It’s My Party and I’ll Vote How I Want to: Experimental Evidence of Directional Voting in Two-Candidate Elections

The competing theories of proximity and directional voting have long been used to model voting behavior. Empirically evaluating these theories, however, requires knowledge of voters’ utilities, which are inherently unobserved. Empiricists have generally dealt with this by using self-reports of utility. Yet, self-reports are likely biased, leaving experimental predictions at odds with real-world election outcomes. We improve upon this method by constructing a discrete choice model which is able to measure the likelihood of any one voter exhibiting proximity voting behavior as opposed to directional voting behavior, without needing to know voters’ utilities. We subsequently conduct a voting experiment with over 1,800 participants to estimate the parameters of the model. Our results suggest that, among voters whose expected behavior differs across the two theories, there is an approximately even split between voting behaviors, and the probability of exhibiting proximity voting behavior decreases by roughly 10 percentage points for each step away from the midpoint of a (−10 to +10) political spectrum. Our results are robust across two measures of preferences and four candidate pairings. The outcomes of our experiment are also consistent with the results of the 2012 Presidential Election, which took place the day after our experiment closed.