Table of Contents
Fetching ...

On likelihood of a Condorcet winner for uniformly random and independent voter preferences

Boris Pittel

TL;DR

The paper addresses the probability of a Condorcet winner under impartial culture as both the number of voters $m$ and candidates $n$ grow. It builds on prior work that analyzed fixed $m$ or fixed $n$ regimes and introduces a representation of random preferences via an $m\times n$ matrix of i.i.d. $[0,1]$ uniforms, enabling a combinatorial-analytic treatment of pairwise contests. Central to the method is Esseen’s inequality applied to sums of independent but non-identically distributed Bernoulli indicators, yielding precise control over the probability that a given candidate defeats all others in pairwise contests. The main results show that, uniformly in $n$, the Condorcet-winner probability $Q_{m,n}$ satisfies $Q_{m,n}=O(\exp(-n^{\varepsilon(n)})+n^2/m^{1/2})$ when $m\gg n^4$, and that for fixed $n$, $Q_n=O(n^{-\ extell})$ for any $\ell>0$. Consequently, with a very large electorate relative to the candidate pool, the probability of a Condorcet winner vanishes, strengthening prior bounds and providing a nuanced view of how $Q_{m,n}$ decays in the large-$m$, large-$n$ regime.

Abstract

We study a mathematical model of voting contest with $m$ voters and $n$ candidates, with each voter ranking the candidates in order of preference, without ties. A Condorcet winner is a candidate who gets more than $m/2$ votes in pairwise contest with every other candidate. An ``impartial culture'' setting is the case when each voter chooses his/her candidate preference list uniformly at random from all $n!$ preferences, and does it independently of all other voters. For impartial culture case, Robert May and Lisa Sauermann showed that when $m=2k-1$ is fixed ($k=2$ and $k>2$ respectively), and $n$ grows indefinitely, the probability of a Condorcet winner is small, of order $n^{-(k-1)/k}$. We show if $m, n\to\infty$ and $m\gg n^4$, then for each fixed $\ell$ the probability of a Condercet winner is at most of order $n^{-\ell} + n^2/m^{1/2}$, thus converges to zero.

On likelihood of a Condorcet winner for uniformly random and independent voter preferences

TL;DR

The paper addresses the probability of a Condorcet winner under impartial culture as both the number of voters and candidates grow. It builds on prior work that analyzed fixed or fixed regimes and introduces a representation of random preferences via an matrix of i.i.d. uniforms, enabling a combinatorial-analytic treatment of pairwise contests. Central to the method is Esseen’s inequality applied to sums of independent but non-identically distributed Bernoulli indicators, yielding precise control over the probability that a given candidate defeats all others in pairwise contests. The main results show that, uniformly in , the Condorcet-winner probability satisfies when , and that for fixed , for any . Consequently, with a very large electorate relative to the candidate pool, the probability of a Condorcet winner vanishes, strengthening prior bounds and providing a nuanced view of how decays in the large-, large- regime.

Abstract

We study a mathematical model of voting contest with voters and candidates, with each voter ranking the candidates in order of preference, without ties. A Condorcet winner is a candidate who gets more than votes in pairwise contest with every other candidate. An ``impartial culture'' setting is the case when each voter chooses his/her candidate preference list uniformly at random from all preferences, and does it independently of all other voters. For impartial culture case, Robert May and Lisa Sauermann showed that when is fixed ( and respectively), and grows indefinitely, the probability of a Condorcet winner is small, of order . We show if and , then for each fixed the probability of a Condercet winner is at most of order , thus converges to zero.

Paper Structure

This paper contains 2 sections, 4 theorems, 44 equations.

Table of Contents

  1. Condorcet winner
  2. Proofs

Key Result

Theorem 1.1

For $Q_n$ defined in 0.9: (a) If $m\gg n^4$, then $Q_{m,n}=Q_n+O(n^2m^{-1/2})$. (b) If $n\to\infty$, then, for every fixed$\ell> 0$, $Q_n =O(n^{-\ell})$; in words, $Q_n$ tends to zero super-polynomially fast. So, $Q_{m,n}\to 0$, if $m\gg n^4$ and $n\to\infty$.

Theorems & Definitions (7)

  • Theorem 1.1
  • Lemma 1.2
  • proof
  • Lemma 1.3
  • proof
  • Lemma 1.4
  • proof