Abstract Several unbiased ratio estimators (URE) in stratified random sampling are presented. The “separate” URE is a direct application of the Hartley-Ross URE within strata. The “combined” URE, for equal sample size k per stratum, is obtained by means of a complete replacement, repeated k times, of a sample of size one per stratum. The relative efficiencies of these estimators with respect to the usual biased RE are discussed. An approximately URE is developed and it is shown that, for large samples, it is as efficient as the “combined” biased RE. Finally, an approximately unbiased estimator r* of the population “rate” is given, and its efficiency relative to the “rate-of-means” estimator is investigated.