Retirement Withdrawal Calculator

The 4% rule is a convention with a specific meaning

The initial-withdrawal-rate convention works like this: in the first year of retirement the withdrawal is a chosen percentage of the starting portfolio — $1,000,000 at 4% is $40,000, about $3,333.33 a month. In every later year the same dollar amount, adjusted for inflation, is taken — not the percentage again. The 4% figure traces to William Bengen's 1994 study in the Journal of Financial Planning, which replayed U.S. stock-and-bond history over rolling 30-year periods, and to the 1998 Trinity study (Cooley, Hubbard, and Walz), which tested a grid of withdrawal rates, portfolio mixes, and payout periods. Both are studies of what survived history under a specific allocation and a 30-year horizon — a research finding, not a guarantee about any future. The companion guide, The 4% Rule: Where It Comes From and What It Assumes, walks through the origin, the mechanic, and the assumptions the back-tests held fixed; this page stays on the arithmetic the calculator performs.

What the withdrawal-rate mode actually computes

The first mode is a single multiplication — first-year income = portfolio × rate — with the monthly figure being that amount divided by 12. That is the entire computation, and it is worth being precise about what it leaves out. It does not project the balance in year two, it does not apply the inflation adjustment the convention calls for in later years, and it does not know how long the money is meant to last: the 30-year horizon lives in the research, not in the formula. The result box reports four things — the first-year dollar figure, its monthly equivalent, the rate and portfolio it was computed from, and a reminder of the convention's second step (adjust the dollar amount; do not re-take the percentage). A month-by-month drawdown that escalates the withdrawal for inflation and reports when the balance is exhausted is a different model, run by the Retirement Calculator.

The rate field accepts any value from 2% to 6% and defaults to 4% when left blank. That band is a limit of this tool, not a research finding: it keeps the input in the neighborhood of the convention the tool is built around, and the engine returns an error for anything outside it rather than computing a number the convention was never about. Within the band the arithmetic is linear, so a short table shows the whole range at once. On a $1,000,000 portfolio:

Two readings follow from the table. First, moving the rate by half a point moves the income by $5,000 a year on this portfolio — $45,000 at 4.5% against $40,000 at 4% — a strictly proportional change, because the formula is a straight multiplication with nothing else in it. Second, the band's edges bracket a wide spread: $20,000 at the 2% floor to $60,000 at the 6% ceiling, 3 times the income from the same balance. The rate is the only lever besides the balance, and the formula treats every value in the band identically; nothing in the arithmetic marks 4% as special. Its standing comes from the research, which is why the guide linked above matters more than the multiplication.

A second portfolio, off the default rate

The same multiplication on different inputs: a $750,000 portfolio at a chosen 3.5% initial rate produces $26,250 in the first year, $2,187.50 a month. Across the full band the same balance spans $15,000 at 2% to $45,000 at 6%. Set beside the first example, the two inputs pull in exactly the directions the formula dictates: a smaller balance lowers the income in proportion, and a lower rate lowers it again. Nothing about the second portfolio's owner, horizon, allocation, or account type enters the calculation — the mode is indifferent to all of it, which is precisely why its output is a first-year figure and not a plan.

The runway formula, shown in full

The savings-runway mode answers a different question: spending a level W per year (in today's dollars) from a portfolio P earning a real return r, when does the balance hit zero? The engine applies growth first and takes the withdrawal at the end of each year, and under that convention the annuity-depletion formula gives the number of years n in closed form:

n = ln( W / (W − P·r) ) / ln(1 + r) valid when W > P·r

Each piece has a plain meaning. P·r is the first year's growth in dollars, so W − P·r is the net drain — how much the balance falls in year one after growth has been credited and the withdrawal taken. The ratio W / (W − P·r) compares the withdrawal to that drain: when growth covers most of the spending the drain is small, the ratio is large, and the runway is long. The logarithms convert that ratio into a count of years by dividing through by the per-year growth factor. In the worked example — $500,000 spending $30,000 a year at an assumed 4% real return — growth alone sustains $20,000 a year, spending is 6% of the balance and exceeds the sustainable figure, and the formula yields about 28 years. A second example with different inputs: $1,000,000 spending $50,000 a year at an assumed 3% real return sustains $30,000 from growth, spends 5% of the balance, and lasts about 31 years. Both figures are computed at build time by the same tested engine as the calculator above, so the copy cannot drift from the tool.

The three cases the formula does not cover

The closed form is valid only when spending exceeds first-year growth. The engine handles the other cases exactly rather than forcing them through the logarithm.

