- What the trade-off actually claims
- The price is the only thing free to move
- Expected return is an average of things that did not happen
- What happens the moment risk becomes a number
- Six different things called risk
- The measure scores a certain loss as the safest thing you own
- Volatility becomes loss only through a sale
- Time changes which risk you are looking at
- The risk nobody pays you to take
- Both sides of the trade cost something
- Looking at the distribution instead of the summary
- Common questions
What the trade-off actually claims
The relationship is not a promise that taking more risk earns you more. It is that an asset with an uncertain outcome has to be sold cheaply enough that somebody agrees to hold it — and a lower price for the same expected cashflow is, arithmetically, a higher expected return.
That is a much smaller claim than the one usually made, and it points the other way. In the popular version, return is a reward that follows risk, like interest following a deposit. In the actual version, the extra return is a discount demanded up front by the person taking the uncertainty on. It exists at the moment of purchase, in the price, and not at the end in the outcome.
Two things follow immediately, and both matter more than the slogan. If the price does not fall, the compensation is not there — the risk can be present and the premium absent. And because the compensation is an expectation rather than a schedule, the realised result is under no obligation to resemble it.
Everything below is that one distinction, worked out. What the price is compensating for, what happens the moment risk is turned into a number, and which risk the number leaves out.
The price is the only thing free to move
Take two claims, each promising ₹100 a year from now. The first is certain. The second pays ₹120 if things go one way and ₹80 if they go the other, the two equally likely and no way to know which — so its expected payment, the average of what it might pay, is also ₹100.
Suppose both traded at ₹94. The certain claim returns ₹100 on ₹94, which is 6.4%. The uncertain one returns 6.4% on average as well, for exactly the same money, with the added feature that you might get ₹80. Nobody holds the second at that price. So it does not stay at that price.
It falls until someone is willing. At ₹88, the expected return is ₹100 on ₹88, or 13.6% — a little over 7 percentage points above the certain claim. Notice what produced that gap. The expected payment never changed; it was ₹100 throughout. The seller gave up ₹6 of price today. The premium is a price concession, not a bonus paid at maturity.
This is why the relationship is usually stated backwards. Return does not rise because risk was taken. Price falls because risk has to be placed with somebody, and a lower entry price is what a higher expected return is. Same cashflow, cheaper entry.
Which gives the first recognisable failure mode: risk without the discount. An asset can carry every bit of the uncertainty and none of the compensation, because the price never fell — it was bid up by people who wanted the exposure. The uncertainty is a property of the asset. The premium is a property of the price you paid for it.
Expected return is an average of things that did not happen
Stay with the ₹88 purchase. Its expected return is 13.6%. Now list what can actually be received: ₹120, which is 36.4%, or ₹80, which is −9.1%.
Neither of those is 13.6%. The expected return is not one of the possible outcomes — it is the average of two outcomes, and the average occurs in no state of the world. It is a summary of a distribution, and the holder gets a draw from that distribution, once.
Extend that from one period to many and it stops being a technicality. A risk premium is earned on average, over many draws, by whoever is present for all of them. An individual holding an asset over one particular decade gets one particular path. The premium can be positive over the long history and negative over the stretch that happened to be your stretch, and both statements can be true at once without anything having gone wrong.
Here is the part worth keeping. That unreliability is not a flaw in the theory; it is the condition that lets the premium exist at all. If holding uncertain assets reliably paid more, buyers would bid the price up until it no longer did — the discount would be competed away, and with it the compensation. The premium survives only because it can fail to arrive.
Which means an argument of the form this asset returned a great deal, so the risk paid has said nothing. One realised path is one draw. It is evidence about what happened, and almost none about what was expected.
What happens the moment risk becomes a number
None of this can be managed until risk has a number attached, and the number almost everyone reaches for is the dispersion of returns — how far the periodic returns sat from their own average, which is the standard deviation. Wider spread, higher risk.
It won for good reasons. It can be computed from a price series with no extra information, it puts every asset on one scale, and it aggregates across a portfolio through the way holdings move together rather than by simple addition. That last property is the one that makes portfolio construction arithmetic rather than opinion, and it is not a small thing.
But it is a summary, and every summary is a decision about what to discard. Four things get discarded here, and each of them matters to somebody reading a factsheet.
