Efficiency measures an estimator’s precision: among unbiased estimators for a given statistical model, an efficient estimator achieves the smallest possible variance. It describes how tightly estimates cluster, not whether they converge to the true parameter. section 3.2.2
Fisher information quantifies how much the data reveal about a parameter. The Cramér–Rao lower bound sets a limit on the precision of unbiased estimators; reaching that bound makes an estimator efficient. Efficiency is often asymptotic: an estimator approaches the bound as the sample size grows. section 3.2.2
Comparing two estimators
Two estimators can both converge toward the true parameter, while one produces estimates more tightly clustered around it. The tighter estimator is more efficient when it uses the available information to achieve lower variance. section 3.2.2
Maximum likelihood estimators often achieve or approach the bound asymptotically under the assumed model. Efficiency is one criterion for evaluating estimators, alongside consistency and sufficiency; it does not by itself guarantee either property. section 3.2 section 3.2.2
