Gauss-Markov Theorem

Also known as · BLUE

Under assumptions A1 to A4 (mean-zero errors, exogeneity, no autocorrelation, homoskedasticity), OLS Estimation has the smallest variance of all linear unbiased estimators, so it is the best linear unbiased estimator (BLUE).

When to use

Use it to say what is lost when an assumption fails: under Heteroskedasticity A4 fails, so OLS stays unbiased but is no longer efficient, which is why Weighted Least Squares can do better.

See Lecture 4 and Lecture 13 and Lecture 12.

Appears in