Endogenous discounting and the domain of the felicity function

The objective is to show that endogenous discounting models should use a felicity function constrained to a positive domain. A variety of articles use the Mangasarian or Arrow and Kurz condition as a sufficient condition for optimality, which restricts felicity to a negative domain. Since the level of the felicity function shows up in the optimal path it leads to qualitatively different solutions when one uses a negative or positive felicity function. We suggest reasons why the domain should be positive. We furthermore derive sufficiency conditions for concavity of a transformed Hamiltonian if the felicity function is assumed to be positive.