Variance vs Standard Deviation

The single biggest mistake newcomers make is treating variance and standard deviation as interchangeable. They are not. Variance is the average of the squared differences from the mean, which puts its unit as the square of the original data. Standard deviation is the square root of variance, which returns the measure to the original units. For any dataset, standard deviation is the square root of the variance. That relationship is the only fact you need to move between them, but knowing when to use each one separates a correct statistical report from a nonsensical one.

Variance is in squared units. If your data is in dollars, the variance is in dollars squared. Standard deviation is in dollars. This is not a minor detail: it is the entire reason standard deviation exists as a separate measure. Variance is the mathematically convenient form used in theoretical derivations and in formulas such as the variance of a sum of independent random variables. Standard deviation is the interpretable form used in reporting, because a reader can understand a number in the same unit as the original measurement.

OpenStax Introductory Statistics 2e section 2.7 defines variance as the average of the squares of the deviations and standard deviation as the square root of the variance. That textbook is the reference for the formulas used here.

The difference between variance and standard deviation is entirely about units. Variance measures spread in squared units; standard deviation measures spread in the original units. If you have a dataset of exam scores from 0 to 100, the variance might be 225 points squared, while the standard deviation is 15 points. The standard deviation tells you that a typical score is 15 points away from the mean. The variance number, 225, is not interpretable in isolation.

This distinction matters because variance is additive for independent variables. If you have two independent random variables X and Y, the variance of their sum is the sum of their variances: Var(X+Y) = Var(X) + Var(Y). Standard deviation does not add this way. The standard deviation of X+Y is the square root of the sum of the variances, not the sum of the standard deviations. That property makes variance the natural choice in theoretical work, including portfolio risk calculations in finance and the derivation of the variance of a sum in probability textbooks such as Blitzstein and Hwang's Introduction to Probability.

In reporting, standard deviation dominates because it is directly comparable to the mean. A 15-point standard deviation on a 100-point exam tells you something meaningful. A variance of 225 does not. This is why textbooks and journal articles report standard deviation in their tables and reserve variance for footnotes or technical appendixes.

To get standard deviation from variance, take the square root. That is the entire conversion. If the variance of a population is σ², the population standard deviation is σ = √σ². If the sample variance is s², the sample standard deviation is s = √s². There is no additional correction or coefficient. The square root operation undoes the squaring that produced the variance in the first place.

This conversion works in both directions. If you have a standard deviation, squaring it gives you the variance. If a TI-84 calculator reports a sample standard deviation Sx of 5.2, then the sample variance s² is 5.2² = 27.04. If Excel returns a VAR.P result of 36, the population standard deviation is √36 = 6. The relationship is exact and deterministic.

The practical failure case occurs when a student or analyst reports a variance in original units. A statement like "the variance is 25 dollars" is wrong. The correct phrasing is "the variance is 25 dollars squared" or "the standard deviation is 5 dollars." The NIST e-Handbook of Statistical Methods and OpenStax both stress the unit distinction in their definitions.

Converting between variance and standard deviation requires no intermediate steps. Square root to go from variance to standard deviation. Square to go from standard deviation to variance. The conversion factor varies depending on whether you are working with a population or a sample because the formulas differ in their denominators, but once you have the correct variance figure, the conversion is identical in both cases.

For a population, the variance formula from OpenStax is σ² = Σ(x - μ)² / N. The standard deviation is σ = √[Σ(x - μ)² / N]. For a sample, the variance formula is s² = Σ(x - x̄)² / (n - 1). The standard deviation is s = √[Σ(x - x̄)² / (n - 1)]. In both cases, the conversion is a square root, not a square root divided by something else or multiplied by a factor. The square root of the variance is always the standard deviation.

The most common error in converting between them is using the wrong formula to compute variance in the first place. If you compute sample variance using n instead of n-1, your variance will be biased low, and the standard deviation you derive from it will also be biased low. That is a failure of the variance calculation, not of the conversion. The conversion itself is reliable.

Yes. Standard deviation is the square root of variance. This is not a rule of thumb or an approximation. It is a definition. OpenStax Introductory Statistics 2e section 2.7 states that the standard deviation is the square root of the variance. Every statistics textbook that defines variance and standard deviation makes the same statement, because it is a mathematical identity.

