Safe Withdrawal Rate and Withdrawal Strategies
Quick answer: "Safe withdrawal rate" is a name for the highest starting withdrawal that passed a particular test, on particular data, over a particular number of years. Change the test, the data or the horizon and the number changes with it, which is why published figures disagree. A strategy for adjusting the withdrawal along the way matters as much as the starting rate. In one simulated example on this page, three strategies that all began at the same $40,000 a year behaved very differently after that.
To try a starting rate of your own, put your numbers into the Monte Carlo simulation retirement calculator and compare a few spending levels side by side. The free Monte Carlo retirement calculator keeps the assumptions in view beside every result, which makes comparisons easier to read.
Education, not financial advice. This page explains a method and shows what a model produced from stated assumptions. It does not tell you what to buy, sell, save or spend. For your own situation, ask a qualified professional.
A name, not a verdict
The phrase sounds like a promise, and it is not one. It does not say that a rate is free of danger. It is shorthand that researchers use for "the largest first-year withdrawal that kept the money going under the conditions we tested." Every part of that sentence is a choice.
There are three choices in every version of the number. The first is the data: which years of returns, from which country, with which asset classes. The second is the horizon: thirty years, forty, to age 95, forever. The third is the pass mark: what counts as having worked. One study may call a rate acceptable if no past start year ran out of money. Another may count the share of start years that did. A third may run thousands of random sequences and ask for a certain probability.
Put two numbers from two studies next to each other and you are usually comparing two sets of choices as much as two rates. The older and better-known figure, the 4% rule, came from the first approach and the second. The page on it goes through what each study said. Here the question is wider: once you accept that a starting rate is only part of the picture, what else determines how a retirement goes?
What the three best-known studies each say
Everything in this section is attributed to the documents we read on 2026-11-04, and each is labelled primary or secondary. Primary means we read the paper's own text. Secondary means we are repeating how another document summarized it.
Bengen (primary, FPA's 2004 reprint of the October 1994 article). The test was historical. In his method, year one's dollar amount is a share of the opening portfolio, and every later year repeats the previous amount with an inflation adjustment. For a 50/50 stock and bond portfolio, a 4% first-year withdrawal was not exhausted before 33 years in any past case. He also notes that 4.25 percent could exhaust a portfolio in as little as 28 years. For a client aged 60 to 65, the rule he offers in practice is "about 4 percent".
Trinity study (primary text, read through a web reading tool). The pass mark here is a share of past payout periods that a portfolio supported. It tested payout periods from 15 to 30 years and annual withdrawal rates from 3 to 12 percent. One secondary source reports that a 50/50 portfolio at 4 percent survived 95 percent of the 30-year periods between 1926 and 1995. We print no other 30-year cell from the study because its main table could not be read.
Guyton and Klinger (primary, March 2006). This paper uses a different test. It ran Monte Carlo trials, 14,000 per scenario, on 1973 to 2004 and 1928 to 2004 data, and tested portfolios with 50, 65 and 80 percent in equities. It concludes that initial rates of 5.2 to 5.6 percent are sustainable over 40 years at a 99 percent confidence standard for portfolios with at least 65 percent equities, rising to 5.7 to 6.2 percent at a 95 percent standard. With 50 percent in equities, the maximum initial rate drops to as low as 4.6 percent.
Those last numbers look much higher than the earlier ones. They are not a disagreement. The Guyton and Klinger rates assume the retiree follows decision rules that change the withdrawal along the way, they use a 40-year horizon, and they use a stated confidence standard. They are an answer to a different question. The paper itself says its approach has limits, including that the historical 1973 start is only one period and that choosing how high a probability of success is high enough cannot be avoided. The paper's introduction also describes the two earlier studies, which makes it a secondary description of each: a range of 4.1 to 4.58 percent for Bengen, and a 4.0 percent starting rate that was nearly always successful for 30-year periods between 1926 and 1995 for the Trinity work.
