A vintage is a complete revised series
The Census Population Estimates Program revises the years back to the last census when it releases a new annual vintage. Updated inputs, methodology, corrections and geographic revisions can change an estimate for a past date. That does not mean residents were counted in real time or that a previous publication silently referred to the new value.
For consistent trend analysis, use all available years from one compatible vintage. This site's initial annual series uses Vintage 2025 for July 1, 2020 through July 1, 2025. Taking a 2021 figure from Vintage 2021, a 2022 figure from Vintage 2022 and so on would create a mixed-release line when a consistent revised series exists.
Revision difference is not annual growth
Suppose two publications give different estimates for July 1, 2023. Subtracting those two numbers measures a revision to the same dated observation. It does not measure growth from 2023 to 2024. Annual growth instead compares distinct observation dates inside a compatible source series.
An original 2020 census count can also differ from the later estimates base for April 1, 2020. That base incorporates documented corrections and revisions before annual estimates are produced. The difference is not an ordinary April-to-July population change, and the estimates base must not be labeled as the original count.
Other reasons figures differ
A city, county and metro can share a familiar name but have different boundaries. A resident population universe differs from household population. An ACS survey-period estimate differs from a PEP July estimate. A state-produced estimate may use a different reference date or method. A modeled current-year total is not necessarily an official annual observation.
Inspect source product, date, geography, universe, measurement type and vintage together. Do not average unlike measurements to invent a compromise answer. Choose the observation that matches the question and document the choice.
Reproducing an earlier report
Retain the file version used in the report, its checksum where available, the relevant geographic code and variable, and the calculation method. This publication's initial downloads are frozen under pep-2025-r1; a subsequent release must not overwrite that archived identifier.
The launch release catalog contains one snapshot, so it does not yet offer a multi-vintage revision chart. A later revision explorer requires at least two verified vintages for the same date and geographic definition. Profile source drawers already show the details necessary to reproduce the initial observation.
Primary documentation: Census PEP revisions FAQ and PEP methodology. Report suspected website errors separately through corrections.