Small sample estimators are developed for two post strata and compared with pooling or collapsing procedures commonly employed in practice. The estimators are not necessarily conditionally unbiased for particular sample split, but their conditional M.S.E. may be smaller than that of the common post stratified estimator. It is proven that it is always possible to construct an estimator superior to the practice of combining two post strata when one contains few sample elements. A procedure for generalizing the two-strata procedures to any number of strata is illustrated. This procedure permits the construction of unbiased estimators for populations divided into a large number of small post strata.