Variance and How to Measure the Spread of Returns
In this chapter
Intermediate
Two funds can both average 4% and give investors very different months. Variance measures how spread out the returns are around their mean. It squares the distance of each observation from the mean before combining the distances. A series that spends more time far from its average usually has higher variance, with large deviations carrying extra weight.
Compare each return with its own series mean
First calculate the arithmetic mean by giving each observation the same weight. Subtract that mean from each return. A return below the mean has a negative deviation, even if the return itself is positive. Those signed deviations add to zero. Squaring makes every contribution nonnegative and gives more influence to the observations farthest from the mean. A four-point distance contributes sixteen squared points. A two-point distance contributes four.
Use equally spaced returns on the same basis. Here the data are hypothetical nominal USD monthly simple total returns, including income and before fees and taxes. The calculations use percentage-point values for deviations. If you enter returns as decimals instead, square decimal deviations throughout. A variance of 5 percentage points squared is \(5/10{,}000=0.0005\) in decimal return units squared. Moving the decimal point as if variance were a return would change the calculation.
Worked case. Find the spread around the same mean
Treat the four months as the full period you want to describe. The mean for each fund is 4.00%.
| Month | A return | A deviation (pp) | A squared (pp²) | B return | B deviation (pp) | B squared (pp²) |
|---|---|---|---|---|---|---|
| 1 | 1% | −3 | 9 | −2% | −6 | 36 |
| 2 | 3% | −1 | 1 | 2% | −2 | 4 |
| 3 | 5% | +1 | 1 | 6% | +2 | 4 |
| 4 | 7% | +3 | 9 | 10% | +6 | 36 |
| Sum | 0 | 20 | 0 | 80 |
Divide each sum of squared deviations by four. Fund A’s variance is 5.00 pp² and fund B’s is 20.00 pp². Every distance from the mean doubled in B, so variance quadrupled. The equal means hide the different spread. A variance of 20 pp² describes the squared distances, with the square root used when you want to express spread in ordinary return units.
Choose the denominator from the question
For a complete population of \(N\) equally weighted observations, use
\[ \sigma^2=\frac{\sum_{i=1}^{N}(r_i-\mu)^2}{N}. \]
\(\sigma^2\) is the population variance, \(r_i\) is return \(i\) and \(\mu\) is the mean of the whole population. If you use \(n\) observations to estimate variance for a wider process, the conventional sample variance is
\[ s^2=\frac{\sum_{i=1}^{n}(r_i-\bar r)^2}{n-1}. \]
\(s^2\) is sample variance, \(\bar r\) is the sample mean, and \(n\) must be at least two. In both formulas the sum adds one squared deviation per row. The sample mean is fitted to the data. After you know it and all but one deviation, the last deviation follows from the rule that they sum to zero. Under independent sampling from a common distribution with finite variance, dividing by \(n-1\) corrects the tendency to underestimate the population variance when the mean is fitted this way.
Worked case. Reuse the same four observations for a different question
| Question about fund A | Squared-deviation sum (pp²) | Denominator | Variance (pp²) |
|---|---|---|---|
| How spread out were these four specific months? | 20 | 4 | 5.00 |
| What does this sample suggest about the broader return process? | 20 | 3 | 6.67 |
The deviations stayed fixed. The first result describes the complete set you named. The second is an estimate for a larger process under the sampling assumptions. A four-month run can leave out whole market conditions. The smaller denominator does not solve that problem. Before comparing variance estimates, check the observation period, sampling window, return basis and units. A return series that is flat throughout your sample has zero measured variance even when future outcomes are uncertain. You can also shift all the returns without changing variance. Add two percentage points to every Fund A observation and its mean rises from 4% to 6%. The first month goes from 1% to 3%, so it stays three points below the mean. Each of the other three distances stays fixed too. Moving the whole series changes its level. Multiplying the deviations changes their spread.
Practice questions
Choose one answer per question. Round only the final result where needed.
1. Calculate descriptive variance
A complete three-month set contains returns of 1%, 3% and 5%. What is the variance in pp²?
2. Switch to a sample estimate
A sample of four months has a squared-deviation sum of 36 pp² around its sample mean. What is sample variance?
3. Use squared units
A complete return series has variance 0.0009 in decimal return units squared. What is the same variance in pp²?
4. Shift a series
You add two percentage points to each of five returns. What happens to variance over those five observations?
5. Choose the descriptive denominator
Your task is to describe all six monthly returns during one already specified reporting window. Which denominator fits?
6. Read a constant series
Four monthly observations all equal +2%. What is the population variance of those four returns?