Given the highly competitive nature of the U.S. credit reporting industry the three major national credit reporting agencies are constantly racing ahead to improve the quality of their data and solutions. The most recent enhancements released by the “Big Three” involves the introduction of a new tradeline data element, payment amount, plus extended fields for current balance, credit limit, and payment amounts fields creating a time series dimension covering several years of monthly tradeline history.
Access to payment amount will now enable users to easily identify consumers that revolve their credit card debt, which should greatly improve estimating account/customer profitability and future credit performance. The inclusion of monthly historical current balance, credit limit, and payment amounts for each tradeline spanning a time horizon of several years (each credit bureau offers a different time horizon) is, in my opinion, a more significant development as it provides insight into two important dimensions of consumer credit behavior, velocity and volatility. The remaining portion of this article and the next few future articles will focus on the significance of the addition of time series credit information.
To incorporate velocity and volatility dimensions into credit behavior models and automated decision support platform users will be challenged on a variety of fronts. This article focuses on just two of the several important analytic considerations pertaining to the use of this new data, namely the testing of new independent variable credit characteristics and the need to evaluate new data analysis techniques traditionally overlooked because the data used in the past lacked a time series perspective.
Testing of time series information can be an expensive and challenging proposition to the user. Because this data is priced as a value added extension to a traditional credit report potential users will want to measure the incremental benefit or “lift” times series data offers from an ongoing ROI perspective. Because migrating to time series information is expensive the Big Three have performed their own generic analysis demonstrating the potential improvement time series data may offer. Although the analysis performed by the Big Three may gain the attention of potential users, potential users will want to perform their own analysis on their portfolios before beginning to seriously consider migrating their underwriting models and decision support strategies to incorporate time series information.
To perform this analysis a user can either rely upon “off-the-shelf” time series credit characteristics developed by each of the “Big- Three”, which may not fully leverage the information value time series data offers. The benefit of using this quick and less expensive approach is that early adopters may gain a competitive edge over their competition by sacrificing speed over substance. However, many potential users of time series information believe that the value of time series credit characteristic comes from “solution” or “target market” specific characteristics, which are engineered to describe the nature of specific credit treatments or segments. According to Ryan Burton, Analytic manager at Capital Card Services, “the isolation of time series activity of specific revolving product type use, not available within the standard time series credit characteristics offered by the three national credit reporting agencies, is expected to provide significant model improvement, beyond the 10% to 15% improvement we’ve already witnessed from standard time series characteristic offerings.”
Similar to the question of using either generic or custom time series credit characteristics to incorporate velocity and volatility credit dimensions, many analysts will want to test analytic approaches not used when developing consumer credit scoring systems based upon “static” information found within traditional credit reports. The reason for testing time series analysis approaches such as ARIMA, Vector Autoregression, and Least Squares Spectral Analysis, to name a few, is that these approaches leverage the direction and speed of data elements over time, which until now have not been present within U.S. consumer credit reports. As Alfred Furth Chief Data Scientist with Capital Card Services explains, “the flexibility of incorporating analytic methods exploiting patterns within past observations, including seasonality, has not been extensively evaluated to explain consumer credit behavior. We expect that these new approaches will provide more stable and reliable estimates of consumer credit behavior. However the complexity of implementing some of the approaches within an existing credit scoring platform may be cost prohibitive. The trade-off between the incremental improvement in risk assessment and potential increased implementation costs needs to be carefully studied and evaluated before adoption of this new data and technology is fully adopted through the lending community.”
Big Three Offer Archived Credit Information to Accommodate Needs of Most Potential Users. Whether a potential user wishes to rely upon either “off-the-shelf’ or “solution/target market specific” characteristics or desires to investigate the potential benefit of time series analysis approaches the Big Three created archived credit files for testing purposes. Ideally, most large potential users will want to obtain time series test data in a “raw” or unaltered format, as they do today for traditional credit information. Data in this format would normally allow them to process archived information through their credit characteristic aggregation system to generate their own custom credit characteristics. Unfortunately, to date there is no commercially available credit bureau aggregation software system available to convert “raw” time series credit information into custom credit characteristics. Although the Big Three offer a vast array of standard time series credit characteristics, they were designed to summarize time series information such that the information can only be used as inputs into traditional credit scoring systems, which will meet the needs of most potential users. Sophisticated users with a need to analyze time series information with either “solution/target market specific” time series characteristics or willing to test various time series analytic approaches will need to go through an expensive and lengthy process relying upon the CRA(s) to design and program custom credit characteristics.
Headed in the Right Direction? The introduction of time series data will radically change the credit scoring and decision support platforms users currently rely upon. Given the lack of understanding of how this data behaves and performs lenders will be looking to the Big Three to take a leadership role in providing insight and user friendly solutions. The solutions from the Big Three, which are just now being introduced, initially appear to be basic credit scoring models estimating standard credit risk measures. While non-risk credit behaviors, such as balance transfer, revolving balance and retention models are also becoming available, it will be interesting to see how quickly the Big Three will leverage this information to create more sophisticated analytic solutions such as loan loss estimation and forecasting portfolio performance.
The availability of this information offers a unique opportunity to the credit reporting industry to offer a wide variety of new solutions focused on data and new information delivery platforms, which have yet to be developed. It will be interesting to see whether the U.S. credit reporting industry can leverage time series information to undergo a metamorphosis distinguishing itself as the leading provider of credit underwriting and risk management platforms.
(Future articles will explore some potential products and services concepts that time series data may facilitate.)
About the Author: Chet Wiermanski is one of BIIA’s contributing editors writing on the subjects of credit scoring and decision systems. He is a Visiting Scholar at the Federal Reserve Bank of Philadelphia researching new applications of consumer credit report information. Additionally, Chet is Managing Director of Aether Analytics which specializes on leveraging hidden data sequences and time series components within consumer credit information typically ignored by traditional credit bureau based solutions. Previously Chet was the Global Chief Scientist at TransUnion LLC. Holding a variety of positions within TransUnion, during his tenure, between July 1997 and February 2012, he was responsible for identifying, evaluating and developing new technology platforms involving alternative data sources, predictive modeling, econometric forecasting and related consulting services.