BACKWARD INDUCTION OR FORWARD REASONING? – AN EXPERIMENT OF STOCHASTIC ALTERNATING OFFER BARGAINING

Bounded rationality questions backward induction without necessarily excluding such reasoning when anticipation is easy. In our stochastic (alternating offer) bargaining experiment, there is a certain first-period pie and a known finite deadline. What is uncertain (except for the final period) is whether there is a further period. Whereas backward induction requires information about all later pie sizes and probabilities, forward reasoning is expected to rely on reasoning for which only the immediate prospects may be relevant. Rather than relying only on decision data, we assess forward reasoning by control of information retrieval before the first move. Participants begin with the shortest games before playing possibly longer games. Surprisingly they rely on forward reasoning already for the shortest game as revealed by their information retrieval and choice data.