Between Theoretical and Applied Approach: Which Compromise for Unit Allocation in Business Surveys?

Neyman’s algorithm for the allocation of sample units in business sampling can result unsatisfactory in domain analysis with imperfect frames and sectorial and/or regional data. Improved estimates can be obtained using stratified estimators combined with an optimal unit allocation. We achieve this outcome by an interdisciplinary approach which leads to a methodological improvement. Starting from Martini’s approach which considers an empirical view of the statistical analysis, we propose the Robust Optimal Allocation with Uniform Stratum Threshold (ROAUST) class of stratified estimators and prove their reliability by using a simulation approach inspired by Magagnoli’s work on this issue. In particular, contrary to Neyman’s stratified estimator with optimal allocation and stratum threshold, our class guarantees better domain representativeness.