Occupational Status Gross Flows Estimation from Repeated Complex Sample Surveys under Differential Nonresponse

Abstract

Occupational status (Occupied, Unoccupied and Inactive) gross flows represent the transitions of the working age population between these categories in two time periods, for a region or population of interest. The analysis of occupational status gross flows is a powerful tool for understanding labour marked dynamics. Estimation of gross flows requires longitudinal measurements at the individual level. We describe a superpopulation modelling approach and corresponding pseudo-maximum likelihood estimation strategy that, combined, enables accounting for both the complex sampling designs typically used in labour force surveys as well as for handling the differential nonresponse often observed in these surveys. This approach compared favourably in simulation studies with simpler methods that ignore the complex design features or the differential nonresponse. It was then successfully applied to estimate gross flows for a large scale monthly labour force survey carried out in Brazil, considering data from one of the six metropolitan areas covered in the survey. The best fitting models for the nonresponse patterns indicate that probability of responding depends on the occupational status on the baseline period. Therefore, estimating gross flows without considering this differential nonresponse would lead to biased estimates of the underlying transition probabilities.