Safe Withdrawal Rate

Key takeaways

• The 4% rule is a historical backtest result, not a guarantee
• The order of returns matters as much as the average return
• A longer retirement requires a more conservative starting rate
• “Safe” means model-tested under stated assumptions — not guaranteed survival
• For Canadians, taxes, account type, CPP, and OAS materially change the calculation

What “safe withdrawal rate” actually means

A safe withdrawal rate, or SWR, is the highest starting withdrawal rate that would have allowed a portfolio to last for a chosen retirement period under a chosen test. Every word in that sentence carries weight.

“Starting” means the rate applies to the first year only. A 4% starting withdrawal rate on $1,000,000 produces a first-year withdrawal of $40,000. Later withdrawals may be inflation-adjusted, changed by a spending rule, or recalculated from the current portfolio value — depending on which strategy is being tested.

“Would have allowed” means the test is historical or simulated. It looks at market returns that happened in the past, or at modelled returns produced by stated assumptions. It is not a guarantee about the future.

“Portfolio” means the result depends on what the portfolio owns. A portfolio of Canadian stocks, GICs, and bonds will not behave the same way as a U.S. stock-and-bond index backtest, even when the long-run averages look similar on paper.

“Last” needs a definition. Some studies define success as ending with more than zero dollars after 30 years. That is portfolio survival, not lifestyle stability. A plan that technically survives on paper — but forces painful spending cuts along the way — may not feel safe to the retiree living through it.

“Chosen retirement period” matters because 30 years is not the same problem as 45 years. More years means more withdrawals, more inflation adjustments, and more chances for bad market sequences to arrive at the wrong time.

A safe withdrawal rate is not a universal number. It is the output of a model — and models are only as good as their assumptions.

Where the 4% rule came from

The 4% rule traces to two pieces of research published four years apart.

The first was William Bengen’s October 1994 article in the Journal of Financial Planning, titled “Determining Withdrawal Rates Using Historical Data.” Bengen, a fee-only financial planner, ran historical U.S. market data through a simple but rigorous test: start with a stock-and-bond portfolio, withdraw a fixed percentage in year one, increase that dollar amount each year for inflation, and see which withdrawal rates survived rolling 30-year periods in the data. His data ran from 1926 to 1992. The worst-case starting point he found — a retiree unlucky enough to retire in October 1968, just before a decade of poor equity returns and high inflation — required a withdrawal rate no higher than 4% to survive 30 years. Bengen called this the SAFEMAX: the highest rate that would have worked even in the worst historical scenario.

The second was the paper now known as the Trinity Study. Published in the February 1998 issue of the AAII Journal under its formal title “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable,” it was written by Philip Cooley, Carl Hubbard, and Daniel Walz — three finance professors at Trinity University in San Antonio, Texas. They built on Bengen’s framework but shifted the framing from worst-case survival to portfolio success rates across many different scenarios, withdrawal rates, and asset allocations. The Trinity Study gave the field a richer vocabulary for thinking about retirement withdrawal probability rather than just a single threshold.

What became famous, as these things tend to go, was the number — not the methodology, the assumptions, or the boundaries. “The 4% rule” lost most of its context as it traveled. The original research was not a promise that every retiree in every country could spend 4% forever. It was not a Canadian tax model. It was not a rule for 50-year early retirements. It was not a guarantee of comfort. It was a U.S. historical backtest, run on a specific dataset, with specific portfolio assumptions. Useful — but bounded.

Bengen himself has since revised his figures upward. Using a broader asset class mix, he now describes a “Universal Safemax” of approximately 4.7% — and notes that the 4% figure was always intended as the floor for the worst-case historical retiree, not the expected outcome. His historical data suggests an average sustainable withdrawal rate closer to 7%.

None of that changes the core insight. The exercise of stress-testing withdrawals against historical return sequences is exactly the right way to think about retirement spending. That 4% number was specific to Bengen’s sequences and assumptions. The same method can be applied to your own withdrawal strategy. The same number? Not necessarily.

