Lying and mistrust in the continuous deception game

I present a novel experimental design to measure lying and mistrust as continuous variables on an individual level. My experiment is a sender-receiver game framed as an investment game. It features two players: firstly, an advisor with complete information (i.e., the sender) who is incentivized to lie about the true value of an optimal investment and, secondly, an investor with incomplete information (i.e., the receiver) who is incentivized to invest optimally and therefore must rely on the alleged optimum reported by the advisor. The extents of lying and mistrust are both measured on continuous scales. This allows observing more differentiated behavior and therefore enables testing of more sophisticated theoretical predictions. I find that the senders lie by overstating the true value of the optimum to an average extent of about 148%, while the receivers suspect them to do so by only 56%. The senders seldomly lie to the fullest possible extent as they correctly expect the receivers to disproportionally mistrust lies of such a high extent. This indicates that people make strategic considerations about their potential to manipulate others when lying. In line with this, I discover that lying and mistrusting behavior can be predicted by first-order beliefs about the other player. Consistent with previous studies, my findings support the conjecture that lying costs increase with the extent of lying. In addition, I provide evidence for some endogenous preference for trust. Both players' behaviors and beliefs are consistent over time. Moreover, my ex ante classification of both players' strategy sets is consistent with their ex post self-assessment of their own behavior within the experiment.