What you are actually watching
While you are saving, the order of your returns barely matters. The moment you start
withdrawing, it can matter more than the average. That asymmetry is the single most
under-explained idea in retirement planning.
What sequence-of-returns risk is
Look at the two retirements in the third panel. They begin with the same portfolio, take the
same income, and receive exactly the same set of yearly returns. The averages are
identical to the last decimal. Only the order differs, and the outcomes are not close.
The mechanism is simple once you see it. When markets fall while you are withdrawing, you
sell a larger share of your portfolio to fund the same income. Those shares are gone, so
when the recovery arrives it lifts a permanently smaller base.
Sequence-of-returns risk is the name for that trap, and it is why the first
five to ten years of retirement carry so much more weight than any later stretch.
While you are still contributing, the same logic runs in reverse. A crash early in your
career is close to a gift, because every contribution afterwards buys more.
Why one smooth line flatters the plan
The dashed red line is what a conventional retirement calculator shows: the same return
applied every year. Compare it to the median simulated life. It sits higher, and the reason
is not sequence risk at all.
It is volatility drag. Gain 20% then lose 20% and you are not back where you
started, you are down 4%. Averages ignore that; compounding does not. A portfolio earning 5%
on average with big swings grows more slowly than one earning a steady 5%, and the wider the
swings the bigger the gap. Set volatility to zero and watch the median path rise to meet the
flat line.
The more serious problem with a single line is not that it is optimistic. It is that it
conveys no range. A plan that looks fine on the average path can still fail in a quarter of
realistic ones, and the average path tells you nothing about that.
Where the 4% rule came from, and what it is not
The 4% rule says you can withdraw 4% of your portfolio in the first year of
retirement, adjust it for inflation each year after, and expect it to last three decades. It
came from studying historical US market data over rolling 30-year windows.
That origin is also its limit. It describes what would have survived the past, in one
country, over a particular period, with a particular asset mix. It was never a law, and the
person who popularized it has said as much repeatedly.
Two things it assumes are worth questioning. It assumes a fixed real withdrawal regardless of
what markets do, which no sensible retiree would actually follow. And it assumes exactly 30
years, so retiring at 55 and planning to 95 is a different question with a different answer.
Push the withdrawal rate here to 5% or 6% and watch how quickly the success rate falls.
Why everything is in today's dollars
A balance of two million in forty years is a meaningless number without knowing what it
buys. This model works in real terms: the return you enter is after
inflation, and the withdrawal holds its purchasing power for life.
That means the figures are directly comparable to money today, and it removes the most
common error in retirement arithmetic, which is applying a nominal return and then forgetting
that the income also has to rise with prices.
How to defend against a bad sequence
The risk cannot be removed, but it can be blunted, and the ways of doing it are all versions
of the same idea: avoid being a forced seller in a downturn.
Holding a couple of years of spending in cash or short bonds means an early crash is
weathered rather than crystallised. Being willing to cut spending in bad years does more than
any asset allocation change, because the fixed-withdrawal assumption is what makes the model
so brittle. And working a year or two longer helps three times over: more contributions, more
compounding, and fewer years to fund.
Try it here. Push the retirement age up by two years and watch the success rate move more
than a full percentage point of return would.
What this model leaves out
A great deal, and some of it is significant. Returns are drawn independently each year from a
normal distribution, which is convenient but wrong in two known ways: real markets show fat
tails, meaning extreme years happen more often than a bell curve predicts, and they show some
mean reversion, meaning a terrible decade is more likely to be followed by a good one than
pure randomness implies. The first makes this model too optimistic, the second too
pessimistic, and they do not neatly cancel.
There is no Social Security, pension, annuity, or inheritance, all of which reduce how much
the portfolio has to carry. No taxes, and the difference between a traditional and a Roth
account is real money. No fees, which compound against you exactly as returns compound for
you. Spending is flat, when real retirement spending usually falls through the seventies and
then rises with medical costs. And there is one portfolio with one volatility, rather than a
mix that shifts as you age.
Use the success rate as a way to compare choices against each other, not as a probability
about your life. A plan that goes from 72% to 91% when you delay retirement two years has
told you something real, even though neither number is a forecast.
Things to try
Each takes a few seconds and lands something the headline number alone will not.
1Turn off the randomness
Set volatility to 0. The bands collapse, the median meets the flat red line, and you
are looking at exactly what an ordinary retirement calculator would tell you.
2Find your breaking point
On any preset, raise the withdrawal rate a tenth of a point at a time and watch the
success rate. The fall is far steeper than the change in the rate.
3Buy safety with two years
Push the retirement age up by two. Compare the improvement against raising the assumed
return by a full point instead, which is a much bigger thing to assume.
4Watch the danger zone move
Load "Retiring next year" and change the simulation code a few times. Notice how much
the outcome swings compared with doing the same on "Starting at 30".
5Save more against earn more
Raise the yearly saving by 20%, note the change, then instead raise the return by 20%
of its value. Which lever moves the plan further depends on how far out you are.
6Read the tenth percentile
Ignore the median entirely and plan against the tenth percentile column. That is the
question a planner actually asks: what if this goes badly.
Common questions about retirement planning
What is sequence-of-returns risk?
It is the risk that the order of your returns, and not just their average, decides whether your money lasts. Once you are withdrawing, a bad stretch early forces you to sell more shares to fund the same income, which permanently shrinks the base that later has to recover. The identical set of returns arriving in a friendlier order can leave you several times better off, which is exactly what the comparison on this page demonstrates.
Why does a flat-return retirement calculator mislead?
Because it draws one smooth line that no real portfolio follows. Applying the same average every year removes both sequence risk and volatility drag, so it tends to be optimistic, and more importantly it gives no sense of the range. A plan that works on the average line can still fail in a quarter of realistic paths, and a single number cannot tell you that.
Is the 4% rule still safe?
It was never a guarantee. It came from studying historical US market data over rolling 30-year windows, so it describes what would have survived the past rather than what must survive the future. Treat it as a starting point, check the success rate for your own horizon and volatility rather than assuming 30 years, and bear in mind that a retiree willing to cut spending in bad years is far safer than the fixed-withdrawal assumption suggests.
Why are the results in today's dollars?
Because a balance decades away is meaningless without knowing what it buys. This model uses a real return, meaning after inflation, and holds the withdrawal constant in purchasing power for life. That prevents inflation being counted twice, which is the most common error in retirement arithmetic, and makes every figure directly comparable to money in your pocket today.
What does the success rate actually mean?
It is the share of simulated lives in which the money lasted to your planning age, so a 90% success rate means one path in ten ran out early. It is not a probability about your life specifically. It depends entirely on the return and volatility you assumed, and real markets can deliver conditions outside anything the model drew. It is most useful for comparing one choice against another rather than as a forecast.
Which matters more, saving more or earning a higher return?
Early on, contributions dominate, because a percentage return on a small balance is a small number and the contribution is entirely within your control. Later, returns dominate, because compounding is acting on a large balance. The practical consequence is that a young saver should concentrate on the savings rate, while someone close to retirement should concentrate on not taking risk they cannot afford to have go wrong.
How many years should I plan for?
Longer than average life expectancy, because planning to the average means roughly half of people outlive the plan. A couple should plan for the second death rather than the first, since at least one of them is likely to live longer than either would alone. Planning into the mid nineties is common, and the cost of being wrong in that direction is far smaller than the cost of being wrong the other way.