Monte Carlo vs the 4% Rule: Simulating Retirement Withdrawals Instead of Assuming Them

Photo by Fumiaki Hayashi on Unsplash

The central question of decumulation is deceptively simple: how much can you spend from a portfolio each year without running out before you run out of years? The 4% rule answers it with a number. Monte Carlo simulation answers it with a distribution. The difference between those two answers is not academic — it changes what you monitor, when you adjust, and how much confidence you are entitled to feel.

This piece is about the methodology, not a product. It matters because a fixed withdrawal number quietly assumes the future will resemble a specific slice of the past, and that assumption is doing far more work than most retirees realize.

Context: where the 4% rule actually comes from

The 4% rule traces to William Bengen's 1994 study and the subsequent Trinity Study (Cooley, Hubbard, and Walz, 1998). Bengen tested rolling 30-year retirement windows against historical US stock and bond returns and found that an initial withdrawal of about 4% of the portfolio, increased annually for inflation, survived every historical 30-year period — including retirements that began just before the 1929, 1937, and 1973 drawdowns. The 4% figure is therefore not an average outcome. It is the withdrawal rate that survived the worst realized sequence in a single country's price history.

That framing is the crux. The rule is a summary statistic of history, and it inherits every limitation of that history: one country, roughly a century of data, a handful of genuinely independent 30-year windows, and a bond regime that for much of the sample paid real yields we have not consistently seen since. As of this writing the 10-year Treasury sits at 4.58% and headline CPI is running near 3.7% (FRED, asof 2026-06-01), so the starting conditions a new retiree faces are not the average of the historical sample — they are one specific draw from it.

What Monte Carlo actually does differently

Monte Carlo simulation does not ask "what happened." It asks "what could plausibly happen, and how often does my plan survive it." The mechanics are straightforward. You specify assumptions for expected return, volatility, and the correlation between assets; you draw thousands of randomized return sequences consistent with those assumptions; you run the withdrawal schedule through each path; and you count the fraction of paths in which the portfolio is still solvent at the horizon. That fraction is the "success rate."

The methodological gain is that you see the whole distribution of outcomes rather than a single pass/fail verdict. The 4% rule tells you a plan would have survived the worst realized past. Monte Carlo tells you it survives, say, 95% of 10,000 simulated futures — and, more usefully, it shows you the 5% where it does not.

Two honest caveats belong here. First, a naive simulation that draws each year's return independently understates real risk, because it discards autocorrelation and the fat-tailed, regime-clustered nature of markets — block bootstrapping from historical data or fatter-tailed distributions partially repairs this. Second, and more fundamentally, the output is only as good as the return and volatility assumptions you feed it. Garbage assumptions produce confidently wrong success rates. This is the live-vs-backtest gap in a different costume: the simulation is precise about a model of the world, not the world.

Neither column is a data table of live funds, and deliberately so — this is a comparison of two methods. For the underlying asset behavior that feeds either approach, the realized returns of common building blocks are documented in what VTI, VXUS, and BND actually delivered.

The insight the success rate hides: failures are not spread evenly

Here is the part most retirement calculators show you and most retirees skim past. When a Monte Carlo run reports a 95% success rate, the intuition is that failure is a small, diffuse background risk. It is not. The failing 5% are overwhelmingly the paths that suffer a large drawdown in the first several years of withdrawals. This is sequence-of-returns risk, and it is the single most important reason two plans with identical average returns can end in wildly different places.

Initially I treated the success rate as the headline number. Then I looked at when the failing paths crossed zero, and the picture changed: the median failure is not a portfolio that slowly bled out over 30 years, but one that took an early hit while withdrawals kept pulling from a shrinking base, locking in losses that a later-arriving bull market could no longer repair. The average path was fine. The order of returns was fatal. That mechanism is worth understanding on its own terms; I have written about it separately in why sequence-of-returns risk is the hidden killer before retirement and in a closer look at protecting a 30-year plan from a pre-retirement crash.

The practical consequence is a second-order effect that a single number obscures: chasing a higher success rate by lowering your withdrawal is an inefficient use of spending. Moving from a 90% to a 99% success target can force a materially lower initial withdrawal, yet the marginal failures you are buying protection against are dominated by extreme early-sequence tail paths — exactly the paths where a disciplined retiree would already be cutting spending in real time. You pay a certain, permanent reduction in lifestyle to insure against a scenario you would respond to dynamically anyway.

