Skip to content

Open data

Every public file from the lab, free to download

datasets
20
data files
24
dates checked
Nov 11, 2026

Downloads need a free account. Create one or sign in; it also gets you a key to the data API.

The US survey

167,000 interviews, Sep 2022 to Jun 2026, in one file, with the exact wording of every question in the codebook, modelled estimates for every state on 25 measures and a short report on each survey.

About 1,000 American adults a survey, interviewed online by YouGov: weekly from September 2022 to mid-2025 and about monthly since. Every survey asks how people feel about the two parties, whether they would support political violence and whether they accept breaches of democratic norms, with rotating questions on policy and values.

  • All survey responses

    One row per completed interview

    Every answer to every published question, with demographics, state, county and congressional district, the survey number and date, and the survey weight. The codebook gives each question’s exact wording and answers; a few modules it describes (the workplace and business items) are not in this file.

    Sources: YouGov, Geocorr 2022.

    Covers
    Sep 2022 to Jun 2026
    Size
    167,000 interviews · 25.4 MB
    Format
    CSV in a ZIP
    Updated
    Last rebuilt Jul 8, 2026

    Dates in the file. The file is rebuilt by hand after survey uploads, and the survey database already runs further (to Oct 22, 2026).

  • State opinion estimates (MRP)

    One row per state and measure

    Modelled estimates of opinion in every state and DC: support for each of the six acts of political violence, agreement with five breaches of democratic norms, and eleven measures of ideology, the MAGA label, immigration, the economy, tariffs and free speech. Each estimate comes with its 95% interval and the number of respondents in the state. Multilevel regression and poststratification pools every survey, so an estimate describes the whole period, not one month. Since November 2026 the model includes party, taken for the population from the Cooperative Election Study, and three measures pool the CES’s own respondents.

    Sources: YouGov, U.S. Census Bureau, Federal Election Commission, Cooperative Election Study.

    Re-estimated Nov 12, 2026 (50 states and DC × 25 measures). The violence items are asked as a ladder: each harsher act is put only to people who supported the milder one. The model now counts everyone not shown an act as not supporting it, so every estimate is a share of all adults. Copies downloaded before Oct 20, 2026 counted only the people asked acts 2 to 6, who had already backed a milder act; replace them.

    Covers
    Sep 2022 to Oct 2026
    Size
    1,122 rows · 64.1 KB
    Format
    CSV (also JSON)
    Updated
    Last rebuilt Oct 20, 2026
  • House district opinion estimates (MRP)

    One row per House district, map and measure

    The same 25 measures for every House district, on the lines used in 2024 and, in the nine states that redrew, in 2026. The model adds a district level to the state model, and its population frame splits each district by party with the Cooperative Election Study respondents who live there. Each estimate comes with its 95% interval and the interviews behind it; one resting on fewer than 50 is marked as mostly modelled, as most are on the questions asked only some months. Also in the House districts bulk file.

    Sources: YouGov, U.S. Census Bureau, Cooperative Election Study, Geocorr 2022.

    Covers
    Sep 2022 to Oct 2026
    Size
    15,225 rows · 1.1 MB
    Format
    CSV (also JSON)
    Updated
    Last rebuilt Nov 7, 2026
  • Survey toplines

    One PDF report per survey

    A short report of the key findings of each survey wave.

    Sources: YouGov.

    Reports
    163, one per survey
    Surveys
    2022, week 38 to 2026, week 41
    Format
    PDF
    Download PDF: survey toplines, survey 195
Dates and sizes

“Covers” is the span of dates in each file, checked every night (last on Nov 11, 2026). Sizes and counts were measured on Oct 20, 2026; files rebuilt nightly grow a little every day.

Back to topNext: International

Surveys in five democracies

50,000 interviews in 5 countries, Jan 2024 to Jan 2026, one file per country, with a report on each round.

