Choose the download that matches your task
A profile's CSV contains its six July 1 annual observations from 2020 through 2025. Directory exports contain only the current visible rows after filtering and pagination. Whole-dataset files contain every supported row for a geography type. State-specific files include typed state, county and whole-place records in that state; these overlapping levels must not be summed into a region total.
The annual dataset page offers long and wide layouts. A long row identifies one geography and one reference date; a wide row stores the six annual populations in separate columns. Long files preserve observation-level metadata, including variable, source URL, release, unit and status. Wide files are convenient for spreadsheets and joining year columns.
Preserve geographic codes as text
Official codes can begin with zeroes. A spreadsheet may convert a quoted CSV code into a number when opening the file directly. Import the file through the spreadsheet's text/CSV importer and explicitly set geographic code and ID columns to text. Quoting by itself does not guarantee that a spreadsheet preserves the leading zero.
Use the type-prefixed internal identifier, such as state:01 or county:01001, for joins inside this snapshot. A place name is not a unique key. Repeated city names and county equivalents require their official codes and geographic type.
Identify the observation and source release
A reproducible citation names the U.S. Census Bureau, Population Estimates Program, Vintage 2025, exact source filename and source variable, geographic code, population universe, reference date and retrieval date. If a calculation or transformation is involved, identify it separately. The site snapshot pep-2025-r1 identifies the retained downloaded files used for this initial publication.
Use the “Copy citation” control on a profile. The full citation remains visible in the source section, so it can also be selected manually. A shared comparison includes geographic IDs, starting and ending years, chart mode and release ID. Save those parameters along with your exported data.
Keep the research package
Download the README, data dictionary, citation text and checksums with the data. SHA-256 identifies a file's bytes and helps detect an accidental change; it does not independently prove that a source selection or interpretation is correct. Historical census counts have a separate source and measurement type and should not be inserted into an annual estimate line.
Read the Census citation guidance and release notes. Basic downloads are direct files. Retain the immutable version when citing a figure, since future annual vintages revise older dates.