The role of inflation

The classic safe withdrawal rate test is built around real spending, not nominal spending. Understanding that distinction matters more than most retirement discussions let on.

Nominal dollars are the dollars printed on your account statement. Real dollars are those same dollars adjusted for purchasing power. When prices rise, a retiree needs more nominal dollars to buy the same basket of goods. A retirement plan that holds nominal spending steady is quietly cutting real spending each year.

Suppose a retiree wants $40,000 of real spending power in year one. With 3% inflation every year, the nominal withdrawal path looks like this:

That year-30 withdrawal is not a raise. It is the same intended spending expressed in later nominal dollars. The retiree begins withdrawing $40,000 and eventually withdraws more than $94,000 per year — and is no better off in real terms. The portfolio has to fund that entire path, through whatever market sequence arrives.

This is one reason retirement planning can feel strange. The dollar amounts keep climbing. The purchasing power stays flat. And the portfolio has to grow fast enough — without growing too recklessly — to sustain both. Compound interest works in the portfolio’s favour during accumulation; in retirement, the same compounding works against a fixed withdrawal plan when inflation is persistently high.

Sequence-of-returns risk

Of all the concepts in retirement withdrawal planning, this one is the most consequential and the most underappreciated. The average return on a portfolio does not — by itself — determine whether a retirement plan survives. The order of those returns is critically important.

Consider two retirees who each start with $1,000,000 and each experience exactly the same three annual returns — just in opposite order. To isolate the sequence-of-returns effect, this example uses a flat $40,000 nominal withdrawal each year, without inflation adjustment. Tax is also set aside. Both simplifications are labeled here so the sequence effect is clear on its own.

Retiree A faces the bad year first:

Retiree B faces the same returns in reverse:

Both retirees withdrew exactly $120,000. Both experienced identical annual returns. Retiree B ends with $887,400. Retiree A ends with $849,600 — a difference of $37,800 after only three years, from the same set of market returns delivered in a different order.

That gap widens over time. Stretch the example across 20 or 30 years and early losses can compound into portfolio failure while the reverse sequence produces a comfortable surplus.

Why early losses are different from late losses

The mechanism behind sequence risk is straightforward once you see it. Early losses hurt because withdrawals keep going while the portfolio is smaller. Capital that has been depleted cannot participate in any subsequent recovery.

Start with a $1,000,000 portfolio, a 4% initial withdrawal, and 3% annual inflation. The first withdrawal is $40,000. If the portfolio then loses 25%, the year-end balance is:

The second-year inflation-adjusted withdrawal is:

That $41,200 is no longer 4% of the portfolio. It is now:

The plan said 4% starting withdrawal rate. After one bad year, the second withdrawal is already 5.72% of what is left. The retiree did not become reckless. The market path did it.

If the portfolio then loses another 15% after the second withdrawal:

Two bad years have turned a 4% withdrawal plan into one drawing 7.36% from a significantly smaller base — and the dollar amount is still rising with inflation. A bad decade near the start of retirement can do more lasting damage than the same bad decade near the end, precisely because early withdrawals remove capital that never gets a chance to recover.

The table below extends the scenario. The return sequence is hypothetical — constructed to reflect the kind of prolonged early damage that real market downturns can produce — but the arithmetic is exact. Withdrawals are inflation-adjusted at 3% annually throughout.

The return sequence above is hypothetical, constructed to illustrate prolonged early-loss damage. Withdrawals are inflation-adjusted at 3% annually. The portfolio is exhausted during year 18. The negative ending balance reflects the arithmetic shortfall under the model assumptions — the withdrawal exceeded what remained.

The chart below adds one more comparison. The green line shows what the same $1,000,000 portfolio would have looked like if it faced the same returns but had no withdrawals. The early losses still hurt. But without forced withdrawals removing capital during the downturn, the portfolio has more capital left to participate in the recovery.

