A Practical Guide for the Computation of Domain-Level Estimates with the Socio-Economic Panel (And Other Household Surveys)

Abstract

There is a huge interest in deriving and comparing socio-economic indicators across societal groups and domains. The indicators are usually derived from population surveys like the German Socio-Economic Panel (SOEP) by direct estimation. Small sample sizes in the domains can limit the precision of these estimates. For example, while SOEP may be a suitable database for determining mean income in Germany, it is unclear whether this also applies to smaller domains (for example, women in Berlin). Here we show SOEP-based applications of Stata’s fayherriot package (Halbmeier et al., 2019). This package implements the Fay-Herriot model (Fay and Herriot, 1979), a small-area estimation technique designed to improve the precision of domain-level direct estimates using domain-level covariates from auxiliary datasets.