Skip to content
Menu
Finabase Learn

Measure and manage portfolio risk Lesson 2 of 14

Portfolio and risk · Measuring return and risk

Volatility and standard deviation: how investment risk is measured

How to calculate the standard deviation of returns, annualize it, and read it on 33 years of SPY, including the months a bell curve says should not happen.

On this page
  1. The idea
  2. How it works
  3. Reading it: the bell curve and where it stops working
  4. Where it fails or misleads
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding
  10. Exercise

Volatility is how widely an investment's returns are scattered around their average. It is measured with the standard deviation, a statistic that turns "this fund jumps around a lot" into a number that can be compared between funds. When a fact sheet says a fund has "risk" of 15%, this is almost always the number it means.

Volatility is the foundation under most of the other measures in this section. The Sharpe ratio divides by it, beta and correlation are built from it, and value at risk scales it. This lesson shows how to calculate it by hand, how to convert a monthly or daily figure into a yearly one, and what it does and does not capture, using SPY's monthly returns since 1993.

The idea

Two funds can both average 1% a month and feel nothing alike. One returns between 0.5% and 1.5% every month. The other swings between −7% and +9%. The average cannot tell them apart. What differs is the spread, and the standard deviation is a measure of spread.

The figure sorts SPY's 404 monthly returns into bars one percentage point wide. Most months pile up near the average of just under 1%. The shaded band is one standard deviation, 4.25 points, on each side of it. A wide pile means high volatility; a narrow one means low.

SPY monthly total returns, 1993–2026: 404 months 0 10 20 30 40 50 −18% −14% −10% −6% −2% 2% 6% 10% 14% average 0.95% −2 SD −7.6% +2 SD 9.5% Months Shaded: one standard deviation (4.25%) either side of the average. 73% of months fall inside it.
SPY monthly total returns, March 1993 to October 2026. The shaded band is one standard deviation either side of the average; the dashed lines mark two.

Three things are visible before any formula. Most months are small. The pile is roughly bell-shaped. And the left side has a longer tail than the right: the worst months are further from the middle than the best months are.

How it works

Squaring does two jobs. It makes every distance positive, so that returns above and below the average do not cancel. And it gives large misses more weight than small ones: a month that is 6 points from the average counts nine times as much as a month that is 2 points away.

A calculation by hand

Take five monthly returns: +2%, −1%, +4%, −3% and +3%. These are illustrative numbers, small enough to check.

  1. The average is (2 − 1 + 4 − 3 + 3) ÷ 5 = 1%.
  2. The distances from the average are +1, −2, +3, −4 and +2 points.
  3. Squared: 1, 4, 9, 16 and 4. Their sum is 34.
  4. Divide by n − 1 = 4: the variance is 8.5.
  5. The square root of 8.5 is 2.92. The monthly standard deviation is 2.92%.

Annualizing

Volatility is nearly always quoted per year, whatever the data it was calculated from.

The five-month example becomes 2.92% × 3.46 = 10.1% a year. The square root appears because variances, not standard deviations, add up over time when one period's return does not predict the next. Twelve months have twelve times the variance of one month, and therefore √12 times the standard deviation. Risk grows more slowly than time, which is why a year is not twelve times as uncertain as a month.

SPY since 1993

The data: 404 monthly total returns of SPY, March 1993 to October 2026, dividends included.

  • Average monthly return: 0.95%.
  • Monthly standard deviation: 4.25%.
  • Annualized: 4.25% × √12 = 14.7%.

Read as a range, one standard deviation either side of the average is −3.3% to +5.2% for a month. On the yearly scale, an average of about 11.4% (0.95% × 12) with volatility of 14.7% puts a one-standard-deviation year between roughly −3% and +26%.

Reading it: the bell curve and where it stops working

If returns followed a normal distribution, the bell curve of statistics textbooks, about 68% of months would fall within one standard deviation of the average, about 95% within two, and 99.7% within three. SPY's actual months:

Distance from the average Normal curve SPY, 404 months
Within 1 standard deviation 68.3% 72.5% (293 months)
Within 2 standard deviations 95.4% 94.3% (381 months)
Beyond 2, on the loss side 2.3% 4.2% (17 months)
Beyond 2, on the gain side 2.3% 1.5% (6 months)
Beyond 3, either side 0.3% 0.7% (3 months)

The middle of the distribution is a little more crowded than the bell curve, and the loss side is heavier. Seventeen months were more than two standard deviations below average against six above. The three worst were October 2008 (−16.5%), August 1998 (−14.1%) and March 2020 (−12.5%). October 2008 sat 4.1 standard deviations below the average. A normal distribution with SPY's own average and volatility gives that a probability of about 1 in 50,000 months, or once in roughly 4,000 years. It happened within 16 years of the fund's launch. Fat tails: why extreme market moves are more common than models say takes up that gap.

