aInstitut für Statistik und Ökonometrie, Humboldt-Universität, Spandauer Str. 1, 10178 Berlin, Germany;bDB Research, Portfolio Strategy, Guiolettstr. 48, 60325 Frankfurt a.M., Germany
Abstract:
VARMA (vector autoregressive moving average) processes are proposed for modelling cointegrated variables. For this purpose the echelon form is combined with the error correction form. Procedures for estimating the Kronecker indices which characterize the echelon form and for specifying the cointegration rank are discussed. The asymptotic distribution of the coefficient estimators is given. An example based o n US macroeconomic data illustrates the procedure and demonstrates its feasibility in practice.