A reader asked me to explain how I come up with my estimates for future investment returns. Specifically, the reader was referring to my Sept. 1 column, where I estimated a future annual return of 12 percent for a Thrift Savings Plan portfolio composed of 42 percent, C Fund; 25 percent, I Fund; 18 percent, S Fund; 10 percent, F Fund; and 5 percent, G Fund. This is the starting allocation for the L 2040 Fund.

This estimate is proprietary to my firm and was produced by looking at the last 80 years of history for a portfolio composed of similar assets – stocks, bonds and cash. We calculated the rate of return, adjusting for TSP’s expenses, for every continuous 30-year period during that 80-year period. Then we ranked these 30-year rates of return and identified the median value – the rate of return in the middle of the list.

The conclusions were based not only on this prediction, but also on an estimate of its reliability. We know that our prediction is likely to be wrong, and we design our plans to account for, and to minimize the threat associated with, this uncertainty.

While a course in developing expectations about future investment performance is beyond the scope of a newspaper column, it is useful to have a general understanding of the way it should be done, particularly if you’re relying on these estimates as part of your financial plan.

Basically, there are three ways that financial planners develop investment return assumptions for planning purposes: specific prediction, historical audit and randomization. The point is come up with an investment return environment against which to test a set of financial goals.

Consider the case for specific predictions, the simplest – and up to a few years ago, the most common – method. You have $100,000 saved in your TSP account and want to begin withdrawals. You want level withdrawal payments that are as high as possible, before taxes, and which will last 30 years. You could make a specific prediction that your investment portfolio, as allocated, will produce a 9 percent return each year, like clockwork. In this case it is fairly straightforward for an analyst to calculate that you can safely withdraw about $8,930 at the beginning of each year and have just enough money to last over the entire period.

But, you likely won’t enjoy an investment return of exactly 9 percent each year for 30 years. This is important, since one bad year could cause you to run out of money too soon, even if the bad year is made up during some other year in the period. This weakness – the risk of a specific prediction being wrong – is inherent in any specific prediction about future investment results.

To overcome this shortcoming, you might decide to replace your series of 30 identical 9 percent return values with a series of return numbers that actually occurred for a similar investment portfolio over a 30- year period in history. This is called historical audit analysis, and it was the state of the art in planning before desktop computing power evolved sufficiently. Depending on the particular 30-year period you choose, the results will vary. Some return sequences will support the $8,930 annual withdrawal rate with money to spare; others will fail miserably and leave you broke with many years to go. Typically, this analysis would be run a number of times – say, using 50 unique 30-year sequences taken from an 80- year period – and the results tracked. If a high number of the simulations result in failure, you would be wise to reduce your expectations and the size of your withdrawals.

The third approach is to calculate the maximum withdrawal rate using randomly created investment return sequences. We create a pool of possible annual returns and draw randomly from this pool to create a large number of 30-year sequences of investment returns. The pool of possibilities is based on the range of historical returns for the allocation to be used, so that the draws are likely to be similar to history, but they can also deviate widely by historical standards. The advantage of this method is that it can test a much larger number of scenarios over a much wider range of outcomes, leading to more reliable results.

There are as many predictions about future investment performance as there are people willing to make them. In the end, it is up to you to make sure that your plans are relying on the most reliable, rational predictions possible.

Written by Mike Miles
For the Federal Times
Publication November 10, 2008