One simulation, many starting rates
The calculator on this site can ask the same question for any starting rate. The table below uses a $1,000,000 balance in a 60/40 mix of stocks and bonds, with the withdrawal held flat in today's dollars. It is an example, and the numbers come from the default assumptions, 1971 to 2025 real returns.
| Starting rate | 20 years | 25 years | 30 years | 35 years | 40 years |
|---|---|---|---|---|---|
| 3% ($30,000 a year) | 100% | 100% | 99% | 99% | 98% |
| 3.5% ($35,000 a year) | 100% | 99% | 98% | 96% | 94% |
| 4% ($40,000 a year) | 100% | 98% | 95% | 91% | 87% |
| 4.5% ($45,000 a year) | 99% | 95% | 89% | 83% | 78% |
| 5% ($50,000 a year) | 96% | 89% | 80% | 73% | 67% |
| 6% ($60,000 a year) | 86% | 71% | 59% | 50% | 44% |
In this example, two patterns stand out. First, the score drops slowly while the rate is low and faster as it rises. For 30 years, a 3.0 percent rate ($30,000 a year) lasted in 99% of the simulated paths, 4.0 percent ($40,000) in 95%, 5.0 percent ($50,000) in 80% and 6.0 percent ($60,000) in 59%. Second, the horizon does just as much work. At 5.0 percent, the plan lasted in 96% of paths over 20 years and 67% over 40.
If a "safe withdrawal rate" in this table were defined as the highest rate whose 30-year score stays at or above 90 percent as a hypothetical pass mark, it would land between two of the rows, and choosing 85 or 95 percent as the pass mark would move it again. That is the whole problem in one picture: the number depends on the mark you draw.
What this number leaves out. It comes from a model with the assumptions shown, in annual steps and after-inflation returns. It does not model taxes, health or long-term care costs, inflation that differs from the model, or how you would act in a bad year. See what a Monte Carlo simulation cannot see.
Three ways to take the money out
So far every withdrawal has been fixed: set a dollar amount, raise it for inflation, and keep going whatever the market does. That is one strategy among several. Here are three, tried on the same simulated paths.
Fixed. The first-year amount, adjusted only for inflation. It is the approach behind the older studies. Spending is perfectly predictable and the portfolio carries all the risk.
Percentage of the balance. Each year, take a set share of whatever the portfolio is worth. Spending now moves with the market, and the portfolio can never hit zero through withdrawals alone, since a share of a positive number is positive. The risk shifts to spending.
Guardrails. Start with a dollar amount, but check it each year against the current balance. If the amount has become too large a share of the portfolio, cut it. If it has become a small share, raise it. Guyton and Klinger's paper restates three earlier rules: spend cash and rebalance excess first and do not sell equities after a down year, raise withdrawals by inflation but capped at 6 percent, and skip the raise after a year with a negative portfolio return. Its capital preservation rule cuts the withdrawal by 10 percent when the current withdrawal rate has risen more than 20 percent above the initial rate, and its prosperity rule raises it by 10 percent when the rate is more than 20 percent below the initial rate.
The simulated example
The table below is a static example, computed once for this page, and it is not a feature of the calculator. The setup: a $1,000,000 portfolio, a 60/40 mix, 30 years, and a first-year withdrawal of $40,000 in today's dollars for all three strategies, run on the default return assumptions with 5,000 simulated paths. The percentage strategy takes 4 percent of the balance each year. The guardrail version is a simplified one written for this page: it uses the 20 percent bands and 10 percent steps from the paper's capital preservation and prosperity rules, but leaves out the other rules and the expiry. Because every figure is in today's dollars, the inflation-raise rule has nothing to do here. It is a sketch of the idea, not a faithful rebuild of the published method.
| Rule | Money lasted | Median spending, year 15 | Median spending, year 30 | 10th percentile, year 30 | Paths with a cut |
|---|---|---|---|---|---|
| Fixed amount, adjusted for inflation | 95% | $40,000 | $40,000 | $40,000 | 0% |
| A set percentage of the balance each year | 100% | $47,000 | $56,000 | $27,000 | not applicable |
| Guardrails (simplified) | 100% | $44,000 | $56,000 | $26,000 | 68% |
Here is what the table shows.
- For example, the fixed strategy left money in about 95% of paths, the same figure as the 4.0 percent, 30-year row above. Spending was $40,000 in year 15 and $40,000 in year 30 by construction, and never moved.
