Dishonesty and justifications: Evidence from the second roll of a dice game

The widely-adopted die rolling experiment measures average lying behaviour. Its original design uses so-called control rolls; subjects should roll twice before reporting their first roll, for which they are paid. The second roll is a control, which is neither paid nor reported. This detail has received little attention in economics, but is the basis for Justified Dishonesty. This popular psychological idea argues observing counterfactuals reduces the internal lying cost.
Specifically, it predicts subjects report the higher of their two rolls, switching relevant and irrelevant rolls if it pays to do so. Initial evidence appears compelling as data resemble its predictions. However current tests cannot distinguish between explanations, as we show other models can make virtually identical predictions without invoking counterfactuals. We test Justified Dishonesty’s mechanisms. First, we conduct a placebo test, finding that Justified Dishonesty’s predictions are accurate even when the proposed mechanism is not present. Second, we record both first and second (control) rolls. This enables a more direct test of the mechanism, which is strongly rejected in preregistered tests. Our results imply that whilst control rolls may slightly encourage cheating, they do so by altering standard lying costs rather than through a mechanism of switching rolls. This result underlines the importance of apparently inconsequential experimental features in influencing levels of lying behaviour.