Modeling the influence of polls on elections: a population dynamics approach

Abstract

We propose a population dynamics model for quantifying the effects of polling data on the outcome of multi-party elections decided by a majority-rule voting process. We divide the population into two groups: committed voters impervious to polling data, and susceptible voters whose decision to vote is influenced by data, depending on its reliability. This population-based approach to modeling the process sidesteps the problem of upscaling models based upon the choices made by individuals. We find releasing poll data is not advan- tageous to leading candidates, but it can be exploited by those closely trailing. The analysis identifies the particular type of voting impetus at play in different stages of an election and could help strategists optimize their influence on susceptible voters.