- Direction — a month that surprised on the way up raises dispersion exactly as much as an equally sized fall. No holder has ever experienced those two events as the same event.
- Sampling error — the figure is estimated from a finite run of history, so it has an error bar of its own. Nobody prints the error bar next to it, and a number quoted to one decimal place invites a precision it does not have.
- Overlap — the error is understated further whenever the observations share data. Three-year returns computed every month give windows in which consecutive observations differ by one month out of thirty-six. Treating those as independent readings makes the estimate look far steadier than the evidence supports, and the method is set out in how rolling returns are built.
- Permanence — dispersion describes movement, and movement that reversed is recorded identically to movement that did not.
The ratios built on top — the ones dividing return by dispersion, or by downside dispersion, or measuring sensitivity to an index — each fix one of these and inherit the rest. What each one measures and which of them describes the experience of holding a fund is the subject of the risk measures article; this one is about what kind of object they all are. They are summaries. Read them knowing what each threw away.
Six different things called risk
Most disagreements about whether something is risky are not disagreements at all. Two people are using the word for two different quantities, and each is right about their own. Setting them side by side is more useful than defending any one of them.
| What "risk" means here | How it gets measured | What it captures well | What it discards |
|---|---|---|---|
| Dispersion of returns | Standard deviation of periodic returns | How bumpy the ride was, on one comparable scale | Direction — an upside surprise is scored as risk |
| Dispersion below a line | Deviation of returns falling short of a chosen threshold | Only the half that hurts | Whether any of those falls were ever recovered |
| Depth of the worst fall | Maximum drawdown, peak to trough | The worst moment a holder would have had to sit through | Likelihood — it is one number from one path, with no probability attached |
| Sensitivity to a common driver | Beta against an index | How much of the movement belongs to the market rather than the asset | Everything specific to the asset, which is where a total loss usually comes from |
| Permanent loss of capital | Not measurable from a return series at all | The only outcome that ends the compounding | It has no number, so it drops out of every ratio built from prices |
| Missing the goal | Shortfall against a required amount by a required date | The only definition tied to why the money was invested | Requires the goal to be stated, which every price-based measure lets you skip |
Read down the last column and the pattern is hard to miss. The measures that are easy to compute discard the things a saver cares about most, and the definitions that match what a saver cares about are the ones with no number. That is not a conspiracy. It is what makes a quantity convenient — availability, not correctness, decides which measure ends up on the factsheet.
A goal-linked reading of the same question is the subject of investing against a stated goal, which is the only frame in which "how risky is this" has a determinate answer.
The measure scores a certain loss as the safest thing you own
Here is the sharpest consequence of measuring risk as dispersion, and it lands on the most conservative holdings rather than the racy ones.
A Post Office Savings Account pays 4%. The Public Provident Fund pays 7.1%. The Employees' Provident Fund is credited at 8.25% for FY 2025-26. Each of those rates is fixed by notification rather than set by a price, so a return series built from any of them barely moves. Its dispersion is close to zero, and every dispersion-based measure will therefore rank it as the least risky thing in the portfolio.
Now ask the question the holder actually has. Whether their purchasing power grows depends on the rate relative to inflation over the same period, which nobody knows in advance. If inflation over the holding period runs above the credited rate, the real return is negative — and the measure records that outcome as risk of approximately zero, because the loss arrived smoothly. Certainty is exactly what the measure rewards, whether what is certain is a gain or a shortfall. The arithmetic of nominal against real is set out here.
There is a second thing hiding in those three rates, and it is the more interesting one for an Indian saver. A market price falls until a holder is willing; an administered rate is announced. Both produce something called a return, but only one of them is compensation that holders demanded. When a notified rate changes, no bargain was struck — which is why the reasoning in this article about premiums and price concessions simply does not apply to that part of a portfolio. It is a different kind of number wearing the same units.
Volatility becomes loss only through a sale
A price that falls and recovers has cost a holder who did nothing at all precisely nothing. The fall shows up in every risk statistic and in none of the outcomes. So the practical question is not how much an asset moved, but what converts a movement into a loss that is kept.
Three routes, and they are worth separating because the defences differ completely.
