Revisiting Sample Allocation Methods: A Simulation-Based Comparison

In stratified sampling the problem of optimally allocating the sample size is of primary importance, especially in business surveys when reliable estimates are required both for the overall population and for the domains of studies. To this purpose, in this paper we compare allocation methods via a simulation engine highlighting the effects on the reliability of the estimates due only to the sample allocation design. Allocation methods considered in this comparison are: the Neyman allocation, the uniform and proportional allocations, the Costa allocation, the Bankier allocation, the Interior Point Non Linear Programming allocation and the Robust Optimal Allocation with Uniform Stratum Threshold, an allocation method recently adopted by the Italian National Statistical Institute. The last two methods outperform the others at the stratum level. At the overall sample level they perform better than the others together with the Neyman allocation method.