The same core questions (feelings toward the parties, political violence and democratic norms, adapted to each country’s parties) asked of about 1,000 adults a round in Brazil, Germany, India, Israel and Poland, each quarter from January 2024. The latest round has 2,000 per country.

  • Survey responses by country

    One row per interview, one file per country

    Every published answer with the round, demographics and the survey weight. Each country has its own party questions, so the columns differ from file to file.

    Sources: YouGov.

    Covers
    Jan 2024 to Jan 2026
    Size
    10,000 interviews · 8.6 MB
    Format
    CSV
    Updated
    Last rebuilt Jun 5, 2026
  • Survey toplines by country

    One PDF report per country and round

    A short report of the key findings of each round in each country.

    Sources: YouGov.

    Reports
    5 countries, up to 9 rounds each
    Format
    PDF
    Download PDF: Brazil toplines, round 9

Back to topNext: Congress

What members of Congress say

What members of Congress have said, Jun 2017 to Nov 2026, passage by passage with the model’s labels, and the full text of 213,019 e-newsletters, 142,231 press releases and 119,025 floor speeches.

What members of Congress publish (posts on X, press releases, e-newsletters and floor speeches) is split into passages, and an AI model labels each one: a personal attack, a policy attack, policy discussion and its area, bipartisanship, credit claiming and extreme labels, with the model’s reason for each label.

Labelled statements

Full text

Back to topNext: States

What state legislators say

1,179,536 labelled posts by state legislators, May 2024 to Nov 2026, with a profile of each of 7,417 legislators.

Posts on X by state legislators in all 50 states, DC and Puerto Rico, labelled by the same AI model as Congress. The files carry each post’s id and web address but not its text, which X’s terms do not let us redistribute. Candidates for state office are not included.

Back to topNext: Violence

Political violence in the United States

372 attacks from 1970 to Sep 2025, each with its place, target, motive and toll, and totals by state.

A database of documented acts of political violence in the United States since 1970: when and where each happened, what kind of attack it was, the target and the motive, a short summary, and what is known about the perpetrators. Beside it, separate records of federal prosecutions for threatening public officials and of threats and harassment against local officials.

Protest, a separate record

  • Protests by county and month

    One row per county and month with a protest

    The Crowd Counting Consortium’s in-person protests, rallies, marches and demonstrations, counted by the lab: protests in each county (FIPS code) and month, how many CCC coded left wing, right wing or neither, how many had arrests, injuries or property damage, and the summed low and high crowd estimates of those that report one. A separate record from the violence events, never merged into them.

    Sources: Crowd Counting Consortium.

    Covers
    Jan 2017 to May 2026
    Size
    55,994 county-months · 2.3 MB
    Format
    CSV
    Updated
    Rebuilt weekly

Threats and harassment, a separate record

  • Threats and harassment against local officials

    One row per incident

    The Bridging Divides Initiative’s Threats and Harassment Dataset (Princeton University): every threat or act of harassment against a local official it has recorded since January 2022, from news reports and its partners’ reports, with BDI’s anonymised summary; the date and how exact it is; the county (with its 2020 FIPS code where the name matches) and state; the officials targeted, how many and their gender; the nature and method, any targeting language and the issue at stake; and BDI’s sources. Events BDI has withdrawn are left out. These are incidents reported in the media and by BDI’s partners, not every threat. A separate record from the violence events and the DOJ prosecutions, never merged into them.

    Sources: Bridging Divides Initiative.

    Covers
    Jan 2022 to Apr 2026
    Size
    New; not yet measured
    Format
    CSV
    Updated
    Rebuilt with each new release

Hate crimes, a separate record

  • Hate crimes by state and year

    One row per state (and the nation) and year

    The hate crime incidents law enforcement agencies reported to the FBI each year since 1991, counted by the lab: incidents, victims and incidents in each of the FBI’s bias categories (race, ethnicity or ancestry; religion; sexual orientation; gender identity; disability; gender), with how many agencies took part, the population they cover (2008 on) and the Census population, and rates per 100,000 residents. Reporting is voluntary, so read every count with its coverage. A separate record from the violence events, never merged into them.

