Small area estimation of sickness absence indicators

Using unit-level multilevel time-series models for estimation of detailed figures on sickness absence for the Health and Wellbeing branch.

Within the AZW research program, there is a demand for detailed figures on a number of sick leave indicators for a detailed breakdown by subpopulations within the Health and Wellbeing branch in particular. To this end, unit-level multilevel time-series models are developed that effectively use the information across subpopulations and over time.

The models are estimated on data from the Netherlands Working Conditions Survey (NEA) from 2014 through 2022. This is used to produce estimates that are much more plausible for small subpopulations than direct estimates based on the NEA survey weights. The time-series model estimates are also an improvement over estimates based on cross-sectional multilevel models, which are estimated for each year separately.