Abstract
Official monthly statistics about the Dutch labour force are based on the Dutch Labour Force Survey (LFS). The LFS is a continuously conducted survey that is designed as a rotating panel design. Data collection among selected households is based on a mixed-mode design that uses web interviewing, telephone interviewing and face-to-face interviewing. Monthly estimates about the labour force are obtained with a structural time series model. Due to the COVID-19 pandemic, face-to-face interviewing stopped. It was anticipated that this would have a systematic effect on the outcomes of the LFS and that the lockdown at the same time affected the real monthly labour force figures. The lockdown indeed marked a sharp turning point in the evolution of the series of the monthly labour force figures and strongly increased the volatility of these series. In this paper, it is explained how Statistics Netherlands produced monthly labour force figures during the COVID-19 pandemic. It is shown how the sudden change in the mode effects, because face-to-face interviewing stopped, were separated from real period-to-period changes in the labour force figures. It is also explained how the time series model is adapted to the increased volatility in the labour force figures.
Original language | English |
---|---|
Pages (from-to) | 1560-1583 |
Number of pages | 24 |
Journal | Journal of the Royal Statistical Society Series A-Statistics in Society |
Volume | 185 |
Issue number | 4 |
Early online date | 27 May 2022 |
DOIs | |
Publication status | Published - Oct 2022 |
Keywords
- discontinuities
- Kalman filter
- measurement error
- mode effects
- state space models
- STATE-SPACE MODELS
- ROTATION GROUP BIAS
- DISCONTINUITIES