WEIGHT A
40
Source agreement.
How closely the second source matches the federal release, in standard error units. Exact match within published MOE scores 100; one-sigma drift scores 70; anything wider triggers manual review.
THE EDITORIAL STANDARD
How we verify every Census and ACS number before it goes live.
This is the working notebook behind every figure MyTwoCensus publishes. The protocol below is what allows a journalist to quote a number from us, and what allows a policy researcher to cite us in a footnote without spending three hours re-checking it against the Census Bureau's own tables. Read it once. Cite it as often as you need to.
CHAPTER ONE
The five steps of the dual-source verification protocol.
Every figure we publish moves through the same five steps, in order, regardless of whether the original release is a Census Bureau redistricting file, an ACS 5-year update, or a one-off table from the Bureau of Labor Statistics. The protocol is the product.
-
I
STEP ONE
Capture the raw release.
Within 15 minutes of a federal release, our ingest pipeline downloads the source tables directly from census.gov, the Bureau of Labor Statistics, or the relevant agency's API. We hash the file and store the original — the byte-for-byte federal copy is the canonical artifact and is never modified.
-
II
STEP TWO
Cross-check against a second source.
Every figure is compared against a second federal or peer-reviewed source: the prior vintage, an independent tabulation, or a state demographer's office. We do not consider a number verified until the two sources agree to within the published margin of error. Disagreements trigger a manual review before publication.
-
III
STEP THREE
Score confidence.
Each verified figure receives a confidence score on a 0–100 scale, derived from four weighted components described in Chapter Two. The score is published next to the figure, not buried in a methodology PDF. A score below 70 is flagged inline as "preliminary" and never appears without an editor's note.
-
IV
STEP FOUR
Editorial sign-off.
A named editor — never an anonymous process — reviews the figure, the second source, the confidence score, and any caveats. The editor's initials are attached to the published brief in a small colophon at the foot of the page, so the chain of custody is auditable.
NOTE The protocol is version-controlled in our public methodology repository. Changes to the five steps are announced in the Census Brief and dated in the changelog. No silent updates.
CHAPTER TWO
The confidence score, in four weighted parts.
The score is not a vibe. It is the arithmetic mean of four component weights, each on a 0–100 scale, then re-bounded. A methodologist can audit the math in under a minute.
WEIGHT B
25
Recency of release.
Time elapsed since the federal source was published, normalized against the release's known update cadence. A 2020 Census table ten years later still scores 95 on this component.
WEIGHT C
20
Coverage completeness.
Share of the relevant universe (national, state, county, ZCTA, tract) for which we hold a verified cross-checked figure. Full coverage is 100; a single gap drops the component proportional to its weight in the brief.
WEIGHT D
15
Historical stability.
Whether the figure has held within expected bounds across the prior three vintages. A sudden, unexplained move reduces this component even when both current sources agree.
CHAPTER THREE
What we will and will not publish.
These are binding editorial standards, not marketing reassurance. They are the rules a researcher can hold us to when defending a quoted figure in a footnote, a courtroom filing, or a city council packet.
We will publish.
- 01
Census Bureau and American Community Survey figures that have cleared the dual-source protocol and carry a confidence score of 70 or higher.
- 02
Corrections to our own previously published figures, issued within 24 hours of a verified error and dated in the changelog. We do not quietly overwrite a bad number.
- 03
Methodology notes, source pairs, and the working components of every confidence score, linked from the figure they describe.
- 04
The names of the editors who signed off on each brief. An anonymous pipeline is not a sufficient chain of custody for a quoted figure.
We will not publish.
- N1
Modeled or estimated data presented as an official Census figure. If a number is a projection, we say so in the headline.
- N2
Claims of real-time refresh. Our fastest legitimate turnaround is hours, not seconds, and the design of this site will never imply otherwise.
- N3
Endorsements, bylines, or quote attributions from specific named journalists or outlets that have not actually cited our work. The verified citation count is in our about page and grows only when it grows.
- N4
Coverage of international censuses beyond the United States, Canada, and the United Kingdom. We do not extend our verification protocol where we cannot staff it.
- N5
Display advertising on data pages. The site is funded by the Knight Foundation, the Lenfest Institute, and the News/Media Alliance — not by an ad network with a stake in what you click.
Questions about a specific figure, or a correction to send our way? Write to the editors at [email protected]. The inbox is read by a human, every weekday, and replies are signed.