Consensus vs. Conformity in Mixed-Motive Games

A correlation between second-order beliefs and strategies – in social dilemmas – has been interpreted as evidence of guilt aversion (Charness and Dufwenberg [2006]). Ellingsen et al. [2010] hypothesize that such correlation might rather be due to consensus effects. Here we propose an additional explanation, conformity, which involves a similar belief-behavior correlation and we set out to tell these motivations apart by proposing a design such that: (i) we reduce the scope for guilt aversion by eliciting and transmitting beliefs about the behavior of other participants in the same role; (ii) we disentangle consensus from conformity by providing an exogenous variation in collective beliefs. The data show that consensus is present (and predominant) but is not the only force driving the belief-behavior correlation. In fact, we also observe “self-servingly conformist” behavior in that subjects choose to match their strategy to the transmitted information when it is in their interest to do so.