Analyzing the changing gender wage gap based on multiply inputed right censored wages

In order to analyze the gender wage gap with the German IAB-employment sample we have to solve the problem of censored wages at the upper limit of the social security system. We treat this problem as a missing data problem. We regard the missingness mechanism as not missing at random (NMAR, according to Little and Rubin, 1987, 2002) as well as missing by design. The censored wages are multiply imputed by draws of a random variable from a truncated distribution. The multiple imputation is based on Markov chain Monte Carlo (MCMC) technique. We complete the dataset with this technique in order to apply a Juhn-Murphy-Pierce-decomposition. As the main sources for the narrowing gender wage gap from 1991 to 2001 we identify an improvement of women's position within the wage distribution.