Methods and scope
Methods and scope
Who is included
The House records begin with everyone who won a seat in the 2024 general election. Those members stay in the record even if they later left office. The Senate records cover the senators serving at the November 1, 2026 membership snapshot. The two chambers therefore do not represent the same kind of snapshot.
When a special election produced a new House member before the membership snapshot, that member received a separate row beside the original winner. FL-20 and TX-23 were vacant at that snapshot. The history chart leaves these later successors out so that its House comparison follows the same group from year to year.
Member position evidence was searched through November 1, 2026. The membership snapshot, position-evidence cutoff, election-status date, and publication date are maintained separately.
How positions are coded
- Support: affirmatively promotes data center or related AI infrastructure expansion without making development contingent on a material prerequisite.
- Conditional support: supports development while requiring meaningful safeguards, planning, cost allocation, or community protections as conditions.
- Regulation: prioritizes binding safeguards, cost-allocation requirements, disclosure rules, siting standards, or resource protections without affirmatively supporting development in the coded action.
- Opposition: affirmatively rejects a project or covered buildout, or calls for a halt, moratorium, or ban.
- No position observed: the research found no codable position by the relevant cutoff, or none was observed in the selected calendar year. It is not a neutral policy position.
A concern does not make support conditional unless it sets a real condition on development. Regulation is also different from opposition. A vote for H.R. 9340, for example, is coded as regulation because the bill establishes a cost-allocation rule for grid upgrades. The vote does not call for stopping data center construction, but neither does it by itself express support for more development.
The category belongs to the action, not to the member once and for all. Someone may oppose a particular project and also support rules governing data centers more broadly. Both actions stay in the record.
What “latest documented action” means
The main table uses the most recent national or mixed-scope action found by the cutoff. Choosing a year changes the table to the last such action found during that year. The position filter follows the same rule.
Open a member’s name to see the full event history. This is where earlier positions and state or local actions appear. “Latest” refers only to timing; it does not turn the most recent action into a final judgment about the member’s views.
If both support and opposition appear, the record is flagged for closer review. Summer L. Lee, for example, has both categories in her national record, but the actions concern different policy instruments. The tracker treats this as a possible change, not proof of a reversal.
Geographic scope matters as well. Gabe Vasquez has national evidence of conditional support for development with safeguards and a separate opposition event concerning the Socorro County proposal. The local event remains visible without replacing the national label in the table.
Chamber-wide votes and evidence basis
Chamber-wide votes are valid documented actions, but they provide thinner evidence about an individual member’s broader views than a statement, sponsorship, letter, or other member-specific action. The tracker therefore keeps the substantive code and the evidence basis separate.
The September 16 House vote on H.R. 9340 and the September 30 Senate cloture vote are marked as chamber-wide votes. A member who has no other qualifying national or mixed-scope record in the selected time view is labeled “Roll-call vote only · H.R. 9340.” “Additional member-specific evidence” means that at least one statement, sponsorship, signed letter, committee action, or similar attributable record is also present. The Evidence basis filter lets readers separate these groups.
H.R. 9340 votes remain coded as regulation because the bill would require large data center loads to bear specified grid-upgrade costs. This describes the policy instrument in the vote. It does not mean that every member voting for the bill has a generally restrictive position on data center development. Statements made by the bill’s sponsors or committee leaders are not attributed to every member who voted for it.
Reading the history chart
The chart starts each member at “No position observed.” When a national action is found, its category becomes the latest action for that year. If no later action appears, the chart carries the category forward. This keeps the annual comparison continuous, but it is not evidence that the member repeated or reaffirmed the view.
The history switch can remove the two H.R. 9340 chamber-wide votes from the chart. This alternate view retains bill sponsorships, public statements, signed letters, committee actions, and other evidence. It is a sensitivity check, not a recoding of the votes.
The chamber tabs change the chart. The search box and the state, party, position, evidence-basis, and year filters apply only to the table below it.
Sources and quotations
Every quoted passage is checked against a complete PDF in the private research archive. The archived file is either the publisher’s original PDF or a searchable PDF containing the complete main text of the cited webpage. Research summaries, selected excerpt cards, search snippets, and reconstructed wording do not qualify as quotations.
Each quotation is copied from that PDF and carries a PDF page or page-range locator. When a quotation joins nonadjacent passages, the omitted text is marked with […]; each displayed fragment must appear in the same order in the archived file. Audio or video is used only when a complete attributable written transcript has also been archived as a PDF. There is no transcript exception for a short clip or a researcher-created excerpt.
Expanding a record shows the date, category, action, summary, original public source, and the exact passage used to make the coding decision. Structured official records, such as bill-status XML, are shown in a shorter readable form while the underlying official record remains preserved.
The categories are interpretations of public evidence. Readers can compare the summary with the quoted passage and original source when a classification is not straightforward. Archived verification copies are kept in the research repository and are not published with the site.
