How to read a voter file

Each state maintains a statewide voter registration database under federal law, and most states will sell you a copy, or a slice of it, for political use. The seller is usually the Secretary of State or a state election board; in some states the practical route is the county clerk or registrar. NCSL's survey of state list-access rules shows that most states include name, address, party affiliation where the state records one, voting history, and year of birth, and that nearly every state restricts commercial use.

The file arrives as one or more delimited or fixed-width text files, one row per registered voter. Texas, for example, ships its public information file as zipped fixed-width text with a separate voting-history file in which each row is one voter in one election, according to the Secretary of State's request form and record layout. Oregon charges $500 for the statewide list and $25 plus 2.5 cents per 100 records for a smaller extract, per the Oregon Secretary of State (fees checked 2026-09-09).

The file is a snapshot. It is accurate as of the day it was pulled and starts going stale the next morning as people move, register, and die. Note the pull date on every file you receive and plan to refresh before your final get-out-the-vote push.

Layouts differ by state, but the same core fields show up nearly everywhere. The Texas public layout is a good model: county code, precinct, a permanent voter ID (VUID), name parts, gender, date of birth, the residence address broken into house number, street, unit, city and ZIP, a mailing address, registration status, and an effective date of registration. The separate history file adds an election date, election type, the party primary the voter chose, and the voting method.

Party looks different depending on the state. In closed-primary states the file carries a party field the voter selected when registering. In states without party registration, such as Texas, the only party signal is which primary the voter participated in, which lives in the history file rather than the registration record. Treat that as a behavior, not an identity; a voter who pulled one party's ballot in a single primary may have been voting in a local runoff, not declaring loyalty.

Status codes matter more than most first-time candidates realize. Texas uses V for active, S for suspense, and C for cancelled. Other states use active and inactive. An inactive voter is still registered and can still vote; the flag means the county has reason to think the address is stale. The MIT Election Data and Science Lab notes that inactive voters are not barred from casting a ballot because they skipped recent elections.

Opinions are not in the file. Nothing in the raw registration record tells you whether a voter supports you, cares about the pool bond, or has ever heard of the school board. Every support score you see from a vendor is a model built on top of the file, not a field in it.

Phone numbers are missing or spotty in most states. Registration forms usually make the phone optional, and several states withhold it from the public file entirely. NCSL's survey notes that Vermont and West Virginia keep phone numbers confidential and Nebraska lets voters opt out of releasing them. Email addresses are rarer still. If you need to call or text, you will be buying appended numbers from a vendor, and those numbers do not carry consent to be texted; see political texting rules.

Full dates of birth are often redacted. Some states release only the year, and some states redact the month and day. Social Security numbers and driver's license numbers are confidential everywhere. The file also does not tell you how anyone voted; it records participation, never choices.

• Not in the file: candidate preference, issue positions, or any survey answer.
• Usually not in the file: cell phone, email, occupation, income, ethnicity (some states carry a surname flag), homeownership.
• Never in the file: how a person voted, Social Security number, driver's license number.

Commercial data vendors buy the state files, standardize them across all 50 states, and append consumer and modeled data. L2 advertises 217 million voter records with more than 600 consumer, voter and behavioral attributes and sells to campaigns on both sides. i360 describes 300 million-plus consumer and voter profiles with 1,800-plus data points per individual and serves right-of-center organizations. Data Trust is the Republican ecosystem's data exchange, with a national file of more than 300 million individuals, live absentee and early-vote information, and support for over 5,000 Republican campaigns in 2024. TargetSmart has served Democratic campaigns and progressive organizations since 2006. Aristotle has sold voter and consumer data alongside its Campaign Manager software since 1983.

What you are paying for is three things. First, hygiene: addresses standardized to postal format, movers updated against change-of-address data, deceased records flagged. Second, appends: phone numbers, emails, household composition, homeownership, estimated income. Third, models: a turnout score, a partisanship score, and issue scores estimated from the vendor's consumer data and survey panels.

Every vendor licenses data per client. Your campaign's copy is yours to use for your race and not to share with another candidate or upload into a shared pool. Read the license before you hand a file to a consultant or a slate partner.

You do not need a vendor to build a useful turnout score. Pick the four most comparable past elections for your race. For a November city council race in an even year, use the last four November generals. For an off-cycle May election, use the last four May elections, even if that reaches back eight years. Count how many of those four each voter participated in. The result is a 0-to-4 score, and the tiers it creates, 4-of-4, 3-of-4, 2-of-4, 1-of-4, are the language every field operation speaks. How many votes to win a local election walks through how to use those tiers.

Adjust for eligibility. A voter who registered in 2024 could not have voted in 2020 or 2022. Score them on the elections they were eligible for and flag them separately; a 2-of-2 new registrant is a likely voter, not a 2-of-4 swing voter. The registration date field is what makes this possible.

Weight recency if you want more resolution. Counting the most recent election double, so the score runs 0 to 5, separates voters who are drifting away from voters who are showing up more often. Keep it simple enough that a volunteer can explain it.

Movers are the biggest problem. The Census Bureau's mobility estimates put the share of Americans who moved in a year at about 14% in 2019 and roughly 13% in 2021. That means a file pulled 12 months ago has around one in eight addresses wrong. Vendors fix some of this with postal change-of-address data; you fix the rest at the door by marking moved records so they drop out of the next list.

Inactive status and list maintenance follow federal rules. Under the National Voter Registration Act, 52 U.S.C. 20507, a state that gets a signal a voter moved must send a forwardable confirmation notice; the voter can be removed only after failing to respond and then not voting through the second federal general election after the notice. Systematic removal programs must finish 90 days before a federal election. The U.S. Election Assistance Commission's 2024 survey counted more than 211 million active registered voters, 86.6% of the citizen voting-age population, and nearly 40 million confirmation notices sent during the 2024 cycle.

Duplicates come from re-registrations at new addresses, name changes, and data-entry variants. Match on the voter ID, never on name plus birth date; two people named Maria Garcia born in the same year is not a rare event in a county of 500,000. When you import a vendor file on top of a state file, keep the newest record by data date and preserve the older one's contact history instead of overwriting it.

Cancelled and deceased records leak in. The MIT lab estimates that roughly 21 million registrations were removed nationally between 2022 and 2024, mostly for moving, dying, or failing to respond to a notice. A file that predates a county's maintenance run will still list many of them. Filter on status before you cut turf.

• Wrong address: about one in eight records after a year. Mark moved at the door.
• Inactive status: still eligible to vote. Include in GOTV only if you have a reason to believe the address is right.
• Duplicates: match on voter ID. Keep the newest record by data date.
• Cancelled or deceased: filter out before cutting turf. Never knock those doors.
• Blank precinct or split precinct: check against the county's current precinct map after redistricting.

Every campaign tool that accepts a spreadsheet does the same job: it reads your header row and tries to match each column to a canonical field like first name, residence ZIP, or registration date. Clean vendor exports map automatically because their headers are predictable. State files are harder. Texas ships positional fixed-width text with no header at all, so the layout document is the only way to know that columns 13 through 22 are the VUID.

The traps are consistent across states. Leading zeros in precinct codes and voter IDs get stripped when someone opens the file in a spreadsheet first, so open text files in a text-aware tool. Dates arrive as YYYYMMDD, MM/DD/YYYY, or Excel serial numbers. Names come combined in one column or split into four. Addresses come as one line or as nine parts. Vote history arrives either wide, one column per election, or long, one row per voter per election, and the two need different handling.

Do a dry run before the real import. Check the row count against what the state told you, spot-check ten voters against the source, and confirm that the voter ID column is unique. If the tool lets you save the mapping, save it; your next refresh from the same source should take minutes, not an afternoon.

• Preserve leading zeros. Import as text, not numbers.
• Confirm the date format on a record whose day and month differ, such as 03/12 versus 12/03.
• Map vote history to a per-election structure so you can score any set of elections later.
• Keep the raw file untouched. Every transform should be reproducible from it.

Start with your state's chief election official. The state-by-state guides list who sells the file, who may buy it, what it costs, and whether a county-level file is the practical route for a local race. In many states the county clerk will sell you just your district for far less than the statewide file. Political parties often get a free copy and will share it with their candidates, which is the cheapest route of all if you are running with a party label.

Then decide whether you need a vendor. For a council race with 12,000 voters and no phone program, the state file plus your own turnout score is enough. If you plan to text or call, or if you want a partisanship model in a nonpartisan race, budget for a vendor file for just your district. Whatever you buy, remember that the door and the phone produce the only opinion data you will ever have, so build the process that writes those answers back to the file. Door knocking for local campaigns covers what to record.

1. 1.NCSL, Access to and Use of Voter Registration Lists (fields, access, permitted uses by state)
2. 2.Texas Secretary of State, public information request form and voter data file record layout
3. 3.Oregon Secretary of State, Request for Voter List (fees)
4. 4.52 U.S.C. 20507, National Voter Registration Act list maintenance requirements
5. 5.U.S. Election Assistance Commission, 2024 Election Administration and Voting Survey release (registration and list maintenance figures)
6. 6.MIT Election Data and Science Lab, Voter List Maintenance (inactive status, removals 2022-2024)
7. 7.U.S. Census Bureau, share of Americans who moved, 2019 and 2021
8. 8.L2, voter and consumer data products
9. 9.i360, data platform description
10. 10.Data Trust, voter file and early-vote data description
11. 11.TargetSmart, data solutions description
12. 12.Aristotle, voter and consumer data and Campaign Manager