The likes of Stephen James are everywhere these days. Personal bankruptcies are breaking records. Credit-card users skipped out on 4.48 percent of their charges in the second quarter of this year, haft again as much as in 1994, and delinquencies on consumer credits like boat loans and student loans are rising, too. But though lenders may be battered, they’re not defenseless. Sophisticated statistical models have turned collections from a banking backwater into a high-tech business. Don’t ask your bank for details; dealing with debtors is so sensitive many banks won’t even discuss it. But bet on this: if your payment is late, every move your lender makes is guided by computers that estimate when and whether you will pay up. Says Jay Meyerson, head of Key Bank in Cleveland, “The world of collections has changed dramatically.”
Bad debts, of course, are the oldest news in the banking business. But until a couple of years ago, debt collection was still in the Stone Age. Every borrower received the same written notice when a payment was 15 days late, the same phone call if no money had come in two weeks later. “We treated all 30-day accounts the same way,” recalls Don Whittaker of Risk Management Resources, a consulting firm. Now, however, the bank can take special aim at you. Computers slice and dice customer information, blending it with economic data and credit-bureau reports to suggest how to proceed with a particular case, which cases to assign to which collector and even how aggressive the collector should be.
That’s why, from her cubicle in Cleveland, Key Bank collector Jeanetta Price was being firm with Stephen James last week. Price handles overdue car or boat loans. A computer usually selects borrowers who are 15 days late, flashing each one’s record on her screen as it dials the call. But with James, a mere 11-day delay was enough to raise warning flags, Although he hadn’t been late before, Key’s computer saw his mediocre credit rating and lack of insurance as marks of a loan that could go sour. Sure enough, “I can’t pay till October $1,” James said. Price was polite but not sympathetic-and told the computer to dial him again on Oct. 31. If the cash has come in, there’s a November payment to worry about. If it hasn’t, Key can repossess the car before he falls farther behind.
Whether a debtor gets a call, a letter or nothing at all increasingly depends on the numbers. Computers at one California thrift automatically score late payers with software from mortgage insurer MGIC. Using 22 variables, from local unemployment data to the borrowers’ payment records, the software predicts which delinquent mortgages are most likely to end in foreclosure. An owner whose home has risen in value might run late without causing a ripple, while one with the same track record in a weak housing market might hear from a collector the day the check is due. Scoring lets a car lender get extra tough on late payers who borrowed in early 1995, if the computer spots unusual problems among that group of loans. Credit-card issuers use scores to lay off of overdue borrowers who are likely to get back on track without being prompted. “The probability of getting a collection call based on one missed payment is very low, unless you’ve missed payments in the past,” says Atlanta model builder Gary Chandler. And if a collector has left a message on your answering machine, well, that may have been modeled, too. American Express has software to predict whether a particular customer will return a collector’s phone call.
Of course, no technology can force an unwilling customer to pay. In growing numbers, borrowers–thanks mostly to promiscuous use of credit cards–are flummoxing lenders by filing for bankruptcy without ever having missed a loan payment. “There are not many models that are very successful in predicting bankruptcy,” says Ross Waldrop of the Federal Deposit Insurance Corp. Not yet, anyway. But you can bet someone in Silicon Valley is working on the problem.