Both portfolio lines used the same annual return sequence for years 1 through 18. The no-withdrawal portfolio falls during those early losses — down to roughly $580,000 by year 3 — but stays invested. When the recoveries arrive, there is still a full capital base to compound on. The withdrawal portfolio runs out of capital in year 18. By year 30, the no-withdrawal portfolio reaches approximately $6.2 million. The annual withdrawal amount continues rising after exhaustion, reaching approximately $94,000 per year by year 30. The 30-year simple average annual return used in the scenario is approximately 7%. Same returns, same starting amount, very different outcome — because the withdrawals in the early years removed capital that could not participate in the recoveries that followed.

The scenario uses a hypothetical return sequence, not the worst case on record. The worst three consecutive calendar years for the S&P 500 in the modern era occurred from 2000 through 2002 — declines of approximately 10%, 13%, and 23% in successive years, a cumulative loss of roughly 40%. For an investor still accumulating, those years were painful but ultimately recoverable. For a retiree already withdrawing inflation-adjusted spending from the portfolio, the arithmetic was different. Withdrawals continued while the portfolio shrank, permanently removing capital that could not participate in the recovery that followed. The market came back. A withdrawal portfolio subjected to that sequence, depending on its size and withdrawal rate, may not have.

That contrast — recoverable for accumulators, potentially not for retirees — is the practical definition of sequence-of-returns risk. The intuition behind it is simple: when a retiree sells shares during a downturn to fund withdrawals, those shares are gone. They do not participate in the recovery. The portfolio that rebounds is a smaller one — and it is carrying the same withdrawal obligation as before.

Time horizon: 30 years is not 50 years

The 4% rule is usually discussed in the context of a 30-year retirement — roughly the span from the mid-60s to age 95. For many Canadians, that is a reasonable planning horizon. For anyone considering early retirement, it is not.

A longer retirement simply requires more real withdrawals before investment returns are considered. At $40,000 per year in constant purchasing power:

• 30-year retirement: 30 × $40,000 = $1,200,000 in real withdrawals
• 40-year retirement: 40 × $40,000 = $1,600,000 in real withdrawals
• 50-year retirement: 50 × $40,000 = $2,000,000 in real withdrawals

These figures ignore investment returns because they are expressed in year-one purchasing power — they simply show the scale of the commitment. A 50-year retirement requires 67% more real withdrawals than a 30-year one, and extends the window in which bad sequences, persistent inflation, policy changes, and personal spending shocks can materialize.

A historical backtest that covers 30-year periods cannot be automatically extended to 50 years by confidence or optimism. The arithmetic changes. The risks compound. The safe withdrawal rate that would have survived all historical 30-year windows may not survive all historical 40- or 50-year windows — and the historical record has fewer of those longer windows to test against in the first place.

Asset allocation changes the result

The withdrawal rate that survives depends heavily on what the portfolio owns. This is not a trivial observation. Two portfolios with identical long-run average returns can produce dramatically different outcomes for a retiree, depending on how their returns are distributed across time.

Stocks can provide long-term growth — but can also fall 30% or 40% in a bear market, right when withdrawals are pulling capital out. Bonds can reduce volatility and provide income — but their returns depend on starting yields, credit quality, duration, and the path of interest rates. Cash can reduce the pressure to sell equities during downturns — but it may slowly lose purchasing power if its returns fall below inflation. For a broader look at how these asset classes behave, the asset classes article covers the mechanics in more depth.

A portfolio with too little growth may struggle to fund decades of inflation-adjusted withdrawals. A portfolio with too much volatility may be vulnerable to the sequence risk described above. Historical withdrawal studies are trying to measure exactly this tension — not “what return did this portfolio earn on average?” but “what happened when withdrawals collided with this specific return path?”

A portfolio that earns a smooth 6% per year looks entirely different to a retiree than one that averages 6% through a crash, a recovery, a boom, and a stagnant decade. Retirees do not spend averages. They spend dollars from the portfolio that exists in front of them at any given moment.

Bonds and the interest-rate environment

Stock sequence risk gets most of the attention. Bond returns deserve some too.

A bond fund’s future return is heavily influenced by its starting yield. When that yield is high, future income is higher and the return cushion is more substantial. When the yield is low, the cushion is thinner. When interest rates rise, existing bond prices generally fall — so the transition period can hurt current portfolio value even as higher future yields improve the long-run outlook.

