The restricted reduced form equation system derived from a simultaneous structural model is the system used for forecasting and of immediate concern to policy decisionmakers and forecasters. For verifying the significance of restricted reduced form coefficients and for measuring the confidence of forecasts, the covariance matrices of these coefficients are required. However, the computation of the covariance matrix of restricted reduced form coefficients involves a heavy workload for a large size model, and the computer software packages for this computation are usually not available. This paper formulates an operational procedure which can be easily implemented by applied econometricians for computing such a matrix. In particular, the procedure, which computes the covariance matrix in a sequential block by block fashion, is computationajefficient and suitable for programming in computers without larger core capacity. In fact, the limitation of core capacity is a critical factor to consider for programming with the increasing/ use of personal computers. Besides, as discussed later, the estimation procedure could be more accurate than that of the pioneer work developed by Goldberger, et al.(1961).