Economics Seminar Series: Stephen P Jenkins
- Date: Wednesday 24 March 2021, 15:00 – 16:30
- Location: Online
- Type: Seminars and lectures
- Cost: Free
You are invited to a Zoom seminar by Stephen P Jenkins from the London School of Economics on “Reconciling reports: modelling employment earnings using survey and administrative data”
There is an extensive literature that has studied measurement error in household survey data on labour earnings using linked administrative data to provide a benchmark earnings measure that is assumed to be error-free. A small and more recent literature allows for measurement error in the administrative data as well. The current paper belongs to this second generation of studies. We contribute new UK evidence about measurement error in employment earnings to a field dominated by findings about the USA, developing and applying new econometric models that allow for various types of measurement error in both administrative and survey data while also allowing error distributions to vary with observed characteristics.
We generalize the models of Kapteyn and Ypma (Journal of Labor Economics, 2007) to allow for mean-reverting measurement error in the administrative data (in addition to linkage mismatch and survey measurement error), and differences in measurement error distributions with differences in observed characteristics. We fit our models to data from the FRS-P14 Linked Dataset, a new source for the UK. By comparison with earlier studies, we give greater attention to differences in the definition of the reference periods for the earnings measures in the survey and to administrative data. This emphasis reflects the way in which UK surveys collect earnings data.
We find evidence of significant measurement error in the administrative data in addition to linkage mismatch and survey measurement error, but no evidence of mean-reverting errors in either source. Measurement error distributions vary systematically with characteristics. For example, earnings measurement error variances are distinctly lower for respondents showing a payslip to their interviewer.
Please note: we have no written version of the paper yet. For a short introduction to Kateyn-Ypma models and our data set, see Jenkins and Rios-Avila (2020), ‘Measurement errors in survey and administrative data on employment earnings: sensitivity to the fraction assumed to have error-free earnings’, Economics Letters; access the ungated pre-print here.
All are welcome to attend!