Why Covariance and Market Variance Define Beta
An asset can have large return swings without moving in step with the market. Another asset can have smaller swings overall but tend to rise and fall with the market. Beta focuses on that second question: how does the asset’s return co-move with the market’s return? It is not a measure of all the asset’s volatility.
To calculate beta, use returns measured over the same period and on the same basis. For example, do not pair an asset’s monthly returns with annual market returns. Let \(R_A\) be the asset return and \(R_M\) the market return. Treat them as random variables with finite means and variances, and assume \(\operatorname{Var}(R_M)>0\). The positive market-variance condition matters: if the market return never varies, the beta denominator is zero and this definition does not produce a beta.
For a finite set of possible scenarios with probabilities \(p_j\), the population covariance and variance are \(\operatorname{Cov}(R_A,R_M)=\sum_j p_j(R_{A,j}-E[R_A])(R_{M,j}-E[R_M])\) and \(\operatorname{Var}(R_M)=\sum_j p_j(R_{M,j}-E[R_M])^2\), where the probabilities sum to one. The same probability weights must be used to calculate both means and both second moments.
These formulas describe population quantities under a stated return model. In an application, data may be used to estimate them, but an estimate based on a limited sample is not automatically equal to the underlying population beta. The definition also requires returns to be matched over the same periods; otherwise, the covariance would not describe the intended co-movement.
What the Ratio Measures
Covariance is positive when above-mean asset returns tend to occur with above-mean market returns, and below-mean asset returns tend to occur with below-mean market returns. Negative covariance indicates a tendency to move in opposite directions. A covariance near zero indicates little linear co-movement in the specified return model; it does not prove that the returns are unrelated in every possible way.
Because market variance is positive, beta has the same sign as asset–market covariance. A positive beta describes movement in the market’s direction on average; a negative beta describes movement in the opposite direction. A beta of zero means the covariance is zero under the model. A beta greater than one indicates covariance with the market that is large relative to the market’s own variance, not necessarily that the asset has more total volatility than the market.
The units help clarify the interpretation. If both returns are recorded as decimals, covariance and variance are both in squared-return units, so their ratio is dimensionless. Recording both return series in percentage points instead multiplies covariance and market variance by the same scale factor, leaving beta unchanged. Mixing units between the two series, however, changes the calculation and can make it meaningless.
Worked Example: Beta and Total Volatility From Market Scenarios
A hypothetical asset and a market index have four equally likely outcomes for a particular period. Returns are shown as percentages. Find beta, then compare the asset’s standard deviation with the market’s. Use the population scenario probabilities, each equal to \(0.25\).
| Scenario | Market return | Asset return |
|---|---|---|
| 1 | -10% | -20% |
| 2 | -10% | 0% |
| 3 | 10% | 10% |
| 4 | 10% | 30% |
Method. Calculate each return mean, then the probability-weighted covariance and market variance. Use decimals throughout: for example, \(-10\%= -0.10\).
The market mean is \((-0.10-0.10+0.10+0.10)/4=0\). The asset mean is \((-0.20+0+0.10+0.30)/4=0.05\), or 5%. The market deviations are \(-0.10,-0.10,0.10,0.10\), and the asset deviations are \(-0.25,-0.05,0.05,0.25\).
The four products of paired deviations are \(0.025,0.005,0.005,0.025\). Thus \(\operatorname{Cov}(R_A,R_M)=(0.025+0.005+0.005+0.025)/4=0.015\). The market variance is \(\operatorname{Var}(R_M)=(0.01+0.01+0.01+0.01)/4=0.01\).
Therefore \(\beta_A=0.015/0.01=1.5\). For comparison, the asset variance is \((0.25^2+0.05^2+0.05^2+0.25^2)/4=(0.0625+0.0025+0.0025+0.0625)/4=0.0325\). The asset standard deviation is \(\sqrt{0.0325}=0.1803\), or 18.03% to two decimal places. The market standard deviation is \(\sqrt{0.01}=0.10\), or 10%.
Check. The asset’s standard deviation exceeds the market’s, and the beta is 1.5. These are different summaries: the standard deviation measures the asset’s overall spread in this scenario model, while beta uses only its covariance with the market relative to market variance. The method avoids treating beta as the ratio of standard deviations.
Why the Denominator Is Market Variance
The denominator is specifically the market’s variance because beta compares asset–market co-movement with the market’s own variation. This can be seen by writing the covariance as a weighted average of paired deviations and the variance as a weighted average of squared market deviations. Both quantities use the same market-return deviations; the numerator additionally weights each one by the asset’s deviation.
A useful consequence follows from the definition of correlation. If both returns have positive standard deviations, correlation is covariance divided by the product of their standard deviations. Rearranging that definition gives:
Here, \(\sigma_A\) and \(\sigma_M\) are the standard deviations of asset and market returns. This identity shows two ingredients in beta: the direction and strength of standardized co-movement, represented by correlation, and the relative scale of the asset’s and market’s standard deviations. It does not make beta equal to correlation. Correlation is bounded between \(-1\) and \(1\); beta is not so bounded.
