“…Models (1) and (2) bances, the two components of the forecast error are independent. Therefore, the asymptotic covariance matrix of the vector of forecast errors is simply the sum of the covariance matrices of the two components and can be written as where F = I, C3 z;', g = m + n being the number of observable variables, and T-';R. is the rg2 x rg2 asymptotic covariance matrix of restricted reduced-form coefficients, computed using the methods of Dhrymes (1973) or Goldberger, Nagar, and Odeh (1961). [Note that, in the framework of (3) and (9, the distinction between current endogenous and exogenous variables and between ex-ante and ex-post forecasting is no more meaningful.…”