Reduced rank regression with autoregressive errors

In this paper we develop estimation procedures for the components of the regression coefficient matrix in a multivariate regression model, when this matrix is assumed to be of reduced rank and the usual assumption of serial independence of the errors is modified by considering a vector autoregressive structure for the errors. Asymptotic theory and iterative computational procedure for the estimator are discussed. Also suggested is an alternative estimator, which is useful for identification of the rank of the regression matrix. To illustrate the procedures, we consider some macroeconomic data of the United Kingdom.