Ask most people how big a retirement corpus should be and they will say 25 times your annual expenses. It is easy to remember and it travels well on WhatsApp.
Very few people forwarding it can say where the 25 comes from, or why two retirees with the same corpus, expenses and average return can end up one comfortable and one broke.
In my retirement planning class in Ahmedabad, this is the session where the room goes quiet. What surprises them is that many have quoted 25x to clients for years without knowing it rests on American market data from 1926, a 30-year retirement and a strict withdrawal habit. This is that session, written down.
Where the 25x rule comes from
25x is the other side of the “4 percent rule”, which comes from one paper: William Bengen’s “Determining Withdrawal Rates Using Historical Data,” Journal of Financial Planning, October 1994. The link is one line:
Corpus = Annual expenses ÷ 4% = Annual expenses × (100 ÷ 4) = Annual expenses × 25
So 25 is simply 1 divided by 4 percent. Whether 4 percent deserves that place is what Bengen tested.
He took a portfolio of 50 percent US stocks and 50 percent intermediate-term US government bonds, using Ibbotson historical data, and simulated a retiree starting in 1926, another in 1927, and so on through the mid-1970s. Each withdrew a fixed first-year percentage, raised it by actual inflation every year whatever the market did, and he counted how long the money lasted.
- At 3 percent, and up to about 3.5 percent, no historical retiree ran out in less than 50 years. He called this “absolutely safe”, to the extent history is a guide.
- At 4 percent, no case ran out before 33 years, and most lasted 50 or more. This is where the rule comes from.
- At 4.25 percent, the worst case lasted 28 years.
- At 5 percent, people who retired in the late 1960s and early 1970s had only about 20 years of money. In his words, clearly not enough for their lifetime in most cases.
A quarter of a point more cost five years in the worst case. A full point cost about thirteen. The rule sits close to a cliff edge, and the WhatsApp forward never mentions it.
It also came with conditions: a 30-year retirement, US market history, equity between 50 and 75 percent throughout (below 50, portfolios ran out sooner; above 75, more risk without more years), and the discipline to take only the inflation-adjusted amount every year, crashes included. Change any of these and the 4, and so the 25, changes with it.
I tell students to use 25x to find roughly which range of corpus they are in. It is a starting estimate, not the answer.
Why average returns mislead
Bengen’s paper opens with a planner who reasons sensibly. Stocks averaged 10.3 percent, intermediate bonds 5.1 percent, inflation 3 percent. A 60/40 portfolio gives (0.6 × 10.3) + (0.4 × 5.1) = 6.18 + 2.04 = about 8.2 percent, roughly 5 percent after inflation. So the planner says: withdraw 5 percent, raise it with inflation, and the corpus lasts forever.
In history it ruined people. In 1973 and 1974 together, US stocks fell 37.2 percent while cumulative inflation was 22.1 percent. Every ₹100 of stock became ₹62.80, while the ₹5 withdrawal had to become ₹6.11 to buy the same groceries. It was 5 percent of that money going in and close to 10 percent of what was left coming out. No later recovery could rescue those retirees.
His conclusion: average returns and average inflation are not a sound basis for deciding what a client can safely withdraw.
One more point from the paper matters for India. The Depression-era retiree, who saw a huge crash but falling prices, was hurt less than the 1970s retiree, who had a smaller crash with high inflation. Inflation hits spending power and the portfolio at the same time, and in India inflation is the normal condition.
Sequence of returns risk
I make every student work this out by hand at least once.
Sequence of returns risk means the order of your returns can hurt you even when the average is fine. Once you are withdrawing, units sold after a fall are gone and miss the recovery.
An example, illustrative and not a forecast. Retirees A and B each start with ₹1 crore and withdraw ₹6 lakh at the start of every year. Both get the same six returns: minus 15, minus 5, plus 10, plus 10, plus 18 and plus 25 percent. A gets them in that order. B gets them in reverse. Each year, take out the withdrawal, then apply the return to what is left.
| Year | Retiree A: return | A: corpus at year end | Retiree B: return | B: corpus at year end |
|---|---|---|---|---|
| 1 | -15% | (100.0 – 6) × 0.85 = ₹79.90 lakh | +25% | (100.0 – 6) × 1.25 = ₹117.50 lakh |
| 2 | -5% | (79.90 – 6) × 0.95 = ₹70.21 lakh | +18% | (117.50 – 6) × 1.18 = ₹131.57 lakh |
| 3 | +10% | (70.21 – 6) × 1.10 = ₹70.63 lakh | +10% | (131.57 – 6) × 1.10 = ₹138.13 lakh |
| 4 | +10% | (70.63 – 6) × 1.10 = ₹71.09 lakh | +10% | (138.13 – 6) × 1.10 = ₹145.34 lakh |
| 5 | +18% | (71.09 – 6) × 1.18 = ₹76.80 lakh | -5% | (145.34 – 6) × 0.95 = ₹132.37 lakh |
| 6 | +25% | (76.80 – 6) × 1.25 = ₹88.50 lakh | -15% | (132.37 – 6) × 0.85 = ₹107.42 lakh |
Same start, the same ₹36 lakh withdrawn, the same average return of about 6.3 percent a year compounded. A ends with ₹88.50 lakh, B with ₹107.42 lakh. A gap of nearly ₹19 lakh in six years, caused only by the order of the returns.