The historical periods used in Bengen’s and the Trinity Study’s backtests included long stretches of falling interest rates, particularly from the early 1980s onward. Falling rates mechanically boost bond prices, which inflated the bond returns in those historical datasets. A retiree entering a period of rising or already-elevated rates faces a different bond return environment than the historical average implies.

This is not a forecast. It is a reminder that historical safe withdrawal research reflects the bond return environment that existed during the tested periods — and that environment is not guaranteed to repeat. Treating the bond side of a historical backtest as timeless is one of the quieter sources of false confidence in retirement withdrawal planning.

Fixed versus flexible withdrawals

The classic 4% rule uses fixed real withdrawals: choose the first withdrawal, then adjust it for inflation each year regardless of what markets do. That method is easy to model, and it protects real spending power if the portfolio can sustain it. But it also means the withdrawal does not automatically fall when the portfolio falls — which puts more stress on the portfolio after declines.

Flexible withdrawal strategies change spending based on portfolio performance. They may withdraw a fixed percentage of the current portfolio value each year, use guardrails that trigger spending cuts after large declines, skip inflation increases after bad years, or blend a floor-and-ceiling approach. The arithmetic is different — and the trade-off is explicit.

Suppose the portfolio has fallen to $750,000 after a bad year. The inflation-adjusted fixed withdrawal would be $41,200 (assuming 3% inflation on the original $40,000). A flexible 4%-of-current-portfolio approach would instead produce:

The portfolio math is easier. But the retiree’s spending math is harder. That $30,000 withdrawal, expressed in year-one purchasing power (deflated by one year of 3% inflation), is:

Compared with the original $40,000 target, that is a real spending cut of about 27%. The spreadsheet looks safer. The retiree’s standard of living took a significant hit.

This is the difference between portfolio survival and lifestyle stability. A flexible strategy can improve the portfolio’s odds by transferring more of the volatility to the retiree’s spending. Whether that transfer is workable depends on whether the spending is actually flexible — which is a household question, not a mathematical modelling question. If every dollar of spending is essential, the arithmetic flexibility is not real flexibility.

How to think about spending cuts during downturns

A spending cut after a market decline is not a moral judgment. It is a mechanical change in the withdrawal path — and understanding the arithmetic of it clarifies why it sometimes helps and what its limits are.

Suppose the inflation-adjusted withdrawal would have been $41,200, but a retiree cuts spending to $35,000 for the year. The cash left invested is:

If that $6,200 earns 10% over the year, it becomes $6,820. If the portfolio instead loses another 10%, it becomes $5,580. The spending cut does not guarantee anything. It reduces the amount of capital sold or withdrawn during a vulnerable period — which matters because sequence risk is partly about how much capital remains available to participate in any eventual recovery.

A plan with genuine spending flexibility has more tools to respond than one committed to identical real withdrawals no matter what markets do. But the flexibility has to be real. For some retirees it is — they have discretionary travel and entertainment budgets that can be trimmed without touching essential expenses. For others, nearly every dollar is load-bearing, and the spreadsheet flexibility is theoretical.

“Safe” is a modelling word

In safe withdrawal research, “safe” has a narrower meaning than everyday English suggests. It typically means “historically sustainable under the tested conditions.” That is a much more specific claim than safe implies to someone reading a financial planning article.

If a 4% withdrawal rate survived every tested 30-year historical window in a given U.S. dataset, that does not mean a 4% withdrawal rate must survive the next 30 years. It means the tested historical periods did not break it under the specific portfolio assumptions used. The future can differ from those periods in many ways:

Inflation can be higher or more persistent. Stock valuations can start at higher levels, compressing future returns. Bond yields can start lower. Taxes, fees, benefit rules, account structures, currencies, and lifespans can all differ from the tested scenario — and usually do.

Safe is a modelling word. It is not a guarantee of portfolio survival.

Canadian retirement: the portfolio is one piece

For Canadian retirees, the investment portfolio is only one part of the retirement income picture — and in many cases not the dominant one.