So volatility describes ordinary months well and extreme months badly. "A two-standard-deviation loss" is a fair description of a bad month. It is not a ceiling.

Daily volatility of stocks and funds

The same calculation on daily closing prices over two years, November 19, 2024 to November 17, 2026 (500 daily returns, dividends not included):

Symbol What it is Daily standard deviation Annualized (× √252)
AGG US bond fund 0.29% 4.6%
TLT Long Treasury bonds 0.71% 11.3%
SPY S&P 500 fund 1.07% 16.9%
KO Coca-Cola 1.16% 18.4%
JNJ Johnson & Johnson 1.28% 20.3%
QQQ Nasdaq-100 fund 1.43% 22.7%
GLD Gold fund 1.56% 24.8%
MSFT Microsoft 1.86% 29.5%
NVDA Nvidia 2.82% 44.7%
TSLA Tesla 3.55% 56.3%

Two things stand out. Single stocks are more volatile than the index that contains them, even steady ones like Coca-Cola: an index averages away part of each company's own swings. And a number depends on its window. Gold's 24.8% in this two-year stretch is well above its 17.4% over 2005 to 2026, because the window includes a 10.3% one-day fall on January 30, 2026.

Where it fails or misleads

Volatility is not constant

Volatility of the last 12 months, annualized: SPY and AGG, 2003–2026 0% 10% 20% 30% SPY (stocks) AGG (bonds) 2005 2010 2015 2020 2025
Volatility over the previous twelve months, annualized, for SPY and AGG, 2004 to 2026. Monthly total returns.

A single volatility figure is an average over the whole period. Measured over the previous twelve months, SPY's volatility was as low as 3.9% (the year to November 2017) and as high as 30.4% (the year to August 2009), around a full-period figure of 14.5%. AGG's ranged from 1.6% to 9.9%. Calm and stormy periods come in clusters: a volatile month is more likely to be followed by another volatile month than by a calm one. A volatility estimated in a quiet stretch understates what comes next when the regime changes.

It punishes gains and losses alike

The formula squares distances in both directions. A fund that occasionally jumps 10% in a month is "riskier" by this measure than one that never does, though no investor loses sleep over gains. The Sortino ratio: judging a return by its downside only is built on a version that counts only the shortfalls.

It says nothing about the order of returns

Standard deviation does not care whether the losses came one at a time or five in a row. A fund that falls 3% in each of ten consecutive months has a tiny standard deviation, because every month is the same, and a 26% loss. Maximum drawdown (MDD): the worst fall from a peak, and the way back measures that kind of risk.

The estimate depends on the sampling interval

Daily, weekly and monthly data give different annualized numbers for the same fund, because the square-root rule assumes one period's return is unrelated to the next. Prices that are updated rarely, such as private funds valued once a quarter, look far calmer than they are.

Risk is wider than volatility

Volatility is a measure of fluctuation. A permanent loss, such as a company going bankrupt, running out of money in retirement, or losing purchasing power to inflation in a "safe" account, is a different thing and can occur with low volatility.

Compared with its neighbours

Measure What it captures Unit Blind spot
Standard deviation Typical spread of returns around the average Percent a year Tails, order of returns, direction
Downside deviation Spread of the shortfalls only Percent a year Needs a target; fewer data points
Maximum drawdown Worst peak-to-trough loss Percent One event; depends on the period
Beta Sensitivity to the market's moves Multiple of market Risk unrelated to the market
Average true range (ATR) Typical daily price range, gaps included Dollars per share Not comparable between prices
Implied volatility Volatility priced into options, looking ahead Percent a year A market price, not a measurement

Historical volatility looks back; implied volatility is a forecast extracted from option prices. Volatility and the VIX: historical vs implied volatility explained explains the second kind.

How Finabase does it

Finabase does not show a standard deviation or a volatility figure for a stock, a fund or a portfolio. The nearest tools:

  • Bollinger Bands. The chart's one standard-deviation study. The bands sit two standard deviations of the last 20 closes above and below a 20-day average, so their width is a moving picture of volatility in price terms. See Bollinger Bands: formula, the squeeze and walking the bands.
  • ATR and NATR. Average True Range (ATR): volatility, stops and position size measures the typical daily range in dollars, and the normalized version states it as a percent of the price, which makes two stocks comparable.
  • Chart scripts. A custom formula can call stdev(source, length). It uses the population formula (divide by n, not n − 1) and works on prices or on whatever series it is given; it is not annualized.
  • Options. The option chain shows each contract's implied volatility, which is the market's forecast rather than a measurement of the past.
  • Comparison. The comparison chart redraws symbols as percent changes from a common start. It gives no number, but the difference between a bond fund's line and Nvidia's is volatility made visible.

