Start here
The basics
At retirement, the plan records how much the household draws as a share of its portfolio. That starting withdrawal rate becomes the reference point for two guardrails.
If withdrawals become too large relative to the remaining portfolio, the rule trims flexible spending. If the portfolio can support more, the rule raises spending. Between the guardrails, spending follows the plan.
Illustrative numbers
A 4% retirement meeting a 30% fall
Anchor year portfolio$1,000,000
Cards’ bill, other money$60,000, $20,000
Anchor withdrawal rate4.0%
Rails at a 20% band3.2% and 4.8%
Next year’s portfolio$700,000
Rate at unchanged spending5.71%
Spending after the trim$56,000
The rate crossed the upper rail, so the model trimmed the year’s withdrawal by a tenth — $40,000 to $36,000 — and the cards’ bill fell to $56,000, a scale of 0.933 rather than the 0.900 a whole-budget trim would have produced. A third year at $640,000 still read 5.63%, so a second trim compounded onto the first and spending went to $52,400. When the portfolio recovered to $1,100,000 the rate fell to 2.95%, inside the lower rail, and the raise took spending back up to $55,640. No clamp bound in this run; on a portfolio that kept falling a tenth a year, the 60% floor caught the scale on the ninth consecutive trim.
Calculation transparency
How it works in MoneyWhatIf
- 01
The anchor is set on the first retired year in which the portfolio holds something and the year actually needs something from it — a retirement whose income covers every bill for three years anchors in the fourth.
That anchor year always spends exactly what its cards say, because it is the measure rather than a verdict: its rate, the portfolio’s share of that year’s bill divided by the portfolio, becomes the reference every later year is read against.
- 02
“Portfolio” means what the accounts closed the previous year holding, each balance floored at zero and summed — the same statement a required minimum distribution is worked out from. Only a funded projection has that statement to read, so a run without one keeps its spending as written whatever the setting says.
- 03
The rule sizes the portfolio’s share of the bill, not the whole bill.
Before the cards are read, the year settles its other money — wages and rent in; pre-tax and after-tax contributions, a job’s own pension, the payroll tax and the income tax already charged on that income, the mortgage, property tax and upkeep, and the debts’ scheduled payments out — and only what the cards ask beyond that is the withdrawal the rails are read against.
The tax the withdrawal itself will owe is settled later in the year and is not inside the measure. On a plan whose income tax settles the following April, the tax counted out above is only the share the paycheck withheld — and last April’s settlement is netted beside it, out when the year owed and back in when it was a refund.
Medicare and marketplace premiums are priced after this step and stay outside it as well, since a bill counted at the anchor and at every later reading alike cancels out of the comparison.
- 04
Every later year divides the withdrawal at the scale currently in force by the portfolio and compares it with the anchor rate.
Above anchor × (1 + band) the year is leaning harder than the retirement’s own rate can carry and the withdrawal is trimmed by the adjustment; below anchor × (1 − band) the portfolio can pay more than it is being asked and the withdrawal is raised by it; between the rails nothing moves.
- 05
The adjustment moves the withdrawal rather than the whole budget: the new scale is the adjusted withdrawal plus the year’s other money, over the cards’ own bill. For a household living mostly off its portfolio the two readings agree to the digit; for one whose income covers most of the bills, trimming the whole budget by a tenth would fling the rate across both rails at once and saw back and forth for years.
- 06
The scale is carried forward and cumulative, so consecutive crossings compound onto one another, and every crossing’s new scale is clamped between the floor and the ceiling — 60% and 150% of what the cards say, by default. A retired year whose portfolio has emptied reads no rate at all and keeps the scale it already had, which is wherever the last crossing left it rather than the floor itself.
- 07
Only retired years are touched: a working year passes through at a scale of one with the rule’s memory intact, and so does a retired year outside the window, which narrows the rule to part of the retirement and anchors it where it opens rather than at the retirement it skipped.
A year with nothing on the cards the rule may bend is skipped entirely, and a card left out of the rule’s scope becomes a bill it budgets around — that card’s cost comes off the year’s other money instead of being scaled.
Keep in mind
Model limits
The engine works in whole plan years, so the rails are read once a year against the prior year-end statement rather than continuously through it.
This is the published rule’s capital-preservation and prosperity rules only; Guyton and Klinger’s other decision rules — the inflation rule that withholds a raise after a losing year, and their withdrawal-sourcing sequence — are not represented here.
Whether a household would in fact accept the trim the rule prescribes is unknowable to a model, and the projection assumes every adjustment is followed exactly and on time.
The floor and the ceiling are the model’s own honesty rails rather than part of the published rule, and a plan that spends most of its retirement pinned to one of them is no longer being described by the guardrails.
This explanation documents the planning model. It is educational, not individualized tax, legal, Medicare, or investment advice.
Where it appears
Where to use it
See this concept in context, with a guide to each page and its controls.
The words behind it
Related financial terms
Plain-English definitions, with 2026 figures and worked examples, from the financial terms glossary.