Quantcast

Risk Scoring Tool Adapted To Changing Customer Behavior

Changing economic conditions, their effect on consumer behavior and related shifts in credit risk are driving efforts to provide credit scoring models with a more predictive risk assessment edge.
VantageScore Solutions, Stamford, Conn., is one such provider that redeveloped its credit-scoring algorithm.

Sarah Davies, SVP of product management, analytics and research, told this publication VantageScore 2.0 is based on over 45 million data files and expands credit rating analysis beyond the typical two-year period dataset to a three-year window. Changes are expected to increase the score’s predictive power from 10% to about 15%.

Davies argues that to be able to commensurate with “a significant change in consumer credit repayment behavior” all credit models should be updated regularly. VantageScore 2.0 replaces an earlier version first introduced in March 2006 in response to market demand for an inclusive credit score model designed to be consistent with consumer scores across the three major credit reporting companies: Equifax, Experian and TransUnion.

VantageScore 2.0 retrieves data from the 2006 to 2008 and 2007 to 2009 performance timeframes, each representing 50% of the data sample.

According to VantageScore president and CEO, Barrett Burns, the expansion of the data pool aims to increase predictability in times when the housing market faces foreclosures and changes in consumers’ payment priorities. The VantageScore rating system can be efficient only if the algorithm is continuously updated in step with these market developments, he says.

At present one growing concern for the mortgage industry is the scoring of seriously delinquent borrowers who are 90 days or more past due on their mortgage.

Recently Lender Processing Services Inc. reported that the nation’s average mortgage loan delinquency exceeds 500 days in five of the 23 states where foreclosures must be approved by a court.
LPS reports that these “extremely delinquent” loans reported in New York, Florida, New Jersey, Hawaii and Maine are part of the nation’s 4.3 million pool of loans 90 days or more delinquent or in foreclosure.

VantageScore 2.0 offers predictive analytics about customers who most likely will default on their loans or become 90 days or more delinquent. One of the advantages is that it features “the exact same algorithm used by the three credit score bureaus,” Davies says, consequently risk assessment is more consistent. “The only reason why a customer would get a different credit score is if the underlying data is different at the credit bureaus.”

Data can be different for various reasons, including filing errors, credit bureau differences in data interpretation, or discrepancies in the time period lenders report data to credit bureaus. In addition, some other credit scores are based on very different algorithms that change the final interpretation of the information along with the results. (As a rule mortgage servicers rely heavily on statistics from at least two credit scores from different credit bureaus to ensure higher accuracy.)

Right now lender-servicers are looking for accurate, consistent interpretation of the customer’s risk, a credit score that really captures what is going on in the economic environment, and eliminates the need to use credit score averages to assess risk, she says.

Using a standard algorithm helps produce more consistent results as much as the size of datasets. Large data pools, such as VantageScore’s 45 million customer files help capture “all the variations in customer behaviors.” Davies stresses, however, that there is one risk: data accuracy is very time sensitive so even a one-year-old dataset can be ineffective.

Research finds have shown more than one example of how customer behavior has changed. As difficult economic times weigh in, some customers are switching their priorities away from their mortgage. A review of delinquent consumer data from 2006 to 2009 shows this behavior shift has affected 1% of the overall U.S. population. Nonetheless it represents a 90% increase in the number of customers who are late on their mortgage but current on credit card and auto loans.

She argues that this finding warns the financial industry—especially servicers—that they need to update their credit risk management models. Servicers need to inquire whether they are able to valuate this shift, whether “this is a systemic and permanent change,” or something that will eventually disappear as customers revert back to what has traditionally been their normal behavior of keeping their mortgage current first and foremost. It is hard to foresee, Davies says, because the reasons why people are doing that vary. But it definitely is a trend worth monitoring for real estate risk management purposes.

Another “pretty obvious” risk management related finding in her view is that currently customers who are looking for new credit tend to be very high credit risk borrowers “in severe need” of credit. “At this point inquiries and newer trade lines are often signals that the customer is potentially exposed financially beyond what they are able to handle.” So it is more of “a red flag” than it would have been during more normal economic times.

The third emerging issue in customer behavior that has not been researched enough as of yet, is borrowers’ reaction to modified mortgage loans. Since the mortgage industry set up certain codes that would allow the identification and monitoring of government modified loans as late as November 2009. That data has been in the credit file for only six to eight months, she says, “so we’re watching it very, very closely.”

Time will tell whether these modifications were effective enough to allow customers return to financial normalcy or just delay the inevitable. At least one year’s worth of behavior is needed to come up with statistically and socially relevant conclusions. “We’re looking at it and will be ready to put the analytics around it when we have the data.”

The hope is that mortgage servicing risk will improve the same way mortgage loan origination risk has improved. Overall, origination risk dropped by 30% compared to last year even thought it is still very high compared to five years ago. The effect of good risk management and disciplined planning played out well with originations, she says. Now the key is to ensure the same happens with all the matured loans on the books and is applied onto the entire credit spectrum.

VantageScore 2.0 is available for testing on consumer data from Experian and TransUnion. It will be officially accessible for use by January 2011.