For A, the ₹6 lakh is 6 percent of the corpus in year one, 7.5 percent of ₹79.90 lakh in year two and 8.5 percent of ₹70.21 lakh in year three. For B it goes the other way: 5.1 percent, then 4.6 percent. Neither spent a rupee more. The market changed the denominator.
Remove the withdrawals and both end at exactly ₹1.44 crore, because 0.85 × 0.95 × 1.10 × 1.10 × 1.18 × 1.25 = 1.4412 in any order. So sequence risk is a retiree’s problem, not an accumulator’s, and the five years before retirement and the ten after are the riskiest stretch of anyone’s financial life.
Bengen’s 1929 retiree saw his 4 percent withdrawal become 7.6 percent of a shrunken portfolio within four years, while a neighbour who retired three years later into a recovery never noticed a problem. Neither was smarter. The test I apply to any retirement plan is what it does when the bad years come first.
Withdrawal rates for Indian retirees
I would not simply import 4 percent, for three reasons.
The first is inflation. Bengen’s retirees lived with an average of 3 percent. India’s CPI inflation for July 2026 was 4.45 percent year on year (MoSPI, base year 2024), with food at 5.52 percent. The RBI targets 4 percent with a tolerance band of 2 to 6 percent, retained for April 2026 to March 2031.
A retiree’s own inflation is usually higher still, because their spending leans towards food, household help and health care. A retired household can easily be living with 6 percent. On ₹9 lakh of annual expenses, the figure I use through the rest of this article, 30 years at 4.45 percent makes the final year cost about ₹33.2 lakh. At 6 percent it is about ₹51.7 lakh, roughly ₹18 lakh a year apart, from one cell in a spreadsheet. I would rather assume 6 percent and be pleasantly surprised than assume 4 percent and leave a client short in their 80s.
The second is how long the money has to last. Bengen’s 4 percent was built for 30 years. Retire at 60 and live to 90 and you have used all of it. Retire at 50, or plan for a couple where one partner may reach the mid-90s, and you need 35 to 40 years. There his 3 to 3.5 percent range is the better guide, which means 28x to 33x expenses.
The third is what the safe option pays. The Senior Citizen Savings Scheme pays 8.2 percent a year for July to September 2026, capped at ₹30 lakh per person. Against 4.45 percent CPI that is a real return of about 3.6 percent before tax (1.082 ÷ 1.0445 = 1.0359), and SCSS belongs in most retirement plans. But a couple can put in at most ₹60 lakh, which gives ₹4.92 lakh a year before tax, a little over half of ₹9 lakh of expenses, and less once tax is paid and the rate resets. That is why even a 65-year-old needs equity. In Bengen’s simulations, portfolios with less than 50 percent equity ran out sooner.
Put together, this is how I translate it for India:
| First-year withdrawal rate | Corpus as a multiple of expenses | Who it suits |
|---|---|---|
| 4% | 25x | 30-year horizon, disciplined withdrawals, 50%+ equity, some flexibility to cut spending in bad years |
| 3.5% | about 28.6x | Longer horizons, higher personal inflation, or a desire to leave an estate |
| 3% | about 33.3x | Early retirees (45 to 50), single-income longevity risk, low flexibility in expenses |
The shorthand I give students: in India, feel uneasy below 25x, 30x lets you sleep, and 33x is the price of retiring early.
Working out the corpus with the SWP formula
Multiples are shortcuts. To test assumptions, work the corpus out from scratch. An inflation-adjusted SWP (systematic withdrawal plan) is a stream of withdrawals growing at inflation g, paid from a corpus earning r, so the corpus is the present value of a growing annuity:
Corpus = W × [1 − ((1+g) ÷ (1+r))^n] ÷ (r − g)
where W is the first year’s withdrawal, g is inflation, r is the portfolio return, and n is the number of years the money must last.
Illustrative assumptions, not promises: expenses of ₹75,000 a month, so W = ₹9,00,000 a year; g = 6 percent; r = 9 percent; n = 30 years.
