Managing Credit Portfolios Part 3
Ignoring macroeconomic changes will result in greater inaccuracy of bad rate projections over time, and therefore unexpected consequences for portfolio delinquency and sub-optimal pricing strategies. While there is clear value in macroeconomic data for predicting portfolio performance, each organization needs to implement the approach that best meets their needs.
Other approaches also can be effective at improving accuracy as well. Thus, each of us needs to compare what we are doing today to the full range of advanced analytical approaches and tools available to determine the best way forward.
The first step is to evaluate whether we are doing everything we can to maximize the accuracy of traditionally generated forecasts. Some examples to consider include:
1. Frequency of Forecast Update. Most people recommend annual portfolio review and odds chart update of a scorecard. However, what would the benefit be of updating more frequently? Using a percentage error based calculation (WMAPE: weighted mean absolute percentage error), we see a 16% reduction in error when moving from an annual to a quarterly update. Conversely, if you don’t even do annual updates, the consequences can be dire. For example, when using a four year old forecast, the error increased by 61% over the annual error rate.
2. Using Other Scores. If we are trying to create the most accurate assessment of risk associated with an applicant or account, are we looking at all the information available to us? A risk score alone does not tell the whole story. Other possibilities include affordability scores, bankruptcy scores and profit measures. Just as a balanced investment portfolio mitigates risk associated with each investment, balanced decision criteria can mitigate risk associated with each behavior forecast.
3. Vintage analysis on younger accounts. One of the problems with a risk score is the time horizon used to forecast behavior. If a score forecasts bad rate after 24 months, then you have to wait 24 months to see how accurate the forecast was. Meanwhile, decisions are being made without any understanding of how the bad rate may be shifting. But there are other options. By tracking and comparing delinquency levels month-by-month for each month’s (or quarter’s) new accounts, an early read on delinquency changes can be seen. This vintage analysis will allow for a cut-off strategy change in the right direction, but the size of the adjustment will not be precise.
We saw a 16% reduction in WMAPE by moving from annual to quarterly traditional updates. By incorporating macroeconomic data, as discussed in Part 2 of this series, Equifax achieved a further 27% reduction in WMAPE. This indicates opportunity for significant error reduction with the addition of macroeconomic data.
Major errors in forecasted portfolio bad rates can be avoided. The question every risk manager needs to ask is, “Am I doing all that I need to be doing to accurately determine risk management policy, considering the reality of economic changes?” If you are expecting a changing lending environment, whether it be associated with changing consumer behavior, changing lending practices, or macroeconomic changes, then are you using all the appropriate data and tools available to manage the changing environment effectively? The tools are there. Embrace them.
If you are interested in a monthly e-mails about blog updates, sign up for our newsletter
This post was contributed by:
Vice President Modeling and Analysis