Backward Induction Reasoning beyond Backward Induction

Backward induction (BI) is only defined for perfect information games, but its logic is also invoked in many concepts for imperfect or incomplete information games. Yet, the meaning of BI reasoning is not clear in these settings, and we lack a way to capture the essence of BI without assuming equilibrium. We introduce backward rationalizability, a nonequilibrium solution concept for incomplete information games, which we argue distills the logic of BI reasoning. We show several of its properties and discuss a few applications, including a new version of Lipnowski and Sadler’s (2019) peer-confirming equilibrium. (JEL C72, C73, D83)