Date created: | Last Updated:

Category: Project

Description: Administrative register data are increasingly used worldwide to replace or supple- ment the census, and are thought to provide a cost-effective opportunity for longitudinal full-population data analysis in the social sciences. They are also frequently used as “val- idation data” to study measurement error in survey questions. In spite of quality control procedures, however, there are strong indications that administrative register data can themselves contain considerable measurement error. Moreover, typically the error pro- cess does not conform to classical measurement error models. Such errors negate the potential usefulness of administrative data, making it essential to evaluate their extent. This chapter discusses latent variable modeling as a way to estimate measurement error in administrative data by combining error-prone administrative data with an error-prone survey. To demonstrate the approach, a latent class model is applied to linked register- survey residence data from the municipality of Amsterdam.