Appendix · Methodology
How the report gets made.
Same pipeline, on display. The universe of public commentators on the question narrows, stage by stage, into a structured dataset of 8,844 quotes clustered into 25 top-level themes. Every step timestamped. Every quote traceable. No black box.
- Sources
- 98
- Videos
- 203
- Hours
- 258.5
- Quotes
- 8,844
- Reach
- 118M
Cast the net.
A keyword cluster — "Why Democrats lost / 2024 election post-mortem / Harris campaign / coalition collapse / Trump 2024" — runs through YouTube search and podcast directories. The candidate set is deliberately wide. Everything narrows from here.
candidate channels surveyed
Plot every voice on the compass.
Each candidate gets sampled, scored on the Political Compass from its actual content, and audited before inclusion. We sample deliberately across spectrum, audience size, and format — not proportionally. Small-but-distinctive voices don't get drowned out by the megachannels.
sources after audit
spectrum tiers
CNN · audited
economic−2.0
social+3.0
tierCenter → Left
Score from sampled segments. Audited 2026-04.
Lift claim-level quotes from raw transcripts.
Transcripts come back from each video. Models surface claim-level spans — statements with a stance and confidence — and anchor every span to its timestamp. The result: a quote that deep-links back to the exact moment it was said. Human reviewers sweep the marginal cases.
videos transcribed
hours of audio
quotes extracted
"Democrats had a real messaging problem… 'Build Back Better' was very abstract. It didn't tell a voter what was in it for them."
Let the themes emerge.
The 8,844 quotes get clustered into the categories the commentary itself was already organized around — bottom up, not imposed. Top-level themes and sub-themes surface from the data. We author the taglines and edit the prevalence ranking. The shape of the conversation is whatever the conversation made it.
top-level themes
sub-themes
- 01 Flawed Strategy & Tactical Incompetence 871
- 02 Neglected Coalition & Demographic Collapse 756
- 03 Ineffective Economic & Policy Messaging 620
- 04 Flawed Candidacy & Leadership Vacuum 530
- 05 Elitist Culture & 'Woke' Alienation 528
- 06 Internal Party Dysfunction & Organizational Decay 488
- +19 more top-level themes…
Layer the audience back in.
Per-video views, likes, and comments join back to each quote's source — so we can show engagement across the spectrum, not just by channel. Daily timelines reconstruct from publication dates. Comment sentiment is scored for the next iteration. Volume is one story; the lean of the volume is another.
views
likes
comments
- farLeft 2.28M
- left 45M
- center 43M
- right 1.07M
- farRight 26M
Dataset
2024 Election Analysis
| Range | 2024-11-05 → 2025-05-05 |
|---|---|
| Sources | 98 |
| Videos | 203 (258.5 hours) |
| Quotes | 8,844 |
| Themes | 25 top-level / 126 hierarchical |
| SHA | 185f02c64ecf |
Frequently asked
A few questions worth answering before they get asked.
- What dates does the study cover?
- The study samples content created between 2024-11-05 and 2025-05-05.
- How is engagement data reported?
- View, like, and comment counts are snapshot from YouTube at the time of analysis. The numbers shift after that.
- How are political leanings determined?
- Each source is scored on the Political Compass by sampling transcripts of its actual content. We audit before inclusion — no self-reported labels.
- How is quote extraction done?
- Transcripts are processed by a combination of Gemini, Claude, and OpenAI models, plus human review. Each quote and timestamp lives next to the underlying claim, on the record.
- What was the primary research question?
- "How do political commentators across the spectrum explain the 2024 election outcome — what they blame, who they think should change, and what they think comes next?"
- Is the full source and content list available?
- Yes — see the Sources page for all 98 channels. Per-video citations live inside each PullQuote's "Why we trust this" expansion.