How important is asset allocation versus withdrawal rates in retirement?
I have this intuition that withdrawal rates are far more important than asset allocation. Whether you withdraw 3% versus 4% is far more important than whether you have an 80/20 portfolio versus, say, a rising equity glidepath.
But how to test that intuition? How to quantify it and puts some bounds on it? Is having a “good” asset allocation (whatever that means) more important that going from 4% withdrawals to 3.9% withdrawals?
When I say “going from 4% withdrawals to 3.9% withdrawals” there are two ways to think about. One, sure, we can find a way to cut our spending in retirement and live more frugally. But, alternatively, we could save more before retiring. If 4% withdrawals require 25x annual spending in savings before retiring, then 3.9% withdrawals require 25.6x annual spending in savings before retiring. So really I think the question we’re asking is, at what point does having over-saved mean your asset allocation decisions are (relatively) meaningless.
First we need a metric. I’ll be using “Years Sustained” which comes from Estrada’s paper “Replacing the Failure Rate”. It measures not just binary failure or success but the magnitude of failure or success. We measure the number of years the portfolio lasted. A “0” means it lasted 30 years. Negative numbers are bad. Higher numbers are better.
Of course, no single metric captures everything important about retirement. Lots of people care about the pain of draw downs, for instance. Still, we’ll stick with a single metric for now and see how it plays out.
“Perfect” Asset Allocations
Given our “Years Sustained” metric, we can calculate the “best” asset allocation for every retirement cohort. That is, the asset allocation that maximizes Years Sustained.
For now, we’ll assume static asset allocations rather than anything fancier like glidepaths or Prime Harvesting or whatnot.
It may be surprising to some, but most of the time, the best asset allocation is 100% stocks. There are handful of times when having 10–20% bonds was nice. And two times where having 50% bonds was best. That huge diverge is what drives most online arguments over asset allocations. ~98% of the time an equity-heavy portfolio is best. But ~2% of the time having a massive amount of bonds was best.
Now we know the “best” asst allocation for each retirement cohort. We can use that as a best case scenario for the importance of asset allocation. Now we have our baseline comparison. We’ll use standard 4% inflation-adjusted withdrawals and compare the “best” asset allocation versus just using a dumb, old 60/40 asset allocation. This shows us how much an impact asset allocation can make. The difference between “perfect” and “average”.
We can see that (unsurprisingly) using the Best Asset Allocation each cohort improves Years Sustained substantially in most cases. But that brings us to our first problem. For most people, there’s a diminishing utility to huge portfolio values in retirement. We care more about those cases where we’re close to failure.
Let’s look at only the cases where our Years Sustained is 10 or fewer; that is we either failed or we succeeded but our remaining portfolio value is relatively small (and we are probably a bit stressed about it).
Now things are a little bit clearer. One surprisingly revelation: knowing the Best Asset Allocation didn’t “save” very many retirements. Only 4 retirements (out of 118, 3.3%) went from “failure” to “success” by virtue of knowing the Best Asset Allocation. And 4 retirements were failures even if you knew the Best Asset Allocation to use.
Clearly asset allocation isn’t a panacea…but we knew that coming in. And the goal here is to try to quantify the trade-off between saving more and trying to pick the perfect asset allocation.
Saving more versus the best asset allocation
Now we can compare the Best Asset Allocation versus various levels of “just save more”. We can chart 28x, 33x, 40x, 50x, 66x, and 100x savings when using a vanilla 60/40 asset allocation. And compare that to using the Best Asset Allocation with just 25x savings.
We can see that at 50x savings using the dumb asset allocation is virtually always better than using the best asset allocation. So we’re starting to get some parameters on our initial question of “how important is asset allocation”. But when we look closer at the 40x savings chart we notice it does much better in the worse scenarios (which, again, are the ones we care about the most). This remains true when we look at the 33x savings chart, too.
These charts are pretty noisy so let’s try to simplify them a bit, just as we did previously. We’ll just focus on the cohorts where things we close to (or at) the edge of failure.
I’ve switched from using “25x savings” phrasing to “4% withdrawal rate” phrasing because we’re using much finer slices and I think the withdrawal rate phrasing makes things clearer in this context.
In our base case, 4% withdrawals for both portfolios, there are 21 borderline failure scenarios where knowing the Best Asset Allocation improved things. In a few cases the improvement was marginal; in others substantial. Still, 21 cases of improvement. That’s our base case with 4% withdrawals on the 60/40 portfolio.
Reducing our withdrawal rate on the 60/40 portfolio to 3.9% (equivalent to saving an extra 7 months of expenses before retiring) reduced that to just 12 scenarios where knowing the Best Asset Allocation improved our Years Sustained.
Reducing our withdrawal rate on the 60/40 portfolio to 3.8% (equivalent to 26.3x years of savings; that is, an extra 15 months of expenses saved up) has almost completely removed the advantage of knowing the Best Asset Allocation. There are still 6 scenarios where it improved our Years Sustained. But if we look closer only 2 of those scenarios (1899 and 1965) were a significant improvement. And now there are scenarios where the 60/40 portfolio is substantially better (1966, 1968, 1969, 1973).
Things are already looking bad for the Best Asset Allocation but we’ll keep going.
At 3.7% withdrawals for the 60/40 portfolio, there is no longer any advantages for the Best Asset Allocation. (In 1965 they are essentially tied.) And there are many scenarios where the 60/40 portfolio has a significant advantage. 3.7% withdrawals are equivalent to an extra 2-years of savings (that is, 27x instead of 25x).
Utility functions provide a concise way of discounting very large positive results (dying with $10 million is only slightly better than dying with $5 million) while providing steep penalties for failure. Estrada & Kritzman’s “Coverage Ratio” takes the raw “Years Sustained” metric above and applies a utility function to it.
We can see that the huge peaks are smoothed down and the failures stand out more. This maps closer to our intuition on these different scenarios. And we can see that when looking at the Coverage Ratio (this time of all cohorts) by the time withdrawal rates drop to 3.7% most of the advantages of the Best Asset Allocation are gone and serious disadvantages have appeared. We don’t capture all of the upside in the best scenarios — a lower withdrawal rate can only do so much — but it is very effective at avoiding the worst scenarios.
In favor of One More Year
With a whole bunch of caveats (we only used one metric; we didn’t consider any kind of dynamic asset allocation) we’re at last able to start putting some bounds on the importance of asset allocation versus withdrawal rates.
Even with perfect hindsight, choosing the best possible asset allocation is only equivalent to going from a 4% withdrawal rate to a 3.7% or 3.8% withdrawal rate. In other words, saving 1 or 2 extra years of expenses dominates getting the asset allocation decision perfectly correct. In reality, we don’t have perfect hindsight and our asset allocation will be sub-optimal.
How long does it take a person to save that extra money? Obviously, it will depend on your personal savings rate. But if you’ve already got 25x saved up, then 4% growth with give you 1 extra year of savings. For many people that means a single year of market returns will give them the extra savings they need. If you’re actually contributing money yourself as well, it will happen even faster.
Instead of stressing about trying to pick “the right” asset allocation, you’re better off picking anything reasonable and ignoring every other asset allocation internet discussion for the rest of your life…and then working an extra six or twelve months to pad out your retirement fund before retiring.
And, as a corollary, if asset allocation is this unimportant…all of the other, even more esoteric decisions (including ones I’ve written about…!) are even less important.