by Peter Lucas
With the confluence of plummeting housing values, the credit crunch, rising oil prices and core inflation dragging the economy into uncharted territory, lenders are reevaluating whether existing analytical models used to predict risk and delinquencies are adequate.
In most cases, current risk models likely will not suffice. The data used to build them does not reflect new financial stresses facing consumers. Consequently, lenders will need to use new data that drills deeper into consumer behavior patterns and, in some cases, relies less on models of older vintage – such as linear regression models that follow monthly aging of delinquencies – until they are tested under the latest risk conditions.
Among the new data that can be tapped to build analytical models to fit the times are whether consumers are suddenly taking cash advances on their credit cards, have their mortgage payment increase due to a resetting of the interest rate or are making higher monthly payments.
At the same time, lenders need to be more proactive with the technological tools at their disposal to manage risk, such as automated phone systems that rely on algorithms used to trigger actions based on events. Actions can include making a phone call or sending a letter to a customer who shows a declining credit score, but whose account is current, to remind them of an upcoming payment. By doing this, lenders can more effectively manage account risk before delinquency.
"Every accountholder has a probability of default and lenders need to understand the changes in risk so the appropriate steps can be taken and tools can be used to manage the risk," says Ezra Becker, principal consultant of financial services for Chicago-based TransUnion. "Risk management is not just a front-end focus. It has to be managed throughout the account life cycle."
Credit card issuers are well positioned to predict if customers are succumbing to economic pressures by leveraging more of the data captured by their transaction processor.
What makes processors a good source of risk data is that they have access to the UPC codes that identify the item purchased. When examined over time, the data can indicate whether a cardholder is moving away from discretionary purchases such as home electronics to buying staples such as food and gas. Banks with both a debit and credit card relationship with a consumer can use processors to identify whether the cardholder is shifting purchases of staples off their debit card and onto their credit card.
Ironically, most transaction data captured by processors is rarely used by issuers. "Banks manage an enormous amount of customer data but don't always model transaction data because they rely on credit bureau scores to indicate account risk," says Mike Geppert, president of First Data Solutions in Greenwood Village, Colo. "Credit bureaus don't have the depth of purchase data that a processor does for analysis and modeling."
Recognizing card issuers need information that reaches deeper into cardholder behavior patterns, First Data developed analytical models using transaction data. In addition, metrics from unstructured data such as notes from cardholder conversations with customer service representatives about their current financial status, cardholder notification of a change in address and severing of a joint account also can be fed into the model.
Including unstructured data enables risk managers to build more variables to predict a delinquency. "An address change indicating the cardholder has gone from being a homeowner to a renter or has divorced impacts the cardholder's lifestyle and financial situation," says Geppert.
The sooner that type of data can be gathered and analyzed, the better. "Waiting for an account to go delinquent before analyzing this information can be too late," says Geppert.
While linear regression models worked well historically for predicting risk on a delinquent account, lenders need to remember that each segment of their customer base will behave differently as economic conditions vary by state – and even within portions of a state.
"There is a corresponding link between the rise in card delinquencies and declining housing values, but that doesn't mean an account has to get shut down based on a risk score developed for different economic times," says Brian Riley, a senior analyst for Needham, Mass.-based TowerGroup. "Once an account is shut down, there is no incentive to pay off the debt, especially if it is unsecured."
Instead, card issuers are better served by developing models that identify the amount at which a credit line can be frozen on a high-risk account and still allow for a manageable open to buy. The same also can be done for a delinquent account to support a treatment strategy.
The rationale for either is that consumers are showing signs of moving away from tapping home equity as their primary credit line to finance purchases.
Despite the shift in the use of credit, many consumers still have the capacity to pay on their credit card and will make payments a priority to continue financing their lifestyle.
Consumers in the latter segment also have little motivation to make mortgage payments on a house in which they have little or no equity (see "Reckless Abandonment," April 2008).
"With negative home equity building, more consumers are moving toward using cards to access credit," says Scott Hoyt, director of consumer economics for Moody's Economy.com in Westchester, Pa.
Determining if consumers in this demographic actually can pay on a card balance requires deeper analysis of their behavior. A cardholder who is chronically late, but pays in full each month or more than the minimum, is more apt to be a better risk than one who is never late and suddenly becomes delinquent.
"The customer with the best payment record is not always the least risky when they become delinquent," says First Data's Geppert. "Analytical models need to provide a more holistic view of the account, because occasional late payments are not necessarily a sign of worsening behavior."
Building models with data reflecting current economic conditions is just one piece of the puzzle. As economic conditions worsen, lenders must become more proactive in reaching out to account holders – to prevent delinquencies and to develop treatment strategies.
"Establishing a customer relationship reinforces the value of the account and the brand of the lender or creditor with the account holder," says Brian Moore, executive director of collection solutions for Varolii Corp., a Seattle-based provider of hosted communications platforms. "One of the best ways to do this is through proactive communications with the account holder."
In 2007, New York-based JPMorgan Chase & Co. began contacting credit cardholders who are chronically late or frequently reviewing their credit limit to offer assistance, measures that include setting up automatic payments or changing the due date to short- and long-term payment programs.
Chase officials have said they feel good about the company's business-as-usual risk management approach, even though card charge-offs in the fourth quarter of 2007 were higher than the previous quarter thanks to a continued normalization of loss rates after bankruptcy reform of two years ago. Chase did not make executives available for comment.
Such efforts reflect a growing trend among lenders and creditors, says Moore. "Some mortgage lenders and wireless carriers are proactively communicating with their customers through automated platforms to go over the terms of the contract, monthly payment and fees and verify addresses," he says.
Some insurance companies go so far as to offer a discount to set up automatic payment plans to reduce the chance of a delinquency on a new account. The discount usually reflects the cost savings of having to send a monthly paper bill.
Because automated communication platforms are software, algorithms can be written into the programming code that will automatically contact customers based on events or types of data. For example, a lender that spots one of its customers has a declining credit score or is showing delinquencies on accounts with other lenders, automatically calls or sends a letter to that customer reminding them of the due date for their next payment.
As helpful as automation is in proactively managing delinquencies, lenders still must balance it with human judgment when evaluating risk across an entire portfolio, so as not to lose opportunities to book profitable customers, who still exist despite the shaky economy.
Charlotte, N.C.-based Bank of America's credit card unit still makes judgment decisions during account origination, because a FICO score of 700 is not always counted equally, Ric Struthers, senior vice president and card services executive told analysts in February.
Struthers contends that while California has experienced a huge loss in home values that is impacting consumer bill payments, the trend is not affecting the ability of all consumers in that state. The dividing line is drawn as much by geography as by blocks of accounts with like behavior patterns. "You can't treat [the state] all as one problem," he told analysts.
To that end, BofA plans to eventually share account data across all units to identify red flags before a customer becomes delinquent and extends the behavior to other BofA accounts.
Collections will be privy to all account data to ensure a single treatment, rather than one for each account. The aim is to treat an account as a single relationship rather than a product relationship. "Managing account holders as a customer gives them more reason to pay us," Struthers told analysts.
With predictions that an economic recovery will not start until after the federal government's stimulus program kicks in during the second quarter, improved data, proactive customer communications and a dash of judgmental account management will be the recipe for successfully navigating current risk conditions.