🤖 AI Summary
This study systematically characterizes measurement error in individual labor income using exact probabilistic linkage between the German Socio-Economic Panel (SOEP) and administrative Integrated Employment Biography (IEB) data. Methodologically, it innovatively constructs a causal identification framework grounded in real-data linkage, combining reliability ratio estimation, first-difference error amplification analysis, and selection bias diagnostics. Key contributions: SOEP income measurement error is nonclassical—exhibiting systematic underreporting, temporal autocorrelation, and dependence on both true income and observable covariates; consent to data linkage introduces observable selection bias, undermining the assumption of random sampling. Results show that the reliability ratio for level income exceeds 0.94, implying minimal attenuation bias in single-period linear regressions; however, measurement error in income changes (first differences) is substantially amplified, necessitating careful modeling in dynamic analyses.
📝 Abstract
This paper exploits the linkage of German administrative social security data (GER: Integrierte Erwerbsbiografien) and survey data from the socio-economic panel (GER: Sozio-""okonomisches Panel, SOEP) for the characterization of measurement error in metrics quantifying individual-specific labor earnings in Germany. We find that survey participants' decision whether to consent to linkage is non-random based on observables. In that sense, the studied sample does not constitute a random sample of SOEP. Measurement error is not classical: we observe underreporting of income on average, autocorrelation, and non-zero correlation with the true signal and other observable characteristics. In levels, calculated reliability ratios above 0.94 hint at a relaitvely small attenuation bias in simple linear univariate regressions with earnings as the explanatory variable. For changes in income, i.e. first differences, the bias from measurement error is exacerbated.