An augmented first-order approach for incentive problems

Incentive constraints are constraints that are optimization problems themselves. If these problems are non convex then the first order approach fails. We propose an alternative solution method where we use the value function as an additional constraint. This ensures that all solutions are incentive compatible. To get the value function we use a function interpolator like sparse grids. We demonstrate our approach by solving two examples from the literature were it was shown that the first order approach fails.