A question that continually intrigues me is what we can do to get more out of our withdrawals in retirement. How can we “push the envelope” to get more in withdrawals and make them sustainable?
In September 2016, I wrote an article for Financial Advisor (“Small-Cap Withdrawal Magic”) exploring the dramatic increases retirees could theoretically make in their withdrawal rates if they used extreme asset allocations, particularly those concentrated in U.S. small-cap stocks. Recently I revisited this topic, with the aid of more asset classes (seven instead of the original three), and saw even more favorable results. In this article, I will report the results of this research, as well as adding something lacking in the original article: a concrete, immediately applicable recommendation for accessing at least a portion of these higher withdrawal rates, without any additional risk.
My current research calls for a “worst-case” 4.7% inflation-adjusted maximum safe initial withdrawal rate (what I call “SAFEMAX”) from a portfolio invested in seven asset classes: an 11% allocation to U.S. large-cap stocks, an 11% allocation to U.S. small-cap stocks, 11% to U.S. micro-cap stocks, 11% to U.S. mid-cap stocks, 11% to international stocks, 40% to U.S. intermediate-term government bonds, and 5% to U.S. Treasury bills. I assume the portfolio will be tax-advantaged, that there’s a 30-year time horizon with a zero portfolio balance at the end, that the portfolio will be annually rebalanced, and that the retiree will make annual midyear withdrawals. The worst-case scenario applies to the unlucky individual who retired October 1, 1968.
The equal-weighted equity allocation seems to work well in my research on sustainable withdrawals, and it’s recommended in books such as Living Off Your Money by Michael McClung. However, as I did in my 2016 article, let’s ask, “What if we depart from conventional wisdom? Could other, more extreme asset allocations improve withdrawal rates?”
We’re all familiar with the concept that asset class performance occurs in waves, and that the best-performing asset classes vary considerably from year to year. Figure 1 presents the annualized geometric returns of six of my asset classes (excluding Treasury bills) for the first 10 years of retirement of 349 individuals, retiring on the first day of each quarter from January 1, 1926, through January 1, 2013. I chose the first 10 years for this chart, as that period is the most important for determining the eventual “safe” withdrawal rate.
The chart demonstrates clearly how asset returns fluctuate over time, both in absolute terms, and relative to one another. Historically, each asset class has spent some time at the “top of the heap” (as the best-returning class) and some time at the “bottom of the heap” (as the worst-returning).
To understand the chart better, let’s follow one of the asset classes—international stocks (on the green line)—across history, beginning at the far left. In the 1920s and early 1930s, international stocks were consistently one of the best-performing classes. Then, beginning in the middle 1930s and extending through the 1940s, the class’s performance began to lag against all other assets, even bonds. It consistently generated negative 10-year returns. This was presumably connected with World War II.
Over the next several decades, the international class saw varying levels of success where it was either lagging or it returned to a leadership position. But since the late 1980s, it has remained at or near the bottom of all five stock classes through to the right end of the chart, a 30-year period of underperformance.
Just about every stock asset class is familiar with this roller-coaster ride. Bonds behave somewhat differently (as you can see from the intermediate-term bond line in the chart); they are less volatile and more likely to produce consistent (if lower) returns. However, they can be affected by long-term (secular) bull and bear markets stretching over decades.
The outcome of this inquiry is depicted in Figures 2 and 3, which split the 88-year period into two parts for clarity. These are the only two charts I have ever constructed that might be considered “wall art.” (It’s very colorful, indeed!)
Each column in Figures 2 and 3 represents the asset allocation of a single retiree producing his or her highest safe withdrawal rate. Each asset class is represented by a bar of its own color, and the size of each bar matches the percentage that asset represents of the total portfolio.
Here are some trenchant observations we can take from these charts:
1. Over time, different asset classes dominate the portfolios. In many cases only one asset class (other than the invariant 5% Treasury allocation) makes up the entire portfolio.
2. There is a close correlation between these two charts and our line graph. The asset classes with dominant 10-year returns in Figure 1 usually dominate the asset allocation of the corresponding retiree in Figures 2 and 3.
3. Rarely do more than two asset classes (aside from U.S. Treasury bills) appear in the same allocation.
4. Bonds contribute virtually nothing to the allocation after the early 1930s. Perhaps this is not surprising, given their low returns compared with stocks. However, this is a dramatic contrast with our standard allocation of 40% to bonds.
5. The two most dominant asset classes overall are U.S. small company and U.S. micro-cap stocks. Again, this is perhaps not surprising, as they are the asset classes with the two highest long-term returns.
I hope you agree with me that Figures 2 and 3 are quite extraordinary. They indicate that the optimal allocation is almost never the 55%/45% standard I have assumed in previous discussions, and that rarely are more than two stock asset classes required to permit the optimum result. Table 1 lists the average unconstrained allocation across all retirees.
Overall, the average allocation is 91.2% stocks and 8.8% fixed income. That’s a long way from our standard 55%/45% portfolio!
