Hypothesis about Lying
Hypothesis.
Conditional on preferring the outcome from lying over the outcome from truth-telling, a person is sensitive to neither her own [monetary] gain from lying, nor how much [monetary] harm she causes the other side.
Notice that aside from the preface of “preferring the outcome from lying”, the rest of the Hypothesis seems quite at odds with Gneezy’s Result. The reconciliation is that a significant fraction of Gneezy’s subjects prefer the outcome from lying to truth-telling in some experimental treatment(s), but not in others. But this by itself is merely evidence of social preferences, which as noted earlier, is a well-established phenomenon. It has nothing to do inherently with lying aversion, since it applies just as well when the allocations are chosen directly in the dictator game rather than through communication in the deception game. This highlights our analytical point: it is important to condition on preferences over allocations when interpreting lying behavior. Indeed, we will show that Gneezy’s data is consistent with an even stronger version of the Hypothesis, viz., one cannot reject that 50 percent of people lie whenever they prefer the outcome from lying versus truth-telling, and 50 percent of people never lie.
That the Hypothesis is consistent with Gneezy’s data does not imply that it is an accurate description of people’s behavior. It is an important hypothesis to test because if it is right, it means that people can be categorized as one of two types: either they are “ethical” and never lie, or they are “economic” and lie whenever they prefer the allocation obtained by lying. If it is wrong, then a richer model of aversion to lying is needed.1
Accordingly, we ran experiments broadly similar to Gneezy’s with the primary objective of testing the Hypothesis. Following Gneezy’s design, we ran treatments of both the dictator game and deception game, but used a within-subject design so that players played both games (unlike Gneezy, where players
Conditional on preferring the outcome from lying over the outcome from truth-telling, a person is sensitive to neither her own [monetary] gain from lying, nor how much [monetary] harm she causes the other side.
Notice that aside from the preface of “preferring the outcome from lying”, the rest of the Hypothesis seems quite at odds with Gneezy’s Result. The reconciliation is that a significant fraction of Gneezy’s subjects prefer the outcome from lying to truth-telling in some experimental treatment(s), but not in others. But this by itself is merely evidence of social preferences, which as noted earlier, is a well-established phenomenon. It has nothing to do inherently with lying aversion, since it applies just as well when the allocations are chosen directly in the dictator game rather than through communication in the deception game. This highlights our analytical point: it is important to condition on preferences over allocations when interpreting lying behavior. Indeed, we will show that Gneezy’s data is consistent with an even stronger version of the Hypothesis, viz., one cannot reject that 50 percent of people lie whenever they prefer the outcome from lying versus truth-telling, and 50 percent of people never lie.
That the Hypothesis is consistent with Gneezy’s data does not imply that it is an accurate description of people’s behavior. It is an important hypothesis to test because if it is right, it means that people can be categorized as one of two types: either they are “ethical” and never lie, or they are “economic” and lie whenever they prefer the allocation obtained by lying. If it is wrong, then a richer model of aversion to lying is needed.1
Accordingly, we ran experiments broadly similar to Gneezy’s with the primary objective of testing the Hypothesis. Following Gneezy’s design, we ran treatments of both the dictator game and deception game, but used a within-subject design so that players played both games (unlike Gneezy, where players