Vanguard's dynamic spending rule is the most reasonable-sounding withdrawal strategy in retirement planning. Each year you take a target percentage of what the portfolio is actually worth, then you clamp the change so your income cannot jump more than 5% or fall more than 2.5% against last year in real terms. It fixes the obvious flaw in the 4% rule, which keeps writing the same cheque into a crash, and the obvious flaw in a pure percentage rule, which hands you a 30% pay cut the year the market halves.
It is also the rule most likely to be sold to you with the wrong number. Search for a comparison and you will be shown success rates: this rule survives 99% of the time, the 4% rule only 75%. That comparison is close to meaningless, and this article is mostly about why.
The number that cannot be wrong
Start with the rule Vanguard dynamic is built on top of: spend a flat 4% of the current balance every year, no clamp. Across 5,000 simulated 30-year retirements it ran out of money 0.0% of the time.
That is not a finding. It is arithmetic. Four percent of a positive number is a smaller positive number, so the balance can be halved forever without reaching zero. A percentage-of-balance rule cannot fail the "did you run out" test any more than a thermometer can fail to report a temperature. If you rank spending rules by survival, this one wins before the simulation starts.
So ask what it cost. In the same 5,000 lifetimes, that rule pushed real spending below $30,000 a year — three quarters of the $40,000 it started at — in 72.6% of them. In the worst tenth of worlds its leanest year came in at $10,576. The money never ran out. The retirement did.
This is why every number below is a spending number. A withdrawal rule that is allowed to cut without limit will always look immortal, and the only honest way to compare it against a rule that refuses to cut is to ask what each one actually paid out over a life.
Four rules, 5,000 identical worlds
The specimen is a household with $1,000,000 retiring at 65 with a 30-year horizon, holding 60% stocks, 35% bonds and 5% cash, rebalanced. No Social Security and no other income: the portfolio funds every dollar of spending, which is how Bengen, Guyton–Klinger and Vanguard all frame the problem, and the only way the rule is the thing being tested.
All four rules start at the same 4% draw, so what moves between them is the rule and nothing else. Every rule sees the same 5,000 market histories in the same order — the same crash in the same year of the same retirement — so when two rules finish in different places, the gap is the rule and not the luck.
| Spending rule | Ran out | Spent under $30,000 | Lifetime spend, median | Lifetime spend, worst 10% | Leanest year, worst 10% |
|---|---|---|---|---|---|
| Constant dollar 4% | 25.7% | 25.8% | $1,183,173 | $883,023 | $0 |
| 4% of balance | 0.0% | 72.6% | $1,079,969 | $672,324 | $10,576 |
| Guyton-Klinger guardrails | 0.0% | 58.7% | $1,077,574 | $698,244 | $11,027 |
| Vanguard dynamic (+5% / −2.5%) | 10.0% | 48.9% | $1,148,305 | $840,101 | $18,215 |
Real dollars. 5,000 shared paths, seed 20260622, 5 bps trading cost, pre-tax. "Leanest year" is the worst single year of real spending, at the 10th percentile across worlds.
Constant-dollar 4% delivers the most lifetime spending in the median world and ruins 25.7% of the time. Its floor column and its ruin column are the same number, 25.8% against 25.7%, because it never voluntarily cuts: the only way it spends less than $30,000 is by having nothing left. It pays full freight until the account is empty and then it pays nothing.
Vanguard dynamic sits between. It cut ruin from 25.7% to 10.0% and holds the highest floor of the three flexible rules: $18,215 in its leanest year at the 10th percentile, against $10,576 for unclamped percent-of-balance and $11,027 for guardrails.
The clamp is the whole rule
Those two facts are the same fact. The −2.5% floor on annual cuts is a promise that your income will not fall off a cliff, and a promise like that has to be paid for out of the portfolio. Unclamped percent-of-balance can cut as fast as the market falls, which is exactly why it cannot ruin and exactly why its spending collapses. Vanguard dynamic refuses to cut that fast, so in a bad enough sequence it keeps drawing more than the portfolio can support, and 10.0% of the time it runs out.
The clamp is not a safety feature bolted onto a percentage rule. It is a deliberate transfer of risk from your standard of living to your balance. That is a defensible trade and it is the opposite of what "never runs out" implies.
It also cuts constantly. Vanguard dynamic reduced real spending at least once in 99.5% of retirements, and the median retirement spent 18 of its 30 years below the income it started with. A rule that adjusts a little every year adjusts almost every year.
Paired against the 4% rule, world by world
Two medians side by side are a weak comparison. Because every rule ran the same worlds, we can do better: take each of the 5,000 retirements one at a time, subtract what constant-dollar 4% delivered from what the other rule delivered, and look at the distribution of the difference.
| vs constant-dollar 4% | Worst 10% of worlds | Median world | Best 10% of worlds | Share of worlds ahead |
|---|---|---|---|---|
| 4% of balance | −$297,017 | −$88,111 | +$641,634 | 41.2% |
| Guyton-Klinger guardrails | −$278,423 | −$88,766 | +$577,534 | 40.4% |
| Vanguard dynamic (+5% / −2.5%) | −$198,351 | +$17,834 | +$580,331 | 54.3% |
Difference in lifetime real spending, computed per world, then summarised. Positive means the rule delivered more than constant-dollar 4% in that world.
Vanguard dynamic is ahead of the 4% rule in 54.3% of worlds, worth +$17,834 in the median one. With 5,000 paired worlds that margin is far outside chance. The other two flexible rules are behind in most worlds: 41.2% and 40.4%.
