Introduction
The classic Condorcet Jury Theorem posits that a group of voters, each possessing a noisy signal of the truth, can collectively identify the "correct" choice with high probability. However, this optimistic view relies on two fragile assumptions: that all voters share the same goal and that they vote honestly. In reality, voters often have conflicting interests and act strategically. While individual voters may lack the power to sway an election, they frequently form coalitions—facilitated by social media, political parties, or local networks—to influence outcomes.
Previous research has explored "strong equilibria" where coalitions of any size can form. However, this condition is often too restrictive for real-world scenarios where coordination is bounded by organizational capacity. This paper addresses a fundamental question: How much coordination can a minority group achieve before they successfully prevent a majority from reaching an informed decision?
Methods: Modeling Bounded Deviations
The authors employ a game-theoretic approach centered on a novel solution concept: the $ε$-ex-ante Bayesian $k$-strong equilibrium. This model introduces two critical refinements to standard voting games:
- Bounded Deviations: It only considers deviations by groups of size up to $k$, reflecting realistic limits on coordination.
- Ex-ante Perspective: Agents evaluate their strategies before receiving private signals. This approach mitigates "rare event" paradoxes common in voting games, where agents might otherwise condition their behavior on highly improbable ties.
The theoretical framework is grounded in a payoff structure where agents' preferences depend on both the state of the world (High or Low) and the final decision (Accept or Reject).
The core contribution is the derivation of a closed-form threshold curve, $ξ^*(α)$, where $α$ represents the fraction of majority agents and $ξ$ is the maximum fraction of agents allowed in a deviating coalition.
Key Insights: The Multi-Segmented Threshold Curve
The most significant finding is that the stability threshold $ξ^*(α)$ is not a simple linear function. Instead, it is a complex, multi-segmented curve that defines the system's "deviation capacity."
As shown in Figure 1, the curve can consist of up to four distinct segments:
- "Flat" Segment (Seg1): The threshold remains constant even as the majority size changes.
- "Steep" Segment (Seg2): The threshold declines linearly with the majority size.
- "Non-Linear" Segment (Seg3): A complex region where the relationship between majority power and coalition resistance is non-linear.
- "Tail" Segment (Seg4): A final linear stretch as the majority approaches a bare 51%.
Stability and Information Dynamics
The analysis reveals several counter-intuitive properties of voting stability:
- Stability at the Limit: Even when the majority has a paper-thin advantage (approaching 50%), the system remains stable against small, unorganized minority deviations. The threshold $ξ^*$ does not drop to zero.
- The Accuracy Paradox: As private information becomes more accurate, the range of majority sizes where a stable equilibrium exists actually contracts. However, the value of the threshold within that range increases. Essentially, better information makes the "informed" outcome easier to protect but harder to establish if the groups are too evenly matched.
- Minority Splitting: In certain equilibrium states near the 50% majority mark, minority agents maximize their disruptive power by splitting their votes between alternatives rather than voting as a monolith.
Key Results
The authors provide the exact mathematical boundaries for $ξ^*(α)$ in Theorem 2. If the coalition size remains below this curve, an equilibrium exists where the "correct" (informed majority) decision is reached with a probability approaching 1 as the population size $n$ grows.
Furthermore, the paper proves that for a stable equilibrium to exist, the majority must maintain a margin of expected votes large enough to absorb a deviation by any group of size $k$ in both possible states of the world. The successful characterization of the "Non-Linear" segment provides a closed-form solution for strategic behaviors that were previously uncharacterized in social choice theory.
Limitations and Future Work
While providing a robust theoretical foundation, the model has several limitations. The current model is limited to binary choices; extending this to plurality voting with multiple candidates would significantly increase strategic complexity. Additionally, the proof assumes independent and identically distributed (i.i.d.) signals, whereas real-world signals are often correlated. Finally, the model assumes any group of size $k$ can coordinate; future work could investigate deviations on social graphs where only specific sub-structures (like cliques) can coordinate.
By mapping the limits of strategic disruption, this research offers a deeper understanding of how democratic institutions can remain resilient in the face of organized minority interests.
Summary updated