Treffer: ON THE CONTROL OF STRUCTURAL MODELS--COMMENT.

The article presents comments of the author on the paper "On the Control of Structural Models," by Alfred Norman. In this paper, Norman has. presented an alternative derivation of an approximately optimal control law of the author for linear structural models to minimize the expectation of a quadratic loss function, and compared the computational efficiencies of the two procedures. The author comments on the nature of the alternative derivation, the relative computational efficiency, and the potential usefulness of the control law. The derivation of the control law starting from Norman's equation employs the standard method of dynamic programming and is basically the same as the derivation of the author. The main difference lies in the writing of the covariance matrix of the vector of state variables which satisfy the linear structural equations. First, how feasible is Norman's algorithm when the number of unknown parameters is large, say over 300, as in the Federal Reserve Board econometric model. Second, how does one deal with a nonlinear structural model when Norman's paper is concerned only with a linear model. A partial answer to this question can be found in the author's where the control of a nonlinear structural model with uncertain parameters is studied.