The Problem with Models

In her Notebook column in Saturday’s Financial Times,¹ Gillian Tett notes that auto loans, including “subprime” auto loans, have risen sharply over the past year.  In addition, the number of delinquent car loans, especially in the subprime category, are rising, too.  In fact, a million more Americans are delinquent on their car loans now than in 2010.

In trying to come up with a reason for the increase, which has apparently left some observers baffled, Ms. Tett says:

“But there is yet another possible explanation: consumer behavior might be shifting and, in the process, upending economists’ projections about default, as subprime mortgages did a decade ago.”

Ms. Tett notes that economists missed the subprime mortgage collapse because, inter alia, their models assumed consumers would default on the loans in a particular way: first, credit cards; second, auto loans; and last, mortgages.  But that wasn’t what happened.  People defaulted on the mortgages first.  Ms. Tett suggests this was because it was more acceptable to default on a mortgage but people needed their credit cards to live.  I would say it slightly differently: People started defaulting on their mortgages first because they realized that, in many states, it took a long time for people to get evicted from their house, while the effects of defaulting on their credit card or auto loan were more immediate.

In any case, and more importantly, as Ms. Tett says: “Most economists missed it [i.e., the subprime mortgage collapse], since their models assumed that the future would behave like the past.”  Which is not only a warning to economists, but also to the rest of us, about economic models.  Models are simplifications of the real world and are only as good as the assumptions in them.  If behaviors change, if the models don’t include the appropriate behaviors, or if the models are just simplified too much, the models may give the wrong answer as to what is going to happen in the future or why something happened in the past.  If the models aren’t right, their answers may not be right.

Which is something to remember when economists, or any experts of any kind, tell us what policy to follow or what we should do based on what their models say.  They might be right – or their models might be wrong.
--------
¹ Gillian Tett, “What is behind the spike in subprime auto loans,” Financial Times, March 16, 2019.  I would link to it, but the Financial Times has a very high paywall