The ISAE manufacturing survey sample : validating the Nace Rev.2 sectorial allocation

After the full implementation of the new EU Standard Classification of Economic Activities (Nace Rev.2) in 2008, statistical agencies have increasingly dealt with the problem of redefining sampling designs and estimation techniques, especially in the case of stratified surveys with NACE codes as stratification variables. In light of this changes, the Italian Institute for Studies and Economic Analysis (Istituto di Studi e Analisi Economica - ISAE) is currently updating the sample design of its Business Tendency Survey (BTS). The focus of this paper is on finding a strata allocation methodology suitable to overcome the NACE Rev.2 changes. The analysis is carried out by considering two opposite needs: i) the strata allocation must retain multiple information; ii) the strata allocation must retain the optimality of the estimates. The allocation methods considered are: i) the classical Neyman x-optimal allocation, ii) the Neyman allocation used by ISAE, i.e. with direct application to areal stratification, iii) the multivariate Neyman allocation on qualitative variance according to Bethel formulation, iv) the Robust Optimal Allocation with Uniform Stratum Threshold (ROAUST). The ROAUST is a new allocation method which generates a new class of stratified estimators. Comparison among these methods is carried out via a simulation device - the Sequential Selection- Allocation (SSA). This simulation device constructs a new population list with units re-labelled within each stratum, such that the new labels corresponds to the order of selection in a SWOR resampling of the stratum units. This process is repeated a certain number N (N=1,000 in the simulation presented in this paper) of times. From this new labelled population, all the allocation algorithms can be evaluated simultaneously