    Sources: FBI Uniform Crime Reporting Program, U.S. Census Bureau.

    Covers
    Jan 1991 to Dec 2025
    Size
    1,820 state-years · 118 KB
    Format
    CSV
    Updated
    Last rebuilt Nov 3, 2026

Back to topNext: Bulk files

This site’s bulk files

22 datasets in 81 files, built Nov 12, 2026, each with its codebook.

Separate from the public files above: the profiles, votes, bills, ideology scores and statement rates behind this site, rebuilt on schedule as CSV and Parquet. Like every download here they need a free account, and they are CC BY 4.0 too. One exception: caption text from broadcasters and video platforms in the state floor speeches file, and campaign-website text in the websites file, remain under their sources’ terms, not CC BY 4.0. They never include donor-level campaign-finance records or data on candidates for state office.

Back to topNext: Sources

Sources

The files draw on 38 outside sources: records of the federal and state governments, open data projects, research data sets, the platforms where officials publish and the firm that fields the lab’s surveys.

Each entry says what the lab takes from the source, which files and pages use it, the terms the source publishes under and the credit it asks for. The labels, scores, estimates and survey questions are the lab’s own work.

Federal government

Federal statistics by place

  • County Business Patterns

    U.S. Census Bureau

    Business establishments, mid-March employment and annual payroll in 2023 for every congressional district (119th Congress lines), state and the nation, in total and by industry sector. The Bureau adds noise to each cell instead of withholding it; figures it flags as carrying five percent noise or more are marked.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105). The Bureau asks to be cited as the source. Read them: the terms of County Business Patterns (opens in a new tab)
    Credit
    U.S. Census Bureau, 2023 County Business Patterns by Congressional District (119th Congress), released June 2025.
  • Rural-Urban Commuting Area Codes

    USDA Economic Research Service

    The 2020 commuting-area code of every census tract, from which the lab gives each district and state the share of its residents in metropolitan, micropolitan, small-town and rural tracts.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105).
    Credit
    USDA Economic Research Service, 2020 Rural-Urban Commuting Area (RUCA) Codes.
  • FHFA House Price Index, census tracts

    Federal Housing Finance Agency

    The annual repeat-sales house price index of every census tract with enough sales, from which the lab gives each district and state the change in house prices over five years and over the last year.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105). FHFA labels its tract indexes developmental.
    Credit
    Bogin, A., Doerner, W. and Larson, W. (2019). Local House Price Dynamics: New Indices and Stylized Facts. Real Estate Economics; FHFA annual House Price Index, census tracts (developmental index).
  • SOI Tax Stats, ZIP code data

    Internal Revenue Service, Statistics of Income

    Individual income tax returns filed for tax year 2023 by ZIP code: the number of returns, adjusted gross income and returns claiming the earned income tax credit, which the lab splits among districts by population.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105).
    Credit
    Internal Revenue Service, Statistics of Income, Individual Income Tax Statistics, ZIP Code Data (SOI), Tax Year 2023.
  • PLACES: Local Data for Better Health

    Centers for Disease Control and Prevention

    Model-based estimates of eight adult health measures (obesity, diabetes, high blood pressure, depression, frequent mental distress, smoking, no leisure-time physical activity, no health insurance) for every census tract, 2025 release, which the lab weights to districts and states. The release has none for Kentucky or Pennsylvania.

    Terms
    Public domain, as data.cdc.gov lists it. The CDC cautions that the model cannot see local programmes: the figures describe places, not the effects of interventions.
    Credit
    Centers for Disease Control and Prevention. PLACES: Local Data for Better Health, census tract data, 2025 release.