- The forced sale — money was needed on a date nobody chose, and the date fell inside the drawdown. Units get sold at the low, and the paper loss is realised. This is the risk that a holding period and a cash reserve actually address, and the drawdown-phase version of it is worked through in how a withdrawal rate is tested.
- The voluntary sale — nothing forced it except how the fall felt. Mechanically it is identical to the first, which is what makes the behaviour expensive rather than merely uncomfortable, and it is the subject of what happens to instalments during a crash.
- The issuer that fails — no sale required and no recovery available, because the asset stopped being worth something. A default, a fraud, a business that ends. Dispersion has nothing to say about this; the price series was often perfectly calm right up to the point it was not, which is the argument running through credit risk in debt schemes.
Only the third is permanent by nature. The first two are conversions — temporary movement turned into kept loss by a transaction — and that is the useful reframing, because it tells you where the defence sits. It is not in choosing a less volatile asset. It is in not being the seller, which is a matter of when the money is needed and how the holder behaves, neither of which appears in any risk ratio ever printed.
Time changes which risk you are looking at
The most repeated consolation in investing is that equities are less risky over long periods. There is real arithmetic underneath it, and the arithmetic says something more double-edged than the consolation does.
Assume for a moment that each year's return is independent of the year before it. Two consequences follow from that single assumption, and they follow together. The spread of the average annual return narrows as the holding lengthens — good and bad years increasingly offset within the average. The spread of the final rupee amount widens, because each year's uncertainty compounds on a larger base than the last.
Both are true. Only the first one gets quoted, and it is the one that describes a statistic rather than a bank balance. The reader hears that the outcome becomes more predictable; what became more predictable is the annualised average, while the thing they will actually spend became less so.
The assumption deserves the same scrutiny as the conclusion. Independence between years is a modelling convenience, not an observed fact, and whether long-horizon returns revert toward an average is contested rather than settled — a question that needs far more independent history to answer than any market has yet produced. If returns do revert, the narrowing is stronger than the arithmetic above implies. If they trend, it is weaker. The honest position is that the direction is not known, and an article that resolves it for you has overstepped.
What is not in dispute is the effect of a long horizon on the conversions from the previous section. A holder who does not need the money for a long time is far less likely to be the forced seller. That is a real reduction in a real risk, and it is a different claim from saying the asset became safer.
The risk nobody pays you to take
If the naive reading were right — more risk, more return — then the surest way to raise expected return would be to concentrate a portfolio into one company. It obviously raises dispersion. The reading falls apart on exactly this case.
Holding a single share exposes a portfolio to two separate things: whatever moves the market as a whole, and whatever happens only to that company. The second one shrinks as holdings are added, because unrelated company-specific events partly offset each other. It can be reduced without giving anything up, and nobody pays you to bear a risk you could have removed for free. So concentration reliably raises measured risk and does not reliably raise expected return.
That is the single most useful correction to the slogan, and it belongs in the pillar rather than in a footnote. Risk is not a dial that pays out proportionally. Some of it is compensated and some of it is simply borne, and the difference is whether it could have been diversified away.
Which risks the market actually pays for beyond the market-wide one is a judgement question, not a settled one. A large research literature argues that several other exposures — company size, valuation, past strength — have carried compensation too, and an equally serious literature argues that some of those findings do not survive out of sample. Anyone stating that list with confidence is telling you about their priors.
Diversification is the closest thing here to a free lunch, and it is still not free. Spreading holdings gives up the outcome where the one concentrated idea was right. It also depends on holdings not falling together, and that offsetting weakens in a general panic — the reduction shrinks precisely when it is most wanted. Building the mix is the subject of asset allocation, and keeping it is rebalancing.
Both sides of the trade cost something
Paying for certainty costs expected return; accepting uncertainty costs control over the outcome. Both bills are real, and choosing one side does not avoid the other's.
The first is visible in advance — it is the higher price paid for the more predictable claim, the ₹94 rather than the ₹88. Nothing about it is hidden. It is simply paid at purchase rather than felt later.
Accepting uncertainty buys the lower price and the higher expected return, and costs an outcome you do not control. The distribution may be attractive and your single draw from it may still be poor, over precisely the years you needed it not to be. No holding period converts an expectation into a schedule.
And the thing neither side buys is the removal of the risks that never had a premium attached — concentration, an issuer that fails, a sale made at the wrong moment. Those are not priced into anything. They are avoided or they are borne.
