Graph-Based Audits for Meek Single Transferable Vote Elections

In the context of election security, a Risk-Limiting Audit (RLA) is a statistical framework that uses a minimal partial recount of the ballots to guarantee that the results of the election were correctly reported. A generalized RLA framework has remained elusive for algorithmic election rules such as the Single Transferable Vote (STV) rule, because of the dependence of these rules on the chronology of eliminations and elections leading to the outcome of the election. This paper proposes a new graph-based approach to audit these algorithmic election rules, by considering the space of all possible sequences of elections and eliminations. If we fix a subgraph of this universal space ahead of the audit, a sufficient strategy is to verify statistically that the true election sequence does not leave the fixed subgraph. This makes for a flexible framework to audit these elections in a chronology-agnostic way.

💡 Research Summary

The paper introduces a novel graph‑based framework for conducting risk‑limiting audits (RLAs) of elections that use the Meek variant of the Single Transferable Vote (STV). Traditional RLAs have been successful for single‑winner contests because they can directly test whether the reported margin could be overturned by the observed sampling error. STV, however, is algorithmically complex: the order in which candidates are elected or eliminated can dramatically affect the final committee, making a full recount appear necessary to certify the result.

To overcome this, the authors model the entire space of possible election histories as a universal directed graph Ω. Each vertex represents a state (H, W) where H is the set of “hopeful” (still‑in‑the‑running) candidates and W is the set of already elected candidates. An edge connects two vertices if, given a bounded discrepancy (e.g., up to 40 votes) between the recorded Cast Vote Records (CVRs) and the true ballots, the election could legitimately transition from one state to the next in a single round.

The auditor pre‑selects a subgraph G ⊆ Ω that contains all states reachable under the assumed error bound and that leads to the reported winner set. The audit then tests the null hypothesis H⋆: “the true election path leaves G.” By randomly sampling physical ballots and comparing them to the CVRs, the auditor estimates the probability that the observed discrepancies would force the true path outside G. If this probability is below a pre‑chosen risk limit α (e.g., 5 %), H⋆ is rejected and the audit succeeds: even though the exact chronology is unknown, every admissible path within G ends with the same committee, confirming the reported outcome.

A key technical contribution is the formalization of the Meek rule as a monotone decreasing map that removes one candidate per round without requiring explicit “election” steps. Meek’s defining features—continuous recalculation of transfer values after each election and dynamic quota adjustment—make the rule inherently chronology‑independent. This contrasts sharply with the Weighted Inclusive Gregory Method (WIGM), where transfer values are fixed after a candidate’s election, rendering the outcome sensitive to the precise order of eliminations and elections.

The authors illustrate the framework with two real‑world elections. In the 2024 Portland City Council three‑seat STV contest, two candidates tied for last place in round 6; regardless of which was eliminated first, the final winner set remained unchanged. The audit graph G captured this redundancy, allowing a small‑sample audit to certify the result without a full recount. In the 2012 Perth‑Kinross Ward 9 election, the WIGM rule produced a razor‑thin margin (0.37 votes) that could be flipped by a tiny perturbation, making the election effectively unauditable under WIGM. Under Meek, however, the same profile yielded a stable margin of 46.87 votes in the final round, and the corresponding audit graph showed that all admissible paths converged to the same winner set.

The paper demonstrates that Meek STV’s chronology‑independence translates into audit‑friendly stability: the subgraph G can be constructed with far fewer vertices and edges, and the statistical test gains power because the worst‑case margin to audit is larger. Moreover, the framework is flexible: auditors can adjust the error bound, the risk limit, or the granularity of G to suit different jurisdictions or resource constraints.

Future work suggested includes optimizing graph construction to handle larger candidate pools, improving sampling schemes to reduce ballot‑draw requirements, and extending the approach to other STV variants (e.g., Droop‑quota based methods) or to hybrid systems that combine STV with other proportional representation mechanisms. Overall, the study provides a concrete, mathematically rigorous pathway to bring RLAs to complex multi‑winner elections, enhancing election security and public confidence while avoiding the prohibitive cost of full hand recounts.