Federal spending and benefits

  • USAspending.gov

    U.S. Department of the Treasury, Bureau of the Fiscal Service

    Federal obligations in fiscal year 2025 by the congressional district the work is placed in, in total, by type of award and for Medicare on its own (the API places Medicare at the contractors and insurers that receive it, not where beneficiaries live), read through the USAspending API.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105).
    Credit
    USAspending.gov, spending by geography (place of performance, congressional district), fiscal year 2025, accessed through the USAspending API.
  • Geographic Distribution of VA Expenditures

    Department of Veterans Affairs, National Center for Veterans Analysis and Statistics

    What the VA spent in fiscal year 2025 in every congressional district and state, direct benefits and indirect spending (health care, education, loans and the rest), by program.

    Terms
    Public domain: a work of the U.S. government, not subject to copyright (17 U.S.C. § 105).
    Credit
    Geographic Distribution of VA Expenditures for Fiscal Year 2025, the National Center for Veterans Analysis and Statistics, Department of Veterans Affairs.
  • Congressional Statistics

    Social Security Administration, Office of Research, Evaluation, and Statistics

    Social Security beneficiaries by type and their monthly benefits, and SSI recipients by category and age and their monthly payments, in every congressional district and state, December 2025.

    Terms
    CC0 1.0, as Data.gov lists it: dedicated to the public domain.
    Credit
    Social Security Administration, Congressional Statistics, December 2025 (2026).

Crime and justice

State governments

  • State legislatures

    The legislatures of the states

    Floor transcripts and journals published by eight legislatures (Connecticut, Hawaii, Illinois, Maine, Nebraska, Nevada, New York and Pennsylvania) and the captions in the video archives other legislatures run themselves, the source of state floor speeches since 2020.

    Used in
    Official profiles, Bulk files
    Terms
    Public records; each state sets its own rules on copyright and reuse.
  • Public-affairs broadcasters and video platforms

    The Florida Channel, TVW, The Ohio Channel, KET, Gavel Alaska (KTOO), Golden Dome VT, and legislature video on Invintus and Vimeo

    Captions of floor sessions in Florida, Washington, Ohio, Kentucky and Alaska; Golden Dome VT’s transcripts of the Vermont Legislature’s video; and captions of Arizona, Georgia and Utah floor video, the source of those states’ floor speeches.

    Used in
    Official profiles, Bulk files
    Terms
    Each source’s own terms; most bar commercial or political use, or redistribution without permission, and Golden Dome VT states no licence. This caption text remains under those terms, not CC BY 4.0.
    Credit
    Courtesy KTOO/Gavel Alaska.
  • State campaign-finance agencies

    The campaign-finance agencies of the 50 states and DC

    Contributions to candidates for state office in every state and DC since 2019, the source of the money totals on state legislators’ profiles and of the donors behind CSIP. Only totals and scores are published.

    Used in
    Official profiles, Bulk files
    Terms
    Public records. Several states bar using contributors’ names for commercial purposes (Kansas, Maine and Minnesota among them); the lab publishes no contributor names.

Open data projects

Research data sets

Platforms and publishers

Election results

Survey fieldwork

Back to topNext: Cite

Citing and reusing the data

Every public file here is free to reuse under CC BY 4.0: copy, share and adapt it, commercially or not, as long as you credit the lab.

Cite the dataset as below in any paper, article or product that uses it, and say which file and the date you downloaded it, since most files are rebuilt as new data arrive.

Citation
Westwood, S. J., & Lelkes, Y. (2024). America’s Political Pulse [Data set]. https://americaspoliticalpulse.com/data

License

Creative Commons Attribution 4.0. The survey toplines and codebooks are covered too. One exception: caption text from broadcasters and video platforms in the state floor speeches file, and campaign-website text in the websites file, remain under their sources’ terms, not CC BY 4.0. Each source lists its own terms.

Programmatic access

The data API serves statements, officials, bills, votes, surveys and violence as JSON, with a free key from your account.

Counting downloads

Each download link passes through the lab’s counter, which adds one to a tally for that file and then sends you to the file. The tally stores nothing else.

Back to top