Calculating Portfolio Standard Deviation
Calculating Portfolio Standard Deviation
How Do You Measure the Total Risk of Your Entire Portfolio?
Individual security volatility tells only part of the story. What matters more is the volatility of your entire portfolio—how much your net worth fluctuates based on market movements. Portfolio volatility depends not just on the individual volatility of holdings but also on how those holdings move together. Two portfolios with identical holdings' individual volatilities can have dramatically different total volatility if one portfolio's holdings are highly correlated (moving together) and the other's are less correlated (moving independently). This is the power of diversification: lower portfolio volatility without sacrificing expected return. Calculating portfolio standard deviation requires understanding both individual security volatility and the correlations between holdings—a more complex calculation than single-security volatility but essential for realistic risk assessment. This article walks through the mathematics, provides practical examples, and shows how to implement calculations in spreadsheets or portfolio-analysis tools.
Quick definition: Portfolio standard deviation is the square root of the weighted variance of all holdings, adjusted for correlations between holdings. It represents the expected annualized price fluctuation of your entire portfolio based on historical volatility and co-movement patterns.
Key takeaways
- Portfolio volatility depends on individual volatilities, weights, and correlations
- Perfectly correlated holdings do not reduce portfolio volatility; uncorrelated holdings do
- A two-asset portfolio's standard deviation includes a correlation adjustment term
- Spreadsheet-based calculations are practical for portfolios up to 20–30 holdings; larger portfolios need portfolio-analysis software
- Historical correlation may diverge from future correlation, especially during market crises
The Basic Two-Asset Formula
The simplest portfolio volatility formula applies to a two-asset portfolio:
Portfolio SD = sqrt(w1^2 * sd1^2 + w2^2 * sd2^2 +
2 * w1 * w2 * rho * sd1 * sd2)
Where:
- w1, w2 = portfolio weights (as decimals; must sum to 1.0)
- sd1, sd2 = standard deviations of each asset
- rho = correlation coefficient between the two assets (ranges from -1 to +1)
Let's break this down:
First term: w1^2 * sd1^2 = weight-squared times volatility-squared for asset 1
Second term: w2^2 * sd2^2 = weight-squared times volatility-squared for asset 2
Third term: 2 * w1 * w2 * rho * sd1 * sd2 = the correlation adjustment
The third term is crucial. It adjusts downward (when rho is positive and less than 1.0) for the covariance between assets. When assets are perfectly uncorrelated (rho = 0), the third term vanishes, and volatility is lowest. When assets are perfectly correlated (rho = 1.0), the third term equals 2 * w1 * w2 * sd1 * sd2, producing a linear sum of weighted volatilities with no diversification benefit.
A Concrete Two-Asset Example
Suppose you build a two-asset portfolio:
Portfolio composition:
- 60% in Stock A: volatility 18%, current price $100
- 40% in Stock B: volatility 12%, current price $50
Historical correlation: You observe that Stock A and Stock B returns correlate at +0.5 (moderate positive correlation; they move together somewhat, but not perfectly).
Calculation:
Portfolio SD = sqrt((0.60)^2 * (0.18)^2 + (0.40)^2 * (0.12)^2 +
2 * 0.60 * 0.40 * 0.5 * 0.18 * 0.12)
First term: (0.36) * (0.0324) = 0.011664
Second term: (0.16) * (0.0144) = 0.002304
Third term: 2 * 0.24 * 0.5 * 0.18 * 0.12 = 0.010368
Sum: 0.011664 + 0.002304 + 0.010368 = 0.024336
Portfolio SD = sqrt(0.024336) = 0.156 = 15.6%
Interpretation: This portfolio has a standard deviation of 15.6%. If you invested $100,000 in this allocation, you would expect annual price movements of roughly ±15,600 (one standard deviation) as a typical range. Using the normal distribution, about 68% of annual outcomes fall within ±15,600; 95% fall within ±31,200.
The Impact of Correlation
To illustrate how correlation changes portfolio volatility, let's recalculate with different correlations while holding weights and individual volatilities constant:
If correlation = +1.0 (perfect correlation):
Portfolio SD = sqrt(0.011664 + 0.002304 + 2 * 0.24 * 1.0 * 0.18 * 0.12)
= sqrt(0.011664 + 0.002304 + 0.020736)
= sqrt(0.034704) = 18.6%
If correlation = 0 (no correlation):
Portfolio SD = sqrt(0.011664 + 0.002304 + 0)
= sqrt(0.013968) = 11.8%
If correlation = -0.5 (negative correlation):
Portfolio SD = sqrt(0.011664 + 0.002304 + 2 * 0.24 * (-0.5) * 0.18 * 0.12)
= sqrt(0.011664 + 0.002304 - 0.010368)
= sqrt(0.0036) = 6.0%
Notice the dramatic difference:
- Correlation +1.0: 18.6% (no diversification benefit)
- Correlation +0.5: 15.6% (some benefit)
- Correlation 0: 11.8% (more benefit)
- Correlation -0.5: 6.0% (substantial benefit)
This is why diversification works. Assets moving independently (or inversely) dampen portfolio volatility far more than assets moving in lockstep.
