Non-Response in Dynamic Panel Data Models

This paper stresses the links that exist between concepts that are used in the theory of model reduction and concepts that arise in the missing data literature. This connection motivates the extension of the missing at random (MAR) and the missing completely at random (MCAR) concepts from a static setting, as introduced by Rubin (1976), to the case of dynamic panel data models. Using this extension of the MAR and MCAR definitions, we emphasize the limits of some tests and procedures, proposed by Little (1988), Diggle (1989), Park and Davis (1993), Taris (1996) and others, to verify the ignorability of the missing data mechanism.