Portfolio variance: understanding return dispersion and investment risk
Risk in investing is fundamentally about return dispersion: how widely an investment's returns fluctuate around the average. Portfolio variance is the standard measure of this—the squared average deviation of returns from their mean, and the mathematical foundation on which volatility, covariance, and modern portfolio theory are built. Understanding variance is the first step to understanding how diversification reduces risk and how portfolio optimisers work.
What variance measures
The mathematical framework for measuring portfolio variance was established by Harry Markowitz (1952) in Portfolio Selection, Journal of Finance—the paper that founded Modern Portfolio Theory and demonstrated how correlation between assets determines the risk-reduction benefit of diversification. Variance is the fundamental risk input in mean-variance optimisation, and Markowitz's insight that portfolio variance depends on asset correlations, not just individual variances, remains the foundation of quantitative portfolio construction.
Variance quantifies how much an investment's returns are dispersed around their mean. A high variance means returns swing widely from the average; a low variance means returns are tightly clustered. Like volatility, variance captures dispersion in both directions—large positive returns contribute as much as large negative returns. It does not distinguish between upside and downside variability.
Variance is expressed in squared units (for example, percentage squared), which makes it less intuitive to interpret directly than volatility. For direct risk assessment, standard deviation—the square root of variance—is the more practical figure. Variance retains its importance in portfolio construction, however, because it is additive under certain assumptions and can be decomposed across assets and factors in ways that standard deviation cannot.
The formula
σ² = T/n × Σ(rᵢ − r̅)²
Where:
- σ² = annualised variance
- T = number of periods per year (252 for daily data, 12 for monthly)
- n = number of observations
- rᵢ = return for period i
- r̅ = mean return
The formula computes the mean squared deviation of each period's return from the overall mean return, then scales to an annual figure. Taking the square root of this result gives the annualised standard deviation (volatility). Variance and volatility are therefore two representations of the same underlying data—one in squared units, one in the original percentage units.
How to interpret variance
Variance is most useful in a comparative context: a higher variance means greater return dispersion. A portfolio with an annualised variance of 0.04 (4% squared) has an annualised volatility of 20%; a portfolio with a variance of 0.01 has a volatility of 10%.
The squared unit makes direct interpretation unintuitive, but the relationship between variance and volatility makes the figure easy to anchor: take the square root to obtain the more familiar volatility figure. Where variance becomes directly useful is in portfolio construction. The variance of a two-asset portfolio is not simply the weighted average of the individual variances—it also includes a covariance term that accounts for how the two assets move together. This is the mathematical mechanism through which diversification reduces portfolio risk.
A portfolio's variance decreases as the correlation between its holdings decreases. Combining assets with low or negative correlation can reduce portfolio variance below the weighted average of the individual variances—the core principle of the mean-variance optimisation underlying pfolio's portfolio construction.
Rolling variance
The scalar variance figure summarises the full measurement period. Rolling variance computes the same metric over a sliding window, showing how return dispersion has evolved through time. Since variance is the square of volatility, rolling variance moves in the same direction as rolling volatility but is more sensitive to extreme observations (because deviations are squared).
Limitations
Variance treats upside and downside deviations symmetrically, which may overstate the risk perception of investors who primarily worry about losses. Downside-focused measures such as downside volatility are an alternative for investors who wish to focus only on negative return episodes.
Variance also assumes that returns are stationary—that the distribution of returns does not change over time. In practice, return distributions shift with market regimes. A variance figure calculated over a calm period will understate the risk of a period that includes a market crisis.
Finally, variance captures only the second moment of the return distribution. It says nothing about skewness (asymmetry) or kurtosis (tail thickness), which are material properties for many investment strategies.
Variance in pfolio
Variance is not exposed as a standalone metric in pfolio. Instead, pfolio surfaces volatility—the square root of variance—as the primary measure of return dispersion. Volatility preserves the same information as variance in the original percentage units, which makes it more straightforward to interpret and compare across portfolios.
The practical difference is presentational: a portfolio with an annualised variance of 0.04 has an annualised volatility of 20%. pfolio reports the 20% figure. The underlying calculation is the same; only the unit differs.
For a full description of how pfolio calculates volatility and all other metrics, see the metrics we use.
Related metrics
Disclaimer
Get started now
