Homophily, Peer Effects, and Dishonesty - Liza Charroin Bernard Fortin Marie Claire Villeval

In the EXO treatment, they were exogenously matched with two peers from the Baseline who selected either the Automatic or the Manual mode, and they were informed of the mode and the average performance of these peers at the beginning of each period. [...] The expressions λ3/(1 + λ3) and λ5δ/(1 + λ5) represent the effects of the average actual and reported performance of peers on the individual’s actual performance, in the case of peers choos- ing the Automatic or the Manual mode, respectively. [...] In 91% of the cases, individuals in ENDO observed an average reported performance of the pair that selected the Manual mode greater than the average actual performance of the pair that selected the Automatic mode. [...] Thus, a simple strategy is pooling the data of individuals who observed the performance of peers who chose the Manual mode and testing whether the significance and magnitude of the peers’ reported performance on the size of lies differ in EXO and ENDO. [...] Table A.4 in Appendix displays the marginal effects of random-effect (RE) Tobit regressions in which the dependent variable is the (latent) size of lies in each of the five periods of part 1.26 Although the estimators of the RE Tobit model may be subject to a selection bias, the regressions on the mode choice reported in Table A.3 in Appendix suggest that it is not likely to be the case, because n.