The new science of borrower behavior

Information technology is taking portfolio modeling to the next stage in its evolution. Behavioral models show promise in helping to predict borrower behavior and control the risk of prepayment. Credit-scoring systems attempt to make lending a science by using sophisticated mathematical analysis to help define the level of risk on any given application. While credit scores recognize the value of external data and incorporate some of it into the default model, the results are still insufficient. Credit scoring does not take into consideration all of the aspects of good lending contained in the classic 5 C's of lending defined by P. Beares: 1. character, 2. capacity, 3. capital, 4. collateral, and 5. conditions. Credit-scoring systems are limited by their purpose as to the type of information they are allowed to gather. The cost to update credit scores on a $25 billion servicing portfolio can approximate $30,000 to $50,000. New tools are needed to help portfolio managers more accurately measure risk under changing economic conditions. Behavior modeling will do this.