Bonus-Hungry Number-Crunchers Are No Substitute For Traditional Analysis

Mike at Rortybomb has a very interesting post looking at some data that, as he says, “should cast some doubts on how much large banks can provide social utility by replacing the soft skills, the comparative advantage of small and mid-sized banks, with number-crunching computers, Ivy-Leaguers and giant bonuses.”

The background, as I’ve noted earlier, is that when you try to assemble a really big financial institution it’s not possible to rely on detailed monitoring to assess the creditworthiness of borrowers. Instead, you’re relying on formulae and securitization to “automate” the formerly laborious process of small-scale banking. If the automation can be made to work, the benefits are potentially large because a bigger institution should, if it can be managed well, be able to bear more risk and thus help more capital get put to use. But can such institutions actually be managed well? He observes:

One interesting test, using a regression discontinuity method, is to see how the behavior of defaults looked around the securitization checkoff line. A FICO score of 620 was the cutoff for most loans – if your loan was 619, you couldn’t be resold to an investment bank in a CDO. At 621 you could. This number was picked arbitrarily by the GSEs in the 1990s, and kept by the hedge funds and investment bankers in the 2000s who were trading the stuff. Since it is arbitrary, a FICO score at 621 is just marginally better than 619. There is no magical jump there in terms of how FICO is measured at that point.

Now check out these charts from “Did Securitization Lead to Lax Screening? Evidence From Subprime Loans” (by Benjamin J. Keys, Tanmoy Mukherjee, Amit Seru, and Vikrant Vig):

In general, higher FICO scores are associated with lower default rates, which is as it should be. But around the 620 cutoff between the loans you can resell and the loans you have to keep, the default rate actually leaps up.

Here again we see that borrowers with FICO scores just above the threshold are more likely to default than are borrowers with FICO scores just below the threshold, contrary to what we would expect based on the scores.

For the low-scoring borrowers, a bank has to “do the homework” in a traditional sense. The slightly higher-scoring borrowers are slightly more creditworthy, but a lender doesn’t need to “do the homework” he can just use the number-crunching computers and bonus-obtaining Ivy Leaguers. Except it seems that the number-crunching does a worse job.