- (1+g) ÷ (1+r) = 1.06 ÷ 1.09 = 0.97248
- 0.97248^30 = 0.4329
- 1 − 0.4329 = 0.5671
- r − g = 0.09 − 0.06 = 0.03
- 0.5671 ÷ 0.03 = 18.90, your personal multiple
- Corpus = ₹9,00,000 × 18.90 = about ₹1.70 crore
The formula cares about the gap between r and g far more than either number. Drop the return to 8 percent with the same inflation and the multiple becomes about 21.5, taking the corpus to about ₹1.93 crore. One point of spread is worth roughly ₹23 lakh here, more than two years of the client’s spending. That is why costs matter once withdrawals start, why asset allocation matters more than fund selection (see Why Asset Allocation, Not Fund Selection, Decides Most of Your Returns), and why an all-FD retirement fails quietly. FDs are not bad products, but after tax they earn close to personal inflation, and as r − g approaches zero the multiple you need keeps rising without limit.
The formula says 18.9x and the rule says 25x. Both are right, because the formula assumes 9 percent arrives every year like a pension, and markets deliver a 9 percent average as minus 18 one year and plus 31 another. The distance between about 19x and the historical worst-case range of 25x to 33x is what you pay for volatility and bad luck. On ₹9 lakh of expenses that is ₹1.70 crore against ₹2.25 crore, roughly ₹55 lakh held as insurance against the order of returns. When a client asks why they cannot retire on the formula number, I tell them the formula assumes the market has read their plan.
The same ₹9 lakh at r = 9 percent, showing the multiple of first-year expenses needed (same formula, illustrative):
| Personal inflation | 25 years | 30 years | 35 years |
|---|---|---|---|
| 5% | 15.2x | 16.9x | 18.2x |
| 6% | 16.7x | 18.9x | 20.8x |
| 7% | 18.5x | 21.3x | 23.9x |
Getting inflation two points wrong on a 35-year horizon, 18.2x against 23.9x, is a corpus error of more than ₹51 lakh. No choice of fund moves the result that much.
The bucket strategy
The maths says hold 50 percent or more in equity for three decades and keep withdrawing calmly through every crash. The retiree says, “My corpus fell ₹40 lakh this year and I cannot sleep.” Buckets are a way to live with both.
Take an illustrative retiree with ₹2.25 crore (25x of ₹9 lakh) and 6 percent personal inflation.
Bucket 1 holds years 1 to 3 of expenses in near-cash: liquid funds and sweep FDs, and for those over 60, SCSS at 8.2 percent fits here and in Bucket 2. Size: ₹9,00,000 + ₹9,54,000 + ₹10,11,240 = ₹28.65 lakh, about 13 percent.
Bucket 2 holds years 4 to 8 in high-quality debt and conservative hybrid funds: ₹9,00,000 × (1.06³ + 1.06⁴ + 1.06⁵ + 1.06⁶ + 1.06⁷) = about ₹60.42 lakh, roughly 27 percent.
Bucket 3 is the rest, in equity: ₹2.25 crore − ₹28.65 lakh − ₹60.42 lakh = about ₹1.36 crore, roughly 60 percent.
That 60 percent sits inside Bengen’s 50 to 75 percent range, and not by coincidence. Mathematically, a bucket plan is just an asset allocation with a better story attached. Some researchers say the cash bucket drags returns against a plain rebalanced portfolio, and over some periods they are right.
I still use it in my own practice as a distributor, because real plans rarely fail on arithmetic. They fail when a client sells ₹1.36 crore of equity in the third week of a crash. Buckets 1 and 2 together hold ₹89.07 lakh, about 40 percent of the corpus and eight years of spending, outside the stock market. A retiree who can see that behaves differently in a 2020-style fall from one staring at a single figure down 25 percent.
The protection lives in the refill rule, which most people skip. Each year, refill Bucket 1 from Bucket 2 and Bucket 2 from Bucket 3, but only after a reasonable year for equity. After a crash, stop refilling and let Buckets 1 and 2 run down while equity recovers. Most historical falls have healed well within eight years. After a strong year, refill fully, which is simply selling equity high. Done properly, it is ordinary rebalancing in a form a retiree can live with.
Putting it together for a client
Start with the real expense figure, including what clients forget: health insurance premiums that rise with age, home maintenance, and the yearly money to the children that nobody budgets for but everybody spends.
Choose personal inflation honestly. The official 4.45 and 5.52 percent adjusted upwards for what a retiree actually buys. Work out the corpus from the formula, then test it with the return one point lower, inflation one point higher and five more years. If the plan survives all three, it is the plan. Set the withdrawal rate below the formula’s answer, using 25x to 33x as the check.
Put the corpus in buckets, with the refill rule written down and agreed while markets are calm. Then review every year. If markets have been kind, do not raise withdrawals. Bengen warned about exactly this. If they have been harsh, a 5 to 10 percent spending cut for two years does more than any fund switch. On ₹9 lakh that is ₹45,000 to ₹90,000, aimed straight at the problem that hurt Retiree A.
