Doing the homework

Sorry for the lax updates, I’ve been sick and chugging Wall-flu, the Walgreens generic Theraflu, like mad the past several days. I missed this from Matthew Yglesias:

But like Ryan Avent I have to wonder if this is really a good thing:

A lot of recent financial innovation has been defended on grounds that it improved the flow of credit or made credit easier to obtain. But increasingly it seems that it did this by allowing everyone to stop doing their homework. Magical de-risking processes made the need to do homework before investing unnecessary. ….

…You can’t really have a homework-based system at a giant institution. Things need to be handled through bureaucratic processes and rules and formulae…..It also leads to a larger amount of safe leverage if you assume you’re not increasing the amount of risk by doing less homework.

The homework they are talking about about is the soft-skills part of mortgage applications, the part of the mortgage application that isn’t just a function of your FICO score and a few quantifiable variables. The soft-skills part has almost entirely been replaced by quantification over the past decade. Part of the move in this direction is the (rightly at a certain point) fear that those soft-skills were cover for racism. Part of the move, as Yglesias suggests, is that the costs of deploying soft-skills when handling loans at large firms is too high once banks get too big. And of course, another suggestion is that the quantification process, FICO and the like, had gotten so good that they could save the costs on this homework part.

How can we tell if the third part is true? 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.

So check out these two charts from the paper Did Securitization Lead to Lax Screening? Evidence From Subprime Loans (Benjamin J. Keys, Tanmoy Mukherjee, Amit Seru, Vikrant Vig), reproduced in their powerpoint presentation:

The only difference between the 619 and the 621, besides a marginal increase in credit quality, is that the 619 had more soft skills used in it and it was very likely that the loan would not be resold but instead would stay on the originating bank’s balance sheet. So in the first chart we should expect 621 to have a slightly lower delinquency rate than the 619. If your FICO is 621, you are 2 points more credit trustworthy than 619. Instead we see a much higher rate. It is even more dramatic with the 615-619 versus the 620-624; the 620-624 should have a lower rate of delinquency, since they have a higher FICO score and are less of a credit risk. Instead we see the exact opposite.

The powerpoint is very interesting, and worth a few minutes of your time. This should cast some doubts on how much large banks can provide social utility by replacing the soft skills, the comparable advantage of small and mid-sized banks, with number-crunching computers, Ivy-Leaguers and giant bonuses.

Quick Update: There are some concerns, in the comments and at Business Insider, that one can game one’s FICO scores. Three quick points. The bank lending can’t alter the FICO score – they just type in your SSN and it shows up.

Two, picture you are a borrower without any qualms who wants to game the loan procedure, and the broker says “We need you to state your income and occupation, but we won’t verify it. Also you need to make $100,000 a year to get this loan. What do you make?” You would respond “What a coincidence! I’m a football player/astronaut, and my salary is $101,000 a year.” Gamed!

If your FICO is 619, it isn’t really easy to do some quick things to make it 621, provided you don’t rob a liquor store to pay off your credit card balance. There’s a time lag, and as anyone who wants to improve their FICO can tell you, the movements can feel arbitrary in the short run. Now if you do pay off your credit cards and make consistent payments on your bills for 6 months, and your FICO goes up, you haven’t gamed the system – you are actually more credit reliable (according to the algorithm).

This entry was posted in Uncategorized. Bookmark the permalink.

10 Responses to Doing the homework

  1. Tyler says:

    I’m not convinced that your explanation for the higher default rate amongst the 620-624 bracket is necessarily the explanation, or at least the whole answer. Isn’t an equally likely reason that brokers were fudging the numbers for applicants they wanted to see approved to (just) exceed the 620 cut-off, so that those in the 620-624 bracket contained a significant number of people who were actually lower credit applicants than those in the (truly) 615-619 band?

  2. Mike says:

    FICO is third party, it is not a score that is computed by the bank, so brokers can’t fudge it. It is very difficult for consumers to manipulate FICO easily in the short term. From the paper:

    == p.5 ==
    Though there is no difference in contract terms around the cutoff, screening is more lax above the 620 score than below it, which may create an incentive for borrowers to manipulate their credit scores. Aside from outright fraud, it is difficult to strategically manipulate one’s FICO score in a targeted manner and any actions to improve one’s score take relatively long periods of time, on the order of three to six months (Fair Isaac). Nonetheless, we investigate further using a natural experiment in the passage and subsequent repeal of anti-predatory laws in New Jersey (2002) and Georgia (2003). The passage of these laws reduced securitization in the subprime market drastically. However, subsequent to their repeal, the market reverted to pre-predatory law levels over a relatively short time horizon. The results reveal a rapid return of a discontinuity in loan performance around the 620 threshold which suggests that rather than manipulation, our results are largely driven by differential screening on the part of lenders.

