“Safe” withdrawal rate (SWR) and the 4% rule

You have invested capital and need to live off it for the next thirty years. How much can you withdraw each year without the risk of running out of money before the period ends? It's the central question of any retirement plan, and also the hardest one to answer with a single number. The “safe” withdrawal rate, or SWR, is the most common attempt at an answer. The quotes around “safe” mark a factual point: the rate describes what held in the historical cases analysed.

What the “safe” withdrawal rate actually is

SWR is the percentage of your starting capital you withdraw in the first year of retirement, then adjust for inflation every year after that. A $500,000 portfolio, a 4% SWR: you take out $20,000 in year one. If inflation runs at 3%, year two you withdraw $20,600, not 4% of a revalued balance, just the same purchasing power as year one.

The capital stays invested the whole time, typically a mix of stocks and bonds. It's not a checking account draining in a straight line. It keeps generating returns, and in plenty of historical scenarios it ends up larger than it started, even after thirty years of withdrawals.

The 4% rule, and where it came from

In 1994 financial adviser William Bengen published an analysis that reshaped how people think about retirement. He tested different withdrawal rates against every possible 30-year period in US market history going back to 1926. Four percent was the highest rate that held up in nearly every scenario, including the worst one on record: someone retiring in 1966, walking straight into a decade of high inflation and flat markets.

Four years later, three professors at Trinity University published what's now known as the Trinity Study: same idea, a wider dataset, portfolios with different stock-bond splits. They confirmed 4% worked in about 95% of historical 30-year scenarios for a balanced portfolio. From there the number turned into a standard, almost a piece of personal-finance folklore.

Why 4% isn't a magic number

The 4% figure is a reasonable starting point. Three things complicate it.

The first is sequence risk: once withdrawals begin, the order returns arrive in weighs at least as much as their average, and more in the early years. Two retirees with identical average returns over thirty years can end up in completely different places if one starts with a crash and the other doesn't, because withdrawals during the crash lock in losses at low prices, permanently shrinking the base the portfolio needs to recover from.

The second is that 4% is a historical result, not a law of physics. It holds for the sample it was calculated on, US stock and bond markets across the twentieth century. Whether it holds elsewhere, or in a world where expected returns for the next thirty years look lower than the last thirty, is an open question.

The third is that Bengen himself, in a book published in 2025, revised his own estimate upward to 4.7%, adding more asset classes to the original portfolio (small caps, micro caps, international equities). Pulling the other way, Michael Finke, Wade Pfau and David Blanchett, in the June 2013 Journal of Financial Planning (volume 26, number 6, pages 46-55), used Monte Carlo simulations calibrated to the January 2013 real yield on five-year TIPS, −1.4%, and kept the historical equity premium. On a portfolio with 50% stocks, a 4% real withdrawal fails in 57% of cases over thirty years. With a 10% failure rate accepted, the rate they obtain is 2.5%, on an allocation close to 45% stocks. Bengen's 4.7% from the 2025 book and this 2.5% answer different questions: the first widens the asset classes on the historical sample, the second starts from January 2013 bond yields.

The Trinity Study and Bengen's analysis are two different calculations. Trinity's figure of about 95% is a frequency on its sample: 4% held in that share of 30-year scenarios for portfolios with more stocks. Bengen measures the safemax, the highest rate that also held in the worst case of his sample. The bonds and the period do not match Trinity's. For the 1966 retiree, on a 50/50, Bengen's safemax is given at about 4.15%. The 4% of the rule sits close to that case, in Bengen's sample.

Long horizons: the FIRE case

Thirty years covers someone retiring at 65 just fine. Someone chasing FIRE and quitting work at 40 is looking at a potential horizon of fifty years or more, and the math shifts.

A withdrawal rate that's sustainable over thirty years isn't automatically sustainable over fifty or sixty. The longer the money has to last, the lower the starting withdrawal needs to be to absorb an unlucky sequence. The gap between the two horizons is wide: the Early Retirement Now blog, already cited in BacktestFolio's sequence-risk methodology, calculates that a 4% withdrawal on a 50/50 stock-bond portfolio holds up in about 95% of historical scenarios over thirty years, but only 65% over sixty. Hence its entire series of posts dedicated to exactly this gap between the classic 4% and realistic rates for much longer retirements.

SWR, PWR and LTWR: three different thresholds

SWR answers one specific question: what's the highest rate that, in the worst historical scenario, doesn't exhaust the capital before the chosen horizon ends? It accepts that by the end of the period, the balance might be close to zero.

The PWR, or Perpetual Withdrawal Rate, is more conservative. It's the rate that preserves the full starting capital in real terms, even in the worst-case scenario. Useful if you want to leave an inheritance, or if you never want to see your balance drop below where it started.

Sitting between the two is LTWR, the arithmetic average of SWR and PWR: neither bare survival nor full preservation, a middle ground.

A numbers example

Two US retirees, the same inflation-adjusted starting 4%, the same 60/40 allocation. One starts in 1982: the thirty years close in 2012, a finished historical window, inside a long rise in US markets.

The other starts in 2000. Those thirty years end in 2030, so an ending balance is not yet a historical figure. The early years include the dot-com bust and then 2008, when withdrawals weigh more. Sequence risk is the difference in order between these two start dates. The 2000 window is still open, so this section stops at the calendar.

In BacktestFolio

The Withdrawal Rate Analysis calculates SWR, PWR and LTWR on your own portfolio using two methods side by side. The first takes every possible starting month in your historical series, about 360 starts, one per month, when the history covers thirty years. With a thirty-year horizon only one window is fully historical. The others continue with simulated returns once the real data run out, Stationary Bootstrap by default, as the site's methodology describes. It then checks how many of those starts the capital would have survived. It's the same logic as the Trinity Study, applied to the portfolio you actually built. The second generates thousands of forward-looking scenarios with Monte Carlo simulation, using Stationary Bootstrap by default, to explore market conditions that history hasn't produced yet.

You can stretch the horizon out to 80 years, the FIRE case described above, apply a tax rate on withdrawals to see net spending power, and turn on a glidepath that gradually raises the equity share in the early retirement years specifically to cushion sequence risk. Horizon, tax rate and glidepath are parameters you set.

On a backtest you get it with Plus and Pro, applied to your portfolio: there you get both the historical path and Monte Carlo. To see it in action, the 60/40 Demo on the Backtest page already shows the result. For a guided path built around decumulation, with the 25x rule as a quick estimate and then a Monte Carlo stress test in the Planner, there's the FIRE calculator. The Planner is on Plus and Pro. On the Free plan, backtest history is capped at 10 years; Plus and Pro unlock the full available history.

Frequently asked questions

Does the 4% rule hold outside the United States?

The number comes from US market data. Countries with different market histories, ones hit by wars or currency crises that never show up in the American sample, have often produced lower historical SWRs. It's one reason it pays to test the rate against your own actual portfolio rather than apply the number from memory.