Covariate measurement and endogeneity

The effects of improving covariate measurement are investigated when the covariate is endogenous even in the absence of measurement error. Reducing measurement error can exacerbate the finite sample bias of Two-Stage Least Squares. An application reveals this is of practical importance.