Trending
Is sequence-of-returns risk (SORR) a reality we are defenseless against? More bluntly, is SORR an accident waiting to happen with little we can do to avoid it. I will present evidence that we can, in fact, do something to prepare for SORR as we transition into our retirement years.
The general idea of SORR is that when we start pulling money out of a retirement portfolio we are naturally more sensitive to the performance of the portfolio each year. In other words, making our withdrawal at the end of a bad year (or a bad sequence of years) will magnify the pain already inflicted on our portfolio by poor returns.
Consider an example scenario. A person retires at the start of the year 2000 and has a 60% equity/40% bond retirement portfolio. The year 2000 was the start of the “tech wreck”. The calendar year return in 2000 was -0.68% for a 60/40 portfolio. In 2001 things got worse with an annual return of -1.89%. The year 2002 was even worse with a return of -10.20%. If the retiree has to make annual withdrawals (based on the need for income and/or the required minimum distribution) they are faced with the distasteful task of withdrawing money from a portfolio that has been beaten up by bad market returns. This is an example of a bad “sequence-of-returns” for a new retiree (or any retiree). Withdrawals from their retirement portfolio only exacerbate the negative returns experienced by their portfolio. Very simply this is SORR.
An understandable “reaction” to the possibility of being adversely affected by SORR would be to build a more conservative retirement portfolio. Entirely logical—but likely the wrong choice for reasons that are only indirectly related to the “sequence” of returns experienced by a retirement portfolio.
In Figures 1-3 you will see the correlation between the sequence of returns of a retirement portfolio in five 5-year increments with the eventual ending value of the portfolio after 25 years of withdrawals (i.e., after five 5-year increments). In short, we are measuring which 5-year period’s returns had the most impact on the eventual ending value of the portfolio after 25 years of withdrawals. The higher the 5-year correlation, the higher the SORR. This analysis examined every rolling 25-year period from 1926-2023 (all 74 of them) using index-based annual returns for large US stock, small US stock, US bonds, and cash.
Four different retirement portfolio asset allocations were tested: a very conservative model with a 20% equity/80% fixed income allocation, a conservative 40% equity/60% fixed income model, a moderate 60% equity/40% fixed income model, and an aggressive 80% equity/20% fixed income model.
In Figure 1 the 25 annual withdrawals (made at year-end) were determined by the mandated RMD percentage withdrawal starting at age 73 through age 98. As you can see the asset allocation model with the highest correlation in each 5-year period to the ending portfolio value after 25 years was the very conservative 20/80 model (as noted by the dark blue bars). In other words, the 20/80 retirement portfolio presented the highest SORR in each 5-year period. The specific 5-year period that exposed the portfolio to the highest SORR was the 3 five-year period—or during years 11-15. The findings for a retirement portfolio experiencing RMD-based withdrawals were virtually identical to a retirement portfolio sustaining withdrawals of 4% of the portfolio’s ending value each year (as shown in Figure 2).
In Figures 1-2 (RMD-based annual withdrawals and 4% annual withdrawals) the retirement model with the smallest overall exposure to SORR was the 80/20 portfolio. The only slight exception to this was in the 1 five-year period where the 40/60 and 60/40 asset allocations had fractionally lower correlation compared to the 80/20 model.
The results get more interesting when we change how money is being withdrawn from a retirement portfolio. In Figure 3, the first-year withdrawal was set at $40,000 and then in years 2-25 the annual withdrawal was inflated by the actual CPI-based rate of inflation each year in each rolling 25-year period. When money is being withdrawn to keep pace with inflation, the exposure to SORR is highest during the first five year period. In Figures 1 and 2, the performance of a 20/80 portfolio during the first 5 years had a correlation of roughly 50% to the ending value of the portfolio after 25 years. As shown in Figure 3 the correlation of the 20/80 portfolio’s first 5-year returns to the ending value of the portfolio after 25 years increases to over 81%. In fact, the high SORR correlation for a 20/80 and 40/60 portfolio extends into the 2 five-year period as well.
The 60/40 (purple bars) and 80/20 (copper bars) portfolios show very low SORR during the 2 five-year period in Figures 1-3. In the last two 5-year periods (years 16-20 and years 21-25) the correlation between sequence of returns and ending portfolio value after 25 years is negligible when money is being withdrawn based on a CPI COLA (cost of living adjustment). Apparently, the die is cast in the first half of the 25-year period under such a withdrawal method (CPI-based COLA).
It should be noted that each retirement portfolio (20/80 through 80/20) survived intact in each of the 74 rolling 25-year periods being analyzed. The only exception was the 20% equity/80% fixed income model. In four of the rolling 25-year periods the 20/80 portfolio ran out of money before 25 years when money was being withdrawn based on a CPI-based annual COLA. Thus, the 20/80 model had a 94.6% success rate (where “success” means remaining solvent through 25 years of withdrawals).
To this point we have been focused on the sequence-of-returns across four retirement models and the associated impact on the retirement portfolio’s ending value after 25 years of withdrawals. Perhaps contrary to intuition, the conservative 20% equity/80% fixed income portfolio was the most susceptible to sequence of returns risk regardless of how money was being withdrawn (RMD, 4%, CPI-based COLA) despite having the lowest standard deviation of return (see Table 1). Standard deviation is often used as a technical measure of volatility (or risk).
