Why an Average Return Can Mislead You Badly
Add up the yearly returns, divide by the number of years, and you get a figure that is arithmetically correct and financially useless. On the Indian record it overstates what money actually did by enough to more than double the answer over a working lifetime — and the size of the error is not an accident of these particular years. It is a property of any series that moves about.
Updated 10 September 2026
Two numbers that both claim to be the return
Take the 26 complete calendar years in our record, from 2000 to 2025, and ask what the market returned per year. There are two defensible answers and they are not close together.
Add the yearly returns and divide by 26, and the answer is +16.79%. Ask instead what single rate would have turned the starting amount into the finishing amount, and the answer is +13.21%. The gap between them is 3.58 percentage points.
Only the second one describes anybody's money. The first is a summary of a list of numbers that happen to be returns, and it answers a question nobody asked.
What the difference does over a lifetime
Percentage points are easy to shrug at, so here it is as money. Take ₹100 at the start of 2000 and leave it alone.
Compounded at the average, it would have become ₹5,655. What it actually became was ₹2,517. The average overstates the outcome by a factor of 2.2 — over a single working lifetime, using the correct arithmetic on the wrong number.
That is the whole article, and everything below explains why it happens and where you will meet it.
Why it happens, and why it always goes the same way
The reason is not statistical subtlety. It is that losses and gains of the same size do not cancel.
A fall of half needs a doubling to get back to level, not another half. So a year that loses a great deal does more damage than a year of the same size in the other direction repairs — and an average, which treats the two as equal and opposite, cannot see that.
The record contains the cleanest possible demonstration. Our worst year was 2008, at -51.27%. The year that followed, 2009, returned +77.59% — the strongest year in the whole record. Average those two and you get +13.16%, which sounds like a perfectly good couple of years. Actually hold money through both and ₹100 becomes ₹87: a return of -6.97% a year. Compounding the average instead would have told you to expect ₹128.
Those two years were not chosen because they make the point well. They are the worst year in the record and whatever came next — a rule fixed in the code that produces this page, so the example cannot be quietly swapped for a better one.
The general version is that the average always exceeds the rate that actually compounds, and the gap grows with how spread out the returns are. In this record the yearly returns have a spread of 27.7%, and half the square of that is 3.85 percentage points — close to the 3.58-point gap we actually observe, and close for a reason rather than by coincidence. It is the standard approximation to this effect, and its accuracy here is the evidence that the gap is a property of the volatility rather than a quirk of these years.
The years themselves
Show these numbers as a table
| What the market returned in a calendar year | Number of periods | Share |
|---|---|---|
| -51.3% to -39.6% | 1 | 3.8% |
| -27.8% to -16.1% | 1 | 3.8% |
| -16.1% to -4.4% | 2 | 7.7% |
| -4.4% to +7.3% | 5 | 19.2% |
| +7.3% to +19.0% | 6 | 23.1% |
| +19.0% to +30.7% | 5 | 19.2% |
| +30.7% to +42.4% | 3 | 11.5% |
| +54.2% to +65.9% | 1 | 3.8% |
| +65.9% to +77.6% | 2 | 7.7% |
Look at where the average sits relative to the bars. 15 of the 26 years came in below it. That is not a paradox — it is what an average does when a few results are far larger than the rest, and it is why "the average year" describes a year that mostly did not happen.
| Year | Return that year | Average of the years so far | What actually compounded, so far |
|---|---|---|---|
| 2000 | -13.36% | -13.36% | -13.36% |
| 2001 | -15.05% | -14.21% | -14.21% |
| 2002 | +5.34% | -7.69% | -8.14% |
| 2003 | +76.61% | +13.38% | +8.17% |
| 2004 | +13.04% | +13.31% | +9.13% |
| 2005 | +38.63% | +17.53% | +13.57% |
| 2006 | +41.90% | +21.01% | +17.24% |
| 2007 | +56.80% | +25.49% | +21.58% |
| 2008 | -51.27% | +16.96% | +9.84% |
| 2009 | +77.59% | +23.02% | +15.24% |
| 2010 | +19.22% | +22.68% | +15.60% |
| 2011 | -23.81% | +18.80% | +11.65% |
| 2012 | +29.43% | +19.62% | +12.93% |
| 2013 | +8.07% | +18.80% | +12.57% |
| 2014 | +32.90% | +19.74% | +13.83% |
| 2015 | -3.01% | +18.31% | +12.69% |
| 2016 | +4.39% | +17.50% | +12.19% |
| 2017 | +30.27% | +18.21% | +13.12% |
| 2018 | +4.64% | +17.49% | +12.66% |
| 2019 | +13.48% | +17.29% | +12.70% |
| 2020 | +16.14% | +17.24% | +12.86% |
| 2021 | +25.59% | +17.62% | +13.41% |
| 2022 | +5.69% | +17.10% | +13.06% |
| 2023 | +21.30% | +17.27% | +13.40% |
| 2024 | +10.09% | +16.98% | +13.26% |
| 2025 | +11.88% | +16.79% | +13.21% |
The last two columns are the ones to follow down the page. They start together and come apart, and they never converge again, because the gap between them is created by the spread and the spread does not go away.
Where you will actually meet this
A fund or product literature quoting a simple average of yearly returns. It is not usually wrong on purpose; adding and dividing is what a spreadsheet does by default. The check is to ask what a single sum invested at the start would be worth now, and to work the rate back from that. Our page on CAGR against XIRR sets out both calculations properly.
A planning tool that projects forward at an expected return. This is the more expensive case, because the error compounds in the same direction for thirty years. If the expected return was estimated as an average of historical years, a projection built on it will overshoot — quietly, smoothly and by a great deal, in exactly the manner shown above.
Any statement about how a portfolio has done that does not say which calculation it used. The two numbers differ by more than most people's entire equity risk premium. A figure that does not say which one it is has not told you very much.
What this does not mean
It does not mean the average is a mistake. It is the right statistic for a different question — what a single year drawn at random looks like. If you are asking how much a year might vary, the average and the spread are exactly what you want. If you are asking what money does over many years, they are not.
It does not mean the compound rate is safe to project forward. It describes what happened over one particular stretch of one market. What the range of outcomes actually looks like, and how little it narrows, is the subject of why long-term equity returns remain uncertain, and a single compound rate hides that range just as thoroughly as an average hides this one.
And it is not evidence about any strategy. Nothing here says shares are good or bad. It is an arithmetic property of a sequence of numbers, demonstrated on the Indian record because that is the record our readers are planning against.
What to take away
When somebody tells you a return, ask which of the two numbers it is. If they added up years and divided, the figure is too high, and it is too high by more the more the years varied.
For anything to do with your own money over more than one year, the number you want is the one that turns the start into the end. It is less flattering, and it is the only one that describes what you would actually have.
Educational content only. This is not personalised financial, investment or tax advice. Figures quoted are historical or illustrative and are not forecasts. Consult a qualified professional before acting on anything you read here.