Limited and full information estimation of the rational expectations demand for the money model: application to Finnish M1

We compare parameter estimates of the intertemporal money-in-the-utility-function model estimated using the Generalized Method of Moments and the Full Information Maximum Likelihood method.The process driving the forcing variables is approximated with vector autoregression.The FIML estimates of the deep parameters are reasonable, although some of them differ from the corresponding GMM estimates.The simulation experiments suggest that the differences are not very big in practice and that they are connected with adjustment costs.The cross-equation restrictions are clearly rejected, as is typical for these kinds of models; exogeneity restrictions are rejected as well.