Die with Zero Calculator: The Earliest Age to Stop Working

What this die with zero calculator answers

The headline answers one question: given what is saved, what is added monthly, and what retirement will cost per month, what is the earliest age the primary paycheck could stop while the chance of going broke stays inside a chosen tolerance, measured either to an age you pick or across a whole life? The answer is a result about the model, not an instruction: “at a 1 in 20 tolerance, the earliest quit age these assumptions support is 53” is the shape of every headline this page produces.

Everything under that headline answers a question it raises. What each extra year of working buys, both in odds and in monthly spending. Where the balance could be as the years pass. And what you would be left with at a given age, which is what the safety cost: an age that clears the running-out test comfortably is usually an age that leaves a great deal unspent, and no probability on its own shows that half of the trade.

The quit age has one precise meaning here: the age at which the monthly saving amount stops and the monthly spending starts. Nothing else moves with it. Dated entries under Life events, a pension from 67, rent, a one-time sale, keep their own dates whatever the solved age turns out to be, so the headline stays readable even when other income continues long after the paycheck ends.

The nearest neighbours on this site answer adjacent questions: the retirement probability calculator takes a retirement you already have in mind and reports its odds, and the lifestyle change calculator projects a single deterministic path for a life change rather than a distribution of outcomes.

How the earliest quit age is found

Each candidate quit age defines a complete cashflow schedule: the monthly saving lands every year up to that age, the monthly spending is drawn every year after it, and the dated Life-events entries sit at their own years. The probability engine then propagates the full distribution of the balance through that schedule, one year at a time, and reads off the cumulative chance the balance has hit zero by the plan-to age. That chance is computed exactly, the way the retirement probability calculator computes its success odds; there is no Monte Carlo sampling and no run-to-run wobble.

The solve itself rests on a property the tests assert rather than assume: quitting a year later swaps one drawdown year for one saving year, so a later quit never carries more risk of ruin than an earlier one. The earliest age whose risk fits under the tolerance is therefore a well-defined boundary, and the calculator searches straight to it. When no age fits, even quitting the year before the plan-to age, the page says so plainly and reports the best achievable figure instead of showing nothing.

Why the risk tolerance is written as odds

The tolerance presets read as plain odds, 1 in 50, 1 in 20, 1 in 4, a coin flip, because that is the decision actually being made: how many futures out of fifty, twenty, or four are allowed to end with the money gone early. Those four fractions are exact by construction, they are settings rather than estimates, so the odds wording claims no precision it lacks.

Every probability the calculator computes is displayed differently: as a whole percent, never finer. A model estimated from 154 annual observations of one market does not know a ruin probability to a decimal place, and printing one would assert that it does. Probabilities near the edges render as “more than 99%” or “less than 1%” rather than a false certainty of 100% or 0%.

How to choose the plan-to age

The plan-to age is the horizon you want the money to survive, an age you choose rather than a prediction of how long you will live. Nothing is filled in for you unless you ask: the calculator reports the odds of the plan lasting to whatever age you enter, and moving that age up or down simply lengthens or shortens the horizon the odds are measured against.

A period life table is a useful reference for picking that age. It records what share of a population reached each age, using the death rates measured across one recent span of calendar years, so it is a statement about a population and never about one person. This site ships period life tables for 41 countries, covering the calendar years 2010 to 2024; the methodology page names their source and licence.

A period table holds those measured death rates frozen, and real populations have kept outliving the period tables of their day, so these figures understate how long people alive now will go on living. Raising the plan-to age lengthens the horizon the money must cover, so at the same tolerance the model requires a later quit age, or the same one. Nothing here models personal health or circumstances, and no input asks about them.

Ticking the estimate box beside the field puts those tables to work in three ways. It fills the plan-to age with the age half of the chosen group had died by, leaving the number editable. It draws the share of that group still alive at each age over the ruin chart, so the horizon stops being a single line. And it offers a second thing to hold the odds against: instead of one age you picked, the chance of running out can be weighted across every age by how many people lived to see it. That last figure treats how long you live as unrelated to what markets do, which is an approximation, and the assumptions below say so.

What this calculator assumes

A probability invites more trust than a point estimate, so it carries more obligation to say what it rests on. These are all of the model’s assumptions, each with its honest limitation attached.

