Longevity Risk and the Withdrawal Rate: Outliving the Money

Longevity risk is the risk of living longer than the money lasts. A withdrawal rate is the share of a pot taken as income each year, and it is what turns a balance into a monthly figure. Change the rate assumed and the pot needed moves enormously. Nobody can tell any household which rate is the safe one.

Underneath that answer sits one asymmetry that separates this subject from everything else. Every other risk a household meets is an event it would prevent if it could: the work stopping, the illness arriving, the balance climbing past what a month can carry. Each has a cause worth removing, and the whole apparatus of buffers and covers exists to soften causes people wish were not there. Longevity risk has a cause nobody wants removed, and a cause nobody wants removed cannot be argued with. It can only be arranged for.

What is longevity risk, and why does it not feel like a risk?

Start with a kitchen. A household buys rice, dal and oil for what it thinks is a week, and the week runs to nine days because two relatives arrived and stayed. Nothing was wasted, nothing stolen, nobody miscounted the sack. The sack was correct for the week expected and short for the week that happened. The mismatch between the week expected and the week that happened is the entire mechanism, and the finance version differs only in the size of the sack and the length of the guess.

Longevity riskThe risk of living longer than the money lasts. Longevity risk is not the risk of dying but the risk of the money running out first. is the risk that a person outlives the money set aside to pay for the years after the earned income stops. Longevity risk is not a risk about health, not a risk about markets, and not a risk about anybody making a bad decision. The risk is a mismatch between two durations: how long the money was arranged to last, and how long it turned out to be needed for.

Longevity risk is the risk of living longer than the money lasts, so it is a risk about duration rather than about amount. Duration is the one input in the whole calculation nobody can look up.

Here is why it does not feel like a risk while somebody is still working. Every other risk announces itself. A lorry route closes and the takings fall that week. A card balance climbs and the statement says so in bold. A claim is cut and a letter arrives with a clause number on it. Each comes with a date and a sheet of paper. Longevity risk comes with neither, decades after the decisions that determined it were taken, or not taken.

Richard Thaler's work on how people weigh outcomes at different distances explains the rest of the feeling. A cost landing decades from now is discounted so steeply against a bill due on Friday that it barely registers, and the bill due on Friday is genuinely real. Steep discounting of a distant cost is the ordinary shape of human attention rather than a defect in anybody. The discounting is also why this arithmetic has to be written down rather than held in the head. Written arithmetic does not fade with distance and an intention does.

There is one more reason it hides. The consequence is not dramatic on the day it happens. Nothing bounces, nobody calls, no letter arrives. A pot that will run short keeps paying out normally for years and then does not, and by the time the shortfall is visible the years in which it could have been changed have gone. A risk with no early symptom is a risk nobody attends to.

Why is this the only risk here caused by something good happening?

Take the four risks this sequence has already treated and put their causes side by side. Income disruption is caused by work stopping. A cut claim is caused by a term that was in the document and was not read. A rising balance is caused by a card charging on what was left unpaid. A shock the buffer could not absorb is caused by something breaking, somebody falling ill, or a job ending. Every one of those causes is something a household would remove tomorrow morning if it were offered the chance.

Now the cause of this one. The money runs out because somebody lived a long time. There is no version of this risk where the cause is unwelcome, and no household anywhere would accept an arrangement that removed the cause in exchange for removing the risk.

Every other risk in this sequence is produced by something a household would prevent. Longevity risk is produced by something a household would choose, so it cannot be prevented, and arrangement is the only response left.

The distinction is not word play, and it changes what the household can do. With a preventable cause, the sensible first move is prevention: fix the roof, read the clause, clear the balance. With a cause nobody wants prevented, the whole response sits on the arrangement side. There are only three levers there: the amount that goes in, the length of the period being funded, and the amount taken out each year once the drawing starts.

The distinction also changes how the risk is talked about inside a household. A preventable risk invites blame, and blame is often the whole reason a household will not sit down and look at one. Nobody caused this one. A household that has run its arithmetic and found a gap has not made an error; it has found out that a good outcome is expensive. Finding out that a good outcome is expensive is a different conversation at a kitchen table, and the only honest version of it.

What is a Withdrawal Rate, and what does it connect?

