Welfare-improving misreported polls

An often-heard criticism about electoral pollsters is that they might misreport pre-election poll results. We show that this can happen even in the absence of partisan motives, but purely for reputational ones. By underreporting the expected number of supporters of the most preferred candidate, the pollster is able to induce an election result more in line with its report. By doing so, not only victory chances of the most preferred candidate in society rise above 50%, but also total election costs are reduced, thus yielding welfare gains. Our model also allows for the accommodation of both the underdog effect (a feature of pivotal voting models) and the apparently inconsistent bandwagon effect, in the sense that the latter may be an illusion on the part of an observer who disregards the possibility of nontruthful polls. All of these results hold even as the electorate size grows without bound.