PaperlayerView exact revision on arXiv

Source-linked AI summary

Frankly Speaking: The Company Election Integrity Keeps in Congressional Communication

Michael A. Robinette, Christina Barsky, Amanda D. Clark, Helen Flannery, Juniper Lovato

arXiv:2610.36187v1cs.CL

TL;DR

This paper asks how members of Congress construct and reinterpret election-integrity rhetoric over time. Using diachronic word embeddings on congressional newsletters, it finds that shared terms can carry distinct partisan meanings through the company they keep.

  • Problem

    Less is known about how political elites construct and reinterpret election-integrity rhetoric over time, despite research on voter confidence, polarization, legal standards, and administrative performance.

    Source: §1 · pp. 1–2

  • Method

    The authors analyze congressional e-newsletters with diachronic word embeddings to track partisan differences, change over time, and the words surrounding election-integrity terms.

    Source: §1 · p. 2

  • Results

    The parties’ semantic neighborhoods reveal distinct meanings despite shared vocabulary: Democrats connect key terms to access and administration, while Republicans connect them to citizenship verification and border security.

    Source: §5 · p. 12, §6 · p. 14

  • Takeaways & Limitations

    Election officials administering one system inherit tension between these competing meanings of election integrity.

    Source: §5 · p. 12

    Word2Vec’s one-vector-per-word design blends non-electoral senses with electoral usage, so measured divergence may partly reflect topic differences rather than framing.

    Source: §5.1 · p. 13

Abstract

from arXiv · show

Election integrity is a shared commitment in American politics, but Democrats and Republicans in Congress increasingly mean different things by it. Analyzing 10,420 congressional e-newsletters sent from 2010 to 2025, we examine how the two parties invoke election terms such as ``integrity,'' ``security,'' and ``fraud.'' Using diachronic word embeddings, which track semantic change through how words are used in context, we measure how differently each party uses each term, how much each party shifts, and whether they share nearest semantic neighbors. We show that members of Congress do not need different words to tell different stories about elections. They need only place those same words in different company. Across nearly every term and year, the two parties surround election-integrity vocabulary with entirely different words. That divergence is easy to miss, because most election vocabulary is no more divergent than other words used at roughly the same frequency, and neither party's usage shifted more than the other's. Only a subset of terms grew more divergent, most notably ``certification,'' while ``rigged'' and ``stolen'' entered election discourse after 2020 and polarized there. By 2024, Democrats surround ``voter,'' ``election,'' and ``integrity'' with the language of access and administration. Republicans surround the same terms with ``citizenship verification,'' ``noncitizen voting,'' and ``border security.'' For one party, integrity means protecting eligible voters' ability to cast a ballot. For the other, it means protecting the electoral process from illegitimate ballots. Election officials, who must administer a single system under both definitions, inherit this tension.

1 Introduction

The paper examines how members of Congress construct and reinterpret election-integrity rhetoric over time, using e-newsletters and semantic context to compare the parties. It finds that similar vocabulary can carry different meanings through the words surrounding it.

1 Introduction

  • Prior research has focused on voter confidence, partisan polarization, legal standards, administrative performance, or mass beliefs, leaving elite reinterpretation of election-integrity rhetoric less examined.Official, publicly funded e-newsletters offer a lens on how representatives communicate with constituents in their governing.Source: §1 · pp. 1–2
  • The study analyzes congressional e-newsletters from 2010–2025 with natural language processing to measure partisan differences, change over time, and the words surrounding election-integrity terms.The authors call a term’s nearest embedding neighbors its semantic neighborhood.Source: §1 · p. 2
  • The paper situates these rhetorical patterns in election law, public administration, and political behavior, with potential implications for institutional legitimacy and voter confidence.It argues that understanding whether integrity language converges or fragments can illuminate how elite discourse reinforces or erodes trust in electoral processes.Source: §1 · p. 2
  • Democrats associate “voter,” “election,” and “integrity” with access and administration, while Republicans associate them with citizenship verification and border security.The paper argues that the terms’ semantic neighborhoods reveal these different partisan meanings.Source: §1 · p. 2

2 Theoretical Framework and Literature Review

The framework treats election integrity as a contested concept that can signify protection from illegitimate votes or protection of eligible voters’ access. It motivates studying how elites construct that meaning through symbolic language and the contexts in which terms appear.

