The 4% rule assumes a constant real withdrawal. Markets do not. On a 40- to 50-year FIRE horizon that mismatch is the whole problem: a fixed rate that survived every 30-year U.S. cohort can be too tight in most paths and still too brittle in the worst one.
Variable strategies do not make 4% dead. They treat Bengen’s 1994 result as a worst-case starting rate for a 30-year, inflation-adjusted withdrawal from a balanced portfolio. Once spending can move, you can start higher (Guyton–Klinger), spend the upside without cutting the floor (Kitces ratchet), or never deplete by construction (Bogleheads VPW). Those are different claims. Do not mix their “success” numbers.
No 10–20% success-rate lift is cited: that figure is not in the papers. Unknown cells stay unknown.
What the fixed 4% rule actually measured
Bengen, 1994
William P. Bengen, “Determining Withdrawal Rates Using Historical Data,” Journal of Financial Planning (October 1994), used Ibbotson year-by-year stock, intermediate Treasury, and inflation data from 1926. He measured portfolio longevity — years until a first-year withdrawal, then inflation-adjusted, hit zero — not average returns.
On a 50/50 mix:
- 3% to about 3.5% never produced a life under 50 years.
- 4% never exhausted the portfolio in fewer than 33 years; most cohorts lasted 50 years or more.
- 4.25% could exhaust in as little as 28 years if history repeated.
- He called 5% “risky” and 6% or more “gambling” for a new retiree.
Fewer than 50% stocks shortened the worst-case life more than extra stocks did. He recommended 50–75% equities at the start. After year one, the withdrawal is last year’s dollars plus inflation. The 4% figure is not reapplied to the new balance. That is a 30-year worst-case floor, not a law of spending.
Trinity Study, 1998
Cooley, Hubbard, and Walz, “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable,” AAII Journal (February 1998), counted portfolio success rates: the share of overlapping 1926–1995 windows in which the portfolio did not hit zero.
Inflation-adjusted withdrawals, 30-year windows (Table 3; 41 overlapping periods):
| Allocation | 3% | 4% | 5% | 6% |
|---|---|---|---|---|
| 100% stocks | 100% | 95% | 85% | 68% |
| 75/25 | 100% | 98% | 83% | 68% |
| 50/50 | 100% | 95% | 76% | 51% |
| 25/75 | 100% | 71% | 27% | 20% |
| 100% bonds | 80% | 20% | 17% | 12% |
A 95% success rate means “failed in 2 of 41 U.S. cohorts,” not a 95% probability that your retirement works. The authors wrote that 3–4% inflation-adjusted rates produced “high portfolio success rates for stock-dominated portfolios,” and that mid-course corrections would be required.
Why 30-year 4% is the wrong single number for FIRE
Bengen’s 1996 follow-up, “Asset Allocation for a Lifetime,” extended the horizon. As summarized by Kitces, the historical safe initial rate fell from 4.1% at 30 years to 3.5% at 45 years, and rose to 5.1% at 20 years. Pfau (2012) later estimated 3.3% at a 95% Monte Carlo confidence level for 40 years.
The other half of a 3.5% / 50-year / 60/40 path (Kitces): the worst U.S. sequence is barely enough; about 90% of those paths finished above 3× starting principal, and the median finished above 9.3×. A fixed 3.5% protects the left tail and leaves most of the distribution unspent.
Sequence of returns is the failure mode. A bad first decade plus inflation (Bengen’s 1973–74 “Big Bang”) is what kills a constant real withdrawal. Longevity just keeps the clock running after that decade.
Four variable strategies
1. Guardrails (Guyton–Klinger)
Guyton and Klinger, “Decision Rules and Maximum Initial Withdrawal Rates,” Journal of Financial Planning (March 2006), tested Guyton’s 2004 rules with Monte Carlo returns (1928–2004 and 1973–2004) over 40 years.
- Capital preservation. If the current withdrawal rate rises more than 20% above the initial rate, cut this year’s dollar spending by 10%.
- Prosperity. If the current rate falls more than 20% below the initial rate, raise dollar spending by 10%.
A 5.0% start therefore cuts above 6.0% and raises below 4.0%. Guyton’s 2004 tweak — skip the CPI raise in any year the portfolio’s return is not positive — is a small permanent cut that helps a lot in bad sequences.
For portfolios with at least 65% equities, the 2006 paper concluded:
- 5.2–5.6% initial withdrawal was sustainable over 40 years at a 99% Monte Carlo confidence standard.
- That range rose to 5.7–6.2% at 95% confidence.
- At 50% equities, the maximum initial rate dropped as low as 4.6%.
Those are simulated confidence levels under the rules, not Trinity rolling-period success rates. Kitces has warned that repeated 10% cuts can stack in a long bad sequence. The higher start is the reward for agreeing, in writing, to take them.
On $1.2 million at 5.2%: year one is $62,400. The cut trigger is a current rate of 6.24% (portfolio ≈ $1.00 million). The raise trigger is 4.16% (portfolio ≈ $1.50 million).
