A maximum likelihood approach to estimating dynamic models with unobserved independent variables

A maximum likelihood method to estimate the parameters of dynamic models containing unobserved independent variables is proposed. The approach is to maximize the likelihood of the residuals produced by the recursive Kalman filter equations applied to the model in state-space form. A simulation study is presented comparing the proposed method to the instrumental variable approach. An example using real data is given which estimates models of the Permanent-Income Hypothesis.