How to Calculate Variance in Statistics

Variance is the average of squared deviations from the mean. If every value equals the mean, variance is 0. If values sit far above and below the mean, variance grows quickly because distances are squared.

Why square?

Raw deviations above the mean are positive and below are negative; their plain average is always 0. Squaring makes every deviation positive and penalizes large outliers more than small ones. That is useful for measuring spread, but it also means variance is in squared units (dollars², cm², points²) — harder to read than standard deviation.

Intuition with two classes

Scores A: 78, 80, 82. Mean = 80. Deviations are small → low variance.

Scores B: 60, 80, 100. Mean = 80. Same center, much larger squared deviations → high variance. Average alone cannot tell these classes apart; variance (or SD) can.

For a full spread toolkit including SD steps, see our standard deviation guide and the free standard deviation calculator.

The formulas look almost identical. The denominator is the difference that matters.

• Population variance (σ²) — you have the entire population of interest. Divide the sum of squared deviations by n.
• Sample variance (s²) — you have a sample and want to estimate the population’s spread. Divide by n − 1 (Bessel’s correction).

Population: σ² = Σ(xᵢ − μ)² / n

Sample: s² = Σ(xᵢ − x̄)² / (n − 1)

When to use which

Use population (÷ n) when the data set is the whole group you care about (every student in a closed class of 20, every machine in a five-machine plant). Use sample (÷ n − 1) when your numbers are a sample used to infer a larger population — the usual case in surveys, experiments, and most lab work.

Dividing by n − 1 makes sample variance a bit larger, which corrects a downward bias that appears if you use the sample mean and still divide by n. For large n the gap shrinks; for n = 5 it is obvious.

Symbols: μ and σ² usually mark population mean and variance; x̄ and s² mark sample mean and sample variance. Software defaults vary — some functions are sample by default (for example many spreadsheet VAR variants). Read the docs before you paste exam data into a random cell formula.

Work a sample variance example with data: 4, 7, 7, 8, 10 (n = 5).

1. Mean: (4 + 7 + 7 + 8 + 10) ÷ 5 = 36 ÷ 5 = 7.2
2. Deviations: 4−7.2 = −3.2; 7−7.2 = −0.2; −0.2; 0.8; 2.8
3. Squared: 10.24; 0.04; 0.04; 0.64; 7.84
4. Sum of squares: 10.24 + 0.04 + 0.04 + 0.64 + 7.84 = 18.8
5. Sample variance: 18.8 ÷ (5 − 1) = 18.8 ÷ 4 = 4.7
6. Population variance (if these five are the whole population): 18.8 ÷ 5 = 3.76

Computational formula (optional)

An algebraically equivalent form uses Σx² and (Σx)² to reduce rounding of many deviations — useful by hand on longer lists. Tools and spreadsheets still follow the same definition.

Paste the list into the free variance calculator to confirm both population and sample results and related stats in one pass.

With n = 2, sample variance is especially sensitive: a single pair of far-apart points produces a large s² because you only divide by 1. That is mathematically correct and a reminder that tiny samples make spread estimates noisy — report them with caution.

Standard deviation = √variance (population SD = √σ²; sample SD = √s²).

From the example above, sample variance 4.7 → sample SD ≈ √4.7 ≈ 2.17. That 2.17 is in the same units as the original data, so “typical distance from the mean is about 2.2” is easier to say than “variance is 4.7 squared units.”

Which number to report

• Variance — common in ANOVA, some finance models, and theoretical formulas that stay in squared units
• Standard deviation — common in reports, quality control charts, and everyday “how spread out is this?” answers

Never mix them: do not average SDs incorrectly or take √ of the wrong variance type. Match population vs sample consistently through mean → variance → SD.

Once you have mean and SD, standardized scores are next — see the free z-score calculator and our z-score guide.

Same mean, different variance, different decisions.

• Quality control — two machines average 100 units; high variance means more out-of-spec parts
• Risk — two investments average similar returns; higher variance (or SD) often means rockier path
• Grading and tests — class means can match while one class is polarized and another is consistent
• Experiments — variance feeds standard errors and confidence intervals that decide if a difference is noise

Coefficient of variation (bonus)

When you compare spread across different means or units, people sometimes use CV = (SD ÷ mean) × 100%. A SD of 5 on a mean of 20 is huge; a SD of 5 on a mean of 500 is small. Many variance tools also show CV for that reason.

Chance models that sit beside descriptive spread are covered in our probability guide and free probability calculator.

Paste or type the data set, choose population or sample mode, and read variance, SD, and related summaries.

1. Open the free variance calculator.
2. Enter each value (comma, space, or line separated as the tool allows).
3. Select population (÷ n) or sample (÷ n − 1) to match your question.
4. Confirm the mean matches what you expect.
5. Read variance first, then standard deviation.
6. Note min, max, or CV if shown for extra context.
7. Rerun after removing an outlier to see sensitivity — large squared penalties show up fast.

Data hygiene

Exclude blank cells and non-numeric junk. Decide in advance whether a value is a true data point or a data-entry error. Outliers are sometimes the story, not something to delete automatically.

Most mistakes are denominator choice or unit confusion.

• Using n when you meant n − 1 (or the reverse) for the research question
• Averaging absolute deviations and calling it variance (that is a different measure)
• Forgetting to square before averaging
• Comparing variance in different units without converting or using CV
• Reporting variance when readers need SD (or mixing the two in one sentence)
• Computing SD on population formula then labeling it sample SD

This guide is educational, not a substitute for course requirements or formal statistical consulting. Match your textbook or analysis plan’s population vs sample definition.