The once-vaunted FICO credit scoring system is now being blamed for failing to flag risky home-loan borrowers. Will an overhaul be enough to appease angry lenders?
Business Week
February 18, 2008
BYLINE: By Dean Foust and Aaron Pressman, with Brian Grow in Atlanta and Robert Berner in Chicago
SECTION: In Depth -- Special Report; Pg. 38 Vol. 4071
From humble beginnings in 1956, Fair Isaac Corp.s credit score-- developed by engineer Bill Fair and mathematician Earl Isaac to help banks and department stores calculate their customers' creditworthiness--has come to loom over consumer finance like no other statistical measure ever has. The ubiquitous three-digit FICO score now helps determine everything from the interest rates people pay on their credit cards to their attractiveness as job candidates. Some hospitals have even begun checking FICO scores before admitting patients. "FICO is the wizard behind the curtain of the economy," says Matt Fellowes, a scholar at the Brookings Institution, a Washington think tank.
But with mortgage defaults surging and credit-card issuers bracing for more problems, the wizard seems to have lost some of its magic. A slew of unforeseen problems, some of Fair Isaac's making and others not, have combined to weaken the credit-scoring system on which most U.S. lenders and investors rely. The FICO score, last overhauled in 1989, is based on a complex formula using many variables--and yet it can be manipulated fairly easily by ordinary people. In the past few years a group of "credit doctors" and mortgage brokers began devising tricks, some illegal, to help borrowers juice their FICO scores to qualify for credit cards and mortgages on homes they couldn't afford. At the same time new, exotic mortgages were bursting onto the scene and Fair Isaac was slow to keep up with the changes. By the end of the housing boom in 2006, FICO's accuracy in predicting the likelihood of a borrower's repaying a debt had slipped. "The more heavily lenders and bankers relied on credit scores, the more mistakes were made," says Anthony B. Sanders, a finance professor at Arizona State University and former head of asset-backed research at Deutsche Bank in New York.
Yet as FICO was becoming less effective, lenders were relying on it more and more. In earlier times, banks would go to great lengths to vet potential borrowers, checking pay stubs and tax returns, calling employers, poring over investment account statements, and on and on, a process called underwriting. The mortgage boom changed all that: Wall Street investment banks were buying up every loan in sight, and lenders had to race to keep pace with the surging demand. The FICO score became as important as a pitcher's earned run average: It was a single, universal statistic that, in theory, could communicate a loan's quality to lenders, investment banks, and investors. Emboldened by its success, Minneapolis- based Fair Isaac marketed the score for other purposes and began offering new products for different industries.
Now the credit markets are in disarray, and big mortgage players like HSBC, JPMorgan Chase, and Washington Mutual --perhaps opportunistically--are laying much of the blame at Fair Isaac's feet, arguing that its score didn't predict delinquencies as expected. (Meredith Whitney, an analyst at CIBC World Markets, called FICO scores "virtually meaningless" in a December note to clients.) Consumer advocates and state regulators are clamoring for Fair Isaac to disclose its formula. And credit-card providers are beginning to question the score, too. "So many people, I think incorrectly, looked at FICO as being the' measure of risk," Discover Financial Services Chief Executive David W. Nelms told analysts in December.
Fair Isaac vigorously defends its product. "We don't think FICO scores have caused or contributed to the subprime mortgage problem," says CEO Mark N. Greene, a 12-year IBM veteran who took the helm at Fair Isaac last February as its problems were becoming apparent. Lenders that followed traditional underwriting standards, he says, "steered clear of subprime issues."
Despite Greene's assertions, however, Fair Isaac has announced a sweeping overhaul of the FICO score, its most dramatic ever. The firm promises FICO 08 will be a better predictor of consumer behavior. "The version that's out there today does a fine job," says Greene. "But FICO 08 does an even better job." Greene also acknowledges that Fair Isaac had grown insular, even arrogant, over the years as its hold over the credit- scoring market strengthened. "Customers haven't always liked the way we've behaved as a company," he says. "We haven't always been as customer-centric as we should."
It's unclear, however, whether FICO 08 will address all the problems that hampered the previous version. What's more, the firm isn't rolling out the new score until later this year, and banks won't fully integrate it into their lending models until 2009. Even if the new score is all it's cracked up to be, credit doctors will likely try to manipulate this one, too.
Fair Isaac's basic approach to credit-scoring hasn't changed much since 1956. The FICO score assigns consumers a rating between 300 and 850 based on factors like total debt burden, payment history, types of loans, and the number of times they've applied for credit. (Income isn't a factor.) The median credit score is 723; scores above 800 are considered excellent, while scores below 620 are considered poor.
