Head-Count Equivalence Theorem

• Head-Count Equivalence Theorem is a model that defines how voting weights in two-tier systems can achieve one-person, one-vote equality.
• It shows that a delegate’s pivot probability is proportional to the product of its weight and the density of voter ideal points at the common median.
• Finite adjustments via the Shapley value help tailor weights to constituency sizes and voter distributions for practical egalitarian representation.

The Head-Count Equivalence Theorem addresses the allocation of voting weights in two-tier collective decision systems, where each of $m$ constituencies sends a delegate to an assembly, and the assembly applies a weighted voting rule to select a collective policy via the Condorcet winner. The theorem provides necessary and sufficient conditions under which all bottom-tier citizens have approximately equal a priori influence on the collective outcome—also known as the "one person, one vote" criterion or head-count equivalence. The theorem shows that, under broad continuity and asymptotic conditions, the assembly pivot probability for each delegate is proportional to the product of the delegate's voting weight and the density of his ideal point's distribution at the common median. This result unifies and extends classical rules, including the square-root and linear rules, and rigorously incorporates preference aggregation models into the design of egalitarian representation structures (Kurz et al., 2012).

1. Formal Model of Two-Tier Collective Decision Systems

The foundational model consists of $n$ bottom-tier voters partitioned into $m$ disjoint constituencies $C_1, ..., C_m$, each of size $n_i$. Voters have single-peaked preferences over a real-valued policy space $X \subset \mathbb{R}$, with ideal points drawn from continuous distributions. Delegate $i$ represents constituency $C_i$ and adopts the position of its median voter, reflecting perfect median congruence. The top tier consists of these $m$ delegates, each assigned a voting weight $w_i$. The assembly uses a simple majority quota $q^m = \frac{1}{2} \sum_k w_k$. The collective decision $x^*$ is the ideal point of the assembly's "weighted median of medians"—the unique $\lambda_{P}$ such that sufficient aggregate weight is accumulated at or left of $\lambda_{P}$ (Kurz et al., 2012).

An individual voter's (from constituency $C_i$) a priori probability of determining $x^*$ is the product of the probability of being his group’s median ($1/n_i$) and his delegate being pivotal in the assembly (denoted $\pi_i$), so $p^l = \pi_i/n_i$. The system meets head-count equivalence if $p^l/p^k \approx 1$ for all voters $l, k$, or equivalently, $\pi_i/\pi_j \approx n_i/n_j$ for all $i, j$.

2. Statement and Proof Outline of the Head-Count Equivalence Theorem

The Head-Count Equivalence Theorem (Theorem 1 in (Kurz et al., 2012)) specifies that, under the following conditions:

• (a) All delegate-ideal-point distributions ($F_i$) share a common median $M$;
• (b) Each $F_i$ is absolutely continuous near $M$ with a strictly positive and locally continuous density $f_i(M)$;
• (c) The assembly chain grows with $m$ (number of constituencies) and fixed positive weights;
• (d) Each $w_i/\sum_j w_j \to 0$ as $m \to \infty$.

Then, for any pair of delegates $i, j$,

$$\lim_{m \to \infty} \frac{\pi_i(\mathcal{R}^m)}{\pi_j(\mathcal{R}^m)} = \frac{w_i f_i(M)}{w_j f_j(M)},$$

so the assembly-pivot probability satisfies $\pi_i \propto w_i f_i(M)$ as $m \to \infty$.

The proof relies on exponentially fast concentration of delegate medians around $M$ (Hoeffding-type bounds), the local uniformity of ranks for delegates near $M$, and the asymptotic proportionality of the Shapley value to weights in large voting games (Neyman’s theorem). Aggregating the pivot probability over the (small) interval around $M$ and conditioning on which delegates land in this interval yields the main result.

3. Corollaries: Square-Root Rule and Linear Rule under Preference Structures

Under the independent and identically distributed (i.i.d.) assumption where all voter ideal points are identically distributed with density $g$ continuous at median $M=0$, the asymptotic distribution of constituency delegate medians yields $f_i(0) \propto 1/\sqrt{n_i}$ from the classical theorem for the sample median. Thus, the theorem implies

$\pi_i \propto w_i \sqrt{n_i}.$

To satisfy head-count equivalence ($\pi_i \propto n_i$), the choice $w_i \propto \sqrt{n_i}$ is required, reproducing the Penrose square-root rule for aggregating districts whose members are i.i.d. This is a continuous policy-space generalization of the original Penrose rule (Kurz et al., 2012).

If, in contrast, voter preferences exhibit strong within-constituency correlation (modeled as $\nu^l = \mu_i + \epsilon^l$, with dominant constituency "shocks" $\mu_i$), as $t \to \infty$ in $\lambda_i = t \mu_i + \mathrm{median}(\epsilon^l)$, the ordering of delegate medians is governed by $\mu_i$ alone. In this case, the pivotal probability becomes $\pi_i \propto w_i$ and head-count equivalence mandates $w_i \propto n_i$, i.e., a linear rule. More generally, optimal weights may need to be determined via the inverse Shapley-value problem to ensure $\phi_i([q;w]) \propto n_i$ (Kurz et al., 2012).

4. Rate of Convergence and Finite-$m$ Adjustments

The convergence of delegate medians around the global median $M$ is exponentially rapid as $m \to \infty$, with the approximation error in $\pi_i/\pi_j \to w_i f_i(M)/w_j f_j(M)$ vanishing as $o(m^{-1/8})$. For finite (moderate) numbers of constituencies, the head-count equivalence principle can be approximated by directly computing the Shapley value $\phi_i([q;w])$, which captures the true assembly-pivot probabilities. In these cases, weights $w_i$ may be fine-tuned to satisfy $\phi_i \propto n_i / f_i(M)$ for greater accuracy (Kurz et al., 2012).

5. Practical Implications in Representative Systems

The theorem's results have direct application to the design of representative assemblies, especially in federations or unions where constituencies differ sharply in size or heterogeneity. The square-root rule is justified only if citizen preferences are i.i.d. across individuals and constituencies. The presence of significant constituency-level correlation or homogeneity—such as demographic, historical, or cultural uniformity—shifts the egalitarian solution toward linear weighting. Empirically, even modest within-constituency affiliation forces the optimal exponent from $1/2$ (square-root) rapidly toward $1$ (linear) (Kurz et al., 2012).

Preference Regime	Density at Median $f_i(M)$	Egalitarian Weight Rule
i.i.d. individual preferences	$\propto 1/\sqrt{n_i}$	$w_i \propto \sqrt{n_i}$
Strong within-constituency correlation	constant	$w_i \propto n_i$ or via Shapley

For real-world assemblies, this suggests careful empirical assessment of within-constituency correlation before rule selection.

6. Extensions and Context within Voting Power Theory

The Head-Count Equivalence Theorem generalizes classical aggregation principles, bridging simple majority voting and weighted-power index approaches (e.g., Shapley value analysis), and providing a rigorous link between first-principle models of voter preferences and assembly-level design. It demonstrates that egalitarian representation is not universally achieved by a single rule; rather, it is determined by the underlying probabilistic structure of the voters' preferences. The theorem also motivates further study of the inverse Shapley value problem in optimizing voting weights to achieve proportional representation beyond asymptotic regimes (Kurz et al., 2012).