Data literacy · Primary sources
How Do I Verify a Demographic Figure?
Establish three things: which dataset the figure came from, which vintage of that dataset, and which table within it. For US population, income, race, education, and housing, the dataset is almost always the Census Bureau's American Community Survey, and once you have the table reference you can reproduce the number at data.census.gov in about a minute. A statistic published without a dataset, a vintage, and a reference is not yet a checkable figure.
Four questions that settle almost any figure
- Which dataset? US demographics come from a Census Bureau program in nearly every case. Country-level population comes from the UN; country-level economics from the World Bank. If a figure cannot name its dataset, treat it as unverified rather than wrong.
- Which vintage? The ACS 5-Year file is reissued every December. The current release covers 2019–2023. Two sources can both be accurate and still disagree because they are reporting different vintages.
- Which geography? “Chicago” can mean the city, the county, the metro area, or the combined statistical area, and the population differs by millions between them. Census geography levels are precise; casual writing about them is not.
- Measured, reported, or perceived? A count of residents, a count of crimes reported to police, and a survey of how safe people feel are three different quantities. All three are legitimate. Substituting one for another is not.
The references you will need most
Searching a table number plus a place name at data.census.gov returns the estimate and its margin of error. The margin is worth reading for small geographies, where a five-year sample can be thin.
| Figure | Published by | Look up |
|---|---|---|
| Total population (state, county, city, ZIP) | Census Bureau ACS 5-Year | Table B01003 |
| Median household income | Census Bureau ACS 5-Year | Table B19013 |
| Race and ethnicity | Census Bureau ACS 5-Year | Table B02001 |
| Educational attainment | Census Bureau ACS 5-Year | Table B15003 |
| Owner vs. renter, median home value | Census Bureau ACS 5-Year | Tables B25003, B25077 |
| Poverty status | Census Bureau ACS 5-Year | Table B17001 |
| Apportionment and redistricting counts | Decennial census | P.L. 94-171 file |
| World population by country | UN World Population Prospects | 2024 Revision |
| GDP, life expectancy, fertility | World Bank Open Data | Indicator code |
| Reported crime, United States | FBI Crime Data Explorer | Agency-level UCR |
| Reported crime, international | UNODC dataUNODC | Homicide and crime series |
| Cost of living | Bureau of Economic Analysis | Regional Price Parities |
Vintage matters more than people expect
A demographic profile is a photograph, not a live feed. The Census Bureau publishes a fresh ACS 5-Year file each December with a lag of roughly fifteen months, so the newest available picture of a small city is already a year or two old — and a figure copied from an older release can be a decade old without looking it. Over ten years, a fast-growing county can add a fifth to its population, and median household income in nominal dollars can move enough to change the answer to almost any question you were asking.
This is the single most common reason two reputable sources disagree, and it is also the easiest to check: if a page does not say which vintage it is reporting, that is the first thing to establish.
Measured, reported, and perceived
Crime is where this distinction does the most damage. The FBI Crime Data Explorer publishes offenses that agencies reported, and UNODC dataUNODC does the same across countries. Separately, several widely circulated crime indices are built from opinion surveys asking residents how safe they feel. Perceived safety is a real and useful measurement, but it tracks news coverage and visible disorder as much as offense counts, so a city can rank very differently on the two. Check which one a ranking used before citing it.
Cost of living has a milder version of the same problem. The federal index is BEA Regional Price Parities, which the federal government uses for its own regional adjustments. Privately assembled composite indices weight rent, groceries, healthcare, and transportation differently, which is a defensible thing to do but means two indices can rank the same state differently. Ours uses BEA.
Checking the numbers on this site
Every entity page shows its dataset and vintage in a badge near the title — currently Census ACS 5-Year 2023, fetched April 12, 2026. Where a figure comes from the World Bank or the BEA instead, the badge names that source.
The underlying rows are downloadable as free CSV files on our data downloads page — no email gate, CC0 license, same JSON that drives every page on the site — so any figure here can be re-derived rather than taken on trust. Our methodology page lists every dataset and how often each is refreshed.
Frequently Asked Questions
Find out which dataset it came from, then look the figure up in that dataset yourself. For any US geography — state, county, city, or ZIP — the answer is almost always the Census Bureau American Community Survey, and you can pull the number at data.census.gov by searching the geography and the table reference. Total population is Table B01003. If the figure matches the ACS 5-Year estimate for the same vintage, it is correct. If it does not, the figure is either from a different vintage, a different geography definition, or a different program such as the decennial census.
A vintage is the span of years an estimate covers. The Census Bureau releases a new ACS 5-Year file every December, and the current one covers 2019 through 2023. Two figures for the same city can differ by a wide margin simply because one is built on a newer vintage than the other. A demographic statistic published without its vintage cannot be verified, and it may be considerably older than the page it appears on — race, income, and education distributions can move substantially across a decade.
Median household income is Table B19013. Total population is B01003. Race and ethnicity is B02001. Educational attainment is B15003. Housing tenure and value are B25003 and B25077. Poverty status is B17001. Entering the table number plus a geography name into the search box at data.census.gov returns the estimate along with its margin of error, which is worth reading for smaller places.
For the United States, the FBI Crime Data Explorer publishes offenses reported to law enforcement agencies. For international comparison, the UN Office on Drugs and Crime publishes dataUNODC. Both are free and both count reported crimes. Be careful with any crime index built from opinion surveys: those measure how safe people say they feel, which is a genuinely interesting thing to measure but is not the same quantity as reported crime and does not move in step with it.
Regional Price Parities, published by the Bureau of Economic Analysis. RPP measures price levels for a geography against the national average, and it is the series the federal government itself uses for regional adjustments. Composite cost-of-living indices assembled by private publishers weight rent, groceries, healthcare, and transportation differently from one another, so two of them can disagree about the same state. When a cost-of-living claim matters, check it against BEA.
Every entity page here shows the dataset and vintage in a source badge near the title — currently Census ACS 5-Year 2023. We also publish the underlying data as free CSV files on our data downloads page, with no email gate and a CC0 license, so you can re-derive any figure on the site from the same rows we used. Where a number comes from the World Bank or the BEA rather than the Census, the badge says so.
Related
Table references and release schedules on this page come from the publishing agencies' own documentation. Every source listed is free to access and requires no account.
Demographic data you can verify
Get demographics updates by email. No spam, unsubscribe anytime.