A household has two kinds of money number and they never mix directly. One is a balance, an amount sitting still: Rs 4,12,000/- in a provident fund at 31 March of year two. The other is a flow, an amount per unit of time: Rs 30,000/- a month. A balance cannot be spent as a balance, and a flow cannot be saved as a flow. Something has to sit between them and translate.

The withdrawal rateThe share of a pot taken as income each year. The rate translates between an amount sitting still and an amount arriving every month. is that translator. A withdrawal rate is the share of a potThe accumulated amount an income is drawn from, wherever it happens to sit. A pot is a total rather than a place. taken as income in a year, written as a per cent. Three per cent of a pot means three rupees out of every hundred sitting in it, taken across the year. Nothing about that sentence is technical, and nothing about it requires a scheme, a product or an adviser to understand.

A withdrawal rate is the share of a pot taken as income each year, and its only job is to translate between a balance that sits still and an income that arrives every month.

The useful property is that it runs in both directions, answering two completely different household questions. Run it forward and it answers what somebody already has: a pot of known size, a rate assumed, and out comes an annual income and then a monthly one. Run it backward and it answers what somebody needs: a monthly figure, times twelve, divided by the rate, and out comes the pot the arithmetic implies.

Twenty four years before anybody retires, what a household needs is a target, and only the backward direction produces one. The backward direction also produces the alarming numbers, and the reason is worth stating now rather than meeting later as a surprise. Dividing by a small number makes a big answer. A rate is a small number. So the pot always comes out as a multiple of the income, and the multiple is large.

One last thing about the word rate. The word does damage. Elsewhere in household finance a rate is set by somebody else and handed over: the rate on a loan is in the agreement, the rate on a deposit is on the receipt. A withdrawal rate is not like that. Nobody sets it, nobody tells anybody what it is, and no document carries it. The rate is chosen by whoever is doing the sum, and the choice is about the future rather than a fact about the present.

How does a required income turn into a pot?

The Bhosale household, invented, has already put its retirement requirement on paper in an earlier part of this sequence: Rs 30,000/- a month, its own estimate, against the Rs 42,770/- that leaves the household in an ordinary month at 31 March of year two. The requirement is the household's own, not a benchmark and not a suggestion, and it was built line by line from what its own later years would carry.

Rs 30,000/- a month is Rs 3,60,000/- a year. The annual figure is the flow the pot has to produce. Everything in this guide is arithmetic performed on that one number and on nothing else.

The requirement is held in today's moneyPriced at what things cost now, so the question of prices rising is kept separate rather than folded invisibly into the same figure. throughout. Folding two assumptions into one figure is the commonest way a household sheet becomes uninterpretable. Two large assumptions are available here: one about what share of a pot can safely be taken as income, and one about what prices will do over twenty four years. Fold both into one figure and the result is a single alarming number carrying two guesses, neither of which can be inspected afterwards or corrected without redoing the whole sheet.

The requirement stays in today's money so that the withdrawal rate assumption and the price rise assumption stay in separate boxes, where each can be argued with on its own.

So the arithmetic is one line. Take the annual requirement and divide it by the rate, written as a decimal. Rs 3,60,000/- divided by 0.04 is Rs 90,00,000/-. One division is the whole calculation. There is no second step, no adjustment and no software. A person with a sheet of paper and a phone calculator can produce every figure in this section.

Notice what is not in that line. No growth rate for anything. No figure for how long anybody lives. No scheme, product, charge or tax. All of those exist and all matter, and each is a separate question. One variable has been isolated deliberately. The rate, on its own, moves the answer more than a household would ever guess.

How much does the assumed rate move the answer?

Here is where this subject stops being a definition and starts being a problem. Run the same requirement of Rs 3,60,000/- a year through three different assumed ratesA figure somebody chooses so the arithmetic can run. An assumed rate is not a prediction, and none of the three carries any claim to being more likely or more prudent than another. and read the three answers next to each other.

The bottom row is the honest content of this subject, and it is worth sitting with for a moment. Rs 48,00,000/- separates the first answer from the third. Nothing in the household changed between those two rows. The requirement did not move, the years did not move, nobody spent differently and nobody earned differently. An assumption moved by two percentage points and the target moved by Rs 48,00,000/-. The spread is more than nine times what this household spends in a whole year.

