Weighting methodology for the national EVREST survey data

Abstract

The main objective was to describe the weighting methodology used for the national EVREST (Evolution and Relations in Health at Work) survey data. The secondary objectives were on the one hand to assess the extent of the differences between crude and weighted estimates, on the other hand to verify that the two-year gap in the availability of the reference data used does substantially not impact the estimates. The study was based on data collected in 2013 and 2014 (N=26,227). The weighting included 2steps: 1)a first weighing to take into account the probability of participation of each employee; and 2)a calibration on margins to correct the potential distortions of the sample in comparison with the scope of the survey, the reference data used coming from the annual declarations of social data (DADS) of the years 2014 and 2012. The impact of the weighting method was studied using the differences between crude and weighted percentages for the 60variables of the questionnaire. 90% of the differences between crude and weighted estimates were between -2.0% and +2.0% using the 2014 DADS, and 83% using the 2012 DADS. The most overestimated crude estimate concerned full-time work and the most underestimated was contact with the public. The impact of the two-year gap in the availability of the reference data used was weak. A weighting methodology for EVREST survey was define and implement, allowing results to be extrapolated to the scope of the survey.