Pfizer recently decided to move from a DB plan to an enhanced savings plan, which will be completed by 2018. Until then, it needed a sophisticated, reliable model to support planning for pension liquidity and its earnings-per-share impact. The pension world is full of predictive models, but Pfizer wanted something more integrated and customized, so it built its own better mousetrap at a cost of over 500 man-hours.
Pfizer developed the pension liquidity model over several years and implemented it in the second quarter. The model does not produce a single forecast, but instead simulates thousands of equally likely future scenarios with confidence ranges around the scenarios, explains Amit Singh, senior director of capital markets. One big challenge is forecasting interest rates. “Due to their mean-reverting and stochastic nature, interest rates are difficult to simulate,” Singh notes. Pfizer chose the Hull-White method to calibrate to current market expectations for interest rates because it is forward-looking instead of historical, unlike most interest-rate models, he says.
Predictions on asset classes and interest rates had to be integrated and internally consistent. Simulating future values for asset classes and correlating them is common practice, Singh acknowledges, but leaving interest rates out of the correlation mix can result in “nonsensical future scenarios where, for example, interest rates decrease in a simulated year but fixed assets perform badly at the same time, something that would never really happen,” he says. “We decided that rather than parse through all the simulated paths and weed out the unreasonable ones before using the rest, which would make the model tedious and sloppy, we would create a technique that did not generate any unrealistic paths by making interest rates a part of the asset correlations matrix. This was a breakthrough that made our model efficient.”
The project team spent time with members of the controller’s group to be sure they understood the intricacies of GAAP accounting and built them into the model. While Pfizer’s model is independent, treasury tested its output against those of other models to see if there were any wild deviations (there weren’t). The model is built on an Excel spreadsheet loaded with RiskTM, an Excel add-in for Monte Carlo simulations, and uses FINCADTM financial software to build its Hull-White interest-rate model and calibrate it to current market expectations.
While the model’s main purpose is forecasting, it also improves asset allocation, which should bolster plan performance in the long run. It is not an asset optimization model, however.
“Prior to this project, we had to rely on actuaries, bankers and investment managers to give us their separate model outputs,” says assistant tresurer Neal Masia. “This was an unwieldy hodgepodge of data that we had to consolidate in-house to manage more effectively. Now we can take preemptive actions, plan for future liquidity needs or adjust our current asset allocation to enhance the likelihood of plan success.”
See a slideshow of the 2012 Alexander Hamilton Award winners here.