Tuesday, November 9, 2010

Another way to think about pensions

In May 2009 I wrote that pension fund management needed a new analytical framework and asked for input from my readers. Now I see that the Boston chapter of QWAFAFEW (Quantitative Work Alliance For Applied Finance, Education and Wisdom, otherwise known as quaff-a-few) is hosting a talk entitled "Using Factors to Dynamically Manage Pension Fund Risk" on Tuesday November 16, 2010. Here is the abstract:
Using Factors to Dynamically Manage Pension Fund Risk - Pension Funds are complex systems with many moving parts. A sponsor needs to be concerned about the behavior of the assets, the behavior of the liabilities, and how the two interact. We look at representative Canadian plan and its sensitivities on both the asset and liability side and how they interact and then suggest some asset allocation strategies that can be used to better manage the return dynamics and funding stability of the plan.

If you are around in Boston next Tuesday, you may consider attending. Click on the link above to get all the details.


Roger Heath said...


Life and annuity insurance companies utilize dynamic models combined with stochastic methods to establish liabilities and determine optimal asset/investment strategies. The models take into consideration the interaction of assets and liabilities. Not just experiments, the business is managed this way.

Roger Heath

Cam Hui, CFA said...


Thank you for enlightening me. Most ALM approaches are highly interest rate sensitive. It is unclear to me, however, whether the factors underlying these models go beyond interest rate and the shape of the yield curve to other economic factors.

Roger Heath said...

To use similar ALM techniques for pensions, the models would have to be modified to reflect the important economic issues of pension investments. Life and annuity companies in the US have it somewhat easier because they need to invest almost exclusively in bonds or similar securities because the mandated risk-based capital requirements of the insurance regulators for other securities is so high (eg. 100% for stocks). Most models reflect the interaction of policy dynamics, insurance company decisions (like crediting interest rates), policyholder actions, etc. But you are right that it is driven off of the stochastic independent variable, interest earnings. Some have gotten sophisticated enough to include a second order set of independent variables like spreads and yield curve shape (not just assuming that they remain constant).

The model would need to reflect whatever affects the pension investments. Some seem to be mundane; others would require pushing the edge of modeling technology (like if CalsTrs had decided to invest in commodities).