Before we consider the ramifications of these observations, consider Figure 4, which depicts the individual SAFEMAX values generated by the unconstrained allocation. Let’s compare it to the safe withdrawal rate for our “standard” allocation. (As you can see, the unconstrained SAFEMAX is higher for every single retiree.) The difference is quite considerable: The average SAFEMAX is 11.72% for the portfolio with unconstrained allocations, while it’s only an average 7.25% for the standard allocation. This represents an increase of 59% in SAFEMAX!
Two other features of Figure 4 are worth noting. The peak SAFEMAX for the unconstrained allocation is an astounding 49.5%, more than triple that of the standard allocation. It was achieved with an allocation of 95% to micro-cap stocks. Imagine withdrawing at a 50% rate during retirement! This lucky individual was the person who retired on July 1, 1932—almost exactly at the bottom of the market after the 90% stock decline of 1929-1932. Great timing!
Furthermore, the worst-case “universal” safe withdrawal rate, achieved by the person who retired on October 1, 1968, was 6.0% when the allocation was unconstrained, while the standard allocation allowed the 4.7% rate we’d previously determined.
Does this mean we can now talk about “the 6% rule”? Unfortunately, it’s not that easy. Being a retiree myself, I wish it were.
Plus, the stakes are much greater here. If a retiree mistimes the market, they may get another chance to get it right in five to seven years. But if a retiree fails in their choice of initial asset allocation, they may have killed their retirement. That’s because the asset allocation decision is for 30 years, not five to seven. Our “unconstrained allocation” is thus largely an illusion.
Still, the allure of such stellar withdrawal rates is nigh irresistible. Isn’t there a way we can adjust our approach so as to enjoy some fraction of these extraordinary outcomes? Could we, for example, increase our safe withdrawal rate by adopting an allocation that slightly favors the best-performing asset classes, namely U.S. micro- and small-cap stocks, but still appears reasonably balanced? For example, what if we were to change the allocation to what’s shown in Table 2.
The “test allocation” reduces the percentage of low-returning bonds in exchange for an increased fraction of higher-returning stock classes. Overall, the portfolio is converted from a 55%/45% stock/fixed income allocation to 63%/37%. It’s a relatively modest change overall, but in retirement investing, small differences in inputs can lead to large changes in outcomes. The primary uncertainty here is whether the increase in return is offset by a loss of favorable asset class correlation, as both are important to improving portfolio performance.
To evaluate the effectiveness of this change, I recomputed the individual SAFEMAX for each of the 269 people retiring from January 1, 1926, through January 1, 1993, using the test allocation. The results are depicted in Figure 5, and we can compare them with our earlier computations for the “standard” allocation.
Although not as spectacular as the results in Figure 4, in the vast majority of cases (95%), the test portfolio produces a higher SAFEMAX. In many cases, it’s considerably higher. The average safe rate for the test allocation is 7.9%, whereas it’s about 7.4% for the “standard” allocation, an increase of about 7%. Would you like to have 7% more to spend every year in retirement? (That’s a rhetorical question!)
There are limited periods where the test allocation is inferior or offers only a very modest improvement in SAFEMAX. Its results weren’t as good in the late 1920s (just preceding the massive stock bear market of the 1930s) or during the “worst case” scenario of the late 1960s (when there were back-to-back bear markets and high inflation). However, it never falls behind by more than 3%, and the average dip is less than 1.5%. During the early 1980s, the two allocations give virtually the same results.
The last concern I had was the effect of the new allocation on the “universal” SAFEMAX, or the “worst-case” scenario. By adopting the test allocation, we shifted the date of the worst-case scenario from October 1, 1968, retiree to someone with an adjacent retirement date, January 1, 1969. The value of the universal SAFEMAX declined in the process from 4.676% to 4.600%, a reduction of only about 1.5%.
What It Gets Us
To sum up, the test allocation affords a higher SAFEMAX in the vast majority of cases, while suffering only a mild degradation in the rest, including the universal safe withdrawal rate. I would say that this approach looks extremely promising, as it offers the prospect of significantly higher withdrawal rates to a large percentage of retirees.
However, I am not yet ready to give this approach my wholehearted endorsement, because I have a few other concerns. For one, increasing the concentration of small and micro-cap stocks in a portfolio will notably increase its volatility, something that might be upsetting to the investor. Furthermore, if many people were to adopt the test allocation, it would significantly increase the demand for smaller company stocks, which are never in great supply to begin with. That could affect their prices and eliminate their historical edge in investment returns.
Despite my reservations, I rate this experiment a considerable success. It affords a third “free lunch” (improved results with no additional risk) to retirement investors, when it’s considered alongside the general principle of diversification and the concept of “glide path” allocation. It’s made me aware of some intriguing possibilities, which might be further enhanced by additional research.
William P. Bengen is a retired financial advisor, who first conceived the 4.7% SAFEMAX withdrawal rate. He lives north of Tucson, Ariz.