Selecting Winners with Partially Honest Jurors

Updated: February 2012. Available on request.

Abstract: We consider the effect of \partially honest" jurors (along the lines of Dutta and Sen (2011)) in the contestant model of Amoros (2010). We analyze the problem of choosing the w contestants who will win a competition within a group of n > w competitors when all the jurors commonly observe who the w best contestants are, but they may be biased and some of these jurors are partially honest. We use two notions of honesty, one of which is linked to the unbiasedness of the juror. The other stronger notion, requires a partially honest individual to have a strict preference for revealing the true state over lying when truth-telling does not lead to a worse outcome (according to preferences in the true state) than that which obtains when lying. We first look at the many person implementation, when the jury consists of at least one partially honest juror, whose identity is not known to the planner. We find that the socially optimal rule is Nash implementable if for all beliefs of the planner (regarding the unbiasedness of the juror) and for each pair of contestants, there are two jurors who treat the pair in an unbiased manner and one of these jurors is partially honest. We also analyze the problem, when there are only two jurors and consider cases both with and without partial honesty.