Trimming extreme reports in preference aggregation

Abstract

Eliminating – or trimming – extreme reports before aggregating them is usually motivated by the perception that it constitutes a remedy for strategic misreporting. This work focuses on the strategic calculus of voting when using trimmed-mean mechanisms for preference aggregation. Contrary to the above presumption, we show, both formally and experimentally that, under such mechanisms, voters persistently resort to strategic polarization for all but the most extreme levels of trimming. Furthermore, we show that while trimming promotes efficiency (i.e. the average payoff increases), it also leads to more inequality (i.e. the minimum payoff decreases). While experimental results provide some support for the idea that trimming can mitigate strategic misreporting, the extent of this non-equilibrium behavior is not enough to neutralize trimming's predicted effect on efficiency and inequality.