The bout with bankruptcies and delinquencies has at least one executive of the consumer credit industry questioning its dependence on the science of credit scoring.
Jeffrey L. Dodge, senior vice president and general manager-banking at Equifax Inc., said lenders may have gotten so caught up in the science that they lose sight of "the art of risk management." He said it could only be a plus if financial institutions reemphasize "nonquantitative elements of human intervention and judgment." He expressed optimism that these are "starting to come back again."
"I am not saying scientific models should be less important," Mr. Dodge said at a recent conference sponsored by HNC Software Inc., a vendor of some of the science he was talking about. "The models are of increasing importance, but the nonscientific should be even more."
As a provider of consumer and market information to financial institutions, Mr. Dodge's company is deep into the technology of data bases and analytics. The same is true of Atlanta-based Equifax's credit bureau competitors, Experian and Trans Union; of the credit scoring community led by Fair, Isaac & Co.; and, increasingly, of the corporations they provide services to.
At one time, Mr. Dodge pointed out, "a lot of information wasn't accessible. Risk management was all art-subjective and human judgment."
The ensuing explosion of information, the need for lenders in competitive markets to act and react quickly with new credit cards and other offers, bankers' aversion to risk, and the proliferation of systematic evaluation tools caused the art to get "overwhelmed" by scientific approaches, Mr. Dodge said.
"Forty years ago, the first rules-based systems were a catalyst for explosive growth in consumer credit," he said. But over time, the mathematical models that predicted market results, and the information they needed to function well, became more and more complicated. Existing products may not be well suited for the "new paradigm" of a strong economy with bankruptcies at record levels.
Hence the demand for bankruptcy predictors like one that Equifax and HNC teamed up to produce.
The state of conventional credit scoring, wherein a given numerical value determines an applicant's creditworthiness, also troubles Robert Hammer, a specialist in credit card portfolio sales.
Speaking to the same HNC financial services conference in San Diego, the chief executive officer of R.K. Hammer Investment Bankers, Thousand Oaks, Calif., said, "All I know is, credit scoring is not as reliable as it used to be. I think we need and will see a complete overhaul, but now I am worried because I can't depend on those ratios the way I used to."
Mr. Dodge said that because scoring is based on consumers' past ability to repay, it loses relevance when the concern is about sudden failures to repay.
The problem has preoccupied data engineers like Shailesh Mehta, chief executive officer of Providian Financial Corp., San Francisco. He said back in 1996 that scoring models were not effectively weeding out borrowers who jumped without warning from "current" status into bankruptcy-an increasingly common occurrence.
To illustrate his points about art and science, Mr. Dodge surveyed the four major monoline credit card lenders. Providian succeeded by being less dependent than others on mass marketing and employing innovative risk management techniques, he said.
Capital One Financial Corp., especially keen on building and refining its proprietary models, was "willing to accept higher risk but priced accordingly." MBNA Corp. was "probably the best at the art of risk management," with a significant human element in its credit decisions and an unwillingness to "sacrifice credit quality for growth," Mr. Dodge said.
The last monoline, Advanta Corp., was forced to sell its card business after not adjusting to indications that its gold-card-oriented strategy would lose its luster. Mr. Dodge attributed that to "an over-reliance on modeling."
Mr. Dodge said lenders came to rely on science "to handle risks it wasn't designed to handle." Lenders must "redefine and rebalance risk management activities so that they better contribute to the long-term strategic objectives of the organization." That means getting beyond raw information output and actuarial predictions, he said, and infusing more of the human qualities of insight, knowledge, and interpretation.
Paraphrasing a metaphor, he said, "Models (became) hammers, and to them every credit decision was a nail. We need to get away from that .... Even the most powerful microscope lacks the ability to understand what it sees."