The ratio is worth naming plainly. The household's ordinary spending is Rs 42,770/- a month, which is Rs 5,13,240/- across twelve months. The spread between the outer two assumptions is Rs 48,00,000/-, which is 9.4 times that annual figure. A number nobody can settle is nine years of a household's entire spending wide.

All three of those rates are this guide's own assumptions, chosen to show a spread. None is described as typical, expected, historical, standard, conservative or prudent, and none is recommended. The three rates are points on a scale that runs continuously, picked because three is enough to see the shape.

The settings the slider passes through appear here as text too, so they survive with the panel closed. At an assumed 3 per cent, Rs 3,60,000/- a year implies a pot of Rs 1,20,00,000/-, and the household's Rs 6,61,000/- is 5.5 per cent of it. At an assumed 4 per cent, Rs 90,00,000/- and 7.3 per cent. At an assumed 5 per cent, Rs 72,00,000/- and 9.2 per cent. The spread between the outer two is Rs 48,00,000/-. At the ends of the scale, 2 per cent implies Rs 1,80,00,000/- and 8 per cent implies Rs 45,00,000/-: four times as much pot at one end as the other, on a requirement that never moved.

Why can nobody say which rate is the safe one?

A figure is usually named at this point in the subject, and the reason none can be named is worth stating directly. Naming a safe withdrawal rate would require knowing two things about a particular household that nobody knows, so no safe rate can be named.

The first is how many years the money has to cover. The number of years is not merely hard to find, and better data would not improve it. For any single household the duration is unknowable in the strict sense, and a table that describes a population does not describe the person reading it.

The second is absorbable variationHow much a household can take without changing what it does. A household with somewhere to fall back on can absorb more than one with nowhere.: how much movement in the money the household can take without having to change what it does. Two households with identical pots and identical requirements answer this completely differently. One has a paid-for roof, grown children with room in their own sheets, and a habit of spending under what arrives. The other has rent every month, nobody to fall back on, and no line that can be cut without something real going. The second can absorb far less, and so cannot take the same share out each year with the same comfort, on identical money.

So the rate is not a property of the money. The rate states a specific household's tolerance for a specific duration. Both halves of the statement are private, both are unknowable in advance, and both change. Any source that prints a figure has quietly asserted two facts about a reader it has never met.

There is a second reason for the refusal, and it is about who the reader might be. Somebody reading this may have no scheme at all, may be self-employed with no statement to look at, may be fifty with nothing set aside, or may be doing this arithmetic for a parent. A single printed rate reads to every one of those readers as a standard they have failed. A printed rate is not a standard. The rate is a knob on a calculation, and different hands turn it to different places for perfectly good reasons.

Showing the width is what can honestly be done instead. The table, the diagram and the panel above show it: the same requirement, three defensible-looking assumptions, and Rs 48,00,000/- between the outer two. A reader who leaves knowing that the width exists and roughly how wide it is has been given something true. A reader who leaves with a single figure has been given something invented.

Where does this household actually stand today?

Now the uncomfortable part. A household is better served by knowing where it stands than by being reassured about where it stands.

At 31 March of year two, the Bhosale household has three things that count towards this goal and nothing else. Meghna Bhosale's Employees' Provident Fund holds Rs 4,12,000/-, built over eleven years of service at Sahyadri Freight Services Private Limited, invented, at a lower salary for most of them. The household's public provident fund account holds Rs 84,000/-, carried on its sheet from the first sequence onward. And the gratuity accruedEarned so far, whether or not it has actually been received. An accrued amount is a claim that has built up, not cash in hand. on her service so far is Rs 1,65,000/-, which is earned rather than held and therefore not spendable today.

The Bhosale household holds 9.2 per cent of the smallest of the three pots, and there are 24 years before Meghna Bhosale is 60.

Separating those two facts makes the first frightening and the second invisible, so read them as one. Meghna Bhosale is 36 at 31 March of year two. Twenty four years is 288 months, and Rs 6,240/- a month is already going into the provident fund, being Rs 3,120/- from her pay and Rs 3,120/- from the employer side, both invented rupee amounts. The 9.2 per cent is a photograph of a position at one date. The 24 years are the part still moving, and they move whether or not anybody looks at the sheet.

