Volatility: how much a portfolio fluctuates

Basics8 min read · updated 2 October 2026 · 8 sources

Volatility measures how strongly a portfolio's returns fluctuate around their average. Technically, it is the standard deviation of returns, usually scaled up to one year. It is the most widely used risk measure, even though it treats losses and gains alike.

How to calculate volatility in three steps

1. Calculate the return for each period, e.g. monthly returns: the price at the end of the month divided by the previous month's, minus 1.
2. Calculate the standard deviation of these returns (sample standard deviation, i.e. divide by n − 1).
3. Annualize: multiply monthly figures by √12 and daily figures by √252 (trading days in a year).

The standard deviation is the square root of the variance, i.e. of the average squared distance from the mean. Squaring ensures that large swings count for more than small ones and that upward swings count just as much as downward ones.

Volatility translated into dollars

Percentages are abstract. It is more helpful to ask what a given volatility means for your portfolio. With $100,000 in an equity portfolio with 15% annual volatility, one standard deviation over a year corresponds to $15,000. Over one month it is one twelfth of the variance, i.e. 15 / √12 = 4.33% or around $4,330. Fluctuations of this size are no cause for alarm; they are an ordinary part of everyday stock market life.

The calculation assumes a normal distribution and therefore underestimates the extremes (more on this below). It is still useful as a first guide when choosing your equity allocation. How investors feel about losses is a different question: losses feel heavier than gains of the same size, which is why many people underestimate in advance how unpleasant a 15% fluctuation is.

Why multiply by the square root of time?

If the returns of successive periods are independent, their variances add up: twelve months have twelve times the variance of one month, and therefore √12 times the standard deviation. With daily data, you work with around 252 trading days. An equity index that moves by 1% on a typical day thus has an annual volatility of around 15.87%.

The rule only holds approximately. If good days tend to be followed by good days (positive autocorrelation), risk grows faster than the square root of time; if prices merely oscillate back and forth (negative autocorrelation), it grows more slowly. For broad equity indices the effect is small over months, but it is pronounced for rarely traded assets and trend strategies.

What the number means

If returns are roughly normally distributed, around two thirds of all annual returns lie within ± one volatility of the average and around 95% within ± two volatilities. An equity ETF with an expected return of 7% and 15% volatility therefore has a return below −8% in roughly one year out of six.

The figures are orders of magnitude from long data series, not fixed properties. In calm years, equity volatility is well below these levels; in crises it can double or triple for several months.

For advanced readers: volatility is not constant

Calm and turbulent phases alternate, and a large swing is often followed by more of them. This volatility clustering is one of the stylized facts of financial markets compiled by Cont (2001). Engle (1982) introduced the ARCH models to capture it:

Bollerslev (1986) generalized this into the GARCH model, still the standard for volatility forecasts today. In it, today's variance is made up of a long-run level, the latest return shock and yesterday's variance:

A simple special case is the exponentially weighted moving average (EWMA), which risk management practice often applies to daily data with λ = 0.94 (Jorion (2007) describes the method): σ²_t = λ × σ²_(t−1) + (1 − λ) × r²_(t−1).

Fat tails

As early as Mandelbrot (1963), it was shown that extreme price changes occur far more often than a normal distribution allows. With normally distributed daily returns, a day with a loss of five standard deviations would be an event that happens roughly once in 13,843 years. In reality, stock markets have had several such days, for example on October 19, 1987. Volatility therefore underestimates the risk of large crashes; that is what value at risk and a look at skewness and kurtosis are for.

How precise is an estimated volatility?

Volatility can be estimated far more precisely than the expected return. That is one reason why optimizers cope well with volatilities but badly with return estimates. With normally distributed returns, the standard deviation estimated from n observations has a standard error of approximately σ / √(2 × (n − 1)).

So more data quickly helps with volatility but hardly at all with returns: the precision of a return estimate depends on the length of the period in years, not on the number of observations. Daily data improve the estimate of volatility, not that of the return.

Volatility and investment horizon

Over longer periods, good and bad years partly cancel each other out. At 15% annual volatility, the average annual return over 20 years fluctuates by only about 15 / √20 = 3.35% around its expected value. But it does not follow that equities become risk-free in the long run: the range of final wealth in dollars keeps widening, because small differences in return are compounded over many years. For withdrawal plans, the order of returns also matters; see sequence risk.

And one more surprise: in the stock market, higher volatility has not been reliably rewarded with higher returns. Ang et al. (2006) found that stocks with high stock-specific (idiosyncratic) volatility delivered unusually low returns. More on this in the article on the low-volatility anomaly.

Realized and implied volatility

Everything above is realized volatility from past prices. Options, by contrast, contain an implied volatility: the fluctuation the market prices in for the future. For the S&P 500, that is the VIX. On average, implied volatility is higher than the volatility realized afterward: buyers of protection pay a premium, which Carr and Wu (2009) measure as the variance risk premium.

Limits of volatility

• Gains count like losses: a fund that often rises sharply looks risky. The Sortino ratio therefore only considers downside fluctuations.
• Fat tails: real returns have more extreme days than the normal distribution predicts. For crises, drawdown and value at risk are more informative.
• Smoothing: illiquid investments (open-ended real estate funds, private equity) are valued infrequently and therefore look calmer than they are.
• Period: volatility changes with the market environment. The VIX shows the expected fluctuation for the next 30 days.
• Currency: for US investors, what counts is volatility in dollars. A European equity fund fluctuates differently in dollars than in euros, because the exchange rate fluctuates too.

Reducing volatility in the portfolio

The volatility of a portfolio is lower than the weighted average of the individual volatilities as long as the assets are not perfectly correlated. That is the central insight of Markowitz (1952). The article on correlation and diversification shows how strong this effect is. If you deliberately manage total risk, you can reduce the share of risky assets in turbulent phases; this volatility control (volatility targeting) is described in the article on momentum strategies.

How Portfolio Lab calculates it

Portfolio Lab calculates volatility as the standard deviation of the period returns (sample, divided by n − 1) and multiplies it by the square root of the periods per year. The frequency is derived from the average spacing of the data: daily data with √252, weekly data with √52, monthly data with √12. Calculations are in the base currency of the backtest; asset classes and most analysis tools use monthly returns. More in the metrics methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.