A Population Model of Voter Registration and Deadwood
Charles Stewart III
Massachusetts Institute of Technology (MIT) - Department of Political Science
April 3, 2016
MIT Political Science Department Research Paper No. 2016-5
We use a model drawn from population ecology to propose a method to estimate the true number of registered voters in a state, independent of the "deadwood" that may be present on the registration rolls. The paper first presents a conceptual model that distinguishes the "true registration rate" from the official registration rate. A dynamic version of the model is specified that allows for population dynamics, that is, growth in registered voters due to birth, in-state migration, and the registration of the previously unregistered, and decline due to death, out-of-state migration, and the removal of previously registered voters from the rolls. The model is expressed as a Markov process, with registration and deadwood rates estimated for each state. These simulation results produce registration estimates that generally over-estimate true registration rates for most states. This demonstrates a need for further refinement of the model. The model's results are most strongly influenced by the "conditional registration rate" (that is, the rate at which unregistered eligible voters are added to the rolls at each time period), which suggests a priority on fine-tuning estimates of this rate in future iterations of the model.
Number of Pages in PDF File: 53
Date posted: April 5, 2016 ; Last revised: April 26, 2016