The question "is standard deviation the square root of variance" arises from the confusion over units. A student who learns that variance is the average of squared deviations and then hears that standard deviation is the average deviation may suspect they are not directly related. They are. The standard deviation is exactly the square root of the variance, no more and no less. If you compute the variance and take its square root, you have the standard deviation. If you compute the standard deviation and square it, you have the variance.

This identity holds for both population parameters and sample statistics. The population standard deviation σ is the square root of the population variance σ². The sample standard deviation s is the square root of the sample variance s². There is no exception for small samples, large samples, normally distributed data, or non-normal data. The relationship is algebraic.

Take the dataset: 4, 8, 12, 16, 20. This is a small sample from a larger population, so you will compute the sample variance and sample standard deviation. The mean is (4+8+12+16+20)/5 = 12. The deviations from the mean are -8, -4, 0, 4, 8. Squaring those deviations gives 64, 16, 0, 16, 64. The sum of squared deviations (SS) is 64+16+0+16+64 = 160.

The sample variance s² uses n-1 in the denominator. With 5 data points, n-1 = 4. So s² = 160 / 4 = 40. The units are squared. Since the original data was a count of something, the variance is 40 counts squared. The sample standard deviation s is the square root of 40. √40 ≈ 6.32. The standard deviation is in the original units (counts). A typical value in this dataset is about 6.32 units away from the mean of 12.

If this dataset were the entire population, the population variance σ² would use N in the denominator: σ² = 160 / 5 = 32. The population standard deviation σ would be √32 ≈ 5.66. The sample statistics (40 and 6.32) are larger because the n-1 denominator corrects for the fact that a sample tends to underestimate the true population spread. This is Bessel's correction, and it is the reason sample variance uses degrees of freedom equal to n-1.

Standard Error and Coefficient of Variation

Two measures that appear alongside variance and standard deviation are the standard error and the coefficient of variation. The standard error of the mean is the standard deviation of the sample divided by the square root of the sample size. It measures the precision of the sample mean as an estimate of the population mean, not the spread of the data itself. The coefficient of variation, defined by OpenStax as (standard deviation / mean) × 100%, is a unitless ratio that allows comparison of variability across datasets with different means. It is only valid for ratio-scale data with a positive mean. For data with a mean near zero or negative values, the coefficient of variation produces a meaningless or infinite result.

Both measures derive from variance and standard deviation but answer different questions. Standard error addresses estimation accuracy. Coefficient of variation addresses relative variability. Neither replaces variance or standard deviation; they supplement them in specific analytical contexts.

Variance in Theory

Variance is used in theory because it is additive for independent random variables. If X and Y are independent, Var(X+Y) = Var(X) + Var(Y). This property makes variance the natural choice for deriving the variance of a sum, the variance of a portfolio, and the variance of a random variable in probability theory. Standard deviation lacks this additive property and is therefore less convenient in algebraic derivations.

Standard Deviation in Reporting

Standard deviation is used in reporting because it is in the original units. A reader can compare a standard deviation to the mean and understand the spread. A variance in squared units requires mental translation that most readers cannot perform. This is why journal articles, business reports, and textbook examples almost always report standard deviation and sometimes do not report variance at all.

The failure case occurs when a report states a variance without stating the unit. If a financial report says "the portfolio variance is 0.04" without specifying that this is in squared percent, a reader might incorrectly interpret it as a standard deviation of 0.04 or 4 percent. The correct interpretation is that the standard deviation is 0.20, or 20 percent. The CFA Institute Curriculum and Bodie, Kane, and Marcus's Investments both use variance for portfolio calculations and standard deviation for risk reporting, respecting this distinction.

When you compute a variance, immediately take its square root to get the standard deviation. Report the standard deviation. Keep the variance in your working notes for any subsequent calculation that requires additive spread, such as adding two independent variances together. Never report a variance in a business document without also providing the standard deviation. The single thing that most often goes wrong is reporting a variance in original units as if it were a standard deviation, which misleads every reader who sees the number.