2.1 Election integrity, fraud, and trust

  • Election integrity can signify protection against illegitimate votes or protection of access for eligible voters, while elite fraud claims have been linked to lower confidence and perceived fairness.Members of Congress communicate as lawmakers, party actors, and visible messengers, and their rhetoric can frame integrity as fraud prevention or fairness and competence.Source: §2.1 · pp. 2–3

2.2 Symbolic politics, elite communication, and problem construction

  • As a resonant but underspecified symbol, election integrity lets officials signal vigilance and democratic commitment while obscuring policy tradeoffs.Symbolic issues can make complex problems visible to constituents without changing underlying structures; congressional newsletters are one official communication channel to constituents.Source: §2.2 · pp. 3–4

2.3 Text-as-data and computational analysis of political language

  • Word embeddings operationalize the idea that a word’s meaning is reflected in its context, making nearest neighbors useful for comparing how parties use election-integrity terms.Because the parties use these terms at comparable rates, counting them or scoring them against a fixed dictionary cannot reveal whether they mean the same thing.Source: §2.3 · p. 4

2.4 Research questions and expectations

  • The study asks whether parties use election-integrity vocabulary differently, whether divergence and within-party usage change over time, and what themes surround the terms.It expects the run-up to the 2020 presidential election to shift party rhetoric in opposite directions on voter access.Source: §2.4 · p. 4

3 Data and Method

The study builds a corpus of congressional e-newsletters and uses party- and year-specific word embeddings to track semantic change and partisan divergence. It compares term distances with frequency-matched words and semantic-neighbor overlap with within-party baselines.

3 Data and Method

  • The study corpus contains 10,420 congressional e-newsletters from 2010–2025, drawn from the DCInbox archive.The corpus includes Democratic and Republican members; Independents are excluded.Source: §3 · p. 5

3.1 Data filtering and corpus construction

  • A three-tier filter selected 10,420 newsletters, or 5.1% of the archive, including 925 documents that keyword search alone would have missed.The resulting corpus spans January 2010 to September 2025 and is 52.4% Republican and 47.6% Democratic communications.Source: §3.1 · p. 5

3.2 Text processing

  • Text processing lowercased and tokenized documents, removed stopwords and short tokens, and detected phrases while reversing merges that would absorb target terms.For example, “voter fraud” was split back into “voter” and “fraud” so target terms remained available for analysis.Source: §3.2 · p. 5

3.3 Diachronic word embeddings

  • The authors trained separate 50-dimensional skip-gram Word2Vec models for each party and year, then aligned each year’s space to a 2010 anchor with orthogonal Procrustes rotation.Cosine distance between a term’s vectors across years measures semantic shift; rotation preserves each space’s internal geometry while enabling comparison.Source: §3.3 · pp. 5–6

3.4 Partisan semantic divergence and shift analysis

  • Cross-party divergence is the cosine distance between aligned party vectors; each term’s percentile among 100 frequency-matched words benchmarks divergence against a noise floor.The authors also compare overlap in each party’s ten nearest words using Jaccard similarity, treating cross-party overlap as meaningfully low only below both within-party baselines.Source: §3.4 · p. 6

3.5 Within-party shift and term-level asymmetry

  • To compare within-party change, the authors average year-to-year cosine distances across target terms and transitions and examine selected periods and term-usage volume.The selected periods are 2019–2020, 2020–2021, and 2024–2025.Source: §3.5 · p. 6

4 Results

Across 10,420 congressional newsletters, Democrats and Republicans use the same election vocabulary to invoke different concepts of election integrity. Semantic neighborhoods show the sharpest differences, while distance and shift measures suggest stability.

4.1 Cross-party divergence over time

  • “Stolen” and “rigged” entered election discourse in 2021; divergence around both later rose, although “stolen” initially showed nearly its lowest divergence.Before 2021, “stolen” appeared above the five-mention threshold only in 2012, in non-election contexts; the 2021 increase for “rigged” accompanied Republican discussion of a rigged impeachment process.Source: §4.1 · pp. 6–7
  • Certification’s divergence rose from the 18th to the 72nd percentile, a 54-point gain; illegal rose 44 points, while election’s 15-point jump stayed below the 50th percentile.Integrity rose 14 points, rigged 15, and fraud 7.Source: §4.1 · p. 7

4.2 Within-party shift

  • Democrats’ 0.71 shift for “illegal” likely reflects inconsistent yearly usage, while Republicans used the term seven times more and consistently associated it with immigration-related neighbors.Across the 16-year period, its top neighbor was “immigrant,” “immigration,” or “alien.”Source: §4.2 · p. 8
  • At roughly 30 mentions per year per party, “trust” was nearly symmetric in frequency, but Republican usage stabilized sharply after 2020 (0.36) while Democratic usage continued to move.By 2022, Republican nearest neighbors included “weaponized” and “delegitimize”; Democratic usage ranged from electoral confidence to ethical concerns.Source: §4.2 · p. 8