A Monte Carlo success rate is not a promise about the average future; it is a statement about the tails, and the tails are almost entirely a question of when the bad years arrive, not whether they do.

Dynamic withdrawal: where the two methods stop competing

Framing this as "4% rule versus Monte Carlo" is slightly false. The more accurate framing is fixed versus adaptive spending. The 4% rule and a static Monte Carlo plan share the same brittle assumption: that you set a withdrawal at retirement and never look at the portfolio again. No disciplined steward behaves that way.

Adaptive rules — Guyton-Klinger guardrails, Vanguard's dynamic spending bands (Vanguard research, 2024), or a simple rule that skips the inflation raise after a down year — change the problem. When you let spending flex within stated bands, simulated success rates rise sharply for the same starting withdrawal, because the plan self-corrects in precisely the early-drawdown paths that cause failures. The cost is variable income, which is a real behavioral burden, not a free lunch. But it reframes the initial-rate debate: the exact starting percentage matters far less once the plan can respond to what the market actually does. I explored the mechanics of relaxing the fixed rate in re-evaluating the 4% rule with sequence risk, yield, and dynamic withdrawal.

This is where the stewardship instinct and the statistics point the same direction: humility about prediction argues for a plan that observes and adjusts, not one that commits to a number in year one and hopes the sample it was fitted to repeats.

What this comparison can and can't tell you

It can tell you that the 4% rule is a useful anchor and a poor plan — a good back-of-envelope check, a bad substitute for watching sequence risk. It can tell you that Monte Carlo's value is the shape of the failure distribution, not the headline percentage. What it cannot tell you is the correct withdrawal rate for a specific person, because that depends on inputs no simulation supplies: your horizon, your other income, your tolerance for cutting spending in a bad year, and forward return assumptions that are genuinely unknowable. Any simulation quoting a 95.0% success rate to three significant figures is expressing false precision about the model, not the world. Treat the second decimal as noise and the first as a direction.

FAQ

Is the 4% rule still valid in 2026? As a survival threshold from historical US data, it hasn't been invalidated — but starting conditions matter. With the 10-year Treasury near 4.58% and CPI near 3.7% (FRED, asof 2026-06-01), a new retiree faces a specific draw from the distribution, not its average. Use 4% as a reference point, then stress it with simulation and a plan to adjust.

What success rate should I target in Monte Carlo? There is no universal answer, but chasing 99%+ often costs more lifestyle than it buys safety, because the marginal failure paths are extreme early-sequence tails you would respond to dynamically. Many practitioners treat 85–95% as reasonable when paired with flexible spending.

Why do two plans with the same average return end differently? Sequence-of-returns risk. When withdrawals coincide with an early drawdown, losses are locked in from a shrinking base, and a later recovery arrives too late. The order of returns, not their average, drives most retirement failures.

Does Monte Carlo overstate or understate risk? A naive version that draws returns independently tends to understate risk by ignoring volatility clustering and fat tails. Block bootstrapping from historical sequences or using fatter-tailed distributions gives a more honest picture.

Should I use the 4% rule or Monte Carlo? They answer different questions. The 4% rule is a fast sanity anchor; Monte Carlo shows the distribution and timing of risk. Neither replaces a dynamic withdrawal plan that observes the portfolio and adjusts spending within stated bands.

Key takeaways

• The 4% rule is one conclusion from one worst-case historical sequence; Monte Carlo is a method that surveys thousands of possible futures.
• A success rate's real information is in the tails — failing paths cluster in early sequence-of-returns risk, not in the average path.
• Pushing the success target toward 100% buys insurance against tail scenarios you would already respond to dynamically, at a certain cost to lifestyle.
• Simulation output is only as trustworthy as its return and volatility assumptions; precise-looking percentages express confidence in a model, not the world.
• Fixed versus adaptive spending is the decision that matters more than the exact starting percentage — a plan that observes and adjusts dominates one fitted to a single historical sample.

This article is for educational purposes and does not constitute personalized financial advice. See the full Disclaimer.