Managing Credit Portfolios Part 2
Introducing Macroeconomic Data
Lenders have changed their lending criteria, regulators have been encouraging portfolio shifts away from sub-prime, and, it appears consumers have shifted their behavior. All will be related to economic changes in some way. It makes sense to see how much of the changing bad rates can be explained by macroeconomic factors, and how much is left unexplained that must result from other factors.
One of the problems with traditional bad rate projections, is that they are based totally on historical information. With macroeconomic data – for example, personal expenditures, residential housing rates and savings rates — we have past data as actual fact, but we also have future projections that can be incorporated in our bad rate projection analysis by leveraging some advanced analytics. This “forward looking” view has the potential to reap great benefit.
To assess this, Equifax looked at historic macroeconomic data and projections, using relevant variables. The models predict the bad rate on auto loans within 18 months of the application date. Figure 3 shows an example of a model built in September 2009, predicting performance over the following 18 months. The model predicts bad rate with good accuracy, even predicting turning points in bad rate (when the trend changes direction)mid-way through the 18 month period. The blue line shows the predicted bad rate from models using macroeconomic data at the application point in time. The white circles show the actual bad rate.
Figure 3: Bad rate projections based on macroeconomic data
Applying this approach to an 11 year period back to 1998, we observe a predicted bad rate that generally follows the actual bad rate. We even see turning points accommodated well. However, while many of the time points have a remarkably good match of actual bad rate to predicted bad rate, there are several points in time where the fit is not so good. Examples are early 2002 and late 2003. These were times when a changing bad rate was accurately forecasted, but the timing was off by six months, indicating either a delayed or a faster response by consumers to changing economic circumstances. This is consistent with consumers sometimes taking awhile to respond to actual economic changes, while at other times their changing behavior contributes to macroeconomic changes.
While macroeconomic variables can have a very positive effect on predicting delinquency, there are a number of more simple predictive techniques that can be applied in the short term to enhance accuracy. We will discuss some of these in part three of our series, and see how they compare to the improved accuracy that results from incorporating macroeconomic data in the delinquency projections.
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