If we are presented with an estimator that is both linear and unbiased, then we know that the variance of this estimator is at least as large as that from OLS.