Do two bribe less than one? – An experimental study on the four-eyes-principle

ABSTRACT

We incorporate a four-eyes-mechanism on the briber’s side into a bribery game. Our results are mixed. We find no effect of the mechanism in a one-shot setting, but a reduction of bribes when the setting is repeated.

KEYWORDS:

• Corruption
• four-eyes-principle
• experiment

JEL CLASSIFICATION:

• C91
• C92
• D73
• M14

Acknowledgments

This research is substantially based on Anja Bodenschatz’s master thesis (Bodenschatz 2016). Financial support through DFG-FOR 1371 ‘‘Design and Behavior’’ and C-SEB is gratefully acknowledged. Conflicts of interest: none.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ For an overview of group decision-making see Charness and Sutter (2012).

² Siemens, for example, emphasized the four-eyes-principle in internal guidelines after their scandal on paying bribes to win orders (Löscher 2012; Siemens 2017).

³ Instructions were in neutral language. The bribery game is designed to be perceived as a corruptive setting, e.g. because, different to a trust game, the corruptive act entails negative externalities on others and a probability of detection. Abbink and Hennig-Schmidt (2006) found loaded framing of the bribery game to yield results similar to those with neutral framing.

⁴ The low detection probability of 0.003, paired with the high potential penalty of an exclusion from the experiment, was chosen in the original bribery game by Abbink et al. (2002). This ensures that only very risk-averse subjects would refrain from accepting the bribe, because of their fear of being disqualified. In their study, however, punishment with this low detection probability significantly reduced corruption.

⁵ Participants were only informed of negative externalities they incurred by actions of other groups at the end of the experiment, to keep observations independent.

⁶ In IC one money transfer was detected in period two of the repeated game. We exclude this group from data analysis beyond the one shot game and the first period.

⁷ All reported tests are two-sided.

Additional information

Funding

This work was supported by the German Research Foundation (DFG) through the Research Unit “Design and Behavior” (FOR 1371) and through the Center of Social and Economic Behavior (C-SEB) at the University of Cologne (C-SEB Student Grant)