We test the effectiveness of two behaviorally inspired manipulations promoting forgiveness in a lab-in-the-field experiment in Cali, Colombia. Offenders (mostly juvenile) can only participate in a restorative justice program if victims agree to forgive them and let them participate. In our experiment, 756 participants were recruited using a panel maintained by a large public university. Participants are randomly assigned to one condition in a 2 × 2 between-subjects factorial design. In all conditions, restorative justice is introduced to participants using requests sent by real offenders to victims asking for their forgiveness. In one dimension, we manipulate the way the program is presented to them, using equivalent losses (risks of recidivism) or gains (successful rehabilitation). In the other dimension, we manipulate the default option (forgive or not-to-forgive), from which participants can easily opt-out. Decisions are incentivized by a sequence of lotteries. In all lotteries, a certain option is associated with the benefits of conventional justice (the opportunity cost of restorative justice) and a risky choice represents restorative justice, with large potential earnings (if the offender does not relapse into crime) and a chance of null earnings (if the offender does). Our results show that reluctance to forgive significantly decreases in the domain of losses relative to the domain of gains, while similar reluctance rates are observed in both default conditions. Disclosing objective information about the low recidivism rate of offenders also has a large impact, strongly reducing reluctance to forgive. Consistent with attention-based models, the magnitude of expected losses plays no significant role, and victims of crime are more willing to forgive than non-victims.