Expanding to Three or More Assets
For portfolios with three or more holdings, the formula becomes a matrix calculation. The variance-covariance matrix includes every pairwise correlation:
Portfolio Variance = sum of all (wi * wj * cov(i,j))
Where cov(i,j) is the covariance between assets i and j (covariance = correlation * sd_i * sd_j).
For a three-asset portfolio, the matrix looks like:
| w1*sd1 w1*w2*cov(1,2) w1*w3*cov(1,3) |
| w2*w1*cov(2,1) w2*sd2 w2*w3*cov(2,3) |
| w3*w1*cov(3,1) w3*w2*cov(3,2) w3*sd3 |
Each diagonal element is wi^2 * sdi^2 (the variance of asset i weighted). Each off-diagonal element is the covariance adjustment between two assets. Sum all elements, take the square root, and you have portfolio volatility.
Practical reality: Calculating this by hand for 20+ assets is tedious and error-prone. Use spreadsheets (Excel's MMULT function or Google Sheets array formulas) or portfolio-analysis software.
A Three-Asset Spreadsheet Example
Suppose your portfolio holds three assets:
| Asset | Weight | Volatility |
|---|---|---|
| Stock A | 50% | 18% |
| Stock B | 30% | 12% |
| Bond Fund | 20% | 4% |
Correlation matrix:
| A | B | Bond | |
|---|---|---|---|
| A | 1.0 | 0.5 | -0.1 |
| B | 0.5 | 1.0 | 0.2 |
| Bond | -0.1 | 0.2 | 1.0 |
Steps:
-
Create a variance-covariance matrix using cov(i,j) = correlation * sd_i * sd_j
-
Multiply each element by the product of weights: w_i * w_j * cov(i,j)
-
Sum all nine cells
-
Take the square root
The variance-covariance matrix:
| A | B | Bond | |
|---|---|---|---|
| A | 0.0324 | 0.0108 | -0.0072 |
| B | 0.0108 | 0.0144 | 0.0096 |
| Bond | -0.0072 | 0.0096 | 0.0016 |
Weighted by w_i * w_j:
| A | B | Bond | |
|---|---|---|---|
| A | 0.50 * 0.50 * 0.0324 = 0.0081 | 0.50 * 0.30 * 0.0108 = 0.00162 | 0.50 * 0.20 * -0.0072 = -0.00072 |
| B | 0.30 * 0.50 * 0.0108 = 0.00162 | 0.30 * 0.30 * 0.0144 = 0.001296 | 0.30 * 0.20 * 0.0096 = 0.000576 |
| Bond | 0.20 * 0.50 * -0.0072 = -0.00072 | 0.20 * 0.30 * 0.0096 = 0.000576 | 0.20 * 0.20 * 0.0016 = 0.000064 |
Sum: 0.0081 + 0.00162 + -0.00072 + 0.00162 + 0.001296 + 0.000576 - 0.00072 + 0.000576 + 0.000064 = 0.011312
Portfolio SD: sqrt(0.011312) = 10.6%
This portfolio has a 10.6% standard deviation despite holdings with individual volatilities of 18%, 12%, and 4%. Diversification and negative correlation (bonds and stocks) dampened volatility substantially.
Estimating Correlations
Correlation is the hardest input to estimate accurately. Historical correlation can be computed using Excel's CORREL() function or calculated manually:
Correlation = covariance / (sd_1 * sd_2)
Where covariance is the average product of deviations from the mean for both assets.
However, historical correlations are unstable:
- Normal markets: Stocks and bonds correlate at -0.1 to +0.2 (weakly positive or negative)
- Crisis markets: Stocks and bonds correlate at +0.5 to +0.8 (strongly positive—both sell off)
- Low-volatility periods: Stock correlations hover around +0.4
- Crisis periods: Stock correlations jump to +0.7 or higher (all sell together)
This is crucial: during the periods when you most need diversification (market crashes), correlations spike, reducing diversification benefits. A bond allocation that seemed protective because of -0.1 correlation suddenly correlates at +0.3 during a crisis, offering little shelter.
Real-World Example: Assessing a Diversified Portfolio
Suppose you hold a $250,000 portfolio:
- $125,000 in U.S. stocks (Vanguard Total Stock Market ETF, volatility 15%)
- $75,000 in international stocks (VXUS, volatility 18%)
- $50,000 in bonds (BND, volatility 5%)
Weights: 50% U.S. stocks, 30% international stocks, 20% bonds
Historical correlations (3-year average):
- U.S. stocks & international stocks: +0.8
- U.S. stocks & bonds: -0.05
- International stocks & bonds: +0.1
Using the three-asset formula, the portfolio's standard deviation is approximately 11.2%.