CREDIT DOCTORS COME CALLING
The firm's first big deals came in the mid-1980s when Diners Club, Bank of New York, and Mobil Oil approached it to develop an automated scoring system to assess potential credit-card customers and mail pre-approval letters to those who fit certain profiles. The firm's statisticians created customized models for each, later developing a generic one that any lender could use. Fair Isaac went public in 1986; a decade later mortgage giants Fannie Mae and Freddie Mac began requiring FICO scores on every loan they bought from lenders. Now more than 80% of mortgage lenders use the score. (The big three credit bureaus, which collect people's credit histories, each offer their own scores as well.)
Today FICO's applications go far beyond home loans, however. Insurers look at credit histories to set premiums. Companies pull credit reports as part of their background checks of prospective employees. Retailers analyze FICO scores to find neighborhoods for new stores. And in recent years Fair Isaac has marketed other esoteric models to help casino operators predict which customers are likely to be the most profitable and health insurers to predict which patients are least likely to take their medications.
Meanwhile, as the housing market was heating up, borrowers and lenders were figuring out methods, both legal and fraudulent, to game the scoring mechanism--and Fair Isaac
failed to keep pace. Consider the rise of credit doctoring, a legal if highly questionable method for boosting a borrower's credit score. Larry D. Hall, a former drug addict who once lived in a homeless shelter, has made a name for himself on Atlanta's south side as the man to see if you can't get a car loan or mortgage. For $500, Hall arranges for a client to become an "authorized user" on the credit- card account of a retiree with sterling credit, a tactic known as "piggybacking." The method, says Hall, can boost a score by 50 to 100 points within a few months. A study by credit-rating agency Fitch Ratings, which looked at loans with average FICO scores of 686, found that 16% had employed the "authorized user" ruse to boost the applicant's score. Fair Isaac says FICO 08 will close this loophole.
BUSY HACKERS
Yet Hall contends he can raise a troubled borrower's score by 50 to 200 points even without piggybacking. He has figured out another ingenious trick: issuing proprietary credit accounts to customers. For $395, he'll report to the credit bureaus that he's opened an account for a customer with a $5,000 limit. The new account--and its low balance-to-credit-limit ratio--helps improve the customer's score, even though the customer can't actually tap the credit line.
If anything, business has gotten better for credit doctors since banks began raising their lending standards. "It's going to spread and grow more popular now that we're in a tighter lending environment," says Evan Hendricks, author of a book on credit scores and editor of the newsletter Privacy Times.
Cory Lamb, for example, sought out Hall's counsel last fall to spiff up his score so he could qualify for a mortgage. The 26-year-old owner of adult-oriented Internet sites says Hall has already helped him raise his score from 513 to roughly 600. On Hall's advice, Lamb flooded credit bureaus with letters disputing that older accounts shown as delinquent or defaulted were his--and demanded that the creditors produce the original loan contracts, a stipulation of the Fair Credit Reporting Act. Given that defaulted loans can be sold two or three times between collection agencies, the strategy is a bet that the current creditor has no idea where the original paperwork is. Lamb was able to remove 14 different debt blemishes from his record. In a few more months, he says, "I'll have a clean slate." Fair Isaac says it's sending cease-and-desist letters to credit doctors (though it won't comment on individual "doctors") and is reporting them to federal authorities.
Credit doctors weren't the only ones manipulating FICO scores during the housing boom--many independent mortgage brokers found ways to cheat the system, too. One software program, PDF Password Remover 2.5, is easily found on the Web. It's designed to help users override the passwords on protected documents; brokers have misused it to hack into the credit reports being sent from credit bureaus to lenders to boost FICO scores. The tactic "was common enough that everyone on the underwriting desk made sure they pulled credit reports [themselves] so they won't get duped," says one mortgage underwriter in Dallas.
Whole companies have formed to help brokers exploit FICO's flaws. Credit- Xpert, a Towson (Md.) startup founded in 2000, claims to have reverse-engineered the formula. With its tools, which are legal, mortgage brokers can run endless "what-if" scenarios to see which moves would boost their customers' scores enough to qualify for a loan. "Demand for our services has gone up dramatically," says CreditXpert CEO David G. Chung. "We're now getting more requests from brokers for advanced user training." Says Fair Isaac's Greene: "We don't believe it's possible to reverse-engineer the FICO scoring formula."
Until a few years ago, FICO was just one factor in the underwriting process. But as Wall Street grew hungrier for mortgages it could stuff into securities and sell to investors, it came to value FICO as an easily understood risk measure. Lenders were all too happy to use it as a substitute for laborious underwriting. "There were investors around the world demanding more and more deals, with investment bankers happily supplying the business," says Ron Chicaferro, a mortgage consultant in Scottsdale, Ariz. "It trickled down to the lender, who told their sales force, The faster you can get me a score and close a loan, the better. We'll forgo the documentation.'"