Nothing about that 9.2 per cent means the household has done anything wrong. The household started, it is contributing, and it has 24 years. It has also just done the single most useful thing available on this subject, writing the number down while there is still time for the number to change. A household that has never worked the figure out has the same position and does not know it. Knowing is strictly better, and it is uncomfortable, and both of those are true at once.

A share at one date is not a verdict, and it says nothing about where the arithmetic lands. The accumulated amount is the smallest of the three inputs at this point, and the contributions and the years have not done their work yet. Nor is the share a comparison with anybody. Ashok Bhosale's tailoring counter carries no scheme, so nothing on his side counts towards the goal. Carrying no scheme is the ordinary position for a self-employed person rather than an oversight by him.

What happens if the money runs out?

The question is usually left out, and leaving it out is what makes the whole subject feel like a threat with no shape. So here is the plain version, with nothing dressed up and nothing softened.

If the money runs out, the household does not disappear. The income that was arriving from the pot stops, and the monthly requirement has to be met some other way. There are only four other ways, and every household in this position uses some combination of them: spend less than the sheet said, which usually means the health and food lines because the discretionary ones went first; lean on people who have their own sheets, most often adult children who then carry two households on one income; keep working past the point the body wanted to stop, which is what a tailoring counter or a stall usually means in practice; or turn a held asset into money, which for most households means the place they live in.

Running out is not an event with a date and a letter; it is the point at which the requirement stops being met by the pot and starts being met by other people, by continued work, or by spending less on things that were not optional.

Two things follow. The first is that the consequence is shared. A pot that falls short lands on the next generation's sheet, at exactly the age when that generation is running its own version of this calculation, and the sharing is what gets missed when the subject is treated as one person's arithmetic. Ira Bhosale is seven at 31 March of year two. She will be 31 when her mother is 60.

The second is that none of the four routes is shameful. Households have supported their elders for as long as there have been households, and doing so is not evidence of anybody's failure. The arithmetic does not offer a way to avoid all four routes, and for many households some of the four will not be available. The arithmetic offers the chance to know years in advance that the routes are in play, and knowing years in advance is the difference between a decision and a surprise.

Once money is being drawn, does only the total matter?

While money is going in, the household can think in totals. A year's contributions are a year's contributions and the order in which the months arrived makes no difference to the balance. Once money is coming out, that stops being true, and the reason is simple enough to draw.

The pot has a floor. A pot cannot go below zero, and a year in which more is needed than remains is a year the pot cannot fund, whatever the total across the whole period says. So the sequence of outcomesThe order in which things happen. The pot has a floor at zero, so once money is being drawn out the order can matter as much as the total. starts to carry weight of its own, entirely separately from the size of the pot.

Take a deliberately simple illustration with no rates in it at all. A pot holds Rs 6,00,000/-. Across five years, four need Rs 1,00,000/- each and one, a heavy year, needs Rs 4,00,000/-. The total needed is Rs 8,00,000/- against a pot of Rs 6,00,000/-, so the pot cannot cover all five in any order. The order still decides how many of the five years get covered.

Put the heavy year first. The pot pays Rs 4,00,000/- in year one and has Rs 2,00,000/- left, enough for years two and three, and years four and five get nothing at all. Three years covered.

Put the heavy year last. The pot pays Rs 1,00,000/- in each of years one to four, spending Rs 4,00,000/-, and has Rs 2,00,000/- left when the heavy year arrives needing Rs 4,00,000/-. The heavy year is half covered. Four years covered and part of a fifth.

The same Rs 6,00,000/- was drawn in both orders and the total needed was identical, but one ordering covered three years and the other covered four and a half. The gap between three years and four and a half is what it means to say the order matters once money is coming out.

The ordering effect is not a technical point for people with portfolios. The floor at zero is why a household with a heavy year early is in a materially different position from one with the same heavy year late, on identical money, and neither chose which it got. The same floor is also why a rate written once at the start, assuming every year looks like the average year, is a fragile thing to lean on.

What can a household do about this while somebody is still working?

Four things, none of which requires an adviser, a product, a scheme or a rupee that the household does not already have. The honest list is short.

