Experimental Design
To investigate these questions, we design an incentivized experiment which we conduct on the platform Prolific with subjects from the UK that currently work in an organization, thus are real employees. In the experiment, two randomly matched subjects form a team of two and collaborate on a real-effort task to receive income. After two rounds of the task, we introduce a situation in which subjects have to decide whether to commit a wrongdoing and whether to react by whistleblowing. Specifically, one team member (Player A) can take away money earmarked for a charity. The other team member (Player B) can react to Player A’s (possible) embezzlement by either remaining silent about it or by blowing the whistle. If Player B remains silent, Player A increases their payoff by the amount withheld from the charity, while Player B’s earnings stay untouched. In the case of whistleblowing, the charity gets the full donation, but both Players A and B get their payoffs reduced (because of punishment and whistleblowing costs).
Before subjects take their decisions, we elicit participants personal normative beliefs and normative expectations regarding both behavioural options they have (whistleblowing and remaining silent).
Lastly, we include the possibility of social sanctions as subjects can be excluded from an additional team task due to their action in the wrongdoing/whistleblowing situation in a second part of the experiment.
In a Baseline Treatment, the experiment is conducted as explained above. Then, we have three treatments in which we implement different social norm interventions. The information communicated in these social norm interventions is based on the elicitation of personal normative beliefs in the Baseline Treatment. Participants receive the information before they take the decision whether to blow the whistle or to remain silent.
Particularly, in Treatment 1 “Whistleblowing Norm”, we present participants the information that the majority of other participants in the same experiment personally considered whistleblowing to be appropriate in the situation (based on previous research we expect to find this in the Baseline Treatment).
In Treatment 2 “Silent Norm”, we present participants the information that the majority of other participants in the same experiment personally considered remaining silent to be inappropriate in the situation (again, based on previous research we expect to find this in the Baseline Treatment).
In Treatment 3 “Whistleblowing & Silent Norm”, we present participants the information that the majority of other participants in the same experiment personally considered whistleblowing to be appropriate and remaining silent to be inappropriate. Thus, we provide participants with both information used in Treatments 1 and 2.
We will conduct the experiment with 100 participants per treatment, of which 50 will be in the role of Player B (i.e., in the role of potential whistleblowers). Every participant will only be able to participate in one of the treatments.
For answering Hypotheses 4 and 5 clearly, we want to make sure that the obtained findings are not the result of potential methodological issues. Hence, due to robustness reasons, we will conduct another treatment where the norm elicitation (appropriateness of whistleblowing and remaining silent) takes place after and not before the decision.
We expect that the timing of the norm elicitation does not make a difference in behaviour. We will hereby use a Chi2-Test to test whether the share of whistleblowers is higher in our Baseline Treatment than in our new Treatment 4 “Norm Elicitation after Decision”.
If there are no statistically significant differences, we can conclude that the obtained results are not affected by the timing of the norm elicitation so that the initial findings are valid for answering Hypotheses 4 and 5. However, if it turns out that the timing makes a difference, the initial findings for answering Hypotheses 4 and 5 need to be re-assessed. We will then plan to re-conduct the according treatments in the new decision environment to likewise clarify whether the hypothesized relationships depend on the timing of the norm elicitation or not.