Title: Efficient GMM estimation with incomplete observations
Authors: Chris Muris - Simon Fraser University (Canada) [presenting]
Abstract: The standard MAR model has 2 strata of missingness, complete observations and completely missing observations. A GMM missing data model is proposed for settings where observations may be incomplete, where some observations are missing some instruments and other observations are missing different instruments. This model is identified under MAR. The implied estimators attain the semiparametric efficiency bound and are easy to implement.