Paper by Uday Rajan, Amit Seru and Vikrant Vig
Abstract:
Using data on securitized subprime loans issued in the period 1997-2006, we demonstrate that as the degree of securitization increases, interest rates on new loans rely increasingly on hard information about borrowers. As a result, statistical default model fitted in a low securitization period breaks down in the high securitization period in a systematic manner: it underpredicts defaults for borrowers for whom soft information is more valuable (i.e., borrowers with low documentation, low FICO scores and high loan-to-value ratios). We rationalize these findings in a theoretical model that highlights a reduction in lenders’ incentives to collect soft information as securitization becomes common, resulting in worse loans being issued to borrowers with similar hard information characteristics. Our results partly explain why statistical default models severely underestimated defaults during the subprime mortgage crisis, and imply that these models are subject to a Lucas critique. Regulations that rely on such models may therefore be undermined by the actions of market participants.
DD:
Don’t get me wrong, boom time data was a factor, but look at the facts, going by the case-shiller index housing peaked around q2-2006 and had only a very moderate decline of about 3.5% over the following year to q2-2007, yet in the sub-prime sector aberrant patterns of delinquencies/defaults appeared as early as q1-2006, particularly a jump in early pay defaults (people who take a mortgage and default on one the firsts monthly instalment) , that is before any decrease in home prices, this is due to very sloppy underwriting late in the boom cycle and fraud. Conventional mortgages did better as the fraud was mostly sub-prime and alt-a and I will grant you that for these the boom time data is the main culprit.
Now I agree that the crash would have happened anyway simply because there is a limit in how far you can go to create “affordability” products, basically ARM loans, or I/O or other gimmicks that allow people to buy something they can’t afford.
JCK,
I’m willing to accept that at least some of the data was fake or inaccurate, but I still think the main issue was the boom time data. Without that, you wouldn’t have the low default rates.
Let’s assume that all of the data was fake. We’d expect a pool of fake loans (forged identity/credit scores etc) to have default rates that differ from predictions based on the forged credit quality pretty quickly. But instead mortgages did well for quite a while, until prices started to drop.
I’m not saying that forged documents didn’t make the problem worse (by causing even more bad mortgages to be issued) but rather that it would’ve happened anyway.
What a joke. More ‘quantitative’ BS. The truth is these borrows would have been defaulting all along…if not for the rising price of the underlying asset. As soon as prices stopped rising, defaults were guaranteed due to rent arbitrage. How can supposedly smart people be so stinking dumb…even now. Correlation is NOT cause and effect. Sheesh…
DD:
“…the abstract suggests that “soft” info is more valuable than numerical data on the lenders.”
Yes and I agree with it ( I call the current crisis “the super-cruncher crisis” following the title of bad book I read last year), the problem is that the numerical data can be “gamed” and was gamed on a spectacular scale, for example you could “rent” someone credit score, you could get fake certificates of income / employment, fake assessment value etc…and all that “hard” information was fed to a computer without a minimum of checking, so in addition to the point you raise about boom times data and lack of bad times data, there are many other issues, it doesn’t even matter that the rating agencies had only boom times data, their data was basically fake data, particularly for sub-prime and alt-a.
Numerical data is important but it has to be clean and it won’t be incentives are misaligned for ex. if they get a commission to close a trade they will actually help the borrower to game the system.
JCK,
I don’t know if I buy their argument. I haven’t read the paper but the abstract suggests that “soft” info is more valuable than numerical data on the lenders.
I think the obvious blunder on the rating agency side was to use data collected during a housing boom to predict default rates. Of course no one defaulted during boom times, which skewed the data. If you couldn’t afford this month’s mortgage, you just need to refinance. Once prices started coming down, everyone defaulted because they couldn’t actually afford their home.
I guess you Do need actual research to figure that out at least for the benefit of regulators and rating agencies…
In another important finding, when there are clouds the Sun is not visible … duh
You needed to do actual research to figure that out?