Estimating Cointegrated Systems Using Subspace Algorithms

In this paper the properties of so called subspace methods in the context of cointegrated processes of order one are investigated. It is shown that the algorithms deliver consistent estimates of the transfer function in the case of general VARMA models and under mild conditions on the underlying noise process. A procedure for the estimation of the dimension of the cointegrating space is presented and consistency for this procedure is proven. Also the estimation of the order of the system is discussed. Simulation examples demon- strate the usefulness of the subspace algorithms for the estimation of cointegrated systems.