Borrower behavior changes over time and investors and lenders should constantly adjust the way they utilize the different credit score values to accurately predict loan performance, according to officials from FICO and VantageScore.
"The number one message FICO would like to convey is that FICO scores will always rank order risk across consumers, markets and products," said Joanne Gaskin, a director at FICO. "What will change is the ratio of the number of performing and nonperforming loans. What the lender has to understand is that a 720 FICO may not perform exactly the same over each and every market cycle."
For instance, Gaskin said that 2009 and 2010 vintages will look very much like those in 2005 because of the lenders' stricter underwriting criteria.
In her presentation at the recent American Securitization Forum conference, Gaskin said that lenders and investors need to monitor and adjust their score cutoff strategies to maintain the same predicted loss rate over economic cycles. For instance, there will be differences between borrower performance when the economy is in a downturn and when it is in an expansion mode. During a downturn in the economy, the delinquency rates rise and lenders then have to increase their "accepted" FICO score. Meanwhile, in the case of an economic expansion, the repayment rates should increase, which should lead lenders to reduce their "accepted" FICO score.
"We always recommend that lenders track performance internally themselves to make adjustments as needed in keeping with their risk policies," Gaskin said. "At the lender level, FICO provides tools to assess consumer risk under various economic scenarios that then allow them to make proactive adjustments to their scoring strategies leveraging prior performance history."
She mentioned that FICO has a new tool called the FICO Economic Impact Index to help investors and lenders manage risk in a changing economy.
"Through this tool, we interpret how a borrower will perform in the future to help investors and lenders adjust their scoring strategies based upon economic scenarios," she said. "For example, in an optimistic economic scenario, a 720 FICO borrower might perform like a 730 FICO borrower. However, what we do know is that, on average, 720 FICO borrowers have been exhibiting poorer performance compared with the period prior to 2006. The index is a tool to understand the changing borrower dynamics over time."
Gaskin also mentioned during her ASF presentation that there have been variations in borrower behavior in terms of geography as these parlay into the economic cycle in each region. California and New York are the states with the higher-score borrowers, while Florida and Nevada have the lower score distributions.
"Incorporating the economic cycle experienced by the borrower - for instance the economic level in the East or West Coast - helps monitor the borrower's changing payment ability," she said. To illustrate, consumers are most likely affected differently by the economic conditions in New York compared with Ohio, where there are more job losses.
In her ASF presentation, Gaskin added that there has been more pristine credit in the more recent vintages, which has followed the trend of decreasing risk in the higher-score bands, while there has been increasing risk in the lower-score bands. "The industry is moving toward a higher FICO band," she said. This is also boosted by certain trends in the mortgage industry, she added, such as borrowers in the lowest score distribution not having the ability to have second liens.
"Really what we are seeing in the data is that historically, over time, the distribution of FICO scores is relatively stable," Gaskin said. "We will continue to monitor for historical movement, but what we've seen in an economic downturn is that higher-scoring borrowers have tended to exhibit sound financial strategies or behavior, such as credit card utilization tending to go down, while borrowers who are having trouble migrate further downward."
Credit Scores Not Static
In his comment letter to the Federal Deposit Insurance Corp. (FDIC) regarding the agency's proposed rules amending the assessment base for insured depository institutions, Barrett Burns, president and CEO of VantageScore, urged the agency to reconsider its proposed definition of "subprime consumer loan," a key factor in calculating risk.
Burns explained that the nature of credit score values is not static and doesn't always represent the same probability of default over time. He said that a credit score's meaning relies on a number of factors including market volatility or even the version of the algorithm used.
In a separate interview, Burns said, "The probability of default associated with a score value will change over time." He added that the concern was regarding how subprime borrowers were factored into a banking institution's assessment, which the FDIC would have done based partly on credit scores. "As the economy improves, the risk in score band will come down," Burns said. "Regulators have to be careful not to lock in the risk, as the risk in that number can change rather dramatically."
The chart on this page reflects three different time periods, which illustrate the default rates (90+ days past due rate) for several two-year time periods. The timeframe presented by this chart captures a broad range of consumer behaviors before and during the wide swings of recent economic events, such as the more recent credit conditions.
The higher gold line shows increasing risk present for the same score band in the June 2008-June 2010 as well as the June 2007-June 2009 periods over the June 2003-June 2005 period.
Recognizing this phenomenon, VantageScore's latest version, VantageScore 2.0, captures both a broad and recent set of consumer behaviors across the full spectrum of economic events, reducing algorithm sensitivity to highly volatile behavior that can be found in a single timeframe and extending performance.
Burns said VantageScore dedicates extensive research and analytic efforts to ensure that the algorithm reflects current timeframes and maintains predictive performance.