2. Variable Percentage Withdrawal (Bogleheads VPW)
VPW is a Bogleheads method, not a journal paper. Each year you withdraw a percentage of the current portfolio, looked up by age (or the younger spouse’s age) and planned equity share. The percentage rises as the horizon shortens. A smaller balance produces a smaller check, so the portfolio is not depleted before the table horizon.
Official table values from the Bogleheads VPW wiki:
| Age | Years left | 50% stocks | 60% stocks | 70% stocks |
|---|---|---|---|---|
| 40 | 60 | 3.8% | 4.1% | 4.3% |
| 45 | 55 | 3.9% | 4.2% | 4.4% |
| 50 | 50 | 4.1% | 4.3% | 4.5% |
| 55 | 45 | 4.3% | 4.5% | 4.7% |
| 65 | 35 | 4.8% | 5.0% | 5.2% |
On $1.2 million, age 40, 60/40: year one is 4.1% × $1.2 million = $49,200. After a 25% drop to $900,000, age-41 / 60% is still 4.1%, so the check falls to about $36,900. There is no Trinity-style success rate. The question is whether you can live on a paycheck that can fall that far.
Use the VPW Accumulation and Retirement Worksheet if a pension or Social Security is coming; the table ignores those cash flows. The wiki suggests capping the withdrawal percentage (no more than 10%) so late-age cells do not empty the account in one year.
3. Buckets / cash-flow matching
A common setup holds 2–5 years of planned spending in cash or short bonds and the rest in the growth portfolio. Down year: spend from cash. Up year: refill from equities.
There is no Trinity-style success-rate table for buckets as a withdrawal rule. If the long-run mix equals a total-return portfolio, the math is mostly a cash sleeve plus rebalancing. The usual argument is behavioral — you are less likely to sell equities in a crash — not a higher safe rate.
The cost is cash drag. Five years of spending at 4% of $1.2 million is $240,000 sitting out of equities during a recovery. Treat buckets as a wrapper around one of the other three rules, not as a fourth engine.
4. Floor plus upside (Kitces ratchet and spending split)
Kitces ratchet (2015). Keep a Bengen-style floor. Raise real spending 10% only when the portfolio is more than 50% above its start, and at most once every three years. On U.S. 60/40 history, the 30-year initial rate that would have worked ranged from 4% to 10%, median near 6.5%. More than two-thirds of 30-year 4% paths finished with more than double starting wealth; median leftover was about 2.8×. In the bad sequences (1966 is the usual example) the 50% threshold never hit, so spending stayed at the floor. Kitces calls this weakly dominant: historically never worse than fixed 4%, usually better, and designed so the floor does not move down.
Spending split. Fund essentials at a conservative rate (Kitces uses 3.5% for a 40–50-year FIRE horizon). Let travel and upgrades float with the portfolio and with post-FI earnings you have actually earned. His capitalization example: $20,000 a year of side income, at 4%, is $500,000 less portfolio — if that income persists. Do not underwrite the floor with income you have not earned.
Social Security is the cleanest later-life floor. For birth year 1943 or later, SSA delayed retirement credits raise the benefit 8% per year from full retirement age to 70. Model it as a future floor; let VPW or guardrails cover the gap years.
On $1.2 million, 3.5% floor: $42,000 for essentials. First 10% ratchet is $46,200, and it waits until the portfolio is above $1.8 million.
Comparison (sourced ranges only)
| Fixed 4% (inflation-adjusted) | Guardrails (Guyton–Klinger) | VPW | Buckets | Floor + upside (Kitces ratchet) | |
|---|---|---|---|---|---|
| What the source measured | Trinity 1998: 95–98% of 30-year U.S. windows survived at 4% for 50–100% stocks. Bengen 1994: no 50/50 path died before year 33. | 5.2–5.6% start, ≥65% equities, 40 years, 99% Monte Carlo confidence (2006). | No depletion before the table horizon, by construction. | Unknown. No published success-rate table comparable to Trinity. | Historically never worse than fixed 4% on Kitces’s 60/40 paths; usually higher lifetime spending. |
| Year-1 rate, long horizon | 4% of start (3.5% if you adopt Bengen 1996 / Kitces for 45 years) | 5.2–5.6% of start if you accept cuts | Age 40, 60% stocks: 4.1% of current | Same as the rule you wrap | 3.5–4.0% of start as the locked floor |
| Dollar path | Constant real | Constant real until a ±20% rate breach, then ±10% | Floats with the portfolio every year | Smooth if cash sleeve is full | Floor never cut; +10% only after +50% wealth, ≤1 / 3 years |
| Forced cut? | No (failure is depletion) | Yes — that is the mechanism | Implicit every down year | Only if you refuse to refill | Designed not to cut the floor |
| Complexity | Low | Medium | Medium (worksheet if pensions) | Medium | Medium |
| Best for | A simple floor you will not override | Higher starting income and a 10% cut you will actually take | A variable paycheck and a terminal date | Investors who would otherwise panic-sell | A locked essential floor; wait for upside |
Do not write “95%+” in the guardrails or VPW cells. Those figures are not in the sources.
