Title: Statistical properties of measurement error in earnings in labor market survey data
Authors: Stella Martin - University of Muenster (Germany) [presenting]
Kevin Stabenow - University of Muenster (Germany)
Mark Trede - University of Muenster (Germany)
Abstract: Large-scale surveys on income play an important role in empirical economic research while being subject to a measurement error little is known about. Little literature, almost entirely based on one small linkage of employee and employer information on earnings, reveals that there might, in fact, be a systematic bias in self-reports of earnings. We use a novel linkage of survey and administrative data. The German Socioeconomic Panel spanning 37 waves (1984 to 2020) and data from the German Pension Insurance on individuals' entire earnings records provide a comprehensive panel on employment biographies and earnings information. We use these data to investigate the statistical properties of the measurement error in earnings on an individual and household level, where, unlike previous work, the rich panel structure in our data allows us to focus not only on the distribution of measurement error in the cross-section, but also on its autocorrelation and time series properties.