Just 18 months ago, the credit score was the gold standard for measuring a consumer's creditworthiness. Financial institutions counted on it for all types of credit-granting decisions and to calculate the interest rate for loans and credit cards. So great grew the reliance on this magical three-digit number that lenders allowed on it to guide them through the tricky waters of the subprime market.
But credit grantors overlooked a key point: the credit score does not predict customer behavior or delinquencies. Instead, it is a tool that weighs past payment behavior and the data found in consumer credit reports and other variables – and then assigns a score to the analyzed data. The higher the score, the more creditworthy the borrower.
As financial institutions plunged headlong into the subprime market – partly to grow their business and partly to feed Wall Street's ravenous appetite for securitized mortgages and credit card receivables – their reliance on the credit score exceeded the score's purpose. Soon, limitations began to show, slowly at first and then in a massive surge of delinquencies and defaults in the past year.
Mortgage lenders banked on credit scores to qualify applicants for stated-income mortgages, but now find themselves awash in a sea of bad debt. By the end of the year, 4% of consumers with mortgages are projected to be 60-days delinquent, up from 3% in the fourth quarter of 2007, according to Chicago-based TransUnion LLC.
The mortgage industry is not the only one to suffer. Credit card issuers that relied on credit scores to pre-qualify customers also are taking a hit. In the fourth quarter of 2007, 1.36% of credit cardholders were at least 90 days delinquent, up 32% from the third quarter. The rate may hit 1.9% by year-end, predicts TransUnion.
"A lot of lenders came to rely too heavily on the credit score during origination when they shouldn't have," says Eva Weber, a research analyst for Boston-based Aite Group. "There are weaknesses in the current scoring models that needed to be taken into consideration."
Those weaknesses include vulnerability to manipulation by credit doctors to raise scores, failure to weigh a consumer's debt-to-income ratio and changes in consumer credit usage patterns and payment priorities.
Another shortcoming is the inability of the credit score to put delinquencies into perspective. Many lenders complain of being burned by consumers with sterling payment records and high credit scores who suddenly became delinquent and defaulted; while consumers who are occasionally late paying bills receive a lower credit score – only to be proven better credit risks.
With those weaknesses exposed many financial institutions hit hard by the subsequent spike in delinquencies and deteriorating credit quality of their loan portfolios are complaining that credit scoring failed them.
Minneapolis-based Fair Isaac Co., creator of the FICO score – the standard bearer of credit scoring – has taken the brunt of the criticism. So deeply have financial institutions been rocked by the subprime meltdown, many have lost confidence in the power of scoring and current risk management practices, leading the industry to call for improvements in the methodology behind it, says Brian Hart, managing director, risk finance and compliance for BearingPoint Inc., a McLean, Va.-based management and technology consulting firm. "Credit scores are useful, but it has been too long since there has been a major rebuilding of engines that power scoring models," he says.
Fair Isaac, having spent much of 2008 vigorously defending the integrity of the FICO score, agrees that consumer behavior has changed enough to warrant the first major overhaul of its scorecards in about two decades.
FICO 08 is scheduled for release this summer. Still, Fair Isaac counters that lax underwriting practices are the real culprit for the subprime mess.
CEO Mark Greene, a former IBM Corp. executive who took the helm at Fair Isaac in 2007, has publicly stated the company often reminded clients that the FICO score is one of many tools to help make lending decisions, but that the advice was not always heeded. Instead, lenders that got caught up in the subprime feeding frenzy ignored the basics of due diligence in the underwriting process to verify incomes and assess the applicant's capacity to keep their account current.
Aite Group's Weber says, "The subprime mess does not hinge as heavily on the shortcomings of credit scoring as it does on a breakdown in underwriting."
With the initial shock of rising delinquencies and defaults wearing off, the industry is shifting its focus to improving the tools available for assessing risk, which includes credit scoring.
Fair Isaac began the process of reevaluating the FICO score two years ago based on client feedback about changes in consumer credit behavior and reports that consumers with low credit scores were turning to credit doctors to raise their scores. Fair Isaac was also moved to act by the growing popularity of stated-income loans.
"We periodically update the FICO score to reflect new dynamics in credit reporting and how consumers shop for and use credit, but we felt the time had come to test some new fundamentals in the scorecards based on what we were seeing in the market," says Tom Quinn, vice president of scoring for Fair Isaac.
The aim is to bring a greater depth of perspective to the FICO score so lenders can better segment customers during the underwriting process. One new feature will be scorecards that examine the seriousness of any delinquency as there is no strict correlation that occasional late payments and short-term delinquencies automatically lead to defaults.
"With the old scoring model, any delinquency or late payment was classified as bad. Now the scorecard will classify delinquencies by risk characteristics," says Quinn. "What we are doing is spreading the population more finely across the risk pools based on credit characteristics."
FICO 08 will include two new scorecards to measure derogatory characteristics by considering the frequency and severity of the delinquency – and how recent it is. The new scorecards also will bypass small dollar delinquencies of up to $100 in calculating a consumer's score. Many of these delinquencies arise from disputes with merchants, or with medical providers that automatically put accounts receivable into collections after specified periods even though they are still negotiating payment with the consumer's insurance carrier. "Many delinquencies in this range are non-credit related," says Quinn.
