The Smoking Gun of the Credit Crisis: FICO

By 2004, the FICO system was gamed.

It was the final step in the destruction of Wall Street, a destruction engineered by the rapacity and dishonesty of American mortgage and fixed-income firms. Here, on Portfolio.com, Michael Lewis reveals that destruction in the best article I have yet read on the origins of the Credit Crisis. Mr. Lewis hasn't hipped to the FICO story yet, but he'll get it. He covers all the rest. The article is absolutely a must-read for anyone who wants to understand what really happened.

I first realized FICO got gamed while reading a paper by the American Enterprise Institute's Peter J. Wallison. Mr. Wallison is the principal author of the Republican talking points on Fannie and Freddie. Wallison has been carrying Wall Street's water for many years in their jihad against the GSEs. If you think Fannie and Freddie are to blame for it all, Peter J. Wallison earned his money. Like all good con men, Mr. Wallison sticks to the truth as closely as possible for as long as possible, but he slipped up with one simple sentence:

There are many definitions of a subprime loan, but the definition used by U.S. bank regulators is any loan to a borrower with damaged credit, including such objective criteria as a FICO credit score lower than 660.

Emphasis mine, but those words "objective criteria" jumped out at me.

Obviously the most objective criterion of credit risk is the actual default rate. Wallison talks about the default rates in the GSE portfolio coming largely from private-lable RMBS, but he does not cite the default rates for the actual borrowers to whom he objects most. I wondered why. If you read the paper and then look at the data (as best you can), I think you'll find something surprising.

First, it turns out that Mr. Wallison is correct that Fannie and Freddie DID take on borrowers with FICO scores lower than they should have - although not in the way he implies. But I found that the borrowers to which Mr. Wallison attributes so much damage actually defaulted at LOWER rates than that same FICO cohort defaulted for private banks. Moreover, because Fannie and Freddie bought Alt-A mortgage paper, Alt-A mortgages and even guaranteed a few Alt-A borrowers (but a very few), one can glean from the data that the farther away a borrower of a given FICO score was from the GSE's full due diligence process, the more likely that borrower was to default. Relative to the private sector, the GSEs were better judges of default risk and their default rates prove it. Even the private-label Alt-A paper bought by the GSEs has a markedly lower rate of default than the industry average.

Still, it was strange. Relative to how it worked in the private-sector data, FICO significantly OVER-predicted defaults in the GSE portfolios (when taken as a whole). If FICO is an "objective" measure, this should not have happened.

And then there is the question of the vintages.

That question became very intriguing to me when I read a much-quoted working paper published by the St. Louis Fed entitled: "Where’s the Smoking Gun? - A Study of Underwriting Standards for US Subprime Mortgages." I've long ago learned that people who have something to hide generally hide it in the title and the appendices. So, I went looking for the smoking gun the authors said was not there and quickly found it.

The authors boldly claim that:

Our results indicate that there is no evidence of a dramatic change in underwriting standards in the subprime market, particularly for originations after 2004. Given the multidimensional nature of ex ante credit risk, it is difficult to emphasize weakening in terms of some attributes as a decline in overall underwriting standards. The results show that while underwriting may have weakened along some dimensions (e.g. lower documentation), it also strengthened in others (e.g. higher FICO scores).

So, I immediately looked at their FICO data. It was shocking. To make the claim they make, you have to completely ignore the performance of the FICO cohorts and the fact that people in the business know that data. If you look at Table 12 (which divides FICO cohorts very suspiciously, but, that's another question) you will find that for the FICO cohort 660-720, first-year mortgage default rates went up a bit in 2004, doubled in 2005, quintupled in 2006 and were about eight times higher than normal in 2007. The numbers were terrible for all FICO cohorts, but that big one with SO many borrowers was just appalling.

So in 2004, bankers had suspect data and kept lending. In 2005, they had alarming data and kept lending. In 2006 they had shocking data and just kept right on lending faster than ever. In 2007, it was a travesty and they still kept lending until the whole thing blew up. Now these Bozos writing for the St. Louis Fed have the nerve to claim that underwriting standards didn't fall? Lenders knew that FICO has already failed. They knew it was gamed. Their data showed it conclusively. As if the data weren't enough, they could see the sudden profusion of new "Credit Reporting Agencies" and all the "score borrowing" and "rapid readjust" nonsense going on.

