Best Linear Unbiased Estimator

A best linear unbiased estimator is a linear unbiased estimator with the smallest variance among all linear unbiased estimators.

A best linear unbiased estimator, usually shortened to BLUE, is a linear unbiased estimator that has the smallest variance among all linear unbiased estimators in the class being considered.

What each word means

  • Best: lowest variance in the relevant class
  • Linear: built as a linear function of the observed data
  • Unbiased: centered on the true parameter in expectation

In introductory econometrics, the famous statement is that ordinary least squares is BLUE under the Gauss-Markov assumptions.

Why the idea matters

BLUE is about efficiency within a restricted class. It does not mean an estimator is perfect, and it does not automatically mean it is best among all possible nonlinear estimators. It means that, given linearity and unbiasedness, no other estimator in that class is more precise.

Model logic

The concept matters because empirical work balances several estimator properties at once:

  • unbiasedness,
  • variance,
  • consistency,
  • robustness to assumption failures.

BLUE clarifies one important benchmark: what OLS can achieve when the classical linear-model assumptions hold.

Knowledge Check

### In BLUE, what does "best" mean? - [x] Lowest variance among linear unbiased estimators - [ ] Largest sample size - [ ] Highest explanatory power in any model - [ ] Zero residuals in every sample > **Explanation:** "Best" refers to precision within the linear unbiased class, not to every possible desirable property. ### Why is OLS often discussed in connection with BLUE? - [x] Because under the Gauss-Markov assumptions, OLS is BLUE - [ ] Because OLS is always unbiased under any condition - [ ] Because OLS is nonlinear - [ ] Because OLS ignores variance > **Explanation:** The Gauss-Markov result is the classic reason BLUE appears in econometrics courses. ### Does BLUE mean an estimator is best among all imaginable estimators? - [ ] Yes - [x] No - [ ] Yes, because unbiasedness guarantees that - [ ] No, because variance is irrelevant > **Explanation:** BLUE is a restricted comparison. It applies to the class of linear unbiased estimators, not to every possible estimator.