optimal model expected level probability equilibrium assume high section positive case incentive set higher equal number function lower problem