They didn't want to see it. The FICO cohort that was falling apart represented the bulk of their profitable borrowers. This was not the sub-620 extreme or the low-profit Prime book. This was the thick part of their bell curve. FICO was gamed. FICO was junk. But question FICO and you yanked the emergency brake on the Gravy Train. So not only did they not question it, they started to rely on FICO almost exclusively. They started to insist on having as little OTHER data on borrowers as possible. When they asked for the Equifax reports, they stopped asking for the employment and income data. They started pushing "low-doc" and "no-doc" paper out the door as fast as they could. See no evil, hear no evil and make a buttload of money.

You'll find that all lenders - including the GSEs - raised their FICO score standards in the later years of the crisis - to no avail. FICO was no longer "objective". It had become the foundation of a scam. As Michael Lews writes of one of the people who saw through the scam:

[He] knew subprime lenders could be scumbags. What he underestimated was the total unabashed complicity of the upper class of American capitalism.

A crisis like this is not an "'or' crisis". It's not a question of looking back and deciding whether this happened "or" that happened. It's an "'and' crisis". This happened "and" that happened but the "this" - the malplactice and fraud - predominated. Yes, there were stupid borrowers and stupid buyers of these securities (including the GSEs) at both ends of the line. But the cause of this crisis was pervasive malfeasance by the people in the private-sector, non-GSE business of selling American mortgage debt to borrowers and customers. Starting with the FICO, there was fraud at every step of the way up to and including the derivatives on the stuctured products. The willful ignorance and stupidity of borrowers and customers were just the whipped cream and cherry on top. We are living through the results what is the largest case of financial malpractice in human history - "malpractice" at best.

I'd call it "fraud".

This is why Paulson's plan's won't work: You can try every excuse in the world, but you can't fix this problem until you admit what it was and is.

Fraud.

Disclosure: No positions.

Comments

[25 years in...:]

Nailed it! FICO underwritting not only made unworthy borrowers loans it invited a new breed of totally inexperenced unskilled loan originators into the business. The sum total of loan origination and loan qualification became " whats your FICO"?

The originators were grossly overpaid and beat like racehorses to originate faster and more... they never heard the words "orginate better" The biggest shame is the collapse has resulted in the dismanteling of the industry that is most likely able to put this whole mess back together... Lenders need to get back to basics. Equity stems off default. if you want to buy a house save some downpayment money and prove your income... and yes i will check your credit, but i could give a damn about your FICO

2008 Nov 20 11:46 AM

[dlaw:]

Thanks for the comment.

FICO over-predicted and, of course, FICO under-predicted default rates. And yet FICO was totally trusted as the underpinning to all the major securitization markets.

No wonder we have a problem and no wonder our paper is not trusted.

2008 Nov 20 12:31 PM

[NJ market w...:]

Well written and correct. When will the Congressional committees call in the FICO score generators and start questioning there methods?

2008 Nov 20 01:03 PM

[Brill:]

Hey, the computer spit out those scores and it's never wrong. Don't question the computer. Only people are not perfect.

American life is all about "scores", nevermind how those scores are computed. Don't look behind the curtain.

2008 Nov 20 04:17 PM

[gemonk21:]

I have been a mortgage underwriter for many years. FICO is a big issue, but even more so is AUS - the automated underwriting sysytems used by the GSEs. Fannie uses Desktop Underwriter (DU), Freddie uses Loan Prospector (LP), and others have their own in-house systems modelled on the GSEs.

Once the AUS made a buy decision, it was virtually impossible to overturn, even if you knew there was something not right about the findings. Between 2000 and 2006, underwriting standards became slacker and slacker, and a higher and higher proportion of loans were approved by AUS.

I also underwrote sub-primes for the wall street investors. They did not ever want to hear "this is not a good loan". they bought all this trash knowing it was trash - "the model accounts for fraud and lower credit quality" they said repeatedly. Whistling past the graveyard...

