After Felsenthal DS, Rapoport A, Maoz Z (1988) experimental research on Duverger’s Law and Strategic Voting has become commonplace, but the research lacks saliency on one particular metric: Voter Number. We test both of these hypotheses in an environments with “large” numbers of simulated voting participants starting from the standard 24 human subject election and going to the average number of voters in the US 2014 election- 230,000. To protect against the effects of priming the subject’s strategies with the votes of the machine, the behavior of the simulated voters (represented by the voting behavior’s Sincere, Strategic or Dominated) is learned from the participants.We then test the polls and votes of the subject voters in the large scale voting experiments against both the all human subject experiments and the theoretical benchmarks of Duverger’s Law and Strategic Voting. We find that while negligible pivot probabilities drastically reduce the effort and strategic behavior of the voters, there is still a robust relationship between their party preferences, polls and votes.
Sturges, Caleb, "Duverger’s Law and Strategic Voting in Large Scale Elections" (2014). Student Research Day Abstracts and Posters. 32.