Why Biased Endorsements Can Manipulate Elections

We study the electoral impact of biased endorsements in large Poisson elections with two candidates. Under fairly general conditions, we derive analytical approximations for the asymptotic voting equilibria. We show that, when the electorate is sufficiently polarized, manipulating public information about candidates' qualities can hardly affect the expected party vote shares from an ex-ante perspective. Nevertheless, such information manipulation can systematically affect the election outcome and voter turnout. We apply our framework to study the influence of the endorser's bias and the entry of new endorsers on the extent of electoral manipulation. With a single endorser, a rise in bias affects the election outcome in a non-monotonic way and reduces voter welfare. Importantly and surprisingly, manipulating election outcomes becomes easier, the larger is the electorate. Entry of a new endorser can systematically affect election outcomes and improve voter welfare in large elections if and only if the entrant is strictly less biased than all incumbents. Moreover, we show that such entry can either systematically increase or decrease voter turnout, depending on the biases of entrant and incumbent endorsers.