The Measurement Error Problem in Dynamic Panel Data Analysis: Modeling and GMM Estimation

摘要: The Generalized Method of Moments (GMM) is discussed for handling the joint occurrence fixed effects and random measurement errors in an autoregressive panel data model. Finite memory disturbances, latent regressors assumed. Two specializations GMM are considered: (i) using instruments (IVs) levels a differenced version equation, (ii) IVs differences equation levels. Index sets lags convenient examining how potential IV set, satisfying orthogonality rank conditions, changes when pattern changes. with long may sometimes give IV-set too small to make estimation possible. On other hand, problems ‘IV proliferation’ ‘weak IVs’ arise unless time-series length small. An application based on (log-transformed) capital stock output from Norwegian manufacturing firms discussed. sample biases quality illustrated by Monte Carlo simulations. Overall, respect bias strength, inference level seems superior differences.