Fair Isaac also is increasing the number of scorecards for non-derogatory characteristics to eight. Non-derogatory scorecards will weigh such factors as a consumer's debt-to-income ratio and the type of debt carried.
Fair Isaac is betting this deeper level of segmentation will prompt financial institutions to enhance their own underwriting practices, especially for consumers with high scores. "A lot of consumers score well, but lenders need more underwriting to arrive at a well-rounded credit decision," adds Quinn. "A credit score is only as good as the underwriting practices around it."
Some credit experts argue otherwise, citing credit scoring's vulnerability to manipulation by credit doctors, which in recent years have emerged from the shadows and into the mainstream thanks to the Internet. A search on Google turns up more than 900,000 results for credit doctors.
The threat credit doctors pose is that they will use unscrupulous methods to raise a consumer's credit score. One popular trick is piggybacking, in which a consumer with sterling credit is paid to be added to a loan applicant's credit card account.
The method can raise credit scores by as much as 100 points in a few months. A recent study by credit rating agency Fitch Ratings found that of the loans it examined with average FICO scores of 680, 16% involved the piggybacking ruse.
There also have been published reports that some credit doctors have successfully reverse engineered the FICO score, which has prompted comparisons to hackers.
"The current ease with which credit scores can be manipulated is a huge limitation in the score itself," says Pamela Martin, director of regulatory relations at Philadelphia-based Risk Management Association.
TransUnion, which offers a competing scoring model along with Experian and Equifax, insists that the number of consumers using credit doctors is much lower – a position based on a study of 7 million consumer mortgages that underwent a significant score shift before origination. Of that group only 0.5% had suspicious activity in their file.
Fair Isaac says it will close the piggybacking loophole by excluding data for authorized users on an account when calculating a consumer's credit score. TransUnion does not take authorized user activity into account when calculating credit scores. The credit bureau is further enhancing its credit score by including triggers that automatically notify lenders when significant changes take place in a consumer's credit report, such as a sudden jump in payment obligations on an existing account or adding a new credit line and using it heavily.
"Credit scores need to weigh the velocity of the changes in a consumer's credit file," says Chet Wiermanski, group vice president of analytics at TransUnion. "It's important to get an analysis of the credit characteristics of consumers who have undergone significant changes in their credit file."
The credit bureaus say that changes to their scoring models are part of a regular cycle of validating the performance of the scorecards at least once a year or at a client's request.
"We want to know if the scorecard is pushing the highest risk scores into lower risk ranges and, if so, by what degree," says Angela Granger, vice president of analytics at Experian's Decision Analytics unit.
Despite planned changes to the FICO score and the scorecards used by the credit bureaus, risk management experts agree that the new scorecards need to be supplemented with additional analytical models. Fair Isaac and Experian have several offerings (See sidebar, below).
"More specialized scoring models are needed because baseline assumptions do not always hold up," says BearingPoint's Hart. "Just look at the assumption that home values would continue to rise and that homeowners would always pay their mortgage."
With home values declining, many homeowners with mortgages are placing a higher premium on maintaining the open to buy on their credit cards because they need it to buy necessities.
Such changes are prompting speculation as to whether scoring models need to undergo significant enhancements every few years, much like software platforms. "It's up to the banks to push for it," says Weber. "If they are going to push deeper into the thin credit file segment of the population, it would make sense to do it."
As Fair Isaac prepares to roll out FICO 08, lenders have been returning to more conservative underwriting practices. While FICO 08 and specialized scoring models will help improve risk management, Fair Isaac is stressing they are not a panacea for predicting consumer behavior. It is a message that lenders would be well advised to remember.
Special Models for Special Situations
Recognizing that base scoring models cannot accurately cover all the bases in predicting a consumer's creditworthiness, Fair Isaac and Experian have launched specialized scoring models.
Fair Isaac is supplementing FICO 08 with a suite of applications including a credit capacity index score and portfolio stress testing analytics. The capacity index can predict outcomes, such as which borrowers with adjustable rate home loans will handle higher payments once their interest rates reset.
Portfolio stress testing uses analytics to complement the FICO score by applying macro-economic forecasts, such as gross domestic product, unemployment rates, and personal income, against a credit portfolio to help lenders simulate aggregate changes on the default probability for the portfolio. The information is expected to help lenders determine when to increase and decrease available credit to customers in the portfolio.
In March, Experian launched the Emerging Credit Score to assist lenders in evaluating the creditworthiness of unbanked and underbanked consumers. Emerging Credit Score, which was created in conjunction with eBureau, a St. Cloud, Minn.-based provider of scoring and analytics, leverages Experian's credit data and eBureau's alternative consumer credit, payment and identity data to generate scores for the estimated 50 million to 80 million consumers with little or no credit history. Traditional scoring models cannot generate a score for this segment.
BankruptcyPredict, co-developed with San Mateo- Calif.-based Visa USA, helps lenders identify consumers in financial distress and predict bankruptcies up to 24 months out.