The identity also explains why an asset’s volatility alone cannot determine its beta. Two assets can have the same \(\sigma_A\) but different correlations with the market, and hence different betas. Conversely, a beta value by itself does not specify the asset’s total volatility unless additional information, such as correlation or the asset’s standard deviation, is supplied.
Worked Example: Calculating Beta From Covariance and Market Variance
A hypothetical investment model gives an asset–market covariance of \(0.00588\) and a market-return variance of \(0.00490\), with both calculated using decimal returns over the same period. Find beta and interpret its sign and magnitude.
Method. Substitute the stated covariance and market variance directly into the definition: \(\beta_A=0.00588/0.00490=1.20\). The division is exact to the shown precision.
The positive sign indicates positive covariance: returns above the asset’s mean tend to be associated with market returns above the market mean, and vice versa, in this model. The value 1.20 means the covariance is 1.20 times the market variance. It does not mean the asset’s standard deviation is 1.20 times the market’s; that conclusion would require more information about correlation or asset variance.
Check. Multiplying back gives \(1.20(0.00490)=0.00588\), the supplied covariance. The common mistake avoided here is replacing the given market variance with the asset variance or standard deviation.
Correlation and Beta Are Not Interchangeable
Correlation removes the scale of each return series by dividing covariance by both standard deviations. Beta divides only by market variance. As a result, correlation describes how tightly two return series move together on a standardized scale, while beta retains information about the asset’s return scale relative to the market’s.
Worked Example: Using Correlation and Standard Deviations
Suppose a hypothetical asset has return standard deviation \(20\%\), the market has return standard deviation \(12\%\), and their correlation is \(0.45\). Assume both standard deviations are positive and are calculated for the same return period. Find beta and compare it with the correlation and the standard-deviation ratio.
Method. Use \(\beta_A=\operatorname{Corr}(R_A,R_M)(\sigma_A/\sigma_M)\), which follows from the definition of correlation. Convert the percentages to decimals: \(\sigma_A=0.20\) and \(\sigma_M=0.12\). The standard-deviation ratio is \(0.20/0.12=1.6667\) to four decimal places. Therefore \(\beta_A=0.45(1.6667)=0.75\) (using the unrounded ratio gives \(0.75\)).
The beta is \(0.75\), whereas correlation is \(0.45\) and the standard-deviation ratio is approximately \(1.6667\). They are not interchangeable. As a direct check, covariance is \(0.45(0.20)(0.12)=0.0108\), market variance is \(0.12^2=0.0144\), and \(0.0108/0.0144=0.75\).
Check. Both routes give beta \(0.75\). The calculation avoids the trap of reporting correlation as beta or assuming beta equals the asset-to-market standard-deviation ratio.
Common Errors and a Reliable Calculation Process
A frequent error is to divide covariance by asset variance. For the inputs in the previous example, the asset variance is \(0.20^2=0.04\). The incorrect calculation \(0.0108/0.04=0.27\) does not give beta relative to the market. The correct denominator is market variance, \(0.12^2=0.0144\), so beta is \(0.0108/0.0144=0.75\).
Another trap is using the market standard deviation in the denominator. The definition requires market variance, not market standard deviation. A third is calculating covariance around zero when the returns have nonzero means. Covariance uses deviations from each variable’s mean; multiplying raw returns instead generally gives a different quantity.
For a scenario calculation, a dependable sequence is:
Confirm that asset and market returns refer to the same period and use the same scale, such as decimals for both.
Use the stated scenario probabilities to calculate \(E[R_A]\) and \(E[R_M]\).
Center each return on its own mean, use the same probabilities, and calculate the two weighted sums.
Compute covariance divided by market variance, check that the denominator is positive, and interpret the sign as the direction of co-movement.
Beta is useful when distinguishing market-related exposure from total return variation. It supports a focused description of how an asset co-moves with a specified market return. It does not, on its own, describe every source of risk, guarantee a future relationship, or identify the cause of the observed co-movement. The market return used in the definition must be specified, and changing that reference can change beta.
Check Your Understanding
For each problem, use the stated return units and assumptions. Answers are provided for checking.
- An asset has covariance \(0.0060\) with the market, whose variance is \(0.0040\). Calculate the asset’s beta and state whether the covariance is positive or negative. (Answer: \(\beta=1.50\); positive covariance.)
- An asset has covariance \(-0.0030\) with the market and market variance \(0.0060\). Find beta and interpret its sign. (Answer: \(\beta=-0.50\); negative co-movement.)
- For two return series, the correlation is \(0.60\), the asset standard deviation is \(24\%\), and the market standard deviation is \(16\%\). Find beta. (Answer: \(\beta=0.60(0.24/0.16)=0.90\).)
- In three equally likely scenarios, market returns are \(-10\%,0\%,10\%\), and asset returns are \(-15\%,0\%,15\%\). Calculate both means, covariance, market variance, and beta using decimal returns. (Answer: means \(0\) and \(0\); covariance \(0.0100\); market variance \(0.0066667\); \(\beta=1.50\).)
- A model gives market variance \(0.0225\) and beta \(-0.40\). Find the implied asset–market covariance, then state what the sign indicates. (Answer: covariance \(=-0.40(0.0225)=-0.0090\); negative co-movement.)