2008 Nov 20 05:46 PM

[dlaw:]

Gemonk, You still have to account for the fact that Fannie and Freddie have LOWER default rates than the private banks. LOWER.

Here's a quote from an industry group about the new Desktop Underwriter 7.0:

"Stop fraud
Other changes are intended to deter fraud. For example, during the boom times, brokers were known to submit a borrower's application repeatedly, fudging the debt and income numbers each time, until Desktop Underwriter granted an approval. DU 7.0 is believed to limit the number of times that the financial figures on an application can be changed; after that, the application is locked out, similar to the way an ATM will reject your card if you enter the wrong PIN multiple times"

In other words, brokers were gaming the system.

And Gemonk, it is not as if a bank is IN ANY WAY required to lend just because Desktop Underwriter says it's okay. DU does not suddenly take the money out of a bank's Federal Reserve account and do an instant closing. It's a tool - one that was badly, badly misused.


On Nov 20 05:46 PM gemonk wrote:

> I have been a mortgage underwriter for many years. FICO is a big
> issue, but even more so is AUS - the automated underwriting sysytems
> used by the GSEs. Fannie uses Desktop Underwriter (DU), Freddie
> uses Loan Prospector (LP), and others have their own in-house systems
> modelled on the GSEs.
>
> Once the AUS made a buy decision, it was virtually impossible to
> overturn, even if you knew there was something not right about the
> findings. Between 2000 and 2006, underwriting standards became slacker
> and slacker, and a higher and higher proportion of loans were approved
> by AUS.
>
> I also underwrote sub-primes for the wall street investors. They
> did not ever want to hear "this is not a good loan". they bought
> all this trash knowing it was trash - "the model accounts for fraud
> and lower credit quality" they said repeatedly. Whistling past the
> graveyard...

2008 Nov 20 07:45 PM

[dlaw:]

The thing I do blame Fannie and Freddie for is not taking the AUS data and having it analyzed with sophistication. There should have been logs of multiple logins, changes in info, and data that did not fit statistical norms.

2008 Nov 20 07:53 PM

[john s. gordon:]

lots of blame to be shared here.

crooked lenders created lots of toxic waste in order to grab big fat fees & run away with them. wall street whipped up a goulash using these ingredients & passed them on to suckers all over the world in order to grab big fat fees & run away with them.

nuff said.
> jack

2008 Nov 21 08:23 AM

[jbde:]

And now FICO's are the problem. Everyone is now on the 'tightening of lending rules' bandwagon. Thus lenders immediately limit the market of buyers.
Second, as more and more people have problems, the avergae scores are dropping.
At some point creditors [sellers] will have to lower credit standards or the markets for houses, cars, white-goods and furniture will continue to shrink. Of course they can lower prices. Not just offer lower priced options, but lower prices on the same item - costs are dropping.

And what about the car pirces? House prices come down, commodity prices come down - even food us dropping as shipping costs are lower.

But not those car prices. Discounting is not lowering. When someone actually lowers prices on an established model instead of giving rebates, they will set the new trend as telling it like it is. Peopl will flock to the truth-teller.

As for US legacy employment costs, the unions have caused a wage bubble - no forklift operator should be getting $103K. No person who attaches part A to part B should be getting a 73K package.

There is only one solution - you have to get rid of the union contracts and BR is the only way.
The government then has to do two things - guarantee car warrantees and underwrite the mortgages of the autoworkers who have bought based on their inflated wages and will require mortgage assistance,

This is the cheapest way out! Otherwise, they come back to the trough.

2008 Nov 21 09:16 AM

[BerkeleyBob:]

Very interesting analysis. I get elevated blood pressure when I read WS Journal editorial garbage or (mostly) Republican jeremiads against the CRA and Fannie and Freddie. I have long suspected that the issues with Fannie and Freddie were not poor underwriting or appraisal practices but they did practice "creative" accounting and did respond to political pressure because of the inherent conflict between their quasi-public status and obligation to maximize shareholder return. California's quasi-public work. comp insurer, State Compensation Insurance Fund has had strikingly similar issues, investigations, resignations, etc.

2008 Nov 21 01:47 PM