I have finished a new article called, “Capital Market Expectations, Asset Allocation, and Safe Withdrawal Rates”. This paper can be downloaded from RePEc.
Here are the main points:
1. Most withdrawal rate studies use a particular assumption about asset returns and do not incorporate many asset choices. This is frustrating for readers who wish to consider other assets (such as international stocks and bonds, REITs, commodities, etc.) or have different forecasts about stock and bond returns than what was assumed in the study they are reading. This paper attempts to help alleviate the problem by showing more generally how sustainable withdrawal rates relate to assumptions about asset returns and volatilities, acceptable failure probabilities, and retirement durations.
2. Sustainable withdrawal rates and optimal asset allocation strategies really are quite sensitive to the underlying assumptions about portfolio returns and standard deviations, which in turn are based on the returns, standard deviations, and correlations among the assets making up the portfolio. The paper demonstrates this point, and suggests how important it is to make these assumptions very carefully, as it may be too simplistic to assume that the future will behave with the same patterns as the past.
3. Some of the best known withdrawal rate studies suggest stock allocations of 50% or higher for retirees. By looking at the range of asset allocations which perform nearly as well as the “optimal asset allocation,” this paper shows that it is often justifiable for retirees to use lower stock allocations. Sustainable withdrawal rates may be slightly lower, but this is a price probably worth paying to allow for more peaceful sleep at night.
4. Finally, the paper explains a step-by-step framework for how readers can translate their own capital market expectations into a withdrawal rate and asset allocation strategy for their retirements. I should say, the intended audience of the paper is financial planners, and not all the steps (such as the mean-variance optimization) will be easy to do without some background in finance. But, at least, all the steps are here. Though I haven't tried out their software, you might check this site for more background on this issue.
Readers of my blog know that the hot topic here in recent days is incorporating life expectancy into the analysis. I already had this paper planned out before looking into life expectancies. This could easily be incorporated though, and I will do it one of these days for the blog.
Finally, here is an illustration of the main idea. Allowing for a 10% chance of failure and a 30-year retirement horizon, the following figure shows how sustainable withdrawal rates relate to the assumptions about the real arithmetic return of the portfolio and the standard deviation of real portfolio returns. As well, I added in the efficient frontier for a portfolio of stocks, bonds, and bills whose characteristics match the historical averages from 1926-2010. With the historical data, the maximum sustainable withdrawal rate is 4.3%, and this occurs with 45% stocks and 55% bonds. The thick red line shows asset allocations which support a withdrawal rate within 0.1 percentage points of the maximum. In this case, stock allocations between 28% and 69%. This information about asset allocations cannot be seen in the figure, but it comes from the programs which perform these calculations. I show these asset allocation results in Table 3 of the paper for a wide variety of failure rates and retirement durations.
Supplemental figures which show the sustainable withdrawal rates for other probability failures and retirement durations can be found here.