Abstract
Strategic voting in large-scale elections often defies classical game-theoretic predictions due to the "voter's paradox," where individual influence is mathematically negligible. This paper establishes a robust foundation for strategic behavior under strict uncertainty using a non-atomic model. By introducing the "local dominance" heuristic—where voters avoid candidates perceived as unlikely to win based on imprecise poll data—we prove that voting systems under plurality rules consistently converge to stable equilibria. This model incorporates psychological principles like Weber’s Law to better reflect human perception of numerical uncertainty, providing a formal mechanism for observed phenomena such as Duverger’s Law.
Introduction
In classical plurality voting, where the candidate with the most votes wins, predicting strategic behavior is notoriously difficult. In large populations, a single voter's impact is negligible, leading to a paradox where almost any voting configuration can be considered a Nash equilibrium—even everyone voting for their least favorite candidate. To bridge the gap between theory and reality, researchers have turned to models of uncertainty. If voters are unsure of the exact standings, they may strategically shift their support to a "lesser of two evils" to avoid a worse outcome.
This research generalizes previous models to a non-atomic setting, treating the electorate as a continuous distribution. The goal is to prove that stable outcomes exist even when voters have varying levels of confidence and move in large groups simultaneously, reflecting real-world conditions more accurately than previous individual-based models.
Methodology: Modeling Uncertainty and Heuristics
The paper employs a formal game-theoretic analysis focused on how voters perceive and react to election data.
Distance-Based Uncertainty
Voters do not perceive poll results as exact points but as ranges of possible outcomes. The model utilizes a multiplicative distance metric (e.g., a score might be perceived as ±10%). This aligns with Weber’s Law, suggesting that humans are more sensitive to small differences at lower magnitudes than at higher ones; for instance, the difference between 5% and 6% is perceived more clearly than the difference between 45% and 46%.
Behavioral Heuristics
Two primary heuristics are analyzed:
- Local Dominance (LD): A behavioral heuristic where a voter rejects candidate A in favor of B if B is "at least as good" in all possible outcomes the voter considers likely, and strictly better in at least one.
- Worst-Case Regret (WCR): An alternative where voters choose the candidate that minimizes their maximum possible "regret" (the utility difference between their choice and the best possible choice they could have made).
The study examines iterative dynamics, where groups of voters repeatedly update their votes based on the latest poll until the system reaches a steady state.
Key Insights
Robustness of Convergence
A central finding is the inherent stability of local dominance. Regardless of whether voters have different confidence levels or move in large, simultaneous blocks, the system is mathematically guaranteed to reach an equilibrium. This robustness suggests that LD is a highly reliable predictor of stable voting patterns even under "messy" conditions.
Explaining Duverger’s Law
The model provides a formal justification for Duverger’s Law, which states that plurality systems tend toward two-party dominance. As voters observe their preferred candidate trailing, the uncertainty intervals lead them to "locally dominate" that candidate in favor of a more viable front-runner, naturally consolidating the field toward the top two contenders.
Psychology and Regret
By comparing LD and WCR, the paper shows they often yield identical results in simple cases. However, in complex scenarios, WCR can be "too aggressive," potentially leading to infinite cycles where no equilibrium is ever reached. LD, being more conservative, ensures convergence.
Results
The theoretical analysis yields several key theorems:
- Theorem 5 (Convergence): Any sequence of strategic moves based on weak local dominance is finite, ensuring the process settles into an equilibrium rather than looping indefinitely.
- Equilibrium Existence: A voting equilibrium is proven to exist even in the most generalized versions of the non-atomic model.
- Structured Dynamics: When starting from a "truthful" state (voting for favorites), the winner's score is non-decreasing over time, and voters tend to move toward compromise candidates.
Conclusion
While this paper provides a rigorous foundation for plurality voting under uncertainty, several avenues for future work remain. The current proofs are specific to plurality rules; extending these to more complex systems like Ranked Choice or Borda Count presents significant mathematical challenges due to the exponential growth of the action space. Furthermore, while the model is psychologically grounded, empirical validation through experimental lab studies or historical poll analysis is necessary to confirm that "local dominance" accurately captures real-world human strategy at the ballot box.
Summary updated