  3. Taunter says:

    How very Freakanomics of you…

    The one thing this leaves out is the actual amount of leverage. It could be the case that the sub-620 mortgages had to be more conventional (perhaps even >20% down), since they were going to be examined one-by-one and held. The over-620s didn’t really matter, since they were going to the GSEs anyway, so lend the full amount, get the borrower to take PMI, and call it a day.

    Not doing diligence creates a risk, but so does the sheer aggressiveness of the loans in question.

  4. Mike says:

    I know, right? Instruments and RD designs oh my…

    The powerpoint/paper shows that there isn’t a jump on the LTV at 620 FICO under the alternate hypothesis part, so LTV isn’t a major driver here. This paper won a bunch of 2008 Finance and Credit Risk awards – it’s a solid home run paper.

    It’s fun to watch them unpack some of the arguments from the 1980s about how it is impossible, simply impossible, to think that financial markets could ever have some sort of agency problems that weren’t taken care of by incentives and long-term money when it came to securitizing loans and having massive banking institutions. Shocking!

  5. Pingback: Matthew Yglesias » Bonus-Hungry Number-Crunchers Are No Substitute For Traditional Analysis

  6. Mike says:

    I added an small update to more directly address people’s concerns that the FICO score at the margins can or have been gamed.

  7. racerx says:

    Does this take into account the amount of leverage? Could the fact that below 620 required a lower LTV be more of a contributing factor to the number of defaults?

  8. pebird says:

    It’s a great example of what happens when your own money is at risk. The 619’s didn’t get securitized – they stayed on house paper – so house rules (whether soft or otherwise) were better at managing risk than securitized “rules” were for “objectively better” risks. Great graph.

  9. pebird says:

    Of course strange things always happen at the margins. Another scenario is that since there were incentives to securitize, if you got someone just at the margin (621) you worked extra hard to get them credit – so other criteria could be manipulated. Whereas if someone came in at 619, you had less interest and the customer had to work harder to demonstrate they could perform.

  10. Steve Sailer says:

    A reader named Matt R. commented on Yglesias’s blog:

    I understand your thinking, but what I don’t think you realize is that the type of judgments and research which you think might make sense for mortgage underwriting have been severely inhibited by the financial regulators. ECOA (equal credit opportunity act) and concerns about so called red-lining have made the regulators virtually eliminate all credit criteria that can’t be coded into a machine. The main push behind this effort has been due to regulators worry that non-algorithmic decision making is just a mask for rejecting minority and underprivileged applicants. This is especially applicable to any policy which attempts to be forward thinking. For instance, if an underwriter were to put in a policy discouraging loans to folks in the struggling construction industry, the general counsel would immediately reply “show me the precise data that says that there is a historical statistical correlation between construction employment and loan default rates, or else we are going to be accused of implementing a policy of ‘disparately impacting’ Hispanics.” One of the consequences of nervousness regarding “disparate impact” has been the increased use of statistical score models in consumer underwriting. In terms of how business is done, this reduces reliance on a semi-skilled workforce that leverages relationship, intuition, and local environmental factors and opens up new opportunities for technical folks good at mining data, recognizing data relationships, and creating logistic regression models and so forth. And yes, this does encourage centralization of consumer banking and the development of a highly paid and (maybe) skilled technical manager set in underwriting.

    The main downfall of this approach which relies on historical data patterns is that it is easy to lose sight of what is really happening to consumer balance sheets and that recent past results may be anomalous. So in the past situation where consumers could roll over debt due to rising home prices and the increased availability of home equity loans, the banks’ models were “tricked” by borrower profiles who were repaying their loans but yet were in unsustainable consumption and borrowing patterns. And because the top underwriting brass are absorbed in their models and far from the realities of the borrowers’ “real” situations–not being able to put the full picture together while sitting across the table from borrowers daily–a realistic picture that might have stimulated a more conservative underwriting approach did not force itself on credit committees until the loans (thus the data) started going bad.

    And by the way, you can see clear proof of the regulatory impact of ECOA laws demonstrated by their non-applicability in small business lending. Relationship and judgmental underwriting remain prevalent in small business lending, even for comparable loan sizes being made to consumers.
    The last point I’ll make is that the stupid money behind securitization business ruined everything anyway. Try adhering to rigorous underwriting processes and reasonable pricing/underwriting criteria at a bank while finance company next door has agreements to have all of its poorly priced and researched loans purchased the next day no questions asked. The banks are then faced with the decision to either put in an application process and pricing structure which no consumer will actually vie for (thus losing all market share) or following the leader. This is why tough handed regulation is needed to keep the industry in check because all can be led astray too easily. The follow the leader psychology is the source of every financial panic.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s