2026 election data
The election page records whether a tracked member appears on the November 3, 2026 general-election ballot and which office that person is seeking. Campaigns that ended before November remain in the member’s history.
Federal candidates come from the House Clerk’s October 8 nominee list. Gubernatorial candidates are checked against state or local election records. “On federal ballot” and “On state ballot” identify verified nominees. “No election scheduled” applies to Senate seats that are not up in 2026 and have no special election. Everyone else is listed as not on the November ballot, with the reason shown in the table.
The election status was checked through November 1, 2026. It describes candidacy before the election and does not report election results. It also has no effect on position coding.
2026 House race comparison
The House race page starts with the Clerk of the House’s October 8 list of general- election nominees. It covers all 435 districts. In the 411 races with exactly one Democratic and one Republican nominee, those two candidates form the comparison. The remaining 24 races have an open primary, a same-party contest, an uncontested major-party candidate, or another ballot structure. All selected candidates in those races remain visible rather than being forced into a Democratic–Republican pair. Altogether, the page follows 892 candidates.
For the 361 candidates who already appear in the congressional tracker, the race page reuses the member evidence through November 1. The other 531 candidates were searched separately for public actions and statements from January 1, 2024 through November 1, 2026. Searches covered campaign and official websites, news coverage, public records, and policy documents. A missing record means that no codable evidence was found in this search; it does not establish that the candidate has no view. Campaign websites were not separately verified in every case.
For every retained event in the separately researched House-candidate archive, the supporting passage was rechecked against a complete archived source PDF. Records without a complete source or, for audiovisual material, a complete attributable transcript are excluded from the active evidence layer. The archived quotation and the four-category verdict were then reviewed together before the race comparison was rebuilt.
The member and new-candidate records currently share a November 1 evidence cutoff, so the page shows one time view. The underlying comparison keeps separate common-cutoff and latest-available fields so a future incremental update can mark any temporary timing difference rather than silently mixing evidence windows.
Once the congressional and candidate records share the same cutoff, these two views are identical. The page then removes the time switch and evidence-timing column and shows the shared date once as the evidence cutoff.
The displayed category is the candidate’s latest national or mixed-scope action in the selected view. When no such action exists but state or local evidence does, the page shows that action and labels its scope. Opening a race reveals every coded event, so the comparison category should not be read as a complete account of the candidate’s record or as proof of a change in position.
A district is marked Open seat when no candidate in that race was serving in the 119th House at the tracker cutoff. Opening the race identifies the relevant current or departing member, explains why no incumbent appears on the district’s November ballot, and links to the underlying election or vacancy record. This covers members who did not seek reelection, lost a primary, ran for another office or district, and seats that were vacant at the cutoff. A separate Two incumbents marker identifies the one race where two sitting House members appear on the same ballot.
The evidence switch provides a sensitivity check for H.R. 9340. All documented actions includes the chamber-wide House roll call for incumbent members. Exclude chamber-wide votes removes that vote from the displayed comparison while retaining candidate-specific statements, sponsorships, proposals, and other actions. Neither view changes the underlying event record.
The page is a descriptive pre-election record, not an assessment of candidate quality, electoral competitiveness, or likely outcomes. The ballot and evidence files should be refreshed before any analysis of November results.
2026 Senate and governor race comparisons
The Senate comparison covers all 35 seats on the November ballot. Thirty-two races use one Democratic and one Republican nominee. Alaska, Idaho, and Nebraska have different ballot fields, so every candidate selected in the verified federal roster remains visible and no forced two-candidate comparison is calculated.
The governor comparison covers all 36 gubernatorial elections. Thirty-five races use the Democratic and Republican nominees; Alaska retains all four finalists. State candidate lists and election-agency records provide the ballot evidence.
Candidates already covered by the congressional tracker are linked to that existing evidence history. The remaining candidates were searched separately for attributable written positions from January 1, 2024 through November 1, 2026. Search snippets and audio or video without a transcript or attributable written excerpt do not qualify. Every coded event has an exact supporting passage and a source link.
The compact label is the latest dated position by the common cutoff. It does not erase earlier actions: opening a race shows the full coded history in reverse chronological order. A candidate can therefore have more than one category in the record. The comparison is descriptive and does not infer that a later regulatory action supersedes an earlier statement of support or opposition.
“No position observed” is used when the completed search found no qualifying written evidence. It is also used when a reviewed source mentioned data centers without establishing support, conditional support, regulation, or opposition. Those two research outcomes remain separately identified in the underlying data.
Current release
Release 0.7.1 was compiled on November 4, 2026. It brings together the House and Senate position records, the 2026 member-status view, and separate House, Senate, and governor race comparisons. Position evidence and election status in this release run through November 1, 2026.