case is a feature of ISAs, not a bug. Frank got a better ROI from his education than expected given all available information, and is paying for the prior risk of ending up worse off than his cohort average. Grace actually did end up worse off than expected and is benefiting from the insurance mechanic. From the provider’s perspective, as long as our income predictions are accurate, our total revenue will even out in the long run.
We can see here that APR is a pretty meaningless metric for ISAs. “Equivalent APR” is sometimes calculated retroactively by pretending that the initial payment was a loan and the total amount paid by a student is the repayment of that loan, but that misses the uncertainty in the repayment amount that makes ISAs unique in the first place. It’s like saying that paying for insurance is a waste of money just because you haven’t had to use it.
In conclusion, we should generally expect ISA providers to determine ISA%s based on the expected income of the student such that they will, in equilibrium, make their money back from the aggregate of all offered ISAs. Keep in mind that this is a simplified model, it’s plausible that cohorts on the lower end of expected income will face slightly higher pricing because they will probably take longer to repay, and the actual target revenue from an ISA provider is certainly going to be higher than that of the equilibrium of a frictionless market. The key insight here is that we expect the ISA% offered to a student to be roughly inversely proportional to the best prediction of their income after graduating.
This question reveals one of the key differences between loans and ISAs. The lender of a loan always expects to be paid back the entire principal plus interest. After some screening such that the lender is confident that the borrower will actually follow the terms of the loan, a lender can offer the same loan terms to everyone because their expected revenue is (more or less) the same for everyone.
Compare this to insurance, an industry which aggressively personalizes prices. The cost or benefit of offering an insurance policy is dependent on the particular consumer, and insurance providers know that they’re not going to make money on every single policy. (If they did, there would be no need for insurance!)
Let’s imagine an insurance provider decides to offer the same coverage at the same price for all customers. To keep the same expected revenue, the new single price will be the average price of their current customers, resulting in a drastic price increase for low-risk customers to cover the symmetrical reduction for high-risk customers. This will result in an adverse selection death spiral: low-risk customers will choose other providers, pushing the average price (and thus the floor for fleeing) higher and higher.
ISA providers face the same risk, forcing them to personalize prices. If an ISA provider offers the same ISA% to all students, it will either lose too much money on lower-earning cohorts or cause higher-earning cohorts to seek a different provider.
Thanks for reading! If you have any further questions or want to talk about the implications of relaxing some of our assumptions, feel free to contact me.