• Each year is an independent draw. Every year is drawn fresh from the return distribution, remembering nothing about the year before. It is the model’s biggest assumption; the data neither rejects nor proves it, and the measurements are on the methodology page.
• The spread widens with the horizon, and its far ends cannot be checked. Independent draws make the range grow without limit as the projection lengthens: the variation in the log balance grows with the square root of the years covered, so a projection running into the nineties reports outer percentiles many times its own median. Nothing can check those figures. 154 years of returns hold three stretches of fifty years that do not overlap, so a fifty-year percentile is the model extending the independence assumption rather than a figure the record has measured. The middle of the range is the part the data speaks to; the far ends of a multi-decade projection are the weakest output this model has.
• US markets, 1871–2024. The return distribution comes from one country across one stretch of history, and the US spent the 20th century as the best-performing major stock market, something fully visible only in hindsight. Nothing guarantees that a reader elsewhere, or a future US reader, draws from that same distribution; of everything listed here this is the item most likely to be materially wrong, and exact computation cannot repair it.
• Bond-heavy mixes: the bands are likely slightly too narrow. Real bond returns inherit inflation’s streakiness, so at high bond weights true long-run variation is somewhat wider than independent draws imply, and the risk shown there errs on the low side. Stock-heavy and mixed portfolios measure clean.
• Everything is in today’s money. Returns are inflation-adjusted, and spending, the monthly saving amount, and every Life-events entry stay constant in real terms. Entering a salary figure that will not keep pace with inflation as if it were constant purchasing power flatters the plan and pulls the quit age earlier than it should be.
• The plan-to age is yours, not a prediction. The age is the one you chose to plan for, and the odds are reported against it. A period life table says what share of a population reached each age, never what one person will do, so the optional estimate fills the field with a population statistic rather than a forecast about you. A higher age lengthens the horizon the money must cover.
• How long you live is treated as unrelated to what markets do. Weighting the odds by a life table multiplies the chance of dying in each year by the chance of being broke by then, which is only correct if the two are independent. They are not exactly: recessions move mortality, and health shocks move spending and retirement dates. The figure would need a joint model to capture that, and this one does not have it. Two smaller roundings pull against each other: the year of death counts as a year the money had to last, which reads the risk slightly high, while deaths from age 105 onward, past the last year the model carries, are counted at the odds for 105, which reads it slightly low.
• One year, one step, rebalanced annually. A year is a single draw with no within-year sequencing, and applying a weighted return to the whole balance quietly assumes rebalancing back to the chosen stock/bond split every year. Both are real assumptions, not details.
• Spending is fixed in real terms after the quit age. The model spends the same inflation-adjusted amount every month no matter what markets do. Real people usually cut spending after bad years, which is why constant-spending ruin numbers overstate risk for anyone willing to adapt.
• The quit age moves only the monthly saving. Solving for the quit age changes one thing per candidate: the year the monthly saving stops and the monthly spending starts. Dated Life-events entries are never moved by the solve, so a pension entered from 67 still starts at 67 whether the headline reads 48 or 58.
• No taxes, fees, or transaction costs. Index returns are used as-is. Costs reduce returns, so the model is optimistic by roughly the total drag on your portfolio.
• Running out is permanent. Ruin is an absorbing state: once the balance reaches zero it stays there, even if a dated windfall arrives later. The model does not simulate noticing trouble in year 20 and adjusting course, which is one more reason the raw number reads pessimistic for anyone willing to adapt.
• The distribution is fitted, not the history replayed. Outcomes are computed from a distribution estimated over all 154 years, in a choice of the normal fit (the default here, and the bell curve most simulation tools assume), the historical fit (keeps the fat left tail as it actually fell), or a fixed return, which reproduces a deterministic projection exactly and is useful for comparing against simpler calculators.
• The expected return is the biggest lever, and no history pins it down. Every band and every quit age on this page rests on a return level estimated from the past, and the future is not obliged to match it. The expected-return control under Return model and extras moves that level while keeping the shape of the distribution, and watching the solved quit age move with it is the most informative check available here: it shows how much of the answer is the plan and how much is the assumption.

The full statistical evidence behind the return model, provenance, fit diagnostics, and the independence measurements, is on the methodology page.