A Newton-type min-max algorithm for rival models of the same system

The existence of rival models of the same macroeconomic system is well known. In this paper, the basic policy optimization problem is extended to take account of rival models. A min-max problem is introduced and subsequently reformulated as that of the minimization of the "worst" convex combination of the objective functions of the rival models. This yields a saddle-point problem. A Newton-type algorithm is proposed for its solution. The algorithm requires the solution of a quadratic subproblem. The proposed range-space based quadratic programming algorithm for the latter utilizes the fact that the models, and thereby constraints, are far fewer than the number of variables.