The question that decides which formula to use
Before computing variance or standard deviation, one question matters more than any formula detail: is your dataset the entire population you care about, or just a sample drawn from a larger population? Population statistics describe exactly the data you have, in full. Sample statistics use your data to estimate a property of a larger group you didn't fully measure — and that estimation task requires a specific correction.
The two formulas, side by side
Population variance: σ² = Σ(xᵢ - μ)² / n
Sample variance: s² = Σ(xᵢ - x̄)² / (n - 1)
The only difference is the denominator: n for population, n - 1 for sample. This adjustment — dividing by one fewer than the count — is known as Bessel's correction.
Why n-1, specifically
Here's the actual reason, not just "because that's the rule": when you compute a sample's mean and then use that same sample's mean to measure how spread out the data is, you're measuring each point's distance from a value that was itself calculated from the same limited data — which systematically makes the sample look slightly less spread out than the true population actually is. Dividing by n - 1 instead of n inflates the result just enough to correct for this bias, producing an unbiased estimate of the population's true variance. Using plain n on a sample would consistently, systematically underestimate the population's actual variability.
A concrete illustration of the bias
Imagine repeatedly drawing different small samples from the same population and computing variance both ways. Using n as the divisor, the average of all those sample variances comes out systematically lower than the population's true variance — a real, provable bias, not a rounding artifact. Using n - 1 corrects this: the average of many sample variances computed this way converges to the true population variance, which is exactly the property that makes it the mathematically correct choice for estimation.
How to decide which one applies to your data
| Your situation | Use |
|---|---|
| You have data for every member of the group you care about (e.g., every student in one specific class, analyzed as that exact class) | Population (÷ n) |
| Your data is a subset used to infer something about a larger group (e.g., a survey sample used to estimate a national average) | Sample (÷ n-1) |
In practice, most real-world statistical analysis involves samples, not full populations — which is why sample variance (n-1) is the more commonly needed formula in applied statistics, even though population variance is the simpler, more intuitive one to derive first.
Common mistakes
- Using population variance (÷n) on what's actually a sample. This systematically underestimates the true variability of the larger population you're trying to draw conclusions about.
- Assuming the difference is negligible for small datasets. The gap between dividing by n versus n-1 is proportionally largest for small sample sizes — exactly when getting it right matters most.
- Applying sample variance to data that genuinely is the entire population of interest. If there's no larger group being estimated, Bessel's correction serves no purpose and population variance is the correct, unbiased choice on its own.
FAQ
Why does sample variance divide by n-1 instead of n?
Because using the sample's own mean to measure spread introduces a small, systematic underestimate of the true population variance — dividing by n-1 (Bessel's correction) corrects for this bias, producing an unbiased estimate.
Does the n vs. n-1 difference matter much for large datasets?
Less than for small ones — the proportional gap between dividing by n and n-1 shrinks as sample size grows, though the correction is still technically appropriate whenever you're estimating a larger population from a sample, regardless of size.
How do I know if my data is a "population" or a "sample"?
Ask whether your data represents every member of the group you care about (population) or only a subset used to infer something about a larger group you didn't fully measure (sample) — this determines which formula is actually correct.
Compute both population and sample variance/standard deviation instantly with the Statistics Calculator — entirely in your browser.