Todd W. Schneider has a write-up and some analysis.
I decided to dig in with some geographic analysis, an attempt to identify the loan-level characteristics most predictive of default rates, and more. As part of my efforts, I wrote code to transform the raw data into a more useful PostgreSQL database format, and some R scripts for analysis. The code for processing and analyzing the data is all available on GitHub.
I recommend reading the entire post.
It’s interesting that at the end of a write-up that blames bank standards for the housing “bubble”, he doesn’t seem to notice that standards were getting slightly tighter during the “bubble” according to his graph on FICO scores. I bet if FICO scores had gone down, even a little, during the “bubble”, he would have noticed. I’m kind of amazed at how everyone seems to accept this narrative as an explanation for the crisis. He has some great graphs showing how areas where new building was happening had default rates of 30% and more and he shows that defaults were very strongly time-oriented (very few mortgages originated before 2004 defaulted, but a lot of 2006-2008 mortgages defaulted). This is the picture of an exogenous shock. Even if banks wanted to destroy their lending standards, they couldn’t possibly organizationally create that sharp of a change in standards.
Some argue that the banks’ behavior (the behavior that never shows up in FICO scores, but we know must be there) was covered by the high prices that were created by their unsustainable credit, until the bust uncovered the inevitable. Many point to statements in 2001-2003 that there already was a bubble then. But, buyers from that period never saw a loss, so there is no basis for saying there was a bubble in those years. If there was a bubble from 2001-2004 that was covered up by its own unsustainable paper gains, then when the mortgage market collapsed (which the author agrees is still the case), wouldn’t the bust have fallen back to pre-“bubble” levels?
So, homeownership levels changed from about 64% to 69% over the course of a decade, and many parts of the country then had 30%+ default rates as a result? There is a lot of discretionary narrative building going on, and ignoring scale and timing.
I do not equate lending standards with FICO scores alone. The most important lending standard has to do with collateral. I would prefer a low LTV with a low FICO to a high LTV with a high FICO. I would prefer a 95 percent LTV purchase loan to an 80 percent LTV cash-out refi. I would prefer a 90 percent LTV owner-occupied loan to a 75 percent LTV investor loan. (I may not have the right empirical magnitudes, but you get the idea). I stick to the view that the boom in house prices was fueled in large part by lowering standards on collateral. The bust was severe because the boom was so large. That narrative does blame loose lending standards, not an exogenous shock.
LTV was rising among sub-primes, but down payments in absolute dollar amounts were level or rising, and according to the BEA, real housing expenditures (in terms of rent and imputed rent) were declining.
It’s hard to imagine how low LTVs could be the causal factor in pushing up prices in this scenario.
http://idiosyncraticwhisk.blogspot.com/2015/04/housing-tax-policy-series-part-29-down.html
Kevin:
Schneider’s paper doesn’t blame “bank standards” for the housing bubble, as you assert. Nearly the opposite in fact. His paper is focused on the early weaknesses and changes in warranting standards of the agencies – Fannie, Freddie, etc. – over the period in question.
In terms of “bank standards”, throughout the period, he notes in this statement that, “…loans that were made by third party originators, e.g. mortgage brokers, increase the hazard rate by 17% compared to loans that were originated directly by lenders.” That variable alone – the “channel” as he has labeled it – and its predictive value on default rates seems to be one that the agencies (at least) ignored during the bubble period.
Following Arnold’s line of reasoning (and I agree), a mortgage originated by a mortgage broker, as opposed to a conventional commercial bank, had the same default/loss potential as a 96 LTV loan versus an apparent 82 LTV loan nominal. “Bank standards” are now and were then evidently higher than those of mortgage brokers throughout the period. Fannie, Freddie, et. al., seemed to have missed at least that.
Point being, Schneider’s paper isn’t about the “banks”, nor is it a narrative about the “banks” or “bank standards”. It is about a more thorough data availability from the agencies and a more thorough methodology for using that data to evaluate/predict future default rates and potential loan losses – both for the agencies and for those who will be buying their warranted paper.
Hi, author here. Thanks to Arnold for the post
To echo what Shayne wrote: I don’t believe I ever claimed that lending to low FICO borrowers was a leading cause of the mortgage/housing crisis. In fact, to the contrary, in my experience as a mortgage analyst, I found that in the worst subprime deals, individual borrower FICO scores were often mildly *positively* correlated to default rates. I always explained that was because the borrowers who had serially refinanced to cash-out into larger and larger subprime loans artificially inflated their credit scores by apparently successfully paying off their debts, but once the subprime market shut down, they had nowhere to go but default.
It’s true that in agency/prime/Alt-A loans, higher FICO is generally correlated to lower default rates, but I’m 100% with Arnold that I’d take equity over credit score any day of the week. Angelo Mozilo of Countrywide said the same thing in private (though in public he sometimes said the opposite, which got him into trouble):
> In my judgement [sic], as a long time lender, I would always trade off fico for equity. The bottom line is that we are flying blind on how these loans will perform in a stressed environment of higher unemployment, reduced values and slowing home sales.
https://www.sec.gov/news/press/2009/2009-129-email.htm
Thanks for the correction, everyone. I apologize for the mischaracterization.
I do think my basic criticism stands, though, regarding the timing and scale of the problem. This data points to an outside factor (monetary) as the cause, not reckless lending.