Standard Error (SE)

What Is a Standard Error?

A standard error (SE) estimates how precisely a sample statistic reflects the true value in the wider population. When you calculate an average from a sample of stock returns, that average is unlikely to match the population exactly. The standard error indicates how much that sample figure would be expected to fluctuate if you repeated the sampling process many times.

It is closely related to standard deviation, but the two measure different things. Standard deviation describes how spread out individual data points are within one sample. Standard error describes how spread out a statistic, such as the sample mean, would be across many samples.

How It Is Calculated

For a sample mean, the standard error is the sample standard deviation divided by the square root of the sample size:

SE = s / √n

Here s is the sample standard deviation and n is the number of observations. Because of the square root, the standard error shrinks as the sample grows, but more slowly than the sample size itself. Quadrupling the sample size only halves the standard error.

For other statistics, such as regression coefficients, the formula is different. Statistical software typically reports the standard error automatically alongside each estimate.

Example and Use in Finance

Suppose an analyst estimates a fund's average monthly return from a sample of 36 months. The standard error indicates how much that estimate could plausibly differ from the fund's long-run average. A small standard error relative to the estimate suggests the figure is relatively precise, while a large one suggests it is noisy.

Investors often see standard errors in regression analysis output, where they accompany coefficients such as a stock's beta. Standard errors are also used to build a confidence interval, typically by adding and subtracting a multiple of the standard error from the estimate.

Limitations

A standard error reflects only random sampling variation. It does not capture biased data, changing market conditions, or errors in the underlying model. Results based on small samples, or on returns that are not independent of one another, can be misleading even when the standard error looks small.