Spending at or below growth (W ≤ P·r). If the year-end withdrawal never exceeds the year's growth, the balance never falls below where it started: it holds level exactly at the boundary and rises below it. There is no depletion year to solve for, so the engine reports the balance as never depleting instead of returning a number. $1,000,000 spending $30,000 a year at an assumed 4% real return is such a case — growth alone sustains $40,000, spending is 3% of the balance, and the result box reads "never runs out" with the sustainable figure shown beneath it. The boundary itself is sharp. In the $500,000 example, spending exactly $20,000 never depletes, while spending $20,001 — one dollar more — yields a finite runway of about 252.5 years. That number is not a forecast of anything; it is what the logarithm does as the drain approaches zero, and it illustrates why a runway result near the boundary is more accurately read as "growth roughly covers spending" than as a calendar.

Zero real return (r = 0). With no growth the balance falls by exactly W each year, and the runway is the plain division n = P / W. The formula cannot express this directly — ln(1 + 0) is zero — so the engine computes the division outright: $500,000 spending $40,000 a year at a 0% real return lasts exactly 12.5 years, and no sustainable-spending line is shown because growth contributes nothing.

Negative real return. A real return below zero — an assumed nominal return that falls short of assumed inflation — is accepted down to -20%. The formula still applies: with r negative, W − P·r is larger than W, both logarithms are negative, and the quotient is a positive number smaller than the zero-growth division. The engine's test suite pins the case: $100,000 spending $10,000 a year at an assumed −5% real return lasts about 7.9 years, against 10 years at 0%. Growth and withdrawals are draining the balance together, and again no sustainable-spending line appears, because growth is not contributing.

Reading the runway result

When the balance depletes, the result box leads with the runway in years, then the first year's spending as a percentage of the portfolio, then — when the real return is positive — the spending that growth alone would sustain. Three details matter in reading it. The years are counted in full annual withdrawals under the end-of-year convention; a fractional result means the balance covers that many full withdrawals and then runs short partway through the next one, so at a 0% real return $500,000 spending $40,000 covers 12 full withdrawals with the remaining balance meeting only part of the next. The percentage line is the spending rate of the runway inputs — $30,000 on $500,000 is 6% — and it is the figure to compare with the withdrawal-rate mode above, since both are a withdrawal divided by a starting balance. The sustainable-spending line is the boundary value P·r; how far the entered spending sits above it is the whole reason the runway is finite.

The same inputs at several real returns

Because a single real-return number carries the entire growth assumption, it is the input the runway is most sensitive to. Here is the $500,000 / $30,000 example held fixed while the assumed real return moves across five values, each row computed by the engine at build time:

The progression is not linear. Each step up in the assumed return adds more years than the step before, because a larger share of each year's spending is covered by growth and the drain on principal shrinks toward zero; at 6% growth alone sustains $30,000, which equals the spending, so the result crosses to "never depletes." Spending moves the runway the other way. Holding the 4% real return fixed, $25,000 a year lasts about 41 years and $35,000 a year about 21.6 years. None of these rows is more likely than another: the engine has no view on what real return will occur, and it computes only the consequence of the one entered.

Two modes, two different questions

It is tempting to read the two modes as the same calculation approached from two ends, and they are not. The withdrawal-rate mode has no return and no horizon in it — it is a convention's first step, and the 30-year survival the research reported was measured against actual historical return sequences, not against any steady rate. The runway mode has a return and no convention — it never adjusts the withdrawal for inflation, because working in real terms holds purchasing power constant by construction, and it never consults the 4% figure. The link between them is the spending ratio: the 6% of $500,000 spent in the runway example is the same first-year ratio the rate mode reports as $30,000 at a 6% initial withdrawal. What differs is what each mode does with that ratio — one reports the dollars and stops; the other assumes a steady real return and reports the depletion year that follows from it.

What a smooth average hides

Both modes assume a single steady return, and this site's methodology is explicit that real markets do not behave that way. The order of returns matters as much as their average: a deep downturn in the first years of retirement, when assets are being sold to fund the withdrawal, does damage that later good years may never repair. That is sequence-of-returns risk, the standard first criticism of any fixed-withdrawal convention and the stress that determined the historical worst cases in the research the 4% figure comes from. A single-rate runway is therefore a midpoint, not a range; running the same inputs at a weaker assumed return — as the table above does — shows how much of the answer rests on the rate. The 4% rule guide lays out the other boundaries of the research: a specific stock-and-bond allocation held throughout, a 30-year horizon, and no fees or taxes modeled. Neither mode here models taxes, fees, Social Security, or pensions either.

One more mechanism sits outside both models. Once the statutory age arrives, required minimum distributions set a floor under withdrawals from most pre-tax retirement accounts regardless of any strategy — a tax rule, not a spending plan, and one that enters neither formula on this page. The RMD rules guide explains the start age, the balance-divided-by-factor formula, and the penalty, and the RMD Calculator performs the division. For a month-by-month drawdown that escalates spending with inflation in nominal dollars, see the Retirement Calculator; the formulas and conventions behind every tool are documented on the methodology page.

Frequently asked questions

Not financial advice: these are general educational conventions that assume a constant return and exclude taxes, fees, Social Security, and the variability of real markets. Historical survival is not a promise about the future. Values are processed locally in your browser and never transmitted. See the methodology page.