Volatility: what standard deviation captures and what it misses

Calculate annualised volatility and explain why it counts gains and losses the same way.

Volatility is the standard deviation of an asset's returns, and it is a good measure of how much a holding wobbles in ordinary months. It is a poor guide to how much you could lose in a bad year, because it counts gains as risk, ignores fat tails, and understates risk when it is measured over calm stretches of history.

What the risk score is built from

Take Marcus, who holds a world equity ETF. Their broker app gives the ETF a risk score of 6 out of 7, and a robo-advisor they tried once called their chosen portfolio "moderately aggressive". Both labels are built mostly from one number, volatility. Neither tells Marcus what an investor actually wants to know: how much the ETF could fall, how often, and whether they could stand it.

Volatility is still the number every risk report starts with, so you need to know exactly what it counts: how far each period's return tends to sit from the average return. A fund returning close to 1% every month has low volatility. A fund swinging between plus 6% and minus 8% has high volatility, even if both funds have the same average. Both funds are made-up illustrations.

Calculating and annualising it in a spreadsheet

Put the monthly returns for your holding in one column of a spreadsheet and use the STDEV function (STDEV.S in Excel). The result is monthly volatility. If you use weekly or daily returns instead, the result is weekly or daily volatility.

Risk reports quote volatility as a yearly, or annualised, figure. To annualise, multiply the period figure by the square root of the number of periods in a year:

monthly volatility x sqrt(12) weekly volatility x sqrt(52) daily volatility x sqrt(about 252 trading days)

The square root comes from an assumption that each period's return is independent of the one before. Under that assumption, variances add up over time, and standard deviations grow more slowly. The rule is close enough for most uses, but it drifts when returns tend to trend or reverse from one month to the next. That drift is one reason two providers can quote different volatility for the same fund.

Over five years, the monthly returns on Marcus's ETF have a standard deviation of 4.0% (these Marcus figures are made up for the example). Annualised, that is 4.0% x sqrt(12), or about 13.9%.

Volatility also costs you compounded return. Lesson 1.1 showed that the gap between the arithmetic and geometric average return is roughly half the variance, and variance is volatility squared. So of two holdings with the same average return, the one with higher volatility compounds to less.

Four ways the figure misleads

First, it counts gains as risk. Standard deviation measures distance from the average in both directions, so a plus 9% month counts against a fund as much as a minus 9% month. An asset with occasional upward jumps looks riskier than one that drifts gently, though no investor worries about surprise gains. This matters most for holdings with lopsided returns. A stock that sits flat for years and then doubles on a takeover shows high volatility. A strategy that makes small steady gains and then one large loss can show low volatility right up until the loss. Lesson 2.3, "Sharpe and Sortino ratios, and when each misleads", gives a measure that only counts the downside.

Second, it hides fat tails, meaning large moves, especially falls, that happen far more often than a bell curve allows. Standard deviation itself assumes nothing about the shape of returns. The trouble starts when people use it as the whole story, which in effect treats returns as a bell curve. Under a bell curve, moves of more than three standard deviations almost never happen. On Black Monday, 19 October 1987, the Dow Jones Industrial Average fell about 22% in one day, a move a bell curve fitted to the prior years would treat as essentially impossible. In March 2020, several daily falls in US shares were each many standard deviations out. Read any volatility figure as a description of normal months, and expect the worst ones to be worse. Lesson 2.6, "Value at Risk, expected shortfall and fat tails", deals with this directly.

Third, the window decides the answer. Volatility is measured over a chosen stretch of history. Five calm years produce a low figure, and that figure then gets used to size positions just before conditions change.

Fourth, volatility comes in clusters. Volatility clustering means calm months follow calm months and wild months follow wild ones. A single figure from a long, mostly calm window describes the typical year and badly misses the year that tests you.

Checking Marcus's worst year

In the worst of Marcus's five years, the ETF's monthly returns, in order, were minus 4.8%, plus 1.9%, minus 2.6%, minus 7.9%, plus 0.4%, minus 6.8%, plus 5.9%, minus 3.7%, minus 8.6%, plus 4.9%, plus 6.2% and minus 1.5% (also made up). Compounded, the year lost about 16.7%. The monthly standard deviation for that year alone is about 5.3%, which annualises to about 18.3%. Against the five-year figure of 13.9%, the ETF behaved like a much riskier asset in the year that hurt.

Risk desks look at several windows, including the worst year they can find, and you should do the same. Calculate volatility over the whole history window, pick out the worst year in that window, calculate volatility for that year alone, and compare the two.

Volatility is still worth calculating. It is the input for the Sharpe ratio, for volatility-weighted position sizing in lesson 11.1, and for comparing funds that hold similar assets. Use it as a measure of the ordinary wobble, and pair it with the other measures in this module for what it misses.

For the activity you'll need five years of month-end prices or total return index values for one holding. Your broker's chart export or the fund's website will usually have them, and you'll calculate the figure twice, once for the whole period and once for the worst year in it.

Take five years of monthly returns for one holding and calculate its annualised volatility. Then do the same for its worst single year.

Course

Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).