4 Results

  • Semantic neighborhoods show Democrats and Republicans using the same election-integrity terms within non-overlapping conceptual networks, despite modest distance and within-party shift measures.The analysis covers 10,420 congressional newsletters from 2010–2025.Source: §4 · p. 6, §4.3 · p. 9

4.3 Semantic neighborhoods

  • The parties shared an administrative vocabulary: Registration was a shared neighbor of voter in 14 of 16 observed years, while differing in the language they used to build around common terms.In 2020, ballot’s shared neighbors were provisional and absentee.Source: §4.3 · pp. 9–10

5 Discussion

The parties use shared election-integrity language to tell different stories, with meaning carried by the words surrounding common terms. These findings are bounded by the corpus, modeling choices, and the fact that newsletters measure elite framing rather than public reception.

5 Discussion

  • Distance and shift measures suggest stability, but semantic neighborhoods show competing definitions: securing a threatened process versus expanding access.The analysis covers 10,420 congressional e-newsletters from 2010 to 2025 and observes these patterns in one official channel.Source: §5 · p. 12
  • Election-integrity language can carry opposing agendas without either party surrendering the moral high ground, while official newsletters give these messages institutional authority.The authors interpret possible constituent responses through symbolic-politics and framing literatures; they do not measure those effects.Source: §5 · p. 12
  • The parties tell symbolic-issue stories with the same words, while their differing semantic company creates opposing frames around election integrity.For example, Democrats placed “ballot” near “prepaid,” “postage,” “postmarked,” and “box,” while Republicans linked it to “harvesting,” “signature verification,” and “identification.”Source: §5 · p. 12
  • Because the same phrase can justify expanding access or restricting eligibility, election officials inherit tension when administering one system under both definitions.Source: §5 · p. 12

5.1 Limitations

  • Low-frequency terms may appear to shift or diverge for reasons unrelated to meaning, and topic differences may also contribute to measured divergence.The analysis uses fixed model settings, and its semantic queries may admit more fraud- and security-framed content than access-framed content.Source: §5.1 · p. 13
  • Within-party shift values lack a noise-floor benchmark, so the authors interpret them only in relative comparisons across parties and periods.Source: §5.1 · p. 13
  • The study measures elite framing in two parties’ opt-in newsletters, not public reception or communication across other venues.The authors describe audience interpretations as unmeasured and leave whether the same neighborhoods appear in other channels for future analysis.Source: §5.1 · p. 13

6 Conclusion

Congressional election rhetoric can appear stable and bipartisan because the parties use largely shared vocabulary, yet semantic neighborhoods reveal distinct meanings. By 2024, Democrats framed key terms around access and administration, while Republicans linked them to citizenship verification and border security.

6 Conclusion

  • Across 10,420 newsletters, most election vocabulary was no more divergent than similarly frequent ordinary words, and neither party shifted more than the other.The shared vocabulary can make congressional election rhetoric appear stable and largely bipartisan.Source: §6 · p. 13
  • “Certification” rose 54 percentile points in divergence, while “rigged” and “stolen” entered election discourse after 2020 and became more divergent.Semantic neighborhoods reveal a contest over meaning that distance metrics alone cannot capture.Source: §6 · p. 13
  • By 2024, Democrats linked “voter,” “election,” and “integrity” to access and administration, while Republicans linked them to citizenship verification and border security.“Registration” remained a common neighbor of “voter” in 14 of 16 years, alongside distinct partisan conceptual structures.Source: §6 · p. 14

Supplementary Material

The supplementary material documents the corpus filter, its term lexicon, a special matching rule for “DRE,” and the semantic-query templates.

Supplementary Material

  • The ordered, mutually exclusive corpus-filter tiers yielded 10,423 documents; removing three duplicates produced the final 10,420-document corpus.The table reports each tier’s share of the final filtered corpus.Source: Table S1 · p. 22
  • Election-integrity subject-matter experts supplied the filter lexicon, with all Tier 2 terms matched as one pooled set.Full term lists and query templates are documented in separate supplementary tables.Source: Table S2 · p. 23
  • “DRE” is matched case-sensitively on word boundaries, so “DRE” and “DREs” match but “hundred” and “address” do not.DRE refers to a direct-recording electronic voting machine.Source: Supplementary Material · p. 23
  • Each document was encoded with Sentence-BERT (allmpnet-base-v2) and scored by its maximum cosine similarity across 19 semantic queries.The supplementary table lists the query templates.Source: Table S3 · p. 24

Related papers

Loading 2610.36187v1…