One portfolio, four first-year checks
$1.2 million, age 40, 60/40, no pension yet.
| Rule | Year-1 withdrawal | After a 25% drawdown (portfolio $900,000) |
|---|---|---|
| Fixed 4% of start | $48,000 | Still $48,000 plus inflation (now 5.3% of what is left) |
| Fixed 3.5% FIRE floor | $42,000 | Still $42,000 plus inflation |
| Guardrails at 5.2% | $62,400 | Current rate 6.93% > 6.24% trigger → 10% cut to $56,160 |
| VPW, age 40–41, 60% stocks | $49,200 | Next year ≈ 4.1% × $900,000 ≈ $36,900 |
| Kitces ratchet from 3.5% | $42,000 | No raise (threshold $1.8 million). Floor unchanged. |
The 5.2% guardrail start is $20,400 more than the 3.5% floor in year one. If the first decade looks like 1966–1974 or 2000–2009, you take the 10% cuts or you have broken the only thing that made 5.2% defensible. VPW takes the cut automatically.
Implementation
Write the rule before you retire. Put the trigger — current withdrawal rate, portfolio vs. start, or VPW cell — in a one-page IPS. If you will not take a 10% cut, do not start at 5.2%. Start at 3.5–4.0% and use the ratchet.
Tax location is not the withdrawal rule. The spending rule sets how many dollars leave; tax location sets which account. In a low-income gap before Social Security and RMDs, filling ordinary-income brackets with traditional withdrawals or Roth conversions often beats “taxable, then traditional, then Roth.” Do not raid Roths to dodge a cut the rule requires.
Put Social Security and any pension under the floor. The 8% delayed-retirement credit from full retirement age to 70 (SSA, birth year 1943+) is a later-life annuity. VPW’s worksheet is built for it. Ignore that floor and a fixed 4% of the whole portfolio over-saves the gap years.
Do not CPI-adjust healthcare as if it were groceries. CMS National Health Expenditure Accounts: U.S. health spending rose 7.2% in 2024 to $5.3 trillion, or $15,474 per person (18.0% of GDP). That is aggregate spending growth, not your premium. It is enough to show that a CPI-U raise on a 4% withdrawal does not automatically cover ACA premiums, cash-pay care, or later Medicare IRMAA. Keep a separate healthcare line, inflate it faster than the rest of the budget, and put it inside the floor.
Rebalance on the rule’s calendar. Guardrails and VPW are annual. Intra-year panic sales are a behavioral override. If you need cash to avoid that override, size the sleeve to one year of the floor, not five years of peak lifestyle.
Backtest the rule you will follow. FIRECalc and cFIREsim replay U.S. historical sequences. The Bogleheads VPW worksheet and backtesting spreadsheet replay VPW. Do not take a success percentage from a model whose rules you will not obey.
What to do this week
- Pick the failure you will not accept: depletion (3.5–4.0% floor plus ratchet), a permanently lower lifestyle (fixed 4% or ratchet), a variable paycheck (VPW), or a 10% cut in a bad market (guardrails at 5.2%).
- Compute the four year-one numbers on your real portfolio, as in the $1.2 million table.
- Write the trigger in one sentence. If you cannot say it out loud, you do not have a strategy.
Then open the EarlyFIRE Target FIRE calculator on earlyfire.quest and rerun the same portfolio at 3.5%, 4.0%, and 5.2% with your age and horizon. Use the band, not a single safe number.
Sources
- Bengen, William P. “Determining Withdrawal Rates Using Historical Data.” Journal of Financial Planning (October 1994).
- Bengen, William P. “Asset Allocation for a Lifetime.” Journal of Financial Planning (August 1996). 30- / 45- / 20-year rates as summarized by Kitces, “Adjusting Safe Withdrawal Rates To The Retiree’s Time Horizon.”
- Cooley, Philip L., Carl M. Hubbard, and Daniel T. Walz. “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable.” AAII Journal (February 1998), Table 3.
- Guyton, Jonathan T., and William J. Klinger. “Decision Rules and Maximum Initial Withdrawal Rates.” Journal of Financial Planning (March 2006).
- Kitces, Michael. “Ratcheting The Safe Withdrawal Rate For Income Upside.” Kitces.com (2015).
- Kitces, Michael. “Flexible Spending Rules To Avoid FIREing At 4%.” Kitces.com.
- Pfau, Wade D. “Capital Market Expectations, Asset Allocation, and Safe Withdrawal Rates.” Journal of Financial Planning (January 2012), via Kitces.
- Bogleheads Wiki. “Variable percentage withdrawal.”
- SSA. “Delayed Retirement Credits.”
- CMS National Health Expenditure Accounts (2024 highlights).
- FIRECalc and cFIREsim.