Trust, reciprocity and altruism: An impossible addition

This paper attempts to measure conditional and unconditional other-regarding preferences in two versions of an investment game: a canonical one and a cheap-talk variant where some pre-play is also allowed (i.e., non-binding unilateral messages). We find that counter-factual measures, as the well-known triadic design (Cox, 2004 [G&EB]) may systematically fail in distinguishing between conditional and unconditional other-regarding preferences due to the existence of frame effects. Specifically, by using indirect methods, we document conditional other-regarding preferences that are systematically neglected by the triadic approach. By inspecting result from the cheap-talk variant of the game, we also find that messages have no effect on average, but they affect the participants’ behavior leading to a polarization of their choices.