Where the 4% rule comes from
The 1998 Trinity Study examined historical US market data and asked a simple question: what starting withdrawal rate, adjusted for inflation each year, would have survived 30 years across every historical starting point?
The answer was approximately 4%. Withdraw $40,000 from a $1m portfolio in year one, increase that amount with inflation each year thereafter, and in almost every historical 30-year window the money lasted.
That is a finding about the past. It is not a law of finance, and it was never meant to be a guarantee.
Why the average return is the wrong question
Here is the part that ordinary retirement calculators hide. Two retirees can experience exactly the same average annual return over thirty years and end up in completely different positions — one with millions, one broke.
The difference is the order the returns arrive in.
When you are still saving, order barely matters; a crash early is arguably good, because you buy cheaply for years afterwards. The moment you start withdrawing, that reverses completely. A downturn in your first years means selling assets at depressed prices to fund living costs, and those units are gone. They are not there to participate in the recovery. The base that has to grow back is permanently smaller.
This is sequence-of-returns risk, and it is the single largest threat to an early retirement. It is why a projection using one average return — the kind almost every retirement calculator produces — tells you very little about whether a plan is actually safe.
The demonstration on this page
Two of the scenarios above are constructed specifically to isolate the effect. The "crash in year one" and "crash in year twenty" scenarios apply an identical bear market — the same five years of returns, the same magnitude — at different points in the retirement.
Same portfolio, same withdrawal, same shock, same average return. Compare the ending balances. The gap between them is sequence risk made visible, and it is not subtle.
The other scenarios show distinct shapes worth understanding: a strong opening decade that builds a cushion before anything bad happens; a flat decade with no crash at all, just ten years of withdrawals against no growth; and a 2000s-style double crash with a weak recovery in between, which is the shape that made the year 2000 the worst modern date to retire.
How the simulation works
Withdrawing before applying the return is the Trinity convention and the conservative choice: money spent at the start of a year is not invested during it. Reversing the order would understate the damage of an early crash — exactly the effect being measured.
Everything is in real terms, so a $40,000 withdrawal keeps the same purchasing power in year 30 as in year 1, and the scenario returns are already inflation-adjusted.
On the data: these return series are hand-specified and stylised. They are chosen to illustrate distinct shapes, not to reproduce actual index history, and nothing is fetched from anywhere. They show how sequence changes outcomes. They are not a backtest and should not be read as one.
One caveat that matters for reading the results. These sequences describe an all-equity path. The Trinity Study tested mixed stock and bond portfolios, and bonds performed strongly through the 2000s precisely when equities did not — so a real 60/40 portfolio came through that decade considerably better than the "lost decade" scenario here suggests. The scenarios are deliberately unkind, which is what makes them a stress test rather than a projection.
The same caution applies to the "rate all scenarios survive" figure in the results. It is the rate at which all six of these sequences survive, including the harshest, with rigid withdrawals and no bond allocation. Read it as a lower bound under pessimistic conditions, not as a recommended withdrawal rate.
What actually protects a retirement
Every scenario here assumes a rigid plan: the same inflation-adjusted withdrawal every year regardless of what markets do. That is deliberately the worst case, and it is not how most people behave. The things that genuinely improve survival:
- Flexible spending. The single most powerful lever. Cutting withdrawals 10% during a downturn rescues sequences that break a rigid plan entirely, because you stop selling into weakness.
- A cash buffer. Two to three years of spending in cash means you need not sell equities at all during a crash. It costs return in normal years and buys survival in bad ones.
- Any income at all. Even modest part-time earnings during the first few years disproportionately protect the portfolio — see Barista FIRE.
- A lower starting rate. For horizons beyond 30 years, 3% to 3.5% is common. Increase the retirement length above and watch scenarios that survived at 30 years start failing.
- Guaranteed income. Social Security, pensions and annuities cannot run out, so the portfolio only needs to cover the remainder.
The honest summary
4% is a reasonable planning anchor and a poor promise. It came from a specific market history over a specific 30-year horizon, and an early retiree with a 50-year horizon is outside the conditions it was tested under.
The useful takeaway is not a number. It is that the first decade of retirement matters far more than the rest, that a plan you can adjust beats a plan you cannot, and that a projection built on a single average return is telling you about a market that has never existed.
If you are retiring before 59½ in the US, sequence risk interacts directly with account access — see the bridge account calculator, since a bridge fund is spent precisely during the years when sequence risk is most acute. For the other side of the coin, the Die With Zero calculator shows what deliberately spending principal looks like.