What Return Assumptions Are You Using?

Imagine a husband and wife who have saved diligently for retirement and are just on the eve of finally entering that stage of their lives.  They meet with their financial advisor who provides them guidance as to how much they can safely spend in retirement.  They all believed they were taking a very conservative approach only to find years later that the advisor’s investment return assumptions turned out to be far too aggressive, causing the couple to spend much more than they really should have.  How could they have known?  Is it possible to get a perspective on whether the assumptions used in a retirement plan are appropriate?

As part of our continuing study of the impact assumptions can have on a retirement plan, this post takes a deep dive into asset class returns; specifically, considering historical vs. projected means and standard deviations.  Most importantly, we will illustrate their variances in terms of the amount of after-tax income a retiree can sustain through a particular age before running out of assets.  However, unlike similar research on this topic, we will determine the after-tax income amounts across another dimension, their success rates of sustaining that income, giving us a clearer picture of how one factor truly compares to another.

Setting the Stage

To set the stage for an accurate comparison of return assumptions, we need to hold all other variables constant.  In this case we will consider a couple, both 65 and currently retired.  They are receiving a total of $27,378 (pre-tax) in annual Social Security benefits and no other guaranteed income.  We assume a flat 2.6% inflation with a gradual taper in spending between the ages of 75 and 85 of 25%.  The couple has $1MM in retirement savings in a qualified account with a 60/40 (EQ/FI) asset mix and paying 1.5% total in expenses and advisory fees.  We will assume a duration through age 95.

Projected vs. Historical Averages

Much has been written on whether one should use historical return assumptions or projected returns, but usually those discussions are in the context of using a constant rate of return.  By the way, this is how the rule of thumb for withdrawal rates commonly known as the “4% rule” was derived (average of 7% portfolio growth – 3% inflation = 4% safe withdrawal with no asset depletion).  If only it were that simple!  If our clients had a lump sum and were not contributing to nor withdrawing from their account, then yes, using the CAGR or a constant geometric mean would be fine.  But we’ll save that discussion for another post.  For our analysis, we will generate 5000 sets of random annual returns using a normal distribution.

Typically for retirement planning, advisor’s use a projected mean and standard deviation as it represents the researcher’s bias of expected future markets.  We use a blend of several publicly available estimates for JourneyGuide’s default return assumptions.  In this example, the Barclay’s US Aggregate Bond Index is used to model Fixed Income and the S&P 500 Index for Equities.  We will also consider the historical averages of those same indices going back 10 and 40 years which gives us the means and standard deviations in the following table. 


Using the blended market expectations and the case assumptions above, our clients would have a 95% likelihood of not running out of money through age 95 if they spent an initial after-tax amount of $54,204.  This compares to after-tax spending amounts of $57,982 and $70,019 using the historical return assumptions for 10 and 40 years, respectively.  A significant 29% difference in what a financial planner might advise the client to safely spend!

In fact, if we compare across other income levels and success rates, we see a similar result as illustrated in the following graph.  At the 50th percentile, for example, the income levels for the projected, 10 year historical, and 40 year historical are $64,728, $71,271, and $87,370, respectively.  This time a 35% difference!

If the advisor used the 40 year historical assumptions and the client had been advised to spend $70,000, both would have thought they were taking a conservative 95th percentile approach.  But if the projected assumptions were used, then that same $70,000 recommendation would have been a much more aggressive 25th percentile approach.  It is critically important that the advisor understands the impact of certain assumptions when considering retirement planning (and in fact, any financial planning involving simulated asset returns).  With all else remaining constant, the impact of asset class returns alone can be extremely significant.  And can lead to a retiree starting out with a much more aggressive spending pattern than they had thought.

If an advisor says “we use historical returns”, client beware!  One might think the more years used, the more conservative the assumptions.  This is not necessarily true.  And remember, there are no do-overs in retirement that aren’t very painful.

In my next post, we will consider other return assumption factors such as the impact of using correlated vs. uncorrelated returns as well as historical backtesting instead of random generation.

If you’re interested in learning more about JourneyGuide, email us at ContactUs@JourneyGuidePlanning.com.

DISCLAIMER: All content on the JourneyGuide blog is provided for informational purposes only. Any statements or opinions expressed throughout this website are those of JourneyGuide and are subject to change without notice. The information provided on this website is from sources JourneyGuide believes to be reliable, but the accuracy of such information cannot be guaranteed. Nothing provided on this website is intended to be personal investment advice, nor does JourneyGuide evaluate or consider the specific financial circumstances of those who read this website. Any references to securities or insurance products are for informational purposes only. Past performance is not indicative of future results.