Identification of Linear Stochastic Models with Covariance Restrictions

摘要

Identification in systems of linear structural equations where disturbances are uncorrelated is treated. The class of decomposable covariance restrictions that give rise to relationships of exogeneity is defined. The wider class of recursively decomposable restrictions is introduced. If all the covariance restrictions are recursively decomposable, then any set of structural parameters that are identifiable are also globally or uniquely identifiable. Undecomposable covariance restrictions are considered. Such restrictions no longer afford an assurance of global identification. Nevertheless, for one model a simple criterion for discriminating amongst the isolated solutions of the identifying equations is shown.