Wealth explained
Sequence of returns decides retirements that the average return says should have worked
Two portfolios can earn exactly the same annualised return across thirty years and end in completely different places. What separates them is which of those years were the bad ones.
In this story
Illustrative. Bank Season is an editorial prototype. This story, its sources, the issuers named in it and every figure it quotes are invented to demonstrate the publication. Nothing here is reported fact or investment advice. Read the disclosure.
Every retirement projection ever shown to a client uses an average return. The portfolio earns six percent a year, spending grows with inflation, and a line on a chart slopes gently upward for thirty years. The projection is not dishonest. It is simply answering a question nobody asked, because no portfolio earns six percent a year. It earns twenty two, then minus fourteen, then nine, and the order in which those numbers arrive decides whether the retirement works.
The mechanism is easy to state and surprisingly hard to feel. While a portfolio is accumulating, order does not matter at all: the same set of returns compounds to the same place regardless of sequence. The moment money starts coming out, order becomes the dominant variable, because a withdrawal made after a twenty percent fall sells more units than the same withdrawal made after a twenty percent rise, and those units never come back.
Portfolio value with identical returns in opposite order
Thousands of US dollars. Both paths use the same thirty annual returns with the same 6.2% annualised average, drawing $45,000 in year one and increasing it with 2.4% inflation.
The two lines are the same investment. The same thirty returns, the same average, the same spending plan, the same starting million. One household runs out of money in its eighty ninth year and the other leaves nearly twice what it started with. Nobody in either household made a mistake, and no adviser could have told them in advance which path they were on.
Identical thirty year returns, four different orderings
| Ordering | Annualised return | Worst three years fall in | Lowest portfolio value | Value at year 30 |
|---|---|---|---|---|
| Poor returns first | 6.2% | Years 2, 4 and 5 | $0 in year 30 | $0 |
| Poor returns last | 6.2% | Years 26, 28 and 29 | $905,000 in year 1 | $1.92m |
| Actual historical order | 6.2% | Years 9, 14 and 22 | $812,000 in year 16 | $1.24m |
| Poor returns in the middle | 6.2% | Years 13, 15 and 17 | $754,000 in year 19 | $1.09m |
| Returns smoothed to the average | 6.2% | No negative years | $1.00m in year 1 | $1.67m |
Bank Season model. Same thirty annual returns in each row, reordered. Spending starts at $45,000 and rises 2.4% a year. Illustrative only and not a forecast.
$1.92m
Difference in terminal wealth between the best and worst ordering of one set of returns
The two paths contain identical annual returns and identical spending. Only the order differs.
The window where it matters
Sequence risk is not a permanent condition. It is concentrated in roughly the first decade after withdrawals begin, and it fades quickly after that for a simple reason: by year fifteen the portfolio is either comfortably ahead of the spending path, in which case a bad year is survivable, or it is already too small, in which case the damage was done earlier. Modelling across historical return series consistently puts around seventy percent of the variance in outcomes into the first ten years.
That concentration is useful, because it means the problem has a defined duration. A household does not need a permanent defensive posture. It needs one for about ten years, and it needs to be willing to abandon it afterwards, which is the part most people get wrong in the other direction by staying defensive for thirty years and running out of money slowly instead of quickly.
- Flexible spending. Reducing withdrawals by ten percent in any year following a portfolio decline of more than fifteen percent removes most of the failure cases in historical testing, and the cumulative reduction in lifetime spending is smaller than households expect because the cuts are temporary.
- A spending reserve. Holding two to three years of planned withdrawals in cash and short bonds means the equity portfolio is never sold into the worst of a drawdown, and the reserve is refilled in years when markets are up.
- Deferring guaranteed income. Every year that the largest inflation linked income source is deferred raises the floor under the whole plan and shortens the period during which the portfolio has to do the work alone.
- Owning bonds that behave like bonds. Credit and equity fall together. A spending reserve built from high yield or from equity income funds is not a reserve, it is the same risk with a different label.
What the mitigations cost
Nothing here is free, and the honest version of this discussion prices each option rather than recommending all of them. Flexible spending costs the household a variable standard of living. A spending reserve costs expected return. Deferring income costs several years of cash flow that has to come from somewhere. The right combination depends on which of those three a particular household can actually tolerate, which is a question about temperament rather than about finance.
Cost and effect of the three main mitigations
| Mitigation | Cost to the household | Effect on failure rate | Effect on median terminal wealth | Tolerable for most? |
|---|---|---|---|---|
| Spending falls 10% after a 15% decline | Variable standard of living | Large reduction | Higher | Yes, if agreed in advance |
| Three year spending reserve | Roughly 25bp of annual return | Moderate reduction | Slightly lower | Yes |
| Defer the largest guaranteed income to 70 | Four to eight years of cash flow | Large reduction | Higher | Depends on other assets |
| Hold a lower equity weight throughout | Roughly 90bp of annual return | Small reduction | Much lower | Generally counterproductive |
| Annuitise part of the portfolio | Liquidity and any bequest on that portion | Large reduction | Lower | Underused, often appropriate |
Bank Season modelling across historical return sequences. Failure is defined as exhausting the portfolio before year thirty. Effects are directional rather than precise.
The fourth row is the one that surprises people. Simply holding fewer equities for thirty years reduces the failure rate only slightly and reduces median wealth substantially, because the portfolio then fails from insufficient growth rather than from a bad start. Defensive positioning helps when it is targeted at the decade that matters and hurts when it is applied to the whole retirement.
I ask clients to write down, before they retire, what they will cut if the portfolio falls twenty percent in the first two years. Not whether they would cut. What. The holiday, the second car, the gifts to the grandchildren. A plan with that page in it survives things that a plan with a higher bond allocation does not.
Margit Sollers, tax director at Thistledown Advisors
The version that fits on one page
Hold three years of spending outside equities. Agree a rule in writing that cuts withdrawals by a tenth after a serious decline and restores them after two good years. Defer the largest guaranteed income source as far as other assets allow. Then stop adjusting, because the remaining decisions are noise, and reassess when the first decade is behind you and the arithmetic has stopped being dangerous.
70%
Approximate share of variance in retirement outcomes attributable to the first ten years
Bank Season modelling across historical thirty year return sequences for a balanced portfolio with inflation linked withdrawals.
None of this is a prediction about markets, which is what makes it usable. A household cannot know whether it is retiring into the top line on the chart or the bottom one. It can arrange its affairs so that the difference between the two is a smaller holiday rather than a different life, and that arrangement has to be made before the first bad year rather than during it.
.jpg){kind=link}