That judgement is the hard part, and no formula does it for you. The CFP curriculum’s retirement modules teach it properly, with the annuity and tax parts I have left out, in the CFP programme.
But you do not need a course to plan your own family’s retirement. Everything above fits in one spreadsheet, and if your own money is all you want to manage, an afternoon with it may be enough. We have written separately about whether learning financial planning yourself is worth it.
The 25x rule did one useful thing: it got people to multiply their expenses by something instead of guessing. Start there, then do the maths.
Frequently asked questions
Does the 4 percent rule (25x corpus) work in India?
As a starting estimate, yes. William Bengen derived it from US market history, assuming a 30-year retirement, 50 to 75 percent equity and strict inflation-adjusted withdrawals. India runs hotter (the RBI’s own target is 4 percent within a 2 to 6 percent band, and July 2026 CPI was 4.45 percent), and a retiree’s spending usually inflates faster than the CPI. In practice, 25x is a reasonable floor for a 30-year horizon with equity exposure and some spending flexibility, while 28x to 33x suits longer horizons or less flexibility.
How much corpus do I need to retire with expenses of ₹1 lakh a month?
As an illustration, ₹1 lakh a month is ₹12 lakh a year. The growing-annuity formula with 9 percent returns, 6 percent inflation and 30 years gives ₹12,00,000 × 18.9, about ₹2.27 crore. The 25x rule gives ₹3 crore and 30x gives ₹3.6 crore. So the answer is a range, roughly ₹2.3 crore to ₹3.6 crore, and where you sit in it depends on your equity allocation, your horizon and whether you can cut spending in bad years. Anyone who gives you a single number without asking those questions is guessing.
What is sequence of returns risk in simple words?
It is the risk that bad market years come early in your retirement, when withdrawals turn temporary losses into permanent ones. Two retirees can earn the same average return and withdraw the same amounts, yet end up lakhs apart only because the returns came in a different order. In the six-year example above, the gap was nearly ₹19 lakh on a ₹1 crore start. That is why planners keep several years of expenses outside equity around the retirement date.
Is an SWP from mutual funds better than putting everything in SCSS and FDs?
They do different jobs, and most good plans use both. SCSS currently pays 8.2 percent with a ₹30 lakh per-person cap, so a couple that fills it earns ₹4.92 lakh a year before tax. That makes it good for the safe buckets but too small, and too exposed to rate resets, to carry a plan for decades alone. An SWP from a diversified portfolio provides growth that can beat inflation, with market risk that has to be managed through allocation and a cash buffer. The bucket structure is how you combine the two.
Does the bucket strategy guarantee my money will last?
No. Mathematically, a bucket plan is an asset allocation with a withdrawal order attached, and no allocation can guarantee an outcome. What buckets do is stop you selling equity in a crash, both mechanically (you spend from cash and debt first) and psychologically (you can see years of expenses sitting safe). Panic-selling in a downturn is the most common way retirement plans actually fail, so that is worth having.
What withdrawal rate should someone retiring early, at 45 or 50, use?
Lower than a 60-year-old’s, because the corpus may need to last 40 years or more, well beyond the 30 years the 4 percent rule was built for. Bengen’s own data put the never-failed-in-50-years zone at around 3 to 3.5 percent, which means 28x to 33x expenses. Test the plan with higher personal inflation too, and keep a meaningful equity allocation, since four decades gives inflation longer to compound against you.
Sources
- William P. Bengen, “Determining Withdrawal Rates Using Historical Data,” Journal of Financial Planning, October 1994 (FPA reprint): https://www.financialplanningassociation.org/sites/default/files/2021-04/MAR04%20Determining%20Withdrawal%20Rates%20Using%20Historical%20Data.pdf
- Ministry of Statistics and Programme Implementation, Press Release of Consumer Price Index on Base 2024=100 for July 2026 (dated 12 August 2026): https://www.mospi.gov.in/uploads/latestReleases/latest_release_1786529680747_3113661d-1a2b-4b9a-af06-b340193ef9a0_Press_Release_CPI_July_2026.pdf
- Reserve Bank of India, Monetary Policy overview (inflation target of 4 percent with a 2 to 6 percent tolerance band, retained for April 2026 to March 2031): https://www.rbi.org.in/scripts/FS_Overview.aspx?fn=2752
- Upstox News, “Senior Citizen Savings Scheme (SCSS) interest rate July-September 2026 announced” (8.2 percent per annum, ₹30 lakh cap): https://upstox.com/news/personal-finance/investing/senior-citizen-savings-scheme-scss-interest-rate-july-september-2026/article-196140/
