Knowledge that Transforms

Increasing virtuous behaviors, such as initiating healthy habits, is an important goal for policymakers and social scientists. To promote compliance with requests to perform virtuous behaviors, we study “returnable reciprocity.” Whereas traditional reciprocity involves giving people unreturnable unsolicited gifts to encourage compliance, returnable reciprocity involves offering opportunities to return the unsolicited gifts if they choose not to comply. Four studies (and two additional supplemental studies) show that returnable reciprocity (compared to traditional reciprocity) leads to higher enrollment in a hypothetical workplace wellness program (Study 1), as well as greater compliance in an incentive-compatible large-scale field experiment (Study 2) and conceptual lab replications (Studies 3 & S1). Returnable reciprocity may be more effective than traditional reciprocity because it induces increased feelings of guilt for non-compliance (Study 3). Though making an unsolicited gift returnable can be inexpensive, it appears to impose psychological costs that negatively affect the tactic’s overall impact on social welfare (Studies 4 & S2).

Selection biases present a fundamental challenge for research on ethics and misconduct. This issue is well understood at the individual level, where lab studies are often employed to sidestep it at the potential expense of external validity. However, much archival field data on ethics and misconduct are at risk of selection bias originating from within organizations, because organizations are typically responsible for evaluating and ultimately documenting who commits misconduct. In this paper I explore the nature and potential scope of this particular form of selection bias, its potential impact on the interpretation of extant findings from the literature, and how studying allegations may help detect it in specific contexts. Using detailed data on formal allegations of police misconduct in Chicago, I highlight how status characteristics such as race and gender may bias the creation of archival data. For example, black officers received allegations at rates similar to white officers but were more likely to have them sustained, and allegations made by black complainants were less likely to be sustained than those made by white complainants—even when including extensive sets of control variables. These findings indicate that accounting for allegations may be a fruitful methodological approach to better understand the optimal use of archival behavioral field data for research on ethics and misconduct.