Point and Set Identification in Linear Panel Data Models with Measurement Error

The rich dependency structure of panel data can be exploited to generate moment conditions that can be used to identify linear regression models in the presence of measurement error. This paper adds to a small body of literature on this topic by showing how heteroskedasticity and nonlinear relationships between the error-ridden regressors and error-free regressors lead to identifying moment conditions in a static panel data setting, how suitably chosen linear combinations of lagged and lead values of the dependent variable can be used as instrumental variables in a dynamic panel data with measurement errors, and how reverse regression can be generalized to the panel data setting, thereby giving bounds on regression coefficients in the absence of point identification