Datasets:

tapanyemre
/

tech-speech-congress

year int64 2k 2.03k	n_speeches int64 18k 66.6k	mean_score float64 0 0	sum_hits float64 1.99k 5.3k	mean_hits float64 0.05 0.15
1,995	66,594	0.000703	3,937	0.059119
1,996	42,115	0.000792	2,598	0.061688
1,997	42,986	0.000632	2,421	0.056321
1,998	43,708	0.000756	2,415	0.055253
1,999	48,071	0.000717	2,634	0.054794
2,000	41,473	0.000766	2,385	0.057507
2,001	42,057	0.00075	2,632	0.062582
2,002	35,161	0.000976	2,636	0.074969
2,003	47,961	0.000931	3,242	0.067597
2,004	37,088	0.000805	2,145	0.057835
2,005	44,533	0.000864	3,117	0.069993
2,006	35,950	0.000817	2,090	0.058136
2,007	53,390	0.001083	4,441	0.08318
2,008	35,109	0.001826	5,302	0.151015
2,009	48,716	0.001467	5,000	0.102636
2,010	33,993	0.001385	3,019	0.088812
2,011	35,792	0.000953	2,468	0.068954
2,012	28,270	0.001602	3,070	0.108596
2,013	30,698	0.001108	2,263	0.073718
2,014	25,482	0.001344	2,353	0.09234
2,015	30,263	0.00204	3,734	0.123385
2,016	24,533	0.001402	2,037	0.083031
2,017	28,538	0.00131	2,320	0.081295
2,018	23,893	0.001955	2,647	0.110786
2,019	25,773	0.0017	2,589	0.100454
2,020	18,033	0.002226	2,590	0.143626
2,021	20,268	0.001707	2,138	0.105486
2,022	18,859	0.002455	2,473	0.131131
2,023	20,232	0.001873	1,987	0.098211
2,024	19,654	0.002037	2,045	0.10405
2,025	20,070	0.002116	2,216	0.110414

Tech-Speech in the Congressional Record, 1995–2025

Descriptive aggregates from a speech-level analysis of technology-related discourse in the U.S. Congressional Record, 1995–2025. Built from govinfo.gov daily Congressional Record packages, parsed to the individual speech turn with speaker metadata (party, state, chamber, Bioguide ID), filtered for procedural speech, and scored with a TF-IDF technology-intensity index over a 237-term technology vocabulary built from federal and policy-taxonomy sources.

Underlying corpus: 1.07 million substantive speech turns from 1,694 unique legislators (filtered from 2.55M total turns across 11,323 daily packages).

Pipeline code: github.com/tapanyemre/tech-speech-congress (MIT).

Files

File	Rows	Description
`tech_index_per_year.csv`	31	Annual tech-intensity: speech counts, mean score, total and mean vocabulary hits, 1995–2025
`per_year_by_party.csv`	99	Year × party (D, R, I, ID) speech counts, mean score, total hits
`per_year_by_sector.csv`	950	Year × technology sector (long format): speeches with hits, total hits, mean score
`state_overall.csv`	56	State-level ranking by mean tech-intensity score
`top_legislators_cumulative.csv`	50	Top 50 legislators by cumulative technology-vocabulary hits, with Bioguide IDs
`top_sectors_overall.csv`	35	Technology sectors ranked by total hits across the full period
`tech_index_per_term.csv`	237	Per-term document frequency, total hits, and IDF weight
`tech_vocabulary.csv`	237	The technology vocabulary: term, category, federal source, polysemy flags, and verbatim provenance rationale

Column conventions: n_speeches counts substantive speech turns in the cell; sum_hits / total_hits count technology-vocabulary term occurrences; mean_score is the mean per-speech TF-IDF tech-intensity score.

Method

The tech-intensity index follows the TF-IDF template of Engle, Giglio, Kelly, Lee & Stroebel (2020, Review of Financial Studies). The vocabulary is built from federal taxonomies (the OSTP Critical and Emerging Technologies 2024 list, CHIPS Act, Cures Act/FDA digital health, Section 230/platform terms) plus Comparative Agendas Project Topic 17 codes, with per-term provenance recorded in tech_vocabulary.csv. The procedural-speech filter follows the framework of Card et al. (2022, PNAS). Full methodology and the ordered pipeline scripts are in the GitHub repository.

Status and roadmap

V0 — provisional. Results are pending a vocabulary review pass, a 480-sample stratified gold standard, and full-corpus LLM classification.

This dataset is the first public release of a maintained successor to speaker-attributed Congressional Record corpora whose coverage ended with the 114th Congress (January 2017). Planned v1 (target: end of 2026): the full speaker-level corpus with per-speech topic correlates and tech-intensity scores, released as versioned, auto-updated datasets as new Congressional Record issues publish, alongside Python and R packages for research use and a public daily dashboard.

License and citation

MIT. If you use this data, please cite:

@misc{tapan2026techspeech,
  author  = {Tapan, Yunus Emre},
  title   = {Tech-Speech in the Congressional Record, 1995--2025},
  year    = {2026},
  version = {0.1.0},
  url     = {https://github.com/tapanyemre/tech-speech-congress}
}

Downloads last month: 73