Politics of Feelings: Emotional Expression and Legislative Effectiveness in the U.S. Congress
cs.CL
Quick verdict
1.7M-speech study ties discrete emotions to legislative effectiveness; strong trends, but weak classifier validation and tiny effects limit causal reading.
Summary
This paper analyzes 1,766,113 U.S. Congressional Record speeches from the 93rd–118th Congresses (1973–2024), scoring each for eight discrete emotions (anger, fear, disgust, sadness, joy, enthusiasm, pride, hope) with a transformer-based multilabel classifier, plus derived measures of emotional valence, intensity, and diversity. Procedural speeches were removed via a zero-shot Llama 3 model, speeches were assigned to 21 Comparative Agendas Project topics (macro-F1 0.857), and data were linked to member-Congress Legislative Effectiveness Scores (LES). Findings: (1) nearly all emotions show significant increasing monotonic trends for both parties (Mann–Kendall with Benjamini–Hochberg FDR), joy among Democrats excepted; (2) emotion profiles differ across policy domains (Kruskal–Wallis, effect sizes ϵ²=0.013–0.034) and are stable over time; (3) female legislators express more joy, enthusiasm, pride, fear, and sadness, while ideology scores predict more anger; (4) enthusiasm (b=0.016) and pride (b=0.011) are positively associated with LES and anger negatively (b=-0.037), with valence and diversity positive and intensity negative. Limitations include classifier validation restricted to the 50 most extreme-scoring speeches per emotion (AUC 0.92–1.00), no code or data availability statement, an unspecified classifier checkpoint, small uninterpreted effect sizes, and a purely associational design for the effectiveness claims.
Review complete
Assessment at a glance
5
risk flags identified
● high 1
● medium 3
● low 1
5 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Congressional speeches became steadily more emotionally expressive from 1973 to 2024, with nearly all eight emotions showing significant increasing monotonic trends for both parties.
- Emotional expression varies systematically across CAP policy domains (fear highest in International Affairs, anger highest in Immigration, pride highest in Education) with profiles stable over time.
- Enthusiasm and pride are positively associated with member-level Legislative Effectiveness Scores, while anger is negatively associated.
- Emotional valence and emotional diversity correlate positively with legislative effectiveness, whereas emotional intensity correlates negatively.
- Legislator attributes matter: women express more joy, enthusiasm, pride, fear, and sadness, and higher ideology scores predict more anger.
Strengths
Unmatched scale and coverage
1,766,113 speeches across 26 Congresses (93rd–118th, 1973–2024), both chambers, with every session contributing at least 20,000 speeches (Fig. 5), enabling stable party-level trend estimation (Table 3, N=26 series per emotion-party pair).
Rigorous trend inference with honest reporting
RQ1 pairs Mann–Kendall tests with Sen's slope and Benjamini–Hochberg FDR (Table 3; Kendall's τ 0.305–0.877) and candidly notes that hope's significant trend carries a near-zero slope (0.001–0.002 per Congress).
Discrete emotions reveal structure aggregates miss
Conservative LES specification
Triangulated face validity
Weaknesses
Validation design cannot support the measurement
Small effect sizes vs. strong wording
Causal slippage on correlational evidence
Measurement provenance opaque
No reproducibility artifacts
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
The paper analyzes 1,766,113 Congressional Record speeches spanning the 93rd–118th Congresses (1973–2024) to answer four questions: how discrete emotions trend over time (RQ1), vary across 21 CAP policy domains (RQ2), correlate with legislator attributes (RQ3), and associate with Legislative Effectiveness Scores (RQ4). Eight emotions — anger, fear, disgust, sadness, joy, enthusiasm, pride, hope — are scored by a transformer multilabel classifier cited as Widmann & Wich (2023), supplemented by valence (positive minus negative sums), intensity (sum of eight), and diversity (entropy). Headline claims: congressional speech has grown steadily more emotional (Table 3); emotions vary by topic and legislator (Tables 4–6); enthusiasm (b=0.016) and pride (b=0.011) are positively associated with LES while anger is negatively associated (b=-0.037, Tables 7–8); valence and diversity are positive and intensity negative (Table 9). The abstract stays associational, but the Discussion slides toward consequential language.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What did the study find about emotions and legislative effectiveness in Congress?
Using over 1.7 million Congressional Record speeches (1973–2024) scored for eight emotions, the study found that enthusiasm and pride in a legislator's speeches were positively associated with their Legislative Effectiveness Score, anger was negatively associated, and emotional valence and diversity were positive while intensity was negative. All effects were statistically significant but small in magnitude, and the design is correlational.
What data did the paper use?
1,766,113 floor speeches from the Congressional Record covering the 93rd–118th U.S. Congresses (1973–2024), policy topic labels from the Comparative Agendas Project taxonomy (21 domains), and Legislative Effectiveness Scores from Volden and Wiseman at the member-Congress level (13,671 observations in the effectiveness models).
How were emotions in congressional speeches measured?
A transformer-based multilabel classifier assigned each speech continuous scores for eight emotions: anger, fear, disgust, sadness, joy, enthusiasm, pride, and hope. The classifier was validated by hand-annotating the 25 highest- and 25 lowest-scoring speeches per emotion, yielding AUCs between 0.92 and 1.00 — though this extreme-sample design does not test accuracy on typical speeches.
Are congressional speeches becoming more emotional over time?
Yes. Mann–Kendall trend tests with FDR correction show significant increasing monotonic trends for nearly all eight emotions among both Democrats and Republicans between 1973 and 2024; the main exception is joy among Democrats, which shows no significant trend. Both positive and negative emotions rose, indicating broad intensification rather than a shift toward negativity.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- The supplied metadata abstract is cut mid-sentence, but the complete final sentence appears in the full-text body; the extraction ends at the Table 4 caption, consistent with the document's true end — no truncation detected.
- Figures 1–7 are present only as captions and axis labels in the extraction; plotted trajectories (e.g., the Democratic anger peak around 2017 and the final-session enthusiasm spike described in §4.1) could not be visually verified beyond the reported test statistics.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.