Bloc Voting on Single Peaked Preferences

We analyze the winning coalitions that arise under Bloc voting when voters preferences are single-peaked. For small numbers of candidates and numbers of winners, we determine conditions under which candidates in winning coalitions are adjacent. We also analyze the results of pairwise contests between winning and losing candidates and assess when the winning coalitions satisfy several proposed extensions of the Condorcet criterion to multiwinner voting methods. Finally, we use Monte Carlo simulations to investigate how frequently these coalitions arise under different assumptions about voter behavior.

💡 Research Summary

This paper investigates the composition of winning committees under Bloc voting when voters’ preferences are single‑peaked. Bloc voting is a simple multi‑winner rule in which each voter casts a vote for their top k candidates, and the k candidates receiving the most votes become the winners. The authors focus on settings with a small number of candidates (m) and a small number of winners (k), and they explore how the structure of the winning set relates to several extensions of the Condorcet criterion that have been proposed for multi‑winner elections.

The authors begin by formalizing the voting model. Voters submit strict linear rankings of all m candidates; the number of voters N is assumed odd to avoid ties. Under Bloc voting each voter gives one point to each of their top k candidates, and the k candidates with the highest totals are elected. For comparison they introduce Copland’s method, which awards a point to a candidate for every head‑to‑head contest it wins (½ point for a tie). Copland scores are exactly the Copeland scores and, in the single‑peaked setting, they coincide with the order of candidates along the underlying left‑right spectrum.

Single‑peaked preferences impose a linear ordering of candidates on a one‑dimensional policy line. Every voter’s ranking rises to a single “peak” and then falls monotonically on either side. This restriction guarantees the existence of a Condorcet winner (the median candidate) by the Median Voter Theorem. Consequently, the pairwise majority relation is transitive, and the Copland ranking of candidates is simply C₁ ≻ C₂ ≻ … ≻ C_m, where C₁ is the Condorcet winner.

Three notions of Condorcet‑type fairness for committees are examined:

1. Gehrlein‑stable sets – a set W is Gehrlein‑stable if every candidate in W defeats every candidate outside W in a head‑to‑head contest.
2. Condorcet sets (Condorcet‑sets) – a set W is a Condorcet‑set if for each candidate outside W there exists at least one candidate inside W that beats it.
3. Locally stable sets – for a quota q, a set W is locally stable if no group of at least q voters unanimously prefers some non‑member candidate to every member of W.

Under single‑peaked preferences the authors prove two key propositions. First, a set is a Condorcet‑set if and only if it contains the Condorcet winner. Second, a Gehrlein‑stable set of size k must consist of the top k candidates in the Copland ranking, and therefore its members are necessarily adjacent on the policy line. The adjacency result follows from the fact that any non‑adjacent pair would admit an intermediate candidate that defeats the more extreme one, contradicting the assumed ordering.

The paper then conducts a case‑by‑case analysis for m = 4 and m = 5. All possible winning committees are enumerated, and their status with respect to the three fairness notions is recorded. The authors illustrate the “center‑squeeze” phenomenon: a centrist candidate may lose to two moderately extreme candidates that flank it, even though the centrist is the Condorcet winner. This can happen under Bloc voting but not under Copland’s method, which always selects the top k candidates from the Condorcet ranking and thus yields an adjacent, Gehrlein‑stable committee.

For larger numbers of candidates (six and seven) the authors develop general lemmas that extend the adjacency argument and show how the probability of non‑adjacent winning sets grows with k and with the degree of polarization among voters. They prove that any committee produced by Copland’s method is always adjacent and Gehrlein‑stable, establishing a clear contrast with standard Bloc outcomes.

The empirical component uses Monte Carlo simulations to estimate how often Bloc committees satisfy the three fairness criteria under three voter‑behavior models:

• Uniform random single‑peaked model – candidate orderings are drawn uniformly at random, then each voter’s ranking is generated consistent with the single‑peaked constraint.
• Gaussian‑center model – voters’ ideal points follow a normal distribution centered on the median candidate, producing a concentration of moderate voters.
• Polarized bipartite model – the electorate is split into two large groups with opposite extreme ideal points, creating strong left‑right polarization.

For each model the authors simulate 100 000 elections with varying m (5–9) and k (2–4). Results show that in the uniform and Gaussian models the overwhelming majority of Bloc committees are both Gehrlein‑stable and Condorcet‑sets; the adjacency property holds in over 95 % of runs. In the polarized model, however, the frequency of adjacent committees drops sharply as k increases, and the proportion of committees that are locally stable (with quota q = ⌊N/2⌋ + 1) falls below 30 % for k = 4. The simulations confirm the theoretical prediction that polarization amplifies the center‑squeeze effect, leading to non‑adjacent, non‑Gehrlein‑stable outcomes.

In the concluding discussion the authors argue that while Bloc voting can be reasonably fair under single‑peaked preferences, its susceptibility to non‑adjacent, non‑Gehrlein‑stable committees in polarized societies calls for caution. They suggest that election designers who value Condorcet‑type fairness might prefer Copland’s method or other Condorcet‑consistent multi‑winner rules, especially when the number of winners is large relative to the candidate pool. The paper also outlines avenues for future research, including extensions to multi‑dimensional policy spaces, the impact of truncated ballots (where voters rank only a subset of candidates), and strategic voting behavior under Bloc.