Credit bureau data, and in particular credit scores, have come to play an increasingly important role in credit, insurance, employment, and rental transactions. Consumers with high credit scores often enjoy lower interest rates on their credit cards, lower automobile insurance premiums, and easier access to rental housing, mobile telephone contracts, and utilities. Despite credit scoring's increasing impact on consumers, there is little empirical research in the public domain on the efficacy and accuracy of credit scores and their underlying credit bureau data.
A paper released by the Consumer Federation of America (CFA) and the National Credit Reporting Association in December 2002 is an early contribution to consumer credit data research. The paper, "Credit Score Accuracy and Implications for Consumers," was written by CFA analyst Brad Scriber. On March 25, 2003, the Payment Cards Center invited Scriber to lead a workshop at which he described his research and its findings.
Three major data repositories maintain consumer credit data in the U.S. These repositories have nationwide coverage — that is, it's likely that three credit files are maintained for every credit user and that, each month, credit grantors report on their customers' borrowing in triplicate. Each of these repositories can also provide bureau users with a credit score that is calculated based on that specific repository's data. The most widely used of these scores is one calculated by a company called Fair, Isaac. This score is commonly referred to as a FICO score. Given that all three repositories have nationwide coverage and all three provide credit scores, lenders, insurers, employers, and lessors typically request a consumer's file or score from just one repository. Scriber's research suggests, however, that any one person's credit data, and hence the credit score derived from that data, may vary depending on which repository is used.
Using a stratified sample of up to 500,000 mortgage applicants, Scriber was able to analyze differences among the three repositories' data for the same consumer. He was particularly interested in those consumers who were on the cusp of being considered "sub-prime" and who had wide credit score variations, e.g., a consumer who had a FICO score of 700 (a relatively high score) when the data from one repository were used to calculate the score and a FICO score of 620 (a relatively low score) when a different repository's data were used. These consumers, he suggested, had the most to gain or lose from credit file data differences, as they could be denied credit or pay higher prices for credit depending on which repositories' data were used to generate their score.
Scriber's major findings included the following: Scores varied significantly among credit bureaus for a given consumer, with one in three consumers having a variance of 50 points or more between his or her highest and lowest score. In addition, one consumer in 10 was missing a credit score from at least one repository, and one consumer's file in 10 contained additional repository reports (perhaps from another person).
If you are interested in learning more about Scriber's work, his paper can be found on the Consumer Federation of America's website at http://www.consumerfed.org/pdfs/121702CFA_ NCRA_Credit_Score_Report_Final.pdf.