Efficiency and feasibility of sampling schemes in establishment surveys

When conducting firm-level surveys, a relevant aspect in sampling design is to guarantee that the selected sample of business units is representative of the population. Stratified sampling design is used to obtain a representative sample. However, stratification of the main characteristics of establishments may not always be computationally feasible, due to the high number of small and empty strata occurring when the population is highly stratified. A simulation with common sampling designs allows the comparison of the local pivotal method and spatially correlated Poisson sampling in terms of spatial balance and efficiency in estimation. These spatial designs are efficient, whereas stratification is no longer feasible.