By Matt Levine
A derivative is more or less a bet on the future value of some underlying quantity. If you think the underlying quantity will go up, or go down, or stay within a range, or whatever, and I think the opposite (or, at least, I can hedge it), then we get together and make a bet on it. But the underlying quantities -- and the ranges and so forth -- are not just given to us in the world. We have to choose them, and we have to agree. You want to bet on an underlying quantity that you think will go up (or whatever). I want to bet on an underlying quantity that I think will go down (or whatever). For the bet to actually happen, we need to reach agreement on some underlying quantity on which we have opposite views.
Sometimes that is easy and straightforward: You think rates will go up, I think they'll go down, we write each other a swap. Often, though, the work of a derivative designer is to build a bet on some underlying quantity where he has a specific and well-supported intuition about the specific quantity one way, and where his counterparty has a vague and general intuition about the world the other way. This is the central scandal of the Abacus, etc., mortgage-backed securities cases: A bank, or a hedge-fund client, would build a collateralized debt obligation filled with specific mortgage bonds that it thought would perform poorly, and would then sell it to an investor whose intuition was that mortgages generally would do well. This, of course, ended in tears, and in accusations that those investors were tricked into buying bonds that were "designed to fail."
Here is the story of a Utah preschool program that was funded by a social impact bond sold to Goldman Sachs. That bond is basically a derivative whose underlying quantity is the success of the preschool program: If the program's success is above X, Goldman will earn an above-market return; if the success is below X, Goldman will get a below-market return or a loss (the details are unclear). This is a good idea! For Goldman and its "philanthropic partner," the success-based payment provides incentives to make sure that the programs they fund actually work well. For the school district, they're a hedge: If the program works well, the district should be happy to pay a bit more for it; if it doesn't, then at least the district saves some money it can use to try something else.
But what does "success" mean, and what is X? Well, the measure was the percentage of "at-risk" children in the preschool program who avoided special education, and the barrier for success was 50 percent. The program achieved a success rate of "almost 99 percent," so Goldman (and, one assumes, the children) won. But the way that the children were identified as "at-risk" seems to have been over-inclusive: A child is "at-risk" if he scores below 70 on a test called the P.P.V.T., but most children who score that low do not actually end up needing special education anyway.
Before Goldman executives made the investment, they could see that the Utah school district’s methodology was leading large numbers of children to be identified as at-risk, thus elevating the number of children whom the school district could later say were avoiding special education. From 2006 to 2009, 30 to 40 percent of the children in the preschool program scored below 70 on the P.P.V.T., even though typically just 3 percent of 4-year-olds score this low. Almost none of the children ended up needing special education.
When Goldman negotiated its investment, it adopted the school district’s methodology as the basis for its payments.
You see the problem? The measurements -- which were the district's measurements! -- seem to have had the result of rigging the bet in Goldman's favor. The district (apparently) had a general worry about failure, Goldman (apparently) had a specific view about the actual metrics used to measure success, and Goldman capitalized on the district's metrics to make a profit. Or not -- the evidence for all of this is pretty circumstantial -- but it is not hard to believe. (Disclosure: I used to build derivatives at Goldman.) The scandal is that this preschool was designed to succeed.