New research shows the bucket strategy is better than previously thought.
Fill the Buckets and Glide Away
After spending many years accumulating assets for retirement, you need a strategy for retirement withdrawals. The former is easy since you’ve been doing this for many years, and you just save until you have enough. The latter is harder since you need to take money from your portfolio and hope that you can maintain your purchasing power for potentially a very long retirement. How do you choose between a traditional static portfolio, the infamous bucket strategy, an upwards glidepath, or some other strategy?
Retirement Bucket Simulator
The first step is to figure out how these strategies really work and some of their advantages and disadvantages. To assist in this endeavor, a flexible bucket simulator was created using historical return data found in Simba’s backtesting spreadsheet. The data includes yearly inflation values and asset class returns from 1970–2021. The Retirement Bucket Simulator, shown in Figure 1, shows bucket and asset balances over 30 years.
Bucket 1 includes the annual budget allowance, and the minimum amount is always 1 year of expenses, assuming that there are funds available. If bucket 1 falls below its minimum value, it takes assets from buckets 2–4, preferring lower buckets first. The expenses for the current year are assumed to be in cash with no investment return (e.g., checking account), and the remainder of bucket 1 is invested in Treasury Bills.
Buckets 2–4 are invested in a mix of equities and bonds with a default of US Large Cap Equities (e.g., S&P 500 or Russell 1000 index) and intermediate-term Treasury Bonds. Each bucket is rebalanced annually. To simulate a traditional static asset allocation (e.g., 70% equity, 30% bonds), buckets 1–3 are set to their minimum values with the target asset allocation in bucket 4.
The Glide path check box allows the user to specify an upward glide path for bucket 4, as suggested in this Kitces blog post (and others). You can specify a starting equity percentage, ending equity percentage, and the number of years for the ramp. Each year, the portfolio is bucket 4 is rebalanced according to the glide path until reaching its maximum. Buckets 2–3 should be set to 0 when using a glide path.
All calculations are done using inflation-adjusted dollars. For example, if the investment return is 10% and inflation is 3%, then the asset value at the end of the year is the beginning asset value multiplied by 1.1 (100% + 10%) and then divided by 1.03 (100% + 3%), for a net return is 6.8%. After the final bucket balances are calculated, buckets 1–3 are refilled to their minimum balance by moving funds from bucket 2–4. For more information about the refill and spill rules included in the simulator, refer to the Help tab.
Example Return Sequences
In Figure 1, we see the balances for a bucket strategy portfolio starting in 1973, which was a particularly bad starting year for a retiree, according to Guyton and Klinger. The simulation started with $1M, which is 25 years of expenses ($40k or 4% withdrawal rate) and after 30 years, we have $422k remaining (10.5 years of expenses). Remember that all calculations are being done in inflation-adjusted dollars, so $40k is always 1 year of expenses.
Figure 2 shows why 1973 was a tough year to retire. You can see that inflation was high and equity returns were poor. Figure 3 shows portfolio balances for starting years from 1973 to 1980 for a static 70% equity allocation (rebalanced annually). For 1973, the total portfolio dropped by almost 50% (inflation-adjusted) in just 2 years, and despite good returns after that, it’s tough to recover. Figure 3 also shows how much difference only a few years can make. The 1977 retiree also lost a significant amount of purchasing power (about 40%) but over 5 years instead of just 2 years. After 30 years, their inflation-adjusted balance had more than doubled. By 1980, market returns were pretty good for many years (until the tech crash starting in 2000). Even with the housing crash just a few years later, this was one of the better results for the standard 70% equity static portfolio.
Figure 4 compares the standard 70% equity portfolio (Static70) to a bucket strategy (Bucket) and a rising glidepath (Glide). The bucket strategy has 3 years of expenses in bucket 1 and bucket 2 has 30% equity, bucket 3 has 50% equity and bucket 4 has 70% equities. The default refill and spill rules were used, which means that bucket 1 is completely used up, then bucket two and three are spent down sequentially. The rising glidepath uses the minimum value for bucket 1 ($40k, 1 year of expenses), and bucket 4 starts at 30% and increases to 70% after 15 years. Buckets 2–3 are not used for the rising glidepath strategy.
Comparing Static70 and Bucket curves, you can see that for 1973 (poor return sequence), the bucket strategy ends with a higher final balance, giving a bit more breathing room for our retiree. However, this improvement when return sequences are poor come at the expense of lower balances when the return sequences are favorable (1977 or 1980).
Bucket Strategy vs Rising Glidepath
If you look at the actual equity allocation over time (see Figure 1), you will see that the equity percentage for the default bucket strategy is generally rising from about 55% to 70% while buckets 1–3 are being used up. This is consistent with the Kitces blog post describing how a rising equity glide path can increase the probability of retiree success and is the reason why the default bucket strategy seems to work well with unfavorable return sequences. However, Kitces suggested that a glidepath going from 30% to 70% might further increase the probability that the retiree does not run out of money. For the very unfavorable return sequence of 1973, the 1973-Glide curve has more than half of the initial purchasing power remaining after 30 years. While the upside for unfavorable return sequences is significant, the final balance when return sequences are favorable (1977 or 1980) are significantly lower. This tradeoff is something that the retiree is going to need to consider when deciding on a strategy for their retirement withdrawals.
