Imputation of Missing Data in Waves 1 and 2 of SHARE

The Survey of Health, Ageing and Retirement in Europe (SHARE), like all large household surveys, suffers from the problem of item non-response, and hence the need of imputation of missing values arises. In this paper I describe the imputation methodology used in the first two waves of SHARE, which is the fully conditional specification approach of van Buuren, Brand, Groothuis-Oudshoorn, and Rubin (2006). Methods for assessing the convergence of the imputation process are also discussed. Finally, I give details on numerous issues affecting the implementation of the imputation process that are particular to SHARE.Data; Multiple Imputation; Markov Chain Monte Carlo; SHARE