Where that leaves an individual reader depends on when the money is needed, what else is available to cover a shortfall, and how a particular person behaves in a fall — three facts an article does not have, which is why one that ends by recommending a level of risk has substituted its own preferences for the reader's circumstances. What it can do is make sure the question being decided is the right one.
Looking at the distribution instead of the summary
The argument above says a single risk number discards most of what a holder wants to know. The remedy is not a better number. It is looking at the spread of outcomes the number was compressing.
FNOTrader's Mutual Funds app runs on the full AMFI NAV history — around 34 million NAV rows — and reports rolling-return distributions across every start date available, alongside invested against value and maximum drawdown. That surfaces the three things this article argues a ratio hides: how far apart the best and worst start dates were, how deep the fall was that a holder would have had to sit through, and how many of the windows finished below water.
Historical distributions describe what happened; past performance does not indicate future results, and the count of tested windows overstates the independent evidence behind them whenever those windows overlap.
FNOTrader is not a SEBI-registered investment adviser or research analyst, and nothing here is a recommendation to hold any asset or any level of risk.
Common questions
Does taking more risk mean earning more return?
No. The relationship runs through price: an asset with an uncertain outcome has to be priced low enough that somebody will hold it, and a lower price for the same expected cashflow is a higher expected return. The compensation is a discount demanded at purchase, not a reward delivered at the end. If the price never fell, the risk is present and the premium is not.
What is the difference between expected return and realised return?
Expected return is the average of a distribution of outcomes; realised return is the one outcome that occurred. Take a claim bought at ₹88 that pays either ₹120 or ₹80, the two equally likely: its expected return is 13.6%, but the only two things that can be received are 36.4% and −9.1%. The expected figure happens in no state of the world. It is a summary, and a holder gets one draw.
Why is standard deviation used to measure risk if it discards so much?
Because it is computable from a price series alone, puts every asset on one comparable scale, and aggregates across a portfolio through the way holdings move together. Those are genuine advantages. What it discards is direction — an upside surprise is scored exactly like an equally sized fall — its own sampling error, and whether any fall was permanent. Availability, not correctness, decides which measure reaches the factsheet.
Why do overlapping windows understate the error in a risk figure?
Because the observations are not independent of each other. Three-year returns computed every month produce windows where consecutive observations differ by one month out of thirty-six and share the other thirty-five. Statistics that assume independent readings therefore treat heavily repeated data as fresh evidence, which makes the estimate look steadier than the underlying history supports.
Is volatility the same thing as risk of losing money?
No. A price that falls and recovers costs a holder who does nothing at all nothing. Movement becomes kept loss through a sale — either a forced one, where money was needed on a date inside the drawdown, or a voluntary one prompted by the fall itself. A third kind of loss needs no sale: an issuer that does not pay. Dispersion measures none of these directly.
Are equities less risky if held for a long time?
Not in the sense the claim is usually made. Two things happen at once, if each year's return is assumed independent of the last: the spread of the average annual return narrows with the length of the holding; the spread of the final rupee amount widens, because each year's uncertainty compounds on a larger base. Only the first is usually quoted. A long horizon does reduce one real risk — the chance of being the forced seller — which is a different claim from the asset having become safer.
Why does concentrating a portfolio not raise expected return?
Because a single holding exposes a portfolio to two things: what moves the market, and what happens only to that company. The second shrinks as holdings are added, so it can be removed at no cost — and nothing has to pay you for bearing a risk you could have removed for free. Concentration reliably raises measured risk without reliably raising expected return.
Is a fixed-rate scheme with no price movement therefore low risk?
It is low risk by the dispersion definition, which is the one that gets measured. Whether purchasing power grows depends on the credited rate against inflation over the same period. If inflation runs above the rate, the real return is negative — and a dispersion-based measure records that as risk of about zero, because the shortfall arrived smoothly rather than in jumps.
How is a market-priced return different from an administered rate?
A market price falls until some holder is willing, so the return embeds a bargain that buyers and sellers actually struck. An administered rate — a small savings scheme, a provident fund credit — is announced by notification. Both are called returns and both are stated as percentages, but only one is compensation that holders demanded, so premium reasoning does not transfer to the other.
Continue reading
More in Advanced Investing · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article