Comparing all years
Figure 5 shows the liquid asset balance for a 70% equity portfolio for all starting years. Simulations starting in 1993 or later wrap around to 1970 for the remaining years. All the different simulations are shown as different color paths to give you an idea of the simulation variability. The box shows the range from the 25th to the 75th percentile simulations, which means that 50% of the simulations are within the box. The size of the box represents the variability of the asset balance, and you can see that, in general, the variability increases each year as you get further from the starting point of the simulation. The whiskers extend to the minimum and maximum values although they do not include the outliers shown as solid dots. The median simulation value is shown as the solid horizontal line in the middle of the box and 50% of the simulations are above (or below) that value. The 1973 start year is the worst case of all starting years for the static allocation, as we expected.
If you change to the bucket strategy (see Figure 6) with 3 years of expenses in buckets 1–3, the 2000 start year is now the worst (1973 is close), but the minimum values have increased, which is consistent with what we observed in Figure 4. If you look closely at the early years of the simulation, you can see that the size of the box is smaller when buckets are used. This means that the distribution of asset balances is clustered more closely to the median. There is less downside and less upside compared to the median result. After the buckets are used up, the asset allocation in Figures 5 and 6 are the same, so would expect the variability of the balance at 30 years to be smaller because the variability of the balance is smaller at 10 years. The real advantage of the bucket strategy, compared to a static allocation is that it controls the variability of the asset balance during the early years where sequence of returns risk is highest.
In Figure 7, we see what happens if we use a rising equity glide path with a very conservative initial equity allocation of 30% and increase the equity allocation to 70% over 15 years. In the early years, the balance variability is even narrower than the bucket strategy and the ending variability is narrower as a result. The minimum balance at the end of the simulation is higher than the bucket strategy, but the median balance is lower, and the maximum balance is almost the same. Compared to the default bucket strategy, the rising equity glidepath provides even more protection from sequence of returns risk; however, if the return sequence is favorable, the final asset balance is lower. The latter result should be expected because both upside and downside variability are constrained in the early years. In Figure 8, the upwards ramp is shortened to 10 years. The downside is slightly better than the 15-year ramp, which was already better than the bucket strategy. The upside is also improved a little, although the bucket strategy is still better for good return sequences.
Up to this point, we have only considered US Large Cap for equities and US Intermediate-term Treasuries for bonds. The equity allocation is arguably not well-diversified, and you can have long-term periods of poor equity performance (e.g., 2000-2009). Figure 9 shows the 10-year rising equity glidepath using the Merriman 4-fund combo for the equities. This equity allocation is still all-US but includes 25% US Small Cap and 25% US Small Cap Value, which generally gives lower returns when US Large Cap is doing very well (e.g., the 1990s), but better performance when US Large Cap is doing poorly. In the early years, variability is still tightly controlled and median asset balances are almost the same. However, in later years, the minimum and median balances are significantly higher than in Figure 8, and the maximum balances are higher as well. The minimum balance after 30 years is more than 25 years of expenses. In other words, the 4-fund combo is probably just a better portfolio choice than US Large Cap equities.
Kitces suggested that bucket strategies are an asset allocation mirage and there are plenty of other articles saying similar things. Some of the Retirement Manifesto’s most read articles are on his bucket strategy, but he rebalances each year, so his bucket strategy is really a fixed asset allocation in disguise. For some, bucket strategies make sense, and they can be a useful tool to determine an initial overall asset allocation. The whole idea behind the bucket strategy is that if return sequences are bad, using buckets 1-3 gives a chance for bucket 4 to recover. However, the 1973 sequence is bad enough that bucket 4 never gets back to its original level (in inflation-adjusted dollars). In Kitces’ blog article, he suggests that one reason bucket strategies don’t do as well is that there is no mechanism to purchase equities when they are down. However, in the worst case (1973), just spending the lower buckets gives a rising equity glidepath, which is equivalent to buying more equities when they are down. When you look in aggregate, bucket strategies may not perform as well, but they do seem to offer a protective benefit if return sequences are unfavorable.
What other conclusions can we draw from these simulations? Compared to a static asset allocation (with the same equity percentage as bucket 4), the bucket strategy decreases our chances of running out of money later in retirement. However, we can use a rising equity glidepath to get similar and perhaps even better results for unfavorable return sequences. An advantage of the rising equity glidepath is that it is very easy to manage. You choose starting and ending percentages and the number of years and then manage it very much like a static asset allocation.
Managing portfolio variability in the early years can have a significant influence on portfolio longevity and bucket strategies and rising equity glidepaths are ways to manage early portfolio variability. The choice of asset allocation also has a major influence on portfolio longevity, perhaps even more than the choice of withdrawal strategy and this can be the subject of future investigation.
Guest post by: