Data documentation
Sources and methodology
What each value means, which period it represents, and how to compare counties responsibly.
Different questions require different datasets
Current characteristics: use the ACS 2020–2024 five-year product, which provides county-level geography coverage even though individual cells can be unavailable.
Annual population: use Census PEP, not an ACS five-year population estimate.
Annual unemployment: use the BLS LAUS annual average, not the ACS employment estimate.
Long-term social change: compare non-overlapping ACS five-year periods; do not compare overlapping five-year periods.
Historical data availability
Loading the build manifest…
Data inventory
| Source | Used for | Period | Important note |
|---|---|---|---|
| American Community Survey (ACS) ↗ | Current county characteristics and non-overlapping historical period estimates | 2020–2024 five-year estimates for the current snapshot | ACS five-year values pool 60 months of survey responses. They are not a simple five-number average and should not be described as a 2024 annual value. Individual cells can still be unavailable. |
| Population Estimates Program (PEP) ↗ | Annual resident population and population change | Vintage 2025 estimates for July 1, 2020 through July 1, 2025 | PEP is the annual population series used for year-to-year population trends. It is different from the ACS population estimate shown in the current characteristics table. |
| Local Area Unemployment Statistics (LAUS) ↗ | Annual-average county unemployment | Annual averages, 2015–2025 where available | County LAUS values are not seasonally adjusted. Historical estimates can be revised when BLS updates inputs and geography. |
| County presidential results ↗ | Descriptive presidential vote shares and Democratic-minus-Republican margin | Presidential election years through 2024 where available | Compiled by Tony McGovern from state and local election reporting. These results are context, not an inferred ideology score for residents and not an authoritative canvass. |
How to read margins of error
ACS values are estimates based on a sample. When an ACS margin of error is available, County Lens displays it at the Census Bureau's standard 90% confidence level. The interval is the estimate plus or minus that margin.
Two estimates that look different may not be statistically different. County profile trend summaries label an ACS difference as statistically significant only when both non-overlapping estimates have margins of error and the difference passes a two-estimate test at the 90% confidence level. A non-significant result does not prove that no change occurred.
V4 converts its historical ACS median-income series to constant 2024 dollars using Census-published R-CPI-U-RS factors. The download retains each original estimate, original margin of error, and adjustment factor. Other dollar series should not be compared over time unless they are explicitly labeled in constant dollars.
ACS change maps show the difference between reported point estimates and are not filtered for statistical significance. Use the county profile and its margins of error before drawing a conclusion about an individual county.
Population Estimates Program and LAUS values are model-based estimates and can be revised, but they do not use ACS margins of error. Their trend lines are labeled separately so those measures are not mixed.
What County Lens does not claim
- A county percentile ranks the reported estimate among counties; it is not a quality, desirability, or performance score.
- County averages do not describe every person within a county and can hide differences among neighborhoods and groups.
- A scatterplot association does not show that one indicator causes another.
- Election margins describe ballots counted in one contest. They do not measure every resident's ideology or the degree of social polarization.
- Recent headlines are an incomplete view of local coverage and should not be used to characterize the people of a county.