Testing the single-peakedness of other-regarding preferences

We test an implication that is common to all prominent theories of outcome-based other-regarding preferences: the ordinal preference ranking of an agent over a finite number of alternatives lying on any straight line in the space of material payoffs to oneself and some other agent must be single-peaked or single-plateaued. An incentive compatible mechanism is used to elicit strict preference rankings of subjects over a finite set of alternatives in decision tasks based on dictator and trust games. Violations of single-peakedness range from a low of 19 to a high of 46 across the six main treatments. Other-regarding subjects violate single-peakedness significantly more often than self-regarding subjects. The presence of equal-split in the set of alternatives increases the number of violations. We argue that the violations cannot be explained by appealing to the relative ease of reporting certain rankings or as errors from the theoretical benchmark.