Limited backward induction: foresight and behavior in sequential games

The paper tests experimentally for limited foresight in sequential games. We develop a general out-of-equilibrium framework of strategic thinking based on limited foresight. It assumes the players take decisions focusing on close-by nodes, following backward induction – what we call limited backward induction (LBI). The main prediction of the model is tested in the context of a modified Game of 21. In line with the theoretical hypotheses, our results show most players think strategically only on close-by nodes without reasoning backwards from the end of the game. A small fraction of subjects play close to equilibrium, while few others try to exploit the limited foresight of their opponent. The results provide strong support for LBI, and cannot be accounted for using the most popular models of strategic thinking, let alone equilibrium analysis.