Objective functions and the complexity of policy design (Q1097156)

Two methods are analyzed for constructing quadratic objective functions for linear and nonlinear macroeconomic decision models. Both methods are iterative and involve interactions with the decision maker. One method generates optimal solutions, with respect to the current objective function, at each iteration. Optimality at each iteration gives this method certain desirable characteristics. The second method generates feasible solutions at each iteration but can be ensured to converge in polynomial time by involving Khatchiyan's linear programming algorithm. Exploiting the similarity between both algorithms leads to a modification of the first algorithm that generates optimal solutions at each iteration and also converges in polynomial time.