Liberal parentalism

What normative constraints should bind parents (or policy makers) if they intervene in the choices of children (or constituencies) whose preferences evolve over time? For a sophisticated child who anticipates correctly his preference change, we prove that generically there exist parental interventions that are Pareto improving over the backward induction path that the child will follow on his own. If, in contrast, the child misperceives his future preferences, Pareto improving interventions might not exist, and even nudges might be painfully sobering. The parent may then choose to minimize the maximal disappointment along time that her benevolent intervention would cause.