Efficiency and feasibility of sampling schemes in establishment surveys

When conducting firm-level surveys, a relevant aspect in sam-pling design is to guarantee that the selected sample of businessunits is representative of the population. Stratified samplingdesign is used to obtain a representative sample. However, stra-tification of the main characteristics of establishments may notalways be computationally feasible, due to the high number ofsmall and empty strata occurring when the population is highlystratified. A simulation with common sampling designs allows thecomparison of the local pivotal method and spatially correlatedPoisson sampling in terms of spatial balance and efficiency inestimation. These spatial designs are efficient, whereas stratifica-tion is no longer feasible.