Campaign allocations under probabilistic voting

We develop a probabilistic voting model where candidates compete by advertising in different media markets. Ads are viewed by everyone within a market and cannot be targeted to subgroups such as one candidate??s partisans. Candidates estimate the distribution of voter preference intensities in a market, and campaign ads then shift this distribution. Individuals with any intensity vote with some probability for each candidate. We derive comparative static implications of changes in a variety of factors on the advertising decisions of each candidate. Using campaign advertising data from 2002, we find these results to be consistent with actual campaign allocation behavior.