This finding suggests that volatility is not the only driver of SORR. There is another version of “SORR”—namely “size-of-returns risk”. A retirement portfolio is actually exposed to two forms of SORR: sequence-of-returns risk and size-of-returns risk. Both matter and both impact the ability of a retirement portfolio to survive 25-years or longer.
Table 1. Historical Performance of Each Retirement Portfolio Model (1926-2023)
As we examine Figures 1-3, the retirement portfolio asset allocation “sweet spot” appears to be a 60% equity/40% fixed income model (the purple-colored bars in each graph). A 60/40 allocation is the classic “balanced” portfolio. The virtue of a 60/40 retirement portfolio allocation (maintained through annual rebalancing) is clearly illustrated in the 2 5-year period of Figure 2 (withdrawal years 6-10). The correlation of annual returns during that particular 5-year period with the ending balance after 25 years is just under 22%--a very low correlation compared to the very high correlations exhibited by the 20/80 and 40/60 models during that same 5-year period.
Is there a lesson to be learned here? May I suggest that the primary lesson is that hunkering down in a very “low risk” retirement portfolio (meaning 20% or less exposure to equities) exposes a retiree to size-of-returns risk while attempting to avoid sequence-of-returns risk. Size-of-returns risk often masquerades as sequence-of-returns risk. A secondary lesson is that a withdrawal method that attempts to keep pace with inflation is a particularly demanding withdrawal strategy. The size of returns in a 20/80 retirement portfolio (and to a lesser extent a 40/60 portfolio) will likely not be high enough to sustain the demanding withdrawal schedule while simultaneously counteracting the natural performance variability that “sequence-of-returns” is generally blamed for.
There is one more issue to consider. Table 2 highlights the average annualized 5-year return of each retirement model during the first 5-year period of all 74 rolling 25-year withdrawal periods. Also shown is the historical frequency of experiencing a negative 5-year annualized return during each first 5-year period. The likely rationale for a new retiree to build a conservative retirement portfolio (such as 20% equity/80% fixed income model) is that it will be less susceptible to sequence-of-returns risk (that is, less likely to experience a negative 5-year annualized return in the first five years of their retirement). That is indeed true. Since 1926, a 20/80 portfolio experienced a negative 5-year annualized nominal return (not accounting for inflation) a mere 1.35% of the time during the first five years of withdrawals—compared to 9.46% of the time for an 80/20 portfolio. In reality, both percentages are encouragingly low. The problem, however, in attempting to avoid sequence-of-returns risk is the specter of size-of-returns risk. This is evidenced by the average 5-year return of 6.23% for a 20/80 portfolio compared to 10.21% for a 80/20 portfolio.
Table 2. Performance in the First 5-Year Period of Each Rolling 25-year Period
74 Rolling 25-Year Periods from 1926-2023
For those who insist on building a conservative retirement portfolio, a percentage-based withdrawal method (such as the RMD or a 4% withdraw rate) appears to modestly insulate a conservative retirement portfolio (20/80 and 40/60) from both forms of SORR (sequence-of-returns and size-of-returns) during the first 5 years of retirement. We observe this in Figures 1-2 by noting that the dark blue bars (20/80 portfolio) and the bright blue bars (40/60 portfolio) have materially lower correlation with the ending portfolio balance in year 25 in the 1 5-year withdrawal period compared to the 2, 3, and 4 5-year periods.
Interestingly, the performance of a 60/40 and 80/20 retirement portfolio have higher correlation with the ending balance in year 25 during the 1 5-year withdrawal period than in the 2 5-year period when money is being withdrawn based on a percentage of the portfolio’s value (see Figures 1 and 2). However, in both 5-year periods, the correlations are quite low suggesting that a 60/40 and 80/20 portfolio are naturally resistant to both forms of SORR during the first 10 years of annual withdrawals.
For percentage-based withdrawals (Figures 1 and 2) the highest correlations between sequence of return and the ending account value after 25 years occurred in the 3 5-year period (withdrawal years 11-15). This was true for each of the four portfolio models. One of the important attributes of a percentage-based withdrawal method is that it allows the annual withdrawal to decrease if the portfolio suffered a loss that year. This is a form of “portfolio compassion”. Conversely, a CPI-based COLA withdrawal method forces each year’s withdrawal to increase more often than not because inflation is generally positive (during this 98-year period the annual rate of inflation was above zero 80% of the time).
Summary
A new retiree could—in many cases—view themself as a long-term investor and, as such, consider the appropriateness of a portfolio with at least a 60% overall allocation to equities and equity-like asset classes (such as real estate investment trusts). Furthermore, retirees are advised to use a percentage-based withdrawal method where possible. The RMD is, of course, a percentage-based method so that will naturally happen with retirement accounts that are governed by RMD withdrawal rules. The RMD is easy to hate because it is forced upon us, but the percentage-based nature of the RMD is a feature that is genuinely advantageous.
These two modest guidelines, a 60% equity/40% fixed income retirement portfolio (or an overall asset allocation in that ballpark) and a percentage-based withdrawal method, will naturally insulate a retirement portfolio from both forms of SORR (sequence and size) during the early years of portfolio withdrawals.
Conversely, a withdrawal method that forces the withdrawal to increase each year (such as a mandated 3% or 4% annual cost-of-living-adjustment OR a CPI-based COLA (which forces an increase in the next year’s withdrawal a high percentage of the time) exposes a retiree to the highest amount of SORR--regardless of their retirement portfolio asset allocation.
Related: How Long Is the “Long-Run”