How reliable are incidence estimates based on cross-sectional distributions? Evidence from simulations and linked employer-employee data

The paper validates an empirical approach developed by Alvaredo and Saez (2007) which estimates the economic incidence of social security contributions (SSC) on the basis of cross-sectional earnings distributions. The method utilizes discontinuities at earnings caps where the marginal SSC rate drops. It does not rely on policy reforms, panel data, or hours information. We demonstrate on the basis of simulations that this comes at the cost of ruling out optimization frictions and measurement error destroying identification. Behavioral responses and incidence heterogeneity are secondary problems. Exploiting German linked employer-employee data that provide separate measures of gross and net earnings we find substantial negative discontinuities in net earnings. Together with small, erratic discontinuities in gross earnings this provides consistent empirical evidence that legal and economic incidence of SSC coincide in Germany.