On the dynamic programming approach to incentive constraint problems