- The percentage strategy lasted in 100% of paths, but that result is nearly automatic: a share of the balance cannot take the balance to zero. The cost shows up in spending. The median path spent $47,000 in year 15 and $56,000 in year 30, but in one path in ten, spending in year 30 was $27,000 or lower, and in one path in ten, spending dipped to $21,000 or lower at some point.
- The guardrail strategy also finished with money in 100% of paths. Its median spending was $44,000 in year 15 and $56,000 in year 30. In one path in ten, year-30 spending was $26,000 or lower, and in one in ten, spending dipped to $24,000 or lower at some point. In 68% of paths, the rule triggered at least one cut.
Notice the exchange the example is describing. The fixed strategy gives the steadiest spending and the lowest score. The other two give scores near 100 percent, and they get there by letting spending move, sometimes well below the starting figure. Which kind of risk is easier to live with is a personal matter, and nothing in a simulation can settle it.
Notice also what the median hides. In the middle path, both flexible strategies ended up spending more than the fixed one, because in the middle path the portfolio grew. That is the other side of the same flexibility: a rule that follows the balance up also follows it down.
What this number leaves out. It comes from a model with the assumptions shown, in annual steps and after-inflation returns. It does not model taxes, health or long-term care costs, inflation that differs from the model, or how you would act in a bad year. See what a Monte Carlo simulation cannot see.
Why the order of returns shows up here
Flexible rules are largely a response to the problem described on the sequence of returns risk page. A fixed withdrawal in a falling market takes a growing share of a shrinking pot, and a rule that trims the withdrawal in those years slows the drain. The guardrail row of the table is the clearest illustration: most of its cuts would come in the paths that open badly, which is why a majority of paths show at least one.
A related approach is to hold a cash reserve so that withdrawals do not come from stocks after a fall. That is a different mechanism, discussed on the bucket strategy page. It changes where the money comes from, whereas a guardrail changes how much comes out.
Questions to ask any withdrawal figure
When you meet a "safe withdrawal rate" in an article or from a tool, a few questions make it much easier to compare with anything else.
- What horizon did it assume, and does it match yours?
- Is the figure for a fixed, inflation-adjusted withdrawal, or does it assume a strategy that flexes?
- What counted as success: never running out in any past start year, a share of start years, or a simulated probability?
- Which years of data and which country does it draw on, and how many independent periods does that really contain?
- Does it include taxes, fees on the investments and the other income you might have?
For the horizon question, the how long your money will last in retirement page runs the same plan to different ages. For the pass-mark question, the page on what a good Monte Carlo success rate means shows why two reasonable people can look at the same score and read it differently. And the model's blind spots cover the items on the fifth question that no withdrawal study here includes.
What this can't tell you
This page can't tell you what starting rate or strategy would fit your life. It cannot tell you which pass mark is the right one, since that is a judgment about how much risk of running short, and how much risk of lower spending, is tolerable. The strategy table is a single run of a simplified model: one dataset, annual steps, real returns, no taxes, no costs and no change in the mix. The studies we cite are decades old, and we did not look for later work that may have revised them, so a figure here may have been updated elsewhere. We also did not read the earlier Guyton paper that the 2006 paper builds on.
Questions readers ask
Is there one safe withdrawal rate for everyone?
No. The phrase names the highest starting rate that passed a stated test on stated data over a stated horizon. Change any of those and the figure changes. That is why studies with different tests report different numbers without contradicting each other.
Why do flexible strategies show higher starting rates in the research?
Because the retiree agrees to cut spending in poor years, the portfolio can afford a larger first year. The Guyton and Klinger paper reports initial rates above the older fixed-rule figures for that reason, along with a longer horizon and a stated confidence standard. The trade is lower spending in some paths.
Does a percentage-of-balance rule mean I cannot run out?
Within the arithmetic, the balance cannot reach zero from withdrawals alone, since each year takes only a share of what is left. In the example above, it showed full success. Spending can fall far enough to matter, though, so the question shifts from running out to how low spending can go.
Where can I see these strategies in the calculator?
You cannot. The strategies table on this page is a static example computed separately. The calculator on the home page models a fixed, inflation-adjusted spending level, and you can compare starting amounts there, but it does not apply guardrails or percentage rules.