Introduction
In Social Choice Theory, the performance of a voting mechanism is often evaluated through the lens of distortion. Distortion is defined as the ratio between the social cost of the election winner and the social cost of the "optimal" candidate—the one who minimizes the aggregate distance or cost for all voters.
Standard models in this field typically assume that every eligible agent casts a vote. Under this assumption, a well-established and somewhat pessimistic result exists: no deterministic voting rule can guarantee a distortion better than 3, even in a simple one-dimensional political spectrum. This paper argues that this theoretical bound is unrealistic because it ignores the pervasive reality of abstention. In practice, voters often stay home if they are indifferent (the candidates are too similar) or alienated (all candidates are too far from their views). This research explores whether modeling such natural behavior reveals a lower, more optimistic distortion value.
The Spatial Voting Model and Abstention
The authors utilize a Spatial Voting Model where voters and two candidates ($L$ and $R$) are represented as points in a metric space. To account for abstention, they introduce a probabilistic model based on the Weber-Fechner law. This law suggests that a human's ability to distinguish between two stimuli depends on their relative difference rather than their absolute difference.
In the context of an election, the probability $p_i$ that a voter $i$ will participate is determined by the ratio of the distances to the two candidates, modified by a constant parameter $\beta$:
- $\beta = 0$: This represents the classic model where everyone always votes.
- $\beta > 0$: As $\beta$ increases, voters become increasingly likely to abstain if they are indifferent between the choices or alienated from both.
Key Insights: Abstention as a Filter
The core finding of the paper is that abstention acts as a natural "filter" for the electorate. Highly motivated voters—those who perceive a significant relative difference between candidates—are the most likely to participate. Because these voters have a higher "stake" in the outcome, their preferences are prioritized by the mechanism. This naturally pulls the election winner closer to the utilitarian optimum.
By focusing on relative distance (the ratio of costs), the model simultaneously captures both indifference-based and alienation-based abstention. The authors employ a geometric technique called "valid displacements" to prove their bounds, shifting voter distributions within the metric space to identify configurations that produce the worst possible distortion.
Results and Distortion Bounds
The analytical results show a significant reduction in distortion compared to the classic model:
- Reduced Distortion: For a standard relative-distance model where $\beta=1$, the distortion of the expected winner is upper-bounded by 1.522, nearly half the classic bound of 3.
- The $\beta$ Sweet Spot: The relationship between $\beta$ and distortion is non-linear. The authors identified a minimum possible worst-case distortion of approximately $\sqrt{2} \approx 1.414$ when $\beta \approx 0.705$.
- Scalability: In large elections, the "Expected Winner" becomes statistically certain to win, and the distortion stabilizes at these lower bounds. In very small elections, the distortion may still approach 2, but it remains significantly better than the non-abstention baseline.
- Metric Invariance: Crucially, the bounds derived for a 1D line hold true for any arbitrary metric space, including high-dimensional ideological or preference spaces.
Conclusion and Future Work
This research demonstrates that accounting for realistic voter behavior significantly improves the theoretical outlook for collective decision-making. Abstention, often viewed as a failure of democracy, may actually serve to improve the social quality of election outcomes by weighting the preferences of those most affected by the choice.
Future work remains to be done in extending this model to multi-candidate elections, where strategic voting and more complex abstention dynamics (such as the "lesser of two evils" logic) come into play. Additionally, further investigation into "Expected Distortion" in small populations could provide deeper insights into local or committee-based decision-making.
Summary updated