The Middle Column

No side. Just the record.

The Method

How a claim becomes a brief, and how to challenge one. This page does not change quietly; any change is versioned and noted here.

1What we check

Claims that real people are actually repeating: in comment threads, group chats, talk radio, and kitchen-table arguments. Not politicians’ greatest hits, and not strawmen. A topic qualifies only if the claim is checkable, primary data exists to check it, and both sides are measurably missing something. If no primary data exists, the brief is about that absence, and says so.

2The source ladder

  • Tier A1 (weight 1.0): the primary record itself: audited statements, statute text, court rulings, raw federal datasets. Can support a “Verified” rating alone.
  • Tier A2 (0.9): official synthesis by the record-keeping body: CBO, BLS, NAEP, GAO.
  • Tier B (0.7): independent expert synthesis with published methods: KFF, Tax Foundation, Yale Budget Lab, RAND. Two independent Tier B sources that agree can support Verified on a figure neither originated.
  • Tier C (0.4): quality reporting that names its underlying source. We cite the underlying source, not the article.
  • Tier D (0.0): never cited: advocacy shops without published methods, anonymous posts, screenshots of screenshots.

The load-bearing rule: every number a brief stands on rests on one Tier A source, or two independent Tier B sources. Anything else is labeled Unverified, or it does not run. A source that publishes a wrong figure and does not correct it drops a tier.

3Confidence ratings

Every load-bearing figure carries one: Verified Corroborated Inference Disputed Unverified. Model estimates are labeled as estimates, with the model named. Arithmetic is checked in two independent passes before publication.

4Show your work

Every load-bearing number carries a source chip naming the document. Every brief has a Receipt Pack: the source documents or links, the dataset extracts, and the calculation sheet. Where a pack cannot be assembled, a Request the Receipts button takes the request, and we answer with the underlying data, or a plain explanation of why it cannot be shared, within two business days.

5Claims are rephrased, not quoted

Every claim in a brief is rephrased accurately in the form it usually takes. Claims circulate in dozens of wordings, and no single sentence belongs to anyone, so no brief presents any claim in one person’s exact words, and no claim is attached to a name or to wording distinctive enough to identify the person who said it. Each claim is rephrased in full, with its substance intact, and both claims in a brief carry the same provenance note, word for word. Quotation marks appear on this site only where the exact text of a law, a document, or a dataset label is itself the evidence. A reader here can trust that difference, because we keep it even when exact words would have made a better headline.

6How to challenge us

One sentence governs this site: if you believe a figure here is wrong, bring a source of equal or better tier and we will re-run it in public. There are exactly three doors in, and a comments section is not one of them:

Values are not banned here; they are housed honestly. Discussion of what should be done, given the facts, belongs to the Common Ground section of each brief’s resolution path, labeled as values, never dressed as data.

7Debatable / Not Debatable

Cannot be debated, only verified or refuted: what a document says; what a dataset recorded; arithmetic; dates; whether a vote happened and how it came out; the text of a law; a measurement taken with a stated method.

Can be debated: what should be done; which value matters more; how to weigh a trade-off; causes, within what the evidence allows; predictions, labeled as predictions.

“I don’t believe that unemployment number” is not a counterargument. Belief is not a unit of measurement.