The Canada Pension Plan and Old Age Security provide income streams that do not require selling portfolio assets. CPP is calculated based on lifetime contributions and the age at which payments begin, with higher monthly amounts for those who delay past age 65 up to age 70. OAS begins at 65 for most Canadians, with the option to defer up to age 70 for a larger monthly benefit. Both programs are indexed to inflation, which is significant — a guaranteed, inflation-adjusted income stream reduces the real burden placed on the investment portfolio to fund living expenses. When CPP and OAS collectively cover a meaningful share of spending needs, the required portfolio withdrawal rate falls, and sequence-of-returns risk is correspondingly reduced.

The account structure of Canadian retirement savings adds further complexity. RRSP and RRIF withdrawals are generally included in taxable income in the year they are taken. RRIFs carry required minimum withdrawals based on age — the government mandates that a minimum percentage of the account be withdrawn each year, regardless of whether the retiree wants or needs that income. Those required minimums can force taxable withdrawals that push income into higher brackets or trigger benefit clawbacks, even when the retiree would prefer to leave the money invested. Strategic RRSP drawdown before those minimums kick in — sometimes called an RRSP meltdown — is one tool for managing that problem. TFSA withdrawals, by contrast, are generally not included in taxable income and do not affect income-tested benefits.

Non-registered taxable accounts add another layer: interest income is taxed at the full marginal rate, eligible Canadian dividends receive a dividend tax credit, and capital gains are taxed on a fraction of the gain. The adjusted cost base of investments tracks the original purchase price for capital gains calculations — and foreign dividends may face withholding taxes before even reaching the Canadian account. Asset location decisions (which investments to hold in which account type) can meaningfully affect after-tax cash flow.

The practical implication: a 4% pre-tax portfolio withdrawal is not the same thing as 4% after-tax spending. The useful version of Canadian retirement withdrawal arithmetic separates account types, taxes, government benefits, inflation, required withdrawals, and spending flexibility, then shows the after-tax cash flow that actually reaches the retiree’s bank account. The portfolio withdrawal rate is one input into that calculation, not a summary of it. The retirement withdrawal calculator makes these inputs inspectable.

Valuation and the danger of false precision

Historical safe withdrawal rates are sensitive to starting conditions — specifically, to the valuation of stocks and the yield on bonds at the beginning of retirement.

When stocks begin retirement at high valuations relative to earnings, future stock returns have historically tended to be lower than the long-run average. Historical data suggests that retirees who began drawing down portfolios at high valuation levels experienced lower sustainable withdrawal rates than those who retired into more modestly valued markets. When bonds begin at low yields, the income they generate is lower. Neither relationship is precise enough to produce a guaranteed future withdrawal rate, but both are consistent enough to warn against assuming the next 30 years will look like the average of all past 30-year periods.

This is one reason false precision matters. A withdrawal model may output 3.82% or 4.17% to two decimal places — but that precision implies more knowledge than the model actually has. The retiree does not get to draw their future from the exact dataset, fee schedule, tax code, inflation path, valuation level, and lifespan assumed in the spreadsheet. They get to live through one sequence of events, uncertain in advance. A withdrawal rate is a planning input, not a measurement. It should be treated with the appropriate humility — as a useful starting point for a plan that will need to be revisited, not as a precise answer to an inherently imprecise question.

Historical backtests versus Monte Carlo simulations

There are two common methods for testing withdrawal strategies, and both have genuine strengths and real limitations.

A historical backtest uses actual historical return sequences — the real market paths that happened, including all the messy correlations between stocks, bonds, inflation, crashes, recoveries, and long stagnant stretches. Its strength is realism: the returns happened, and their relationships to each other were real. Its limitation is sample size. There are only a finite number of non-overlapping 30-year periods in modern market history — and most well-known datasets are U.S.-heavy, which matters for Canadian and other non-U.S. readers.