To get an actual figure, export closing prices from a symbol's History tab, compute the day-to-day percent changes in a spreadsheet, take STDEV.S of them and multiply by the square root of 252.

Frequently asked questions

Is high volatility bad?

Not by itself. Volatility is the size of the swings in both directions, and assets with higher long-run returns have generally had more of it. It becomes a problem when money is needed at a fixed date, when borrowing is involved, or when the swings push an investor into selling low.

What is a normal volatility for the stock market?

SPY's monthly returns since 1993 annualize to 14.7%. Twelve-month readings between about 8% and 20% have been common, with quiet years near 4% and crisis years near 30%.

Why multiply by the square root of 12 and not by 12?

Because variance, the squared standard deviation, is what adds up across independent periods. Twelve months carry twelve times the variance of one month, so the standard deviation grows by the square root of twelve, about 3.46.

What is the difference between volatility and beta?

Volatility measures all of an investment's movement. Beta measures only the part that moves with the market. Gold has been about as volatile as stocks and has had a beta near zero.

Does low volatility mean low risk?

Only in the sense of small fluctuations so far. A fund can be calm for years and then lose heavily, and a calm asset can still fail to keep up with inflation.

Summary

Volatility is the standard deviation of returns: the square root of the average squared distance from the mean, scaled to a year by the square root of time. SPY's has been about 14.7% a year since 1993, which describes its ordinary months well: 72% fell within one standard deviation of the average and 94% within two. It describes the extremes poorly. The worst month was more than four standard deviations out, and losses beyond two standard deviations came almost three times as often as gains. Volatility also drifts from 4% to 30% depending on the year, and it counts gains as risk. It is the first number to look at and never the only one.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. A fund’s monthly returns have a standard deviation of 3%. What is its annualized volatility?

    1. 36%, because 3% × 12
    2. About 10.4%, because 3% × √12
    3. 3%, because volatility does not depend on the period
    Show the answer

    B. About 10.4%, because 3% × √12 Variance adds across periods, so standard deviation grows with the square root of time. √12 is about 3.46, and 3% × 3.46 = 10.4%.

  2. Two funds have the same average return and the same standard deviation. Fund A lost 3% in each of the last eight months of the period. Fund B’s losses were scattered. What does standard deviation tell you about the difference?

    1. Nothing: it ignores the order in which returns arrive
    2. Fund A is riskier, because its losses were consecutive
    3. Fund B is riskier, because scattered losses raise the standard deviation
    Show the answer

    A. Nothing: it ignores the order in which returns arrive Standard deviation uses each return’s distance from the average and not when it happened. A run of losses is captured by maximum drawdown, not by volatility.

  3. SPY’s worst month since 1993 was 4.1 standard deviations below its average. What does that show?

    1. The standard deviation was calculated wrongly
    2. Extreme months happen far more often than a normal distribution predicts
    3. Such a month cannot happen again for thousands of years
    Show the answer

    B. Extreme months happen far more often than a normal distribution predicts A bell curve with SPY’s average and volatility puts a month like October 2008 at roughly once in 4,000 years, yet it occurred within 16 years of the fund’s launch. Real returns have fatter tails than the normal curve.

  4. A fund has one month of +15% among otherwise quiet months. How does that month affect its volatility?

    1. It lowers it, because gains reduce risk
    2. It has no effect; only losses count
    3. It raises it, because the formula squares distances in both directions
    Show the answer

    C. It raises it, because the formula squares distances in both directions Standard deviation treats a large gain exactly like a large loss. Measures such as downside deviation exist because investors do not.

Exercise

One task to do in Finabase. Nothing is graded or saved.

Open the two-year SPY chart with Bollinger Bands. The bands sit two standard deviations of the last 20 closes above and below a 20-day average, so the gap between them is a picture of volatility. Find where the bands are widest and where they are narrowest over the two years. Is the gap today nearer the wide end or the narrow end, and did the wide stretches come one day at a time or in runs?

Open the SPY chart with Bollinger Bands
What you should find

The gap changes a great deal within two years: volatility is not constant. Wide stretches follow sharp moves and last for weeks before they narrow, the clustering the lesson describes. A narrow gap today describes the recent calm only. A figure estimated in a quiet stretch understates what comes when conditions change.

Sources and further reading

  1. Zvi Bodie, Alex Kane and Alan J. Marcus, Investments, 12th ed. (McGraw-Hill, 2021) — variance, standard deviation and the normal distribution
  2. Benoit Mandelbrot, “The Variation of Certain Speculative Prices,” Journal of Business, vol. 36, no. 4 (1963) — fat tails and volatility clustering
  3. U.S. Securities and Exchange Commission, Investor.gov — Volatility (glossary)
  4. Finabase chart data — SPY and AGG monthly bars and dividends through October 2026; daily closes of ten symbols, November 2024 to November 2026; statistics computed by Finabase