The first is to write the assumption down as an assumption. Not the pot, the rate, and beside it the words this is my own assumption and nobody told me it. A bare figure looks like a fact and a labelled assumption looks like a question that has been parked, so the two behave completely differently on a sheet.

The second is to carry a range instead of a figure. Rs 72,00,000/- to Rs 1,20,00,000/- is more useful than Rs 90,00,000/- and more honest. The width is the real content. A household that knows its target has a Rs 48,00,000/- width knows something true about its own position. A household carrying a single figure knows something false with great precision.

The third is to put a date on the sheet and honour it. Not a vague intention to look again, a date: 31 March, every year, when the statements arrive anyway. The events that move the rate are ordinary household events, and none of them announces itself as relevant to a retirement sheet, so the only thing that reliably brings somebody back to it is a date.

The fourth is the one that actually moves the arithmetic, and it belongs last because the first three are what make it possible to judge. Of the three levers named earlier, the amount going in and the length of the funded period are the two a household has any hold on before the drawing starts. The rate is not a lever; it is a description of a position. How to pull either of the two real levers is a decision for the household itself, about its own money and its own life.

Before longevity risk arrives, a household can make the assumption visible, keep it as a width, put a date against it, and know which two of the three levers it holds.

What do people actually do with a withdrawal rate?

Three uses, and it is worth seeing them because the figure looks academic until somebody watches it being used.

A household uses it to convert a balance it can see into an income it can picture. Somebody at a kitchen table with a provident fund statement in front of them cannot feel what Rs 4,12,000/- means. A balance has no unit a household thinks in. Divided by twelve at some assumed share, it becomes rupees a month, and rupees a month is the only unit a household actually reasons in. The translation is the reason the figure exists at all, and the reason it gets adopted carelessly.

Somebody assessing a loan application uses it in reverse, and this is where a household meets it without knowing what it is. When an applicant has no salary and offers a holding instead, whoever assesses the application has to turn that holding into a serviceable monthly figure, and the share applied for that purpose is a withdrawal rate wearing a different name. An applicant who has never seen this arithmetic cannot understand why the same holding supports very different assessments in different places. Different assessors assume different shares, on their own criteria.

Somebody reading a household's position from outside, whether an analyst looking at a segment or a relative helping a parent, uses the width rather than the figure. The most informative thing about a household's retirement arithmetic is not the target written down but how far apart its defensible targets are. The width says how much of the plan is money and how much is assumption. A household whose target moves by Rs 48,00,000/- on a two percentage point change in an assumption is holding a plan that is mostly assumption, and knowing that is more useful than any of the three targets on their own.

None of those three uses requires anybody to settle on a rate. All three work on a range, and the first and third work better on one.

What can nobody settle about this?

Two things, and naming them is the last part of the mechanism rather than a disclaimer bolted to the bottom.

The first is which withdrawal rate is safe. A safe rate is not a fact about money, so which rate is safe cannot be settled. A safe rate would be a joint statement about a duration nobody knows and a tolerance only the household knows, and any figure printed for it has manufactured both.

The second is whether this household, or any household, will be all right. Nobody can know that in advance. The household holds 9.2 per cent of the smallest of three pots at 31 March of year two, and it has 24 years, Rs 6,240/- a month going in, and two adults whose working lives are not finished. The position, the years, the monthly contribution and the two unfinished working lives are the facts. Where the facts land depends on things nobody has any access to.

The honest output of this whole subject is a width and a set of levers rather than a verdict, and a verdict handed to a household replaces the two things nobody knows with two things invented.

Five things survive and are worth carrying. Longevity risk is caused by something nobody would prevent. A withdrawal rate translates a balance into an income and runs in both directions. On an unchanged requirement of Rs 3,60,000/- a year, three defensible-looking assumptions produce targets Rs 48,00,000/- apart. The rate is a fact about a household rather than about a pot. The rate has to be revisited, and nothing on a sheet reminds anybody to. And once money is coming out, the order matters as well as the total. All five hold whatever rate anybody assumes and whatever duration turns out to be true.

The Bhosale household, Meghna Bhosale, Ashok Bhosale, Ira Bhosale and Sahyadri Freight Services Private Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.