Taking Money Out

A withdrawal rate is a starting rule

The familiar 4% is not a law; it is the answer to a historical question — what withdrawal rate survived the worst starting points in the record, for a portfolio held for about thirty years? It is a useful anchor precisely because it was built from bad sequences rather than average ones. But it depends on things the household controls: the fees it pays, the flexibility of its spending, the length of the horizon, and the valuation it starts from. The mechanics are worth seeing once. A $1,200,000 portfolio at 4% pays $48,000 in year one. If the portfolio falls 20% and spending is unchanged, the same withdrawal is a 5.0% rate — and with an inflation raise it is 5.1%. Nothing about the plan changed except the price of the assets, and the plan is now making a larger demand on a smaller balance. That is why a fixed real withdrawal is not the safe option; it is the rigid one. The same arithmetic in the other direction sets the honest expectation for how long a plan lasts. At a 2% real return, a 4% withdrawal declines by about 2% a year in real terms and lasts roughly 35 years — which is why 4% is described as surviving a long retirement most of the time rather than as a guarantee. At a 4% real return it lasts indefinitely; at a 0% real return, twenty-five years. The first year, then the bad one — Portfolio, then withdrawal at 4%: $1,200,000 → $48,000 · After a 20% fall: $960,000 · Same spending, plus a 2.5% raise: $49,200 = a 5.1% rate ← · Years the plan lasts at a 2% real return: about 35 A higher starting rate is not automatically reckless — it is riskier. Households with a guaranteed income floor covering essential spending can rationally take more from the portfolio, because the portfolio is only funding discretionary spending (PF11 covers the same reasoning for insurance).

Guardrails, agreed in advance

The alternative to a fixed real withdrawal is a spending rule with a band. The common shape: withdraw a starting percentage; if the portfolio falls 20% below the plan, trim spending about 10%; if it rises 20% above, allow a modest raise. The increase matters as much as the cut, because a household that only ever trims ends up spending far less than it could have — the failure mode of flexibility is that it is used in one direction only. The rules have to be written in a good year, when they are easy, so that they exist in a bad one, when they are not. That is the whole point: it converts an emotional decision made at the bottom of a market into a mechanical one made in advance. A guardrail is only as good as the moment it was agreed, and a rule invented after a 25% fall will be invented by somebody who is frightened. The other half of the design is the floor. Essential spending — housing, food, insurance, health costs — should be covered as far as possible by guaranteed income: a pension, an annuity, or simply a smaller withdrawal that leaves the essentials untouched. The portfolio can then be managed aggressively for the discretionary layer, and a bad year means a cancelled trip rather than a missed premium. Where the essentials are exposed to the market, a bad decade is a crisis; where they are not, it is an inconvenience. One guardrail, two directions — Portfolio 20% below plan: trim spending about 10% · Portfolio 20% above plan: raise spending modestly — or bank it · The rule: agreed in advance, applied mechanically · What protects the floor: guaranteed income covering essentials ← A guardrail band is the retirement version of the rebalancing band in PF8: a range instead of a number, because acting on every movement costs more than it fixes.

The mechanics of taking the money out

Which account the withdrawal comes from is a tax decision as much as a cash-flow one. Drawing from the taxable account first keeps the wrappers compounding and can keep taxable income low; but for a household with a large traditional balance and required distributions ahead, deliberate withdrawals from the traditional account in the low-income years before they start can reduce the tax bill later. Roth is usually the last sleeve to spend, and often the right one to use for a large lump — a new roof, a gift — in a year where an extra dollar of taxable income would be expensive. The practical structure is simpler than the tax theory. Hold one to three years of withdrawals in cash and refill that sleeve once a year from whichever account suits the tax position (PF13). Automate the monthly transfer so the withdrawal arrives as income rather than as a decision, and check the whole plan quarterly rather than watching the balance daily — the cadence is part of the design, because a plan reviewed every morning gets traded, and a plan reviewed every quarter gets managed. Three dates belong in the calendar. The sleeve refill, so a bad year does not force a share sale. The guardrail check, so the band is applied while it is small. And the tax review in the low-income years, when traditional withdrawals and Roth conversions are cheapest — which is the same logic as PF9, arriving at the other end of the plan. A withdrawal calendar — Monthly: an automated transfer from the cash sleeve · Annually: refill the sleeve from the tax-appropriate account · Annually: apply the guardrail band to spending · Quarterly: review the plan — not the daily balance ←

Pricing the floor

The rule that essentials should be covered by guaranteed income is easy to state and rarely costed, and the costing is what decides whether the plan is real. Two mechanisms turn capital into an income floor, and they behave differently. A bond ladder — a set of maturities that pay out in turn — is a series of known cash flows with no mortality pooling: the capital is spent down, the income is nearly certain in nominal terms, and it lasts exactly as long as the ladder is long. An inflation-linked ladder does the same in real terms at a lower starting yield. Both are transparent, and neither benefits from the fact that a retiree will not live forever. An immediate annuity adds the one ingredient a ladder cannot price: the mortality credit. The insurer pools the money of everyone in the cohort and pays it out to those still living, so each survivor receives their own capital plus the capital of those who did not survive. That is why an annuity’s payout rate exceeds the yield on a bond of matching duration by a wide margin, and why the margin grows with age: at seventy the credit is a modest uplift, and at eighty it is the dominant term. It is a genuine transfer rather than a fee, and it is the only asset that turns a lump sum into a lifetime income at a better rate than the market’s yield. The cost of the credit is the loss of control and of legacy. Once the money is annuitised it is gone from the estate, the payments stop at death, and the credit is forfeited if the household dies early. Three practical responses reduce that cost without giving up the mechanism. Annuitise in tranches rather than all at once, so the decision is made at several ages and rates instead of one; use a deferred income annuity that begins paying at a later age, which buys the most mortality credit for the least capital; and delay the start of a public pension, which functions as an inflation-linked annuity whose payout rises several percent for each year of waiting and is frequently the best-priced longevity insurance a household can buy. Sizing the floor is then arithmetic. Take essential annual spending in today’s money, subtract whatever a public pension and other guaranteed income already cover, and the remainder is the gap the portfolio must fund forever or an annuity must cover for life. Divide by the annuity’s payout rate to get the capital required, and compare that with the portfolio. If the gap is small relative to the portfolio, the floor is cheap and buys a great deal of behavioural freedom; if the gap is most of the portfolio, the plan does not have an income problem but a spending problem, and no product solves that. Either answer is more useful than a rule about floors applied without a price. • A ladder is certain in nominal terms and spends down: it lasts exactly as long as it is long. • An annuity adds the mortality credit — the capital of the cohort’s non-survivors — which is why its payout beats a bond yield. • The credit grows with age and is forfeited on early death, which is why insurance decisions later are cheaper. • Tranches, a deferred start date and a delayed public pension all buy the credit with less irreversibility. • Size the floor: essential spending less guaranteed income, divided by the payout rate, compared with the portfolio. An income floor is not a return maximiser and should not be judged as one. Its purpose is to make the essentials immune to the sequence risk described in PF13, so that a bad first decade forces a change in spending rather than a failure of the plan — a different objective from beating the market, and it should be measured against that objective.

Sources

• The 4% rule and its caveatsBengen (1994); the Trinity study; subsequent critiques on valuation and flexibility
• Dynamic withdrawal rules and guardrailsGuyton–Klinger decision rules; Kitces retirement research
• Sequencing withdrawals across taxable, traditional and Roth accountsIRS rules on required minimum distributions; Bogleheads wiki — withdrawal methods
• Guaranteed income floors and longevity riskStandard retirement-income research on annuity floors