Instrument Variable Methods for Studying Models of Economic Time Series

Abstract

This paper considers instrumental variables methods for studying linear models-of economic time series. The instrumental variables used in estimating parameters are assumed to be uncorrelated with the disturbance vector but are not restricted to be strictly exogenous. A greatest lower bound on the asymptotic covariance matrices of a family of instrumental variables estimators is derived and characterized. This bound is then compared to bounds derived previously in both the econometrics literature and the engineering literature.