The Effect of Pledges on the Distribution of Lying Behavior: an Online Experiment

Reminding people to behave honestly or asking them to actively commit to honest behavior is an easily implementable intervention to reduce dishonesty. Earlier research has shown that such truth pledges affect lying behavior on a group level. In this study we are analyzing how a truth pledge changes the distribution of lying types which have been established in the literature, i.e. truth tellers, partial liars and extreme liars, to better understand whether truth pledges can affect the decision to lie or merely the extent of lies. For this purpose, we conduct a 2x2 experiment with 484 participants in which we apply a truth pledge in a gain and a loss frame. We introduce a novel "Even-Odd task'' for online lying experiments, which is based on the well-established coin-toss design. The Even-Odd task takes into account that unbiased, physical randomization devices are not always available in online settings, which can be a problem for truth-tellers if they are bad mental randomizers. We therefore ask participants to think of privately known numbers (house numbers, phone numbers) and then determine randomly whether even or uneven numbers result in the higher payment. We find that the truth pledge significantly reduces lying but also that this effect is strongest for extreme liars. The uneven shift in the distribution of liars suggests that truth pledges are effective in decreasing the size of lies but not the number of lies told. This result is robust for both frames.