Skip to content

Covariance of Returns

Covariance of Returns

How two return series move together: the average product of their deviations from their means, using the sample divisor n − 1. Positive when the two tend to rise and fall together, negative when they offset.

When to use: Use as the raw ingredient of portfolio variance and beta. Covariance is in squared-return units, so for a readable strength-of-relationship figure use correlation.

Calculator

Formula

Cov(Ra,Rb)=(Ra,tRˉa)(Rb,tRˉb)n1\text{Cov}(R_a, R_b) = \frac{\sum (R_{a,t} - \bar{R}_a)(R_{b,t} - \bar{R}_b)}{n - 1}

Variables

SymbolNameDescriptionUnit
CovCovarianceSample covariance of the two return seriesvariance
AssetReturnsAsset A ReturnsPeriodic returns of the first asset%
BenchmarkReturnsAsset B ReturnsPeriodic returns of the second asset over the same periods%

Real-Life Examples

Example 1: Fund Versus Benchmark

Five periods of returns: fund 4%, 3%, 5%, 2%, 6%; benchmark 3%, 2%, 4%, 2%, 5%.

Given

AssetReturns = [0.04, 0.03, 0.05, 0.02, 0.06]BenchmarkReturns = [0.03, 0.02, 0.04, 0.02, 0.05]

Step-by-Step

1.Means: fund 0.04, benchmark 0.032
2.Sum of products of deviations = 0.0008
3.Cov = 0.0008 / (5 − 1)
4.Cov = 0.00020000
Result:0.00020000 (return²)

Positive: the fund and its benchmark move together. Divided by the benchmark variance (0.00017) this gives a beta of about 1.18.

Example 2: Volatile Asset Versus Steady One

Asset A: 10%, −5%, 8%, 2%, 12%; asset B: 6%, 1%, 5%, 3%, 7%.

Given

AssetReturns = [0.1, -0.05, 0.08, 0.02, 0.12]BenchmarkReturns = [0.06, 0.01, 0.05, 0.03, 0.07]

Step-by-Step

1.Means: A 0.054, B 0.044
2.Sum of products of deviations = 0.00662
3.Cov = 0.00662 / 4
4.Cov = 0.00165500
Result:0.00165500 (return²)

A larger covariance, driven mostly by asset A's wider swings. Whether the two are more closely related than the first pair needs correlation, which strips out the scale.

Frequently Asked Questions

The series are a sample of the assets' behaviour, not the whole population, and dividing by n − 1 corrects the downward bias of estimating the means from the same data. It matches the sample standard deviation used elsewhere on SnapFi.