How to Win a Large Election

How should a political campaign maximize its chance of winning an election when voters can abstain? We analyze when it is optimal for the campaign to maximize its expected margin of victory (EMV) over its opponent, i.e., the difference between its vote share of the pool of potential voters and its opponent’s share. If campaign decisions are decentralized to independent localities, EMV maximization is optimal when the number of voters is large and the localities assign positive probability to the election being potentially close. When a campaign is centralized, and is uncertain about the limit as population grows of its EMV, then it should maximize the probability that this limit is greater than zero. If the campaign can influence only a fraction of the population, this rule reduces to EMV maximization. The techniques we develop also permit calculation (for nonstationary populations of possibly abstaining voters) of the rate at which the probability of a tied election goes to 0 as population size increases.