Retirement Withdrawal Survival Simulator: the Historical Cohort Replay (1871–2025)

The “safe withdrawal rate” is one of the most searched questions in personal finance — and one of the most poorly framed. US market history does not supply a universal rate: it supplies cohorts. A 1982 retiree and a 1966 retiree, with exactly the same capital, the same withdrawal and the same allocation, met opposite fates: the first ended with a multiple of the starting capital, the second ran out before the finish line.

That is what this simulator makes visible. It projects nothing: it replays history. Your withdrawal plan is applied to every starting year since 1871, on US real returns (Shiller data). Every grey trajectory is a cohort that reached the end; every terracotta one, a cohort whose capital was depleted along the way. The gap between the two owes nothing to the plan or to discipline — only to the order in which returns arrived: sequence-of-returns risk.

What the tool actually replays

The engine runs a single mechanic, cohort by cohort: the withdrawal is taken at the start of each year, then the remaining capital earns the year’s real return on the mix — an equity share (S&P Composite, dividends reinvested, CPI-deflated) and a 10-year US bond share, rebalanced annually. Because the replay uses real returns, a constant withdrawal in the tool corresponds to an inflation-indexed withdrawal in real life — the convention of the reference studies. A cohort “fails” when its capital can no longer fund the year’s withdrawal.

The equity share is an assumption you enter, from 0 to 100% — the tool offers no profiles and recommends no split. Moving the slider shows exactly what it should: every mix produces its own distribution of historical survivals and failures, and none of them makes sequence risk disappear.

The default-settings example

At the tool’s default settings — €500,000 of capital, a €20,000 initial annual withdrawal (4.0%), 30 years, a 50% equity exposure assumption — the replay counts 126 thirty-year cohorts starting between 1871 and 1996, of which 120 reach the end with a positive capital: 95% of starts. All six failures cluster on the 1964–1969 starts: the 1970s stagflation hit the earliest withdrawal years with the double blow of negative real equity returns and inflation-ravaged bonds.

The extremes tell the story better than any average rate: the worst cohort, starting in 1966, saw its capital depleted in 1991, after 25 years; the best, starting in 1982, finishes at €2,977,407 in real terms — nearly six times its stake. In between, the median cohort ends at €510,011, roughly its starting capital preserved in purchasing power. Same plan, same amounts: only the starting year differs.

The 4% rule, put back in context

The 4% reference comes from the Trinity study (Cooley, Hubbard and Walz, 1998), which measured historical success rates across US cohorts — not a guaranteed entitlement to withdraw 4%. This simulator is, at heart, its interactive version: it displays the reference in grey, as a descriptive comparison point, and lets you observe that the success rate of the same withdrawal varies with duration, mix and — above all — cohort. Withdrawal rates and starting valuations extends this reading.

What this replay does not say

• US data only. The US market is the great survivor of the 20th century; replays on other markets (Japan, Germany, Austria-Hungary…) yield lower survival rates. This survivorship bias makes the sample rather optimistic.
• Excludes fees, taxes and behaviour. Management fees, taxation and panic selling all degrade a plan’s real-world survival relative to the raw replay.
• 126 cohorts, but not 126 independent observations. Thirty-year windows overlap: the same 1970s run through dozens of cohorts. The sample describes history, not a statistical distribution.
• Data conventions. Shiller prices are monthly averages of daily closes; the bond total return is Shiller’s approximation from the 10-year yield (GS10), not a market index.
• The past does not predict future cohorts. The replay documents what happened; it does not bound what can happen.

Frequently asked questions

Why do two cohorts with the same withdrawal rate end up so differently?

Because in the withdrawal phase, the order of returns matters as much as their average. Early losses, suffered while money is being withdrawn, destroy a capital base that later good returns never recover: that is sequence risk. The 1964–1969 cohorts are its historical illustration — their 30-year average return was not catastrophic, but their first decade was. The mechanism in detail is here.

What does the 4% reference show historically across US cohorts?

On the data replayed here (Shiller, 1871–2025), a 4% initial withdrawal over 30 years with a 50% equity exposure assumption was sustainable for 120 cohorts out of 126 — an order of magnitude consistent with the success rates published by the Trinity study (1998) on its own period and its own data. Failures cluster on the late-1960s starts.

Why does the 1929 cohort not appear among the failures?

It is the replay’s most counter-intuitive result: in real terms, the 1930–1932 deflation mechanically reduced the withdrawals of an indexed plan, cushioning the crash. The historical worst enemy of a withdrawal plan was not the deflationary depression but stagflation — weak markets and inflation swelling the withdrawals. Hence the 1964–1969 failure cluster, with 1966 as the worst cohort.

How does this tool differ from the capital-for-income simulator?

The capital-for-investment-income simulator answers a perpetuity question: what capital covers expenses without ever touching the principal, at an assumed constant return. This one answers the complementary depletion question: did a given capital, drawn down every year, historically last the distance — cohort by cohort, on real returns as they actually occurred. The first reasons in steady state; the second replays the turbulence.

Go deeper

• Withdrawal rates and starting valuations — what market levels at retirement have historically changed.
• Sequence-of-returns risk in retirement — the mechanism this simulator turns into pictures.
• A realistic retirement savings amount — the upstream accumulation question.
• Common mistakes about retirement.
• All Eco3min financial tools.

An educational tool built on public historical data (Shiller, 1871–2025). Results describe past cohorts and constitute neither investment advice, nor a recommendation, nor a prediction.

Last updated — 13 July 2026