TheMReport — News and strategies for the evolving mortgage marketplace.
Issue link: http://digital.themreport.com/i/59481
FEATURE ANALYTICS Game Before the Numbers Gain The Numbers A strong data platform can turn statistics into success for mortgage professionals. By Brian A. Lee S eizing their Moneyball moment, research directors at mortgage lenders across the country more and more are using predictive modeling to improve both their companies' tactical and strategic decisions related to consumers. While they are not likely to come out of it looking like Brad Pitt, star of the Oscar-nominated film, the statistical results can be quite attractive to a financial institution's bottom line. As a mortgage tool that incor- porates loan history data, detailed borrower information, local economic conditions, and other factors, predictive modeling can be used to forecast the likelihood of key events such as prepayment, delinquency, and default. "Incorporating up-to-date information about property values, borrower health, and borrower behavior into models provides insight across geographic and product-driven segmentation," said Mike Smith, chief technology officer/chief architect at Interthinx, an Agoura Hills, California-based firm that helps financial institutions quantify, price, forecast, and mitigate risk. "Lenders who understand the dynamics driving portfolio performance can make more informed choices for loan origination, loan modification, and account management." Since the housing bubble burst, mortgage companies are leaving no stone unturned to obtain that extra informational advantage in the market and, of course, manage risk. Not surprisingly, they are relying on predictive modeling and other analytics more than ever before. After the downturn, the way predictive analytics are used also changed, according to Dr. Sule Balkan, clinical associate professor, Department of Information Systems, at Arizona State University's W.P. Carey School of Business. "Now more focus is given to minimizing loss, preventing or predicting foreclosures, or even rank ordering them," she said. Current predictive models in the single-family mortgage market analyze many factors in consumer behavior, such as the likelihood of change in the borrower's mortgage status, which calculates the probability of default for different levels of delinquency; the homeowner's likelihood to be foreclosed upon; cash flows at the loan level; and the probability of homeowners qualifying for refinance or other loans. "More sophisticated double- hurdle models estimate two-stage probability such as projecting the dollar amount of loss given that the mortgage will default," Balkan said. The previous loan-forecasting standard was a "hazard model that captured the nonlinear mat- uration dynamics relative to the date the loan originated," Smith said. Today's cutting-edge ap- proaches, like that of Interthinx, extend the hazard model by integrating fresh data during the entire loan maturation process, including borrower information, home values, and changing eco- nomic conditions. "The ability to forecast under multiple scenarios gives lenders and investors a way to optimize decisions against a range of future conditions," Smith added. "This is essential to meeting the growing regulato- ry requirements in stress testing and capital planning." Both lenders and consumers should understand the value THE M REPORT | 63 ORIGINATION SERVICING ANALYTICS SECONDARY MARKET