Most
good insurance companies and every failed insurance company are familiar with a
phenomenon called “adverse selection.” Adverse
selection says that any situation in which the buyer and seller of insurance
have asymmetric information will ultimately create a cycle whereby the low-risk
individuals forego insurance that they perceive as too expensive while the
high-risk individuals aggressively purchase insurance that they perceive as
cheap.
Take
the example of health insurance. In the
U.S., there are strict regulations about discriminating against individuals
with hereditary pre-existing conditions: let’s take hereditary hypertension as
an example. Hereditary hypertension creates
a higher risk for cardiovascular disease and, thus, a higher risk for a health
insurer. However, because the insurer
cannot discriminate for this condition in its pricing model, the insurer is
forced to ignore this predictive factor and price the insurance as if this
information were unattainable. However,
most individuals who are purchasing insurance will know whether or not they
have hereditary hypertension, and those individuals who do will have a great
demand for insurance at a price that is ignoring that risk. For those who do not have hypertension,
however, this insurance will seem unduly expensive. As a result, each year the “healthy”
individuals will tend to opt against purchasing insurance that is unjustly
expensive while the “sick” individuals will aggressively purchase insurance
that is irrationally cheap. The cost of
insurance will increase each time a “healthy” individual leaves the insurance
pool or an additional “sick” person joins at the price that ignores this risk. As this cycle progresses, the risk in the underlying
insurance pool becomes ever more highly concentrated, until eventually you are
left with an insurance pool consisting entirely of sick individuals – which isn’t
insurance at all, it’s just the original risk.
Lending
is a lot like insurance: a lender creates a diversified portfolio of loans
under the assumption that while some may fail, on average the portfolio will
perform to some level of consistency. Like
insurance, a lender is trying to minimize an unpredictable risk (or difficult
to predict risk) through diversification and the law of large numbers. Unsurprisingly then, we find that adverse
selection is also highly relevant for student loans. In the previous post we noted that the top
two causes of student loan default are the failure of the borrower to graduate and
the failure of the borrower to attain gainful employment within six months of
graduating. We also previously noted
that student loans, like mortgages, are freely prepayable – a unique attribute
in the broader capital markets. Further,
we reviewed how the combination of these two qualities introduces meaningful equity
prepayment risk into the system. At its
core, we can show that this equity prepayment risk becomes a textbook example
of adverse selection.
The
primary goal of a student lender is to predict default risk. A lender who is more skilled at predicting
the probability that a borrower will default on their obligation will
experience smaller loan losses, which will allow it to offer lower rates than
its less adept competitors. Consequently,
we would expect student lenders to expend great efforts attempting to predict
graduation rates and post-graduate employment rates. Although there are some lenders partially doing
this, the vast majority of student lenders ignore these risks almost entirely
or else price them very broadly: law students have higher average interest
rates than medical students, but the interest rates between individual law students are
largely ignorant of the individual’s academic prowess. Most importantly, student loans by their very
definition are priced at the time when the uncertainty of the borrower’s
eventual graduation and employment is at its highest – that is, when the
borrower is actually a student. The
result is that student lenders do a very poor job of predicting which borrowers are going to default; they
only consistently predict how many
borrowers are going to default.
This
presents an obvious problem: lenders are pricing loans under the assumption
that the risks of failure to graduate and unemployment are effectively
unknowable to both the lender and the borrower.
More importantly, lenders are assuming that this fact will remain
unknowable throughout the life of the loan.
This is obviously untrue: within six months of entering repayment a
borrower will know with complete certainty whether they have graduated or attained
a job. In a rational market, we would
expect these “healthy” borrowers to view their existing interest rate as unduly
expensive while the “sick” borrowers would view their interest rate as far too
cheap. Further, given that student
loans, like mortgages, are fully prepayable, then we would expect increasing
numbers of healthy borrowers to exercise their prepayment option and opt out of
their high-interest loans. This is just
adverse selection and will ultimately have the same consequences: student
lenders will find that the risk in their loan pools are becoming increasingly
concentrated.
There
is one key difference, however, between adverse selection in insurance and
lending. In insurance, the insurer has
the opportunity each year to withdraw or modify its offer to sell insurance at
the same price, that is: adverse selection in insurance drives up annual
premiums, but it does not create long-term liabilities for the insurer. In lending, the lender has no opportunity to
withdraw or modify their initial loan terms, that is: adverse selection in
lending reduces the value of the existing loan portfolio through the concentration
of default risk, which creates an enormous long-term liability for the
lender. This creates major instabilities
in today’s student loan market and, ultimately, makes the entire practice
unsustainable.
Byline:
Derek
Kaknes
