Multiple Imputation of Missing Income Data in the Survey on Income and Living Conditions

Imputing for missing data in the EU-SILC survey on Income and Living Conditions is a challenging task, due to the large amount of available information and to its complicating features. ISTAT currently imputes one value for each missing income datum, using the sequential regression multivariate approach. However, it is well known that the possible drawback of single imputation is the underestimation of the true estimates’ variability. To avoid this drawback, multiple imputation has been proposed. In this paper a description of multiple imputation under the sequential regression approach is presented, together with a description of the theoretical and applied desirable characteristics of multiple imputation. Moreover, a multiple imputation model taking into account the structure and the characteristics of missing and available EU-SILC data is proposed. The results of this model are compared to those obtained without imputation and using single imputation, to evaluate the impact on the final target income estimates produced using EU-SILC data.