Journal · 14 November 2026 · By Kenneth McDonald · 9 min read

The 4 per cent rule at thirty-two: the number moved, and so did the question

In October 1994 an American financial planner, William Bengen, asked a question that sounds simple: how much can a retiree take out of savings each year without the money running out before the retiree does? He tested it against the American record of shares and government bonds from 1926 to 1992. He took a first-year withdrawal as a share of the starting pot, raised it every year in line with prices, and looked for the highest starting share that had never emptied a portfolio within 30 years. The answer, published in the Journal of Financial Planning, was about 4 per cent. On a pot of 500,000, that is 20,000 in the first year.

Thirty-two years later the figure appears in money books on the bestseller lists of the United States, the United Kingdom and Australia. Chapter 20 of The Money Shelf tested it against the records of all three countries and gave it the verdict "Holds with conditions". Since that chapter was written, the research has moved in two directions at once. This article follows both, and then a third line of work that asks whether the original question was the right one.

The number moved up

In 2025 Bengen published a book, A Richer Retirement: Supercharging the 4% Rule to Spend More and Enjoy More. It raised his own estimate of the safe starting rate from 4.15 per cent to 4.7 per cent, according to Morningstar Australia's report of 21 August 2025.

The mechanism is diversification. The 1994 study held two things: large American company shares and medium-term United States Treasury notes. The new work spreads the money across seven kinds of asset, adding the shares of smaller American companies and of companies outside the United States, and holds about 55 per cent in shares, 40 per cent in bonds and 5 per cent in cash. When one kind of asset falls, another often does not. Smaller falls in the bad years leave more of the pot intact for the recovery, so the pot can carry a larger withdrawal from the first year.

On 500,000 the difference is 23,500 a year against 20,000. The record behind it is still the American one.

The number moved down

The same years produced a much lower figure. In July 2025 the Journal of Pension Economics and Finance published a study by Aizhan Anarkulova, Scott Cederburg, Michael O'Doherty and Richard Sias. They used the returns of 38 developed countries, not one. Their answer for a couple aged 65, willing to accept a 5 per cent chance that the money runs out, is 2.31 per cent a year. On 500,000 that is 11,550.

The gap between 4.7 and 2.31 comes from what each study counts as the past. The United States had one of the strongest share markets of the twentieth century. A study that draws only on American years draws on a winner. A study that draws on 38 countries also includes markets that fell hard and stayed down, such as Japan's after 1989. A retiree cannot know in advance which kind of market lies ahead, so the 38-country study treats that uncertainty as part of the risk. It also models how long a couple actually lives, using tables of death rates from the United States Social Security Administration, so a couple whose members live well past 90 is part of the calculation rather than an exception to it.

The same authors reached a second finding in a 2023 working paper, Beyond the Status Quo: across those 38 countries, a portfolio held entirely in shares, half at home and half abroad, built more wealth by retirement on average than a mix of 60 per cent shares and 40 per cent bonds. Their figures, as reported by Bloomberg in December 2023, were just over 1 million against 760,000. That challenges a second piece of standard advice, that a retiree should hold more bonds as they age, and it comes with a price the study states: deeper falls along the way.

The number depends on where you live

For readers outside the United States, a new figure arrived in 2026. On 2 July 2026 Morningstar's report The State of Retirement Income UK 2026 put the highest safe starting rate for a 30-year retirement at 4.1 per cent, for a globally diversified portfolio leaning towards British bonds and cash, with a 90 per cent chance that the money lasts. That is above the 3.77 per cent that the economist Wade Pfau found for the United Kingdom in 2010 even with the best mix of shares and bonds chosen in hindsight, and it rests on different assumptions: Pfau used the longest record available and the worst starting year, while Morningstar models the years ahead from current market conditions.

Australia already answers the question by law. The Australian Taxation Office sets a minimum yearly payment from an account-based pension of 5 per cent of the balance between the ages of 65 and 74, rising to 14 per cent from 95. A retiree in Melbourne draws more than 4 per cent whether the research approves or not. The Morningstar Australia note of 7 January 2026 says plainly that the 4 per cent rule is "not a substitute for minimum drawdowns from Super".

The question moved

The third line of work asks whether retirees spend the way the rule assumes. Bengen's test raises the withdrawal with prices every year for 30 years, so a retiree spends the same in real terms at 92 as at 65.

In May 2014 the Journal of Financial Planning published a study by David Blanchett, a researcher on retirement income, that followed 591 retired American households through the Health and Retirement Study and its spending survey from 2001 to 2009. Their real spending fell by roughly 1 per cent a year on average, faster in early and late retirement, more slowly in between. He called the pattern the retirement spending smile. A household that spends less each year in real terms needs less from the pot later, which frees more of it for the early years of retirement.

Morningstar's research for 2026 tests the same idea from the other side. With fixed, inflation-linked spending it gives 3.9 per cent for a 30-year American retirement. With rules that let spending move with the markets, cutting back after a bad year and spending more after a good one, it supports starting rates of nearly 6 per cent. Flexibility does the work that a lower starting rate would otherwise have to do.

A pension paid by the state changes the question again. In the United States, each year a retiree waits beyond full retirement age to claim Social Security adds 8 per cent to the monthly benefit, until age 70, according to the Social Security Administration. A retiree who draws more heavily from savings in the early 60s and claims at 70 uses the pot as a bridge to a larger, inflation-linked income for life. The withdrawal rate in those bridge years can sit well above 4 per cent, because the pot is not being asked to last 30 years on its own. The United Kingdom State Pension and the Australian Age Pension reduce, in the same way, what any pot must supply.

What thirty-two years add to the question

The 4 per cent rule began as a test of one country's past against one fixed way of spending. Each line of work since has extended it rather than replaced it:

The table does not pick a winner. It shows that a single number was always an answer to a narrower question than the one a retiree faces. Three questions travel better than any percentage. Which country's past does this figure draw on? Does it assume spending fixed in real terms, or spending that can move? And what income will arrive from outside the pot, and when?

The official tools in each country let a retiree put those questions to their own numbers: the Internal Revenue Service's required minimum distribution tables in the United States (irs.gov/publications/p590b), MoneyHelper's guide to pension drawdown in the United Kingdom (moneyhelper.org.uk), and Moneysmart's account-based pension calculator in Australia (moneysmart.gov.au).

Bengen's 1994 question opened the field to anyone with a calculator and a record of returns. The answers have multiplied since, and each one hands the next retiree a better-drawn map of the same ground: a map to be read with one's own country, one's own spending and one's own years in view.