Home › Learn
Margins of Error Explained in Plain Language
This guide is part of the ZIP Code Demographics learning library, written by our editorial team to help readers use ZIP-level Census data accurately. It explains margins of error in plain language, with the same care we apply to the data pages themselves: no invented figures, no borrowed statistics, and a clear line between what the Census Bureau publishes and what it does not.
Every estimate has a plus-or-minus
Because the ACS interviews a sample rather than everyone, every published estimate comes with a margin of error, or MOE. The Census Bureau publishes MOEs at a ninety percent confidence level: the true value is expected to fall within the estimate plus or minus the MOE about ninety percent of the time. A ZCTA population of 40,000 with an MOE of 1,500 is properly read as roughly 38,500 to 41,500 — a range, not a point.
Why small areas have bigger margins
Sampling error shrinks as samples grow. National estimates rest on millions of responses and have tiny relative MOEs; a ZCTA estimate may rest on a few hundred responses over five years and can carry an MOE of five, ten or even twenty percent on detailed measures like median income for a subgroup. Medians have MOEs too, computed by the Bureau from the distribution of responses. This is the main reason ZIP-level rankings should be handled gently: two ZCTAs whose confidence ranges overlap heavily are statistically indistinguishable on that measure.
What we do about it on this site
ZIP Code Demographics publishes the Bureau's point estimates, because those are the official figures and the ones other sources quote. In the narrative we avoid declaring winners on small gaps, we round derived percentages sensibly, and we warn explicitly — on this page and in the disclaimer on every profile — that the figures are estimates. Where the Bureau suppresses a value entirely, we omit the measure rather than estimating around the suppression. If you need MOE-level precision for a formal study, pull the same tables listed on our methodology page, which publish the MOE alongside every estimate.
A practical rule of thumb
For quick reading, treat differences under about two percent on a ZCTA-level share, or under a few thousand dollars on a ZCTA median income, as noise rather than signal. Treat large gaps — a degree share double the state figure, a median home value triple the national one — as real patterns; gaps of that size dwarf any plausible margin of error. Most local stories live in between, and the honest sentence is the one that names the figures and the uncertainty in the same breath.
Frequently Asked Questions
What should I know about margins of error?
Because the ACS interviews a sample rather than everyone, every published estimate comes with a margin of error, or MOE. The Census Bureau publishes MOEs at a ninety percent confidence level: the true value is expected t...
Where does the data on ZIP Code Demographics come from?
Every figure on this site comes from the U.S. Census Bureau's American Community Survey (ACS) 2020–2024 5-year estimates, published as table-based summary files and retrieved in October 2026. No figures are modeled or taken from commercial sources.
Data: U.S. Census Bureau, American Community Survey (ACS) 2020–2024 5-year estimates, retrieved October 2026. Source files and table numbers are listed on our methodology page; the underlying data is published at census.gov.