A theorem, proved independently by Rao in 1945 and Blackwell in 1947. Let {Xi} be independent identically distributed random variables with θ being an unknown parameter of their common probability distribution. The theorem states that, if T and S are two functions of the {Xi}, with T being an unbiased estimator of θ, and S being a sufficient statistic, then, for all θ, the expected value of T given S is an unbiased estimator of θ with variance never greater than that of T.