Saturday, May 16, 2009

A new paradigm needed for pension management

In light of extremely weak returns for capital, there has been a spate of articles about how pension fund liabilities may be the next shoe to drop. Looking at the some of the sample headlines, the outlook appears dire:

Flying in a snowstorm without a map
Despite the trillions of pension fund assets around the world, what has puzzled me over the years is the utter lack of sophistication in the models used to model DB (defined benefit) pension liabilities. Until plans begin to better understand how their liabilities behave, we are likely to lurch from one crisis to the next.

Unless we understand how liabilities are likely to behave, investing the assets is like flying around in a snowstorm without a map.

A sample defined benefits plan
Typically, the pension plan has a board of trustees. The Board periodically undertakes an actuarial study of the liabilities of the plan. They may, at about the same time, conduct a study of the structure of the assets. At the end of the day, the Board compares the value of the assets to the value of the liabilities and pronounces the plan to be in surplus or deficit (and the value of the surplus or deficit).

A typical DB plan pays benefits based on some formula of final year(s) service, e.g. you get x% of the average of your last five year’s compensation. The amount may be subject to some inflation indexation, e.g. benefits will increase annually up to 60% of inflation, etc.

Total pension liability is the sum of the net present value (NPV) of currently retirees’ benefits, which is based on some mortality assumption of the retiree population, interest rates and inflation rate, and the NPV of the current workforce. The value of currently retirees is relatively simple to model, but I believe that current actuarial models do not accurately model the value or volatility of current workforce’s pension benefits.

To model the NPV of the current workforce’s pension benefits, the actuary looks at the age demographic of the working population and benefit formula. He then makes some assumptions about the workers’ annual salary increases and attrition rates (how many people die, get fired, leave their jobs, etc.) He then takes some interest rate and discounts those pension benefits back to a net present value to arrive at an answer.

Before I get flames and hate emails, I recognize that this is a gross simplification. Nevertheless there are a number of problems with this approach.

The model has few inputs: In classical economics, the three standard factors of production are capital, labor and rent. Workers’ wages are mostly a function of the relative demand for and productivity of labor and capital. Where is the productivity factor in this model?

The model is highly interest rate sensitive: Once you’ve assumed that wage increases are based on some standard formula, the only serious input into the equation is the discount rate and the inflation rate. Interest rates and inflation rates are highly correlated as inflationary expectations is a driver of long rates.

Are pension liabilities just interest rate sensitive?
Because of the assumptions built into these liability models, many pension plans that adopt an asset liability management (ALM) framework wind up with a very high fixed income allocation.

Are pension liabilities just interest rate sensitive?

If so, there would be little need for equities or other asset classes in a pension plan. I believe that pension liabilities are modeled incorrectly. This creates problems not only with the valuation of pensions, but with understanding their volatility and sensitivity to changes in economic conditions. Moreover, it lessens the effects that pension plan sponsors have played using differing return and discount rate assumptions to manage the stated value of pension fund shortfalls.

How not to fix the pension time bomb
If I am right, here are some ways of not fixing the pension plan time bomb:

Asset liability management: The principles of ALM have great merit – which is to model and manage the asset-liability surplus or shortfall. But how can you create an ALM framework when your liability model is broken?

Change from a defined benefits plan to a defined contributions plan: While changing from a DB to a DC plan gets future pension liabilities off the balance sheet of the organization, it doesn’t solve the macro problem if we don’t understand how pension liabilities behave. How do you advise individuals on how to manage their retirement assets if you don’t have the proper framework for modeling their pension liabilities? Is a 60/40 asset mix appropriate? Are life cycle funds even the right paradigm? If not, then will these potential pension shortfalls come back to bite us as a society when retirees wind up on social assistance?

These are all thought provoking issues. I would be interested to hear from any pension fund and actuarial professional who believe that I’ve gone off the deep on this. Please email me your comments at cam at hbhinvestments dot com.

Addendum: While we are on the topic of pensions, David Merkel at Aleph Blog has some interesting, though politically unpalatable, solutions for Social Security and Medicare. These are issues that need to be raised.

No comments: