Draft. Last updated November 12, 2026.

Methods

How we count votes, where every number comes from, and what our results can and can’t tell you. We follow the disclosure list of the American Association for Public Opinion Research’s Transparency Initiative.

The short version

• This is an opt-in sample, not a scientific poll. Anyone can join, so the people voting here aren’t a random slice of Americans.
• Every result shows how many votes it’s based on, split three ways: raw (every account), verified (accounts with a checked phone number) and weighted (adjusted to match the U.S. population).
• Weighted results aren’t available yet. Weighting needs basic facts about voters, such as age group, which we don’t collect. Until we do, we say so instead of guessing.
• Every fact has an official source and an “updated” date. Bills come from Congress.gov, rules from the Federal Register, money from the Treasury and USAspending.gov, and tax law from the IRS.

Who runs BuckBallot and how it’s paid for

BuckBallot is run by [legal name of owner or company], [city], Connecticut. It is not part of any government, party, campaign or advocacy group. It has no sponsors, no paid placement and no sponsored questions. It’s currently paid for by its owner. Questions: buckballot@gmail.com.

Exact question wording

• Budget splits: People divide 100 bucks across federal or state spending areas. Giving one area more means giving another less.
• Bills: “Do you support this bill?” The wording stays the same as a bill moves through Congress. Once it becomes law: “Do you support this law?” For resolutions: “Do you support this resolution?” (or “this budget resolution” for the yearly budget). Answers: Support or Oppose.
• Agency rules: “Do you support this proposed rule?” Answers: Support or Oppose.
• Tax questions: “Should this stay in the tax code as it is?” Answers: Keep it as is, or Change it. “Change it” includes expanding, limiting or ending the provision.

The order of the two answers switches from question to question and person to person, so neither answer always comes first. Bill summaries are written by the nonpartisan Congressional Research Service. Rule summaries are the agency’s own words and are labeled that way.

Bill titles

Every bill has an official title: the long, formal one, such as “To amend the Internal Revenue Code of 1986 to…”. Many bills also have a short title chosen by the bill’s sponsor, and some short titles read like slogans. So we use one rule for every bill, whoever wrote it:

• The card’s main title is the bill’s official title (as introduced), copied from Congress.gov.
• If the bill has a short title, it’s shown underneath, labeled “Short title.”
• We never rewrite, shorten or pick between titles bill by bill.

[Owner: this starts with the next daily data update. Until then each card shows the single title Congress.gov lists first, which is the short title when a bill has one.]

Who can vote, and how often

• Anyone 18 or older with a free BuckBallot account and a display name. [Lawyer: confirm the age requirement.]
• A display name is a handle the person chooses. It doesn’t have to be a real name, and we don’t check it. Results count accounts, not verified identities.
• One vote per account per question. You can change or remove your vote at any time while voting is open.
• Home states are self-declared. We don’t check them.
• Votes on bills, rules and tax questions are private. Only counts are public. If you add a note, your vote and note become a public “take” under your display name.

How people find us (the sample)

BuckBallot is an opt-in, volunteer sample. People choose to sign up, usually because a friend shared a link or they found the site themselves. It is not drawn at random from the population, so results describe the people who took part, not all Americans. Opt-in samples can include people who misreport who they are, and research by the Pew Research Center finds no single method fully fixes that.

Verification

“Verified” counts accounts that confirmed a phone number by text message. Each phone number can belong to only one account. Phone checks are not switched on yet, so verified counts start at zero. Verified results appear once at least 200 verified people have answered a question.

Weighting

Weighting adjusts results so groups that are over- or under-represented among voters count in proportion to the U.S. population, using benchmarks such as the Census Bureau’s American Community Survey. We don’t weight yet, because we don’t collect the needed facts (such as age group, education or state). If we add an optional, private question about them, this page will say exactly which benchmarks, years and methods we use. Until then, every result reads “Weighted: not available yet.”

How we got each number

• Raw counts every vote from accounts in good standing. Votes from accounts the owner removed for abuse are left out.
• Percentages appear once a question has at least 10 votes. Below that, only the number of votes is shown.
• Likely range. Not shown yet. Once weighted results exist and a question has at least 200 verified votes, each weighted result will show a credibility interval (never called a margin of error), and this page will explain how it’s calculated.
• Hourly updates, not live counts. Results are recalculated about once an hour, which makes it much harder to work out one person’s vote by watching a total change. On a question with very few voters it can’t rule that out completely. Below 10 votes, the yes/no split is withheld entirely; below 200 verified votes, the verified split is withheld. [Owner: until database update 006 is run, the server still sends the split from 5 votes and the website holds it back until 10.]
• No state-by-state results yet. We’ll add them only when there are enough verified voters per state that a small group can’t reveal how one person voted.
• Fast-moving results. When more than half of a question’s votes are new in the last 24 hours (and there are at least 20 of them), the result says so. We watch for coordinated voting but don’t publish our exact checks.
• Dates. Voting on a bill stays open while it’s listed. Voting on a proposed rule closes when the agency’s comment period ends. You can remove your public note at any time, even after voting closes.

The research pool

People can choose to add a rounded copy of their tax receipt to a research pool (the privacy policy lists what it holds). Nothing from the pool is shown on the site today. Our rules for it:

• No town-level figure for fewer than 20 people. A town appears in any table or export only when at least 20 people in the same state named it. The database enforces this itself, so an export can’t skip the rule. [Owner: this takes effect when database update 006 is run.]
• No subtraction trick. If the people left out of every town group number fewer than 20, the smallest town shown is held back too, so nobody can work out a small group by subtracting.
• Town is typed by hand. We ignore capital letters, punctuation and extra spaces when grouping (“St. Louis” and “st louis” are one town). We don’t guess at misspellings: a misspelled town counts as its own small group and is held back by the 20-person rule. Receipts with no town are never in a town figure.
• No averages for the whole pool below 20 people, and never a single person’s row.

How bills, rules and tax questions are picked

Each morning an automatic job collects:

• Bills from the Congress.gov API with action in the last 60 days, in the tax or public finance policy areas, or about appropriations, tariffs, rescissions, the debt limit or budget reconciliation.
• Proposed rules from the Federal Register that are open for comment, from the agencies that handle taxes, tariffs, borrowing and the budget, plus rules the White House budget office lists as significant.
• Tax questions about the 12 largest income tax expenditures in the Treasury’s official ranking, plus the three largest refundable credits (child tax credit, earned income tax credit and premium tax credit). For those three, we count both the lower taxes owed and the payments, the same way for each.

How they open for votes. We don’t hand-pick. On November 12, 2026, every bill, proposed rule and tax question the job had collected was opened for votes at once, without a case-by-case review. Each one is shown in official words, not ours: bills by their titles from Congress.gov and, where one exists, the Congressional Research Service summary; rules by the agency’s own title and summary. Tax questions use the Treasury’s name for each provision, with a short plain-English description written by BuckBallot and the Treasury or IRS source beside it.

The site owner can hide an item, for example one that turns out not to be about taxes or public money. Hiding closes voting on it; votes already cast are kept. The site has no tool for editing an item’s title, summary or results.

Items the job collects after that date don’t open by themselves yet. For now the owner opens them by hand. [Owner: update this paragraph if new items start opening automatically.]

Where the facts come from

• Bills: Congress.gov API (Library of Congress). Summaries by the Congressional Research Service. Cost estimates by the Congressional Budget Office. Roll-call totals from the House Clerk and the Senate.
• Rules: Federal Register API (National Archives). Comment counts from Regulations.gov as reported by the Federal Register, with the date they were checked.
• Federal money: Monthly Treasury Statement, Debt to the Penny, Daily Treasury Statement and average interest rates (U.S. Treasury, Fiscal Data).
• Grants by state: USAspending.gov. These are obligations (money committed), not payments. Social Security and other direct payments to people aren’t grants; Medicare payments are often filed under the state where a claims processor sits. Neither is included.
• Tax law: IRS releases, Revenue Procedure 2025-32, Publication 505 and Form 1040-ES. Tax expenditures: Treasury Office of Tax Analysis.

Shares “per $100” are calculated by BuckBallot from the official totals and are labeled that way. When official documents disagree, we show both and say so. When a source is missing or broken, we say that instead of filling the gap.

Limitations

• Results describe BuckBallot users, who are likely more interested in politics and budgets than most people.
• Without weighting, groups that use the site more count for more.
• Self-declared home states may be wrong.
• Some government data arrives late: USAspending agency data lags by a month or more, and defense contract data by 90 days.
• A BuckBallot vote is not an official vote or an official public comment.

Corrections

If you find a wrong number or unfair wording, email buckballot@gmail.com. We fix errors and note the fix here.