babas03/texas-house-votes-2025

Texas House roll-call votes, 2025

Every recorded floor vote of the Texas House of Representatives in the 2025 regular session (89th Legislature): 3,546 roll calls and 488,969 individual yea or nay positions across 150 seats, 3,454 of the roll calls reconciled against the official House Journal. Plus a ZIP code to House district crosswalk and the counties and towns each district's residents live in, both built from 2020 Census blocks.

Public domain (CC0). Use it for anything, including classroom work, journalism and models, with no permission needed.

It is the data behind rightnleft.com, a free bilingual site that shows Texans how their state representative voted. Its open data page describes the same files, and the method and code are in the public repository github.com/marras0914/plainrecord.

Who made it, and why that matters

Marco Arras, a Texas resident. He donates to the Democratic Party and says so first, because a reader is entitled to know it before trusting a dataset about elected officials. That is also why everything here is a record rather than a judgement: positions as cast, counts as the chamber reported them, and every derived number defined below so anyone can recompute it.

Files

File	Rows	One row per
roll_calls.csv	3,546	recorded House floor vote
positions.csv	488,969	yea or nay cast by a member on a roll call
members.csv	150	member who cast votes in the session
zip_to_district.csv	3,268	ZIP code and House district it touches
district_places.csv	1,166	district and Census place (city, town or named community) its residents live in
district_counties.csv	355	district and county
votes_89R.json		the original file the CSVs unpack, unchanged

Join positions.vote_id to roll_calls.vote_id, and positions.member_id to members.member_id.

roll_calls.csv

Column	Meaning
vote_id	Open States vote id
bill_id	e.g. SB 2. Bill titles are deliberately not included; join them from LegiScan or Open States on this id
category	subject area
vote_type	the stage of the vote. Every row in this release is final
substantive	whether the vote was on the bill rather than a procedural motion. Every row in this release is true: procedural roll calls are dropped when the data is ingested, not carried with a flag
yeas, nays	what the chamber recorded. Use these for the result
named_yeas, named_nays	how many of those could be attributed to a named member. See "Known gaps"
source	journal (reconciled against the House Journal) or scrape (Open States only)
journal_record	the record number in the House Journal
republican_yea_share, democratic_yea_share	each caucus's yea share among its members who voted yea or nay; members with no recorded vote are excluded
valence	Republican share minus Democratic share. +1 is a party-line Republican yea, -1 a party-line Democratic yea, 0 no partisan split. Measured, not judged
eligible	the filter rightnleft.com uses before choosing quiz questions: substantive, at least one yea and one nay, a minority bloc of at least 10 members and at least 5%. Shipped as a flag, not applied

positions.csv

position is yea or nay. No row means no vote was recorded for that member on that roll call: absent, excused, present-not-voting, or not attributable. An absence is not a neutral position.

members.csv

Name, district and party (D or R) as of the session. One member cast votes in 2025 but no longer holds a seat, and the public roster sources do not name them, so their name and district are blank rather than guessed. Their votes are kept.

zip_to_district.csv

ZIP codes (2020 ZCTAs) do not nest inside House districts: 913 of Texas's 1,992 ZIPs span more than one. Each row is one ZIP and one district it touches.

• residents_2020 is the number of 2020 Census residents of the part of that ZIP inside that district. It is blank where the ZIP lies wholly in one district.
• zip_land_pct is the share of the ZIP's land area in that district, kept for audit only. Rank by residents, not land: ranking by land picks a different district first in 108 ZIPs.

Built as an ID join through 2020 Census blocks, which nest inside both, so there are no boundary slivers. The method and its known error are written up in ZIP_TO_DISTRICT.md.

district_places.csv and district_counties.csv

Where each district's 2020 residents live. Districts are on the map the current members were elected on.

• share_of_district: the share of the district's residents living in that place or county.
• share_of_place (places only): the share of that place's residents who are in this district. 1 means the whole town is in the district.
• place_type: city/town for incorporated places, census-designated place for unincorporated communities the Census names.

Places are listed when they hold at least 1% of a district's residents, or lie wholly inside it with at least 1,000 residents. Across all 150 districts the residents sum exactly to the 2020 Texas count of 29,145,505.

Known gaps

• 103 roll calls have fewer named positions than the chamber recorded. 92 of them are every scrape-sourced roll call in the file. The cause is Open States truncating surnames by one character, so a vote is recorded but cannot be attributed. Compare named_yeas and named_nays with yeas and nays. Use the chamber totals for what happened and positions.csv for who did it.
• No bill titles or summaries, on purpose: the project does not paraphrase the record in a data file. rightnleft.com carries plain-language summaries for the 67 votes it uses.
• No vote dates in this release.

Sources

Roll calls from Open States, reconciled against the official Texas House Journal. Census 2020 redistricting block populations, the Census 2022 state legislative block equivalency file, and the 2020 place and ZCTA block assignment files. All public record or public domain; CC0 covers this compilation and the derived fields.

Rebuilding

node scripts/export_dataset_csv.mjs <out-dir> in the public repository writes these CSVs from the files the site publishes. It refuses to write unless the positions total matches the JSON's own count.