A Monte Carlo simulation generates a large number of possible return paths by drawing randomly from assumptions about returns, volatility, inflation, and correlations. Its strength is breadth: it can test tens of thousands of possible futures instead of waiting for history to supply them. Its limitation is assumption risk. The precision of a Monte Carlo output is entirely dependent on the quality of its inputs. If the assumed return distribution does not accurately capture the real world’s fat tails, regime shifts, or correlation breakdowns, the probability estimates it produces can be misleading — and the additional decimal places can look more like knowledge than they are.

Neither method eliminates uncertainty. A historical backtest asks: “What would have happened under these past sequences?” A Monte Carlo simulation asks: “What might happen if this statistical model is a reasonable description of future returns?” A careful plan can draw on both — without treating either as a window into the future.

Five common misunderstandings of the 4% rule

The first misunderstanding is that 4% means withdrawing 4% of the current portfolio every year. In the classic rule, it does not. It means withdrawing 4% of the starting portfolio value in year one, then adjusting that fixed dollar amount for inflation in subsequent years. The percentage of the current portfolio being withdrawn changes — sometimes substantially — as markets move.

The second misunderstanding is that the rule is guaranteed. It is not. It is a historical backtest result under specific assumptions about U.S. stock and bond returns, a specific time period, and a specific portfolio structure. Future retirees do not get to draw from that exact dataset.

The third misunderstanding is that “success” means comfort. A portfolio can survive a 30-year test on paper while the retiree living through it experiences large account declines, forced spending cuts, tax surprises, or benefit clawbacks. Portfolio survival and lifestyle stability are related outcomes, but they are not the same thing.

The fourth misunderstanding is that 4% proves someone has “enough.” It only tests one withdrawal strategy against one set of assumptions. Whether a specific household has enough depends on spending requirements, taxes, government benefits, time horizon, account mix, inflation, investment costs, flexibility, and the emotional capacity to stay with the plan through difficult markets.

The fifth misunderstanding is that the average historical return settles the question. Sequence risk is the reason it does not. Two retirees who both earn the same average return over 30 years can end up in completely different financial positions depending on when in that sequence the good and bad years arrived.

The real question is not one number

The safe withdrawal rate is useful because it forces the important variables into the open. It is a framework for asking the right questions, not a formula that produces the right answer.

The right questions are: How large is the portfolio? How much spending must it support, in real after-tax terms? Is the withdrawal inflation-adjusted or nominal? Over what time horizon? With what asset allocation and what fees? What income arrives from CPP, OAS, pensions, or other sources that does not require selling assets? How much spending can actually change after a bad decade? What happens to the plan if the first ten years are poor?

A retirement withdrawal calculator that makes these questions inspectable — showing year-by-year cash flows including opening balance, withdrawals, inflation adjustments, government benefits, required RRIF minimums, taxes, investment returns, and ending balance — does more useful work than any single percentage. The goal is not to find a magic number. It is to show how the plan behaves when assumptions change, so a retiree can understand which variables matter most in their specific situation and make informed decisions accordingly.

Conclusion: Withdrawal planning is probability management

A safe withdrawal rate is not a permission slip. It is a stress test.

The 4% rule became useful because it gave retirees a concrete starting point: take an initial percentage of the portfolio, adjust the dollar withdrawal for inflation, and test whether the portfolio survived historical market sequences. That clarity helped people have more grounded retirement conversations than the purely intuitive guesses that preceded it. The rule earned its popularity.

But the simplicity that made it useful also became the source of its most common misuses. The retirement problem is harder than one number because the risks do not arrive evenly across time. Early losses matter more than late losses. Inflation compounds. Longer retirements require substantially more real withdrawals. Asset allocation changes both growth and volatility in ways that interact with the sequence of returns. Bond returns are not independent of the interest-rate environment in which they occur. Canadian taxes, government benefits, account rules, and currency exposure add layers that a U.S. historical backtest was never designed to capture.

“Safe” in safe withdrawal rate means historically or model-tested under the stated assumptions. It does not mean guaranteed. The future is lived one year at a time, through whatever sequence of market returns, inflation, and personal circumstances actually arrives.

What the arithmetic can do is make the assumptions visible, the risks nameable, and the trade-offs inspectable. That is the work worth doing.

Just arithmetic.

Frequently asked questions