Throughout the housing boom, Fair Isaac promoted FICO's usefulness for other purposes, too. Bond rating agencies relied on it to assign grades to mortgage-backed securities and other more exotic bonds. Over the years, Fitch Ratings upped the score's weighting in its model, reflecting the lending industry's growing reliance on the measure. "Fair Isaac worked with us to develop the [securities] ratings model with FICO," says Glenn Costello, co-head of Fitch Ratings' U.S. residential mortgage-backed securities group.
NO WARNINGS
Now Fair Isaac's Greene takes issue with the way Wall Street applied credit scores to troublesome mortgage-related securities such as collateralized debt obligations. FICO scores, he says, were intended to measure the likelihood that a single borrower, not an entire pool of home loans, would default. And the score, he says, was never geared to exotic loans whose variable rates could soar. "It didn't anticipate a product that would effectively double the interest rate six months down the line," he says. "That's not the way the score is constructed." But his argument rings hollow to Fitch's Costello. "I never heard them warning anyone away from FICO," he says. "I can't say we were ever told it was a bad idea." (Fair Isaac says that over the years it has warned Fitch and the other rating agencies that there are appropriate and inappropriate ways to use FICO scores.)
While Fair Isaac was singing FICO's praises to bankers and ratings agencies, the model was breaking down. According to a Fitch study, the average FICO score of borrowers who stopped making home-loan payments was 589 in 2001, compared with 620 for those who were paying on time--a 31-point difference that pointed to FICO's predictive ability. By 2006, as subprime loan volume was surging, the gap had closed to just 10. Costello says the data suggest "there's something wrong with FICO." Adds Jeffrey E. Gundlach, a mortgage-backed securities expert who runs the TCW Total Return Bond Fund: "There's nothing in the FICO score that worked in terms of predicting the default and delinquency trends." Fair Isaac 's own analysis shows only a "tiny bit" of erosion in the formula's predictive value for subprime mortgages over the past few years, an amount that's "not alarming," says Tom Quinn, vice-president for scoring solutions.
Golden West Financial, a longtime FICO skeptic, is one of the few mortgage lenders to minimize its use in recent years--and it credits that decision for its below-average mortgage losses. Now a subsidiary of Wachovia, Golden West's delinquency rate on traditional mortgages is running at 0.75%, vs. 1.04% for the industry. Richard Atkinson, who oversees part of Golden West's mortgage unit from San Antonio, says the bank calls to verify employment, examines a borrower's stock holdings and other assets, and employs a team of appraisers who are judged not by the volume of loans but by the accuracy of the appraisal over the life of the loan. "The way we do business is a lot more costly, and cost was a big reason many competitors embraced credit scoring," he says. "But some of our best borrowers had low FICO scores and our worst had FICO scores of 750."
James C. Blaine, CEO of the $12 billion North Carolina State Employees' Credit Union, the nation's second-largest, has long eschewed credit scores in favor of a more egalitarian process. He charges those of his 1.2 million members who meet his underwriting standards the same flat rate--now 6.25% --regardless of their credit history. "You shouldn't abuse a good person just because they don't understand the financial system," says Blaine. The former construction worker argues that subprime borrowers default because of high interest rates, not FICO scores. He says there's barely a difference in defaults among his borrowers with the lowest scores and those with the highest. And of his members who would be traditionally classified as subprime, just 1.25% defaulted on their home loans, well below the 7% that analysts expect at lenders like WAMU this year.
A FINE-TUNED UPGRADE
Most lenders, despite their vocal protests, are sticking with the FICO score for now. Many have responded to the subprime debacle simply by raising the score for new subprime loans. Consider Accredited Home Lenders, a San Diego firm that specializes in subprime mortgages. Back in 2004, AHL could arrange 100% financing to any applicant with a 640 score--including people who'd never owned a home. Today, borrowers with a 640 score may be asked to make a down payment of 15% or more just to get the loan.
FICO 08 could well help matters. The new system still spits out scores in the 300-to-850 range, but its analysis goes deeper. Whereas the current score penalizes consumers equally for any type of delinquency, FICO 08 is fine-tuned to reflect how many times a borrower is delinquent and the types of debt involved. Overall, Fair Isaac estimates, FICO 08 will improve the accuracy of lending decisions by as much as 15%, cutting default rates.
But only two of the three major credit bureaus have agreed to use FICO 08 with their consumer data. (Equifax, based in Atlanta, says it won't implement the new score. Fair Isaac sued it in 2006 over a rival credit-scoring product.) Even after the credit bureaus get on board, lenders may need 6 to 12 months to revamp their systems. Meanwhile, credit doctors and other characters will start lining up to take their whacks at the new system. Says Hendricks: "I don't expect any of this to go away."