Interpretation: You expect annual price fluctuations of roughly ±$28,000 (11.2% of $250,000) as a typical range. A severe year (2 standard deviations) would produce drawdowns around ±$56,000, or -22%. This seems reasonable for a diversified portfolio with 50% stocks.
Common Mistakes
Mistake 1: Using only the simple average of individual volatilities. A naive approach would calculate (0.50 * 15% + 0.30 * 18% + 0.20 * 5%) = 13.6% as portfolio volatility. The correct answer (11.2%) is significantly lower because correlations reduce joint volatility. The difference is the diversification benefit, which naive approaches miss.
Mistake 2: Assuming stable correlations. Using 5-year average correlations assumes future correlations match history. During crashes, correlations jump sharply. A portfolio that looks 11% volatile based on normal-period correlations can spike to 16%+ volatility during crashes when correlations rise. Stress-test your portfolio using elevated crisis-period correlations.
Mistake 3: Ignoring tail risk in volatility forecasts. Standard deviation assumes normal-distribution returns; real markets exhibit fat tails (extreme events more frequent than normal distributions predict). A portfolio with 11% volatility might experience 3-standard-deviation events (losses exceeding -33%) roughly every 5–10 years. Assume larger drawdowns than volatility alone suggests.
Mistake 4: Using too little historical data. Correlation estimates require at least 3 years of daily or monthly returns to be reliable. Using one year of data produces noisy, unreliable correlations. Using 10+ years of data may include obsolete regime (e.g., pre-financial crisis when correlations were different). Three to five years is a practical balance.
Mistake 5: Forgetting to account for rebalancing. Portfolio standard deviation assumes static weights. If you rebalance quarterly, your realized volatility may differ from the static calculation because you are systematically reducing concentration in outperformers (selling winners, buying losers). Rebalancing can reduce realized volatility but also reduces expected returns.
FAQ
How often should I recalculate portfolio standard deviation?
Quarterly at minimum. Volatility and correlation estimates change as market conditions evolve. An asset's volatility that was 15% might become 20% during a volatile period. Recalculating every quarter keeps your risk estimates current. During extreme market conditions, monthly recalculation is prudent.
What standard deviation should I target for my portfolio?
This depends on your time horizon and risk tolerance. Young investors might target 12–18% volatility (stock-heavy). Mid-career investors might target 10–12% volatility (balanced). Near-retirees might target 6–8% volatility (bond-heavy). Using standard deviation to size risk exposure is sensible: multiply your annual volatility by the square root of 2 to estimate likely worst-case years (e.g., a 12% volatility portfolio typically experiences 17% drawdowns in bad years).
Can I use standard deviation alone to assess portfolio risk?
No. Standard deviation assumes normal distributions and ignores tail risk. Use standard deviation as one tool alongside maximum drawdown, value-at-risk, and stress tests. A 10% standard deviation portfolio can still experience 40% drawdowns in severe market dislocations. Understand your portfolio's downside scenarios, not just its typical volatility.
How do negative correlations improve diversification?
Negative correlations mean assets move inversely. A stock-bond correlation of -0.2 means when stocks fall, bonds tend to rise slightly. This inverse movement provides a natural hedge—bond gains offset stock losses. Traditional 60/40 stock-bond portfolios rely on stocks and bonds' low (often negative) correlation for volatility reduction.
What if I cannot calculate correlations for my holdings?
Use index correlations as proxies. If you own 10 individual tech stocks, use the correlation of the Nasdaq 100 Index with other holdings instead of calculating 45 individual stock-pair correlations. This simplifies calculation and is reasonably accurate for uncorrelated holdings within a sector.
Does rebalancing change portfolio volatility?
Rebalancing changes realized volatility slightly (typically 0.5–1% reduction) but does not change the theoretical standard deviation calculation. The formula assumes static weights; rebalancing makes weights dynamic (selling winners, buying losers), which is a drag on returns but a modest help on volatility. For long-term planning, use the static volatility formula.
Related concepts
Summary
Portfolio standard deviation quantifies total portfolio risk by combining individual security volatilities with correlations between holdings. The calculation is straightforward for two assets (using the correlation-adjustment formula) but requires matrix algebra for three or more assets. Diversification works precisely because correlation adjustment reduces volatility when holdings move independently. However, historical correlations underestimate crisis-period correlations; real drawdowns often exceed estimates based on normal-period correlations. Use spreadsheets or portfolio-analysis tools to calculate portfolio volatility, recalculate quarterly as volatilities and correlations shift, and stress-test using higher crisis-period correlations. Standard deviation is one of many risk metrics; pair it with maximum drawdown and value-at-risk for a complete risk picture.