Instrumental variables estimation with partially missing instruments

This paper shows that instrumental variables estimators currently in use, require strong but neglected auxiliary assumptions to be consistent in situations with partially missing instruments. We introduce an alternative instrumental variables estimator that does not require auxiliary assumptions.

► This paper examines IV estimation with partially missing instruments. ► One application is in the literature on the relationship between family size and child outcomes. ► We characterize the inconsistency in IV estimation when the instrument is missing non-randomly. ► We show that standard IV estimators require strong auxiliary assumptions to be consistent. ► We introduce an alternative IV estimator that is robust to non-randomly missing instruments.