Abstract
We study measurement error in the simplest dynamic panel data model without covariates. We start by investigating the first-order effects, on the most commonly used estimator, of the presence of measurement error. As was to be expected, measurement error renders this estimator inconsistent. However, with a slight adaptation, the estimator can be made consistent. This approach to consistent estimation is ad hoc and we next develop a systematic approach to consistent estimation. We show how to obtain the most efficient estimator from this class of consistent estimators. We illustrate our findings through an empirical example.