And in the worst tenth of worlds Vanguard dynamic delivers −$198,351 against the 4% rule. It is ahead in the middle and behind in the tail, which is what a rule that cuts in bad sequences must look like when it is measured against a rule that does not. We pre-registered the left tail as the primary test before running anything, and it came back negative, so there is no version of this result that reads "Vanguard dynamic is better." There is a version that reads "it trades tail spending for a lower chance of zero," and that one is true.
The ranking does not survive a different world
The comparison above lives inside one calibration of one simulator. So we moved the calibration and re-ran the identical rules on the identical seed.
| Assumption | Vanguard dynamic vs 4% rule, median world | Share of worlds ahead |
|---|---|---|
| Base calibration | +$17,834 | 54.3% |
| Volatility +20% | +$18,716 | 56.6% |
| Crisis-heavy transition | −$6,272 | 47.3% |
| Stock drift −2pp | −$13,881 | 44.4% |
| Inflation +2pp | −$19,660 | 38.7% |
The advantage survives higher volatility and disappears everywhere else. Lower expected stock returns, a more crisis-prone market, or persistently higher inflation each flip it behind the plain 4% rule. The mechanism is not subtle: Vanguard dynamic targets a percentage of the current balance, so in a world where balances are smaller the target is smaller, the clamp spends its life dragging income down, and a rule that simply keeps paying $40,000 until it dies can deliver more total spending across a fixed 30 years.
This is the part a success-rate table cannot show you. Whether flexible spending is worth it is not a fact about the rules. It is a bet on which world you are retiring into.
What actually happened, 1928 to 2025
Simulated worlds are worlds a model believes in. So we replayed the same four rules on the real US tape: every 30-year retirement starting between 1928 and 1996, 69 of them.
History disagrees with the simulator. On the tape all three flexible rules delivered more lifetime spending than constant-dollar 4% in most windows — Vanguard dynamic in 60.9% of them, percent-of-balance in 58.0%, guardrails in 56.5% — where in the simulator two of the three were behind. Constant-dollar 4% ran out in 4 of 69 historical windows; none of the flexible rules ran out in any.
Those two answers are not averaged into one. The twentieth-century US tape was kinder to balance-following rules than the simulator's crisis and stagflation regimes are, and the disagreement is the information. Anyone quoting one of these numbers without the other is quoting half a study.
1966, the window that breaks things
The single worst year to retire in modern US history is 1966. Here is what each rule did to a household that retired into it.
| Retired in 1966 | Lifetime spending | Leanest year | Left at the end |
|---|---|---|---|
| Constant dollar 4% | $990,652 | $0 | $0 |
| 4% of balance | $761,633 | $15,397 | $788,741 |
| Guyton-Klinger guardrails | $765,666 | $18,614 | $686,141 |
| Vanguard dynamic (+5% / −2.5%) | $830,267 | $18,954 | $501,469 |
Read the first row carefully, because it is the best argument in this article against the metric this article uses. Constant-dollar 4% posted the highest lifetime spending of any rule in 1966 — $990,652, more than Vanguard dynamic's $830,267 — and it did it by spending itself to zero and leaving the household with nothing. On the lifetime-spending column it looks like the winner. It is a catastrophe.
That is why the tables above always print spending, the floor, and ruin together. Every single one of those columns can be gamed by a rule built to game it. Lifetime spending rewards spending everything early. Survival rewards never spending. The floor rewards never promising anything. A rule looks good or bad depending on which column you were already looking at, and the honest version of this comparison shows all three and lets you pick the one you actually care about.
What this does and does not say
Vanguard dynamic spending is a coherent rule that does what it claims: it steadies income against a percentage rule and lowers the chance of zero against a fixed-dollar rule. On this specimen, in this simulator, at this calibration, it delivered more lifetime spending than the 4% rule in the typical world and less in the worst tenth. Change the assumed world and that ordering reverses. On the real tape it looks better than the simulator says.
What it is not is a rule that cannot run out, and it is not a rule you can evaluate with a success rate. If a calculator shows you 99% and stops, it has told you the least interesting true thing about the strategy.
You can run the same clamp on your own numbers with the Vanguard dynamic spending calculator, compare it against Guyton–Klinger guardrails, or read how the 4% rule holds up on its own.
Specimen: one hypothetical household, $1,000,000 at 65, 30-year horizon, 60% stocks / 35% bonds / 5% cash rebalanced, no Social Security or other income. Q1 and Q2: 5,000 paths on a regime-switching generator, seed 20260622, shared market paths across all four rules, 5 bps trading cost, pre-tax, real dollars, engine commit 3e471a24f005. Rules frozen before the run: constant-dollar 4% (Bengen 1994), 4% of balance, Guyton–Klinger guardrails at a 20% band and 10% adjustment (Guyton & Klinger 2006), and Vanguard dynamic at a 4% target with a +5% / −2.5% real clamp (Vanguard, From assets to income: A goals-based approach to retirement spending). Q3: the same frozen rules on the 1928–2025 US tape, bundle ffff7127cae8, 69 overlapping 30-year windows but only 3 independent ones, so those percentages are not independent trials — failures cluster because windows share the same crashes. Windows starting in 2000 and 2008 do not exist at a 30-year horizon on a tape ending 2025; 1929, 1966 and 1973 all ran. Tape months before 1988 are annual figures spread across the year, so within-year swings are understated, which slightly flatters rules that react to the balance. The $30,000 spending floor is an assumption we chose, not a result; a household with a different floor gets different floor-breach numbers. Q4: costs on at 5 bps, taxes not modelled. Pre-registered study card and both frozen datasets are published with this article. One specimen is not a claim about households in general. This article is educational analysis, not investment advice, and does not recommend any security or strategy.