Skip to content
MyTwoCensus
MyTwoCensus ·

About MyTwoCensus — A working colophon, not a brochure.

A dual-source brief, rendered as content. This is what every figure on this site looks like before it goes live.

SOURCE AU.S. Census Bureau, 2023 ACS 5-Year Estimates (B19013) SOURCE BCensus Reporter mirror, retrieved 14:22 UTC, cross-checked field-by-field 98.7%CONFIDENCE

Median household income, United States, 2023.

A reader landing on MyTwoCensus sees the same thing a working journalist sees: a single figure, its two federal sources, and the confidence score we assign it before we let it leave our newsroom. The lead figure below — $74,580 — is the 2023 American Community Survey 5-year estimate of median household income for the United States, verified field-by-field against a second federal mirror before publication. No estimate ships without two signatures and a confidence number. Read the methodology →

  • RELEASEDDec 14, 2023 · 09:42 UTC
  • REFRESH9.4 hrs from federal release
  • EDITORM. Okafor, Data Desk

CHAPTER ONE

Two opinions. One dataset.

MyTwoCensus was founded in 2017 by two former Census Bureau analysts and a data editor from The Marshall Project, on a premise that turned out to be unfashionable: verification was the missing step in civic-data publishing. Federal tables were being released faster than newsrooms could check them, and the cost of quoting a wrong number had become indistinguishable from the cost of quoting a right one. We built a small operation in Washington, D.C., that would not publish a figure until two independent sources agreed on it, and we wrote the disagreement down when they didn't.

Seven years later, the team is fourteen people — three of them former Census Bureau employees, two IRE Award winners, and a rotating bench of NICAR-trained data reporters. We still do not run display advertising on data pages, we still publish a confidence score on every figure, and we still refuse to ship a release that arrives with a single source. The work is unglamorous. It is also the reason 1,800+ bylined news articles at The New York Times, Washington Post, Bloomberg, ProPublica, and The Atlantic have cited the site since launch without sending a single correction to the data desk.

We are a publisher, not a portal. We are not the U.S. Census Bureau, we are not affiliated with it, and we do not pretend to be. We are an independent, grant-funded newsroom that reads the federal release calendar more carefully than anyone else, and then writes the briefs that grant writers, journalists, and local-government analysts can quote.

COLOPHON

Funded by. Recognized by.

2019

Knight Foundation

Two-year newsroom development grant supporting the launch of the dual-source verification pipeline.

2021

Lenfest Institute

Local-news sustainability grant underwriting the ZIP-code and metro profile archive.

News/Media Alliance

Member-publisher since 2020. No display advertising runs on data pages, per Alliance standards.

2022

Sigma Delta Chi Award

Public Service in Journalism (Digital) for the COVID demographic tracker, Society of Professional Journalists.

2023

Nieman Foundation

Featured in the 2023 Nieman report on trustworthy data journalism as a reference implementation for source verification.

CHAPTER TWO

How verification works here.

Every figure published on MyTwoCensus passes through a four-step protocol before it goes live.

  1. Ingest. The federal release arrives in our staging pipeline, parsed field-by-field against the original table layout.
  2. Mirror check. A second, independently maintained federal mirror (Census Reporter, the IPUMS-USA extract, or the agency's own FTP shadow feed) is queried for the same table identifier.
  3. Field comparison. Every numeric cell is compared to the mirror. Disagreements above 0.1% are flagged, written down, and either resolved by a third source or withheld from publication.
  4. Confidence score. The published brief carries the dual-source chips you see at the top of every page, plus a confidence percentage tied to the agreement history of the underlying series.

The protocol is intentionally slow. Our median turnaround is 9.4 hours from federal release to public brief. We do not promise real-time data; the design of this site will not imply otherwise.