Digital behavior / Sources & methods
The definition is part of the finding.
Our digital-behavior collection keeps reported activities, internet adoption and platform audiences as separate measures.
Assisted source review 2026-10-08 · Snapshot digital-behavior-2026-10-08-v1
NTIA asks about internet users aged 15 and older
The November 2023 CPS Internet Use Survey’s activity sequence selects one internet user aged 15+ in a household. A household member may answer for that person. The resulting activity percentages describe people, not households, and the source’s adultInternetUser name must not be interpreted as an age-18+ rule.
NTIA asks respondents to focus on online activities done at least occasionally during the past six months. The sequence covers video, audio, calls, messaging and social networks with separate questions. It does not measure daily activity, viewing time, paid subscriptions or platform-specific audiences.
Sources: original questionnaire and eligibility, PDF page 148; official NTIA tabulation code; published aggregate table.
Weighting, uncertainty and population coverage
The Current Population Survey covers the U.S. civilian noninstitutionalized population. The November 2023 fieldwork period was November 12–18. These selected activity records cover internet users aged 15+ in the 50 states and District of Columbia.
NTIA estimates from one randomly selected internet user aged 15+ per household, using PWPRMWGT; proxy responses allowed. Source standard errors use successive difference replication with 160 respondent replicate weights. We use published weighted aggregates and source standard errors, without processing respondent records. Reporting, nonresponse, coverage and imputation error can remain beyond sampling uncertainty.
Source proportions and their SEs are multiplied by 100, producing percentages and percentage-point standard errors. Source sample sizes and confidence intervals are not supplied for the selected cells. An SE is not a margin of error. We calculate no significance test or causal effect.
What the demographic explorer allows
One activity, November 2023, one original demographic classification. Groups are compared within the same activity and source wave. Different groups have different denominators; percentages are participation within each group, not the group’s share of the audience.
- Age: Internet users in each age group. Ages 15–24 include teenagers and adults; this is not a teen-only survey.
- Family income: Annual family income, not personal earnings. All values refer to internet users aged 15+ in the income group.
- Education: Educational attainment of internet users aged 15+. High school includes GED; some college includes associate degrees and vocational study short of a bachelor’s degree.
- Sex: The source tabulates male and female categories. These are its sex categories, not a gender-identity measure.
- Race and Hispanic origin: Hispanic includes any race. Other groups exclude Hispanic people; Other includes multiracial and other less common responses. Preserve the source grouping.
- Disability status: CPS-reported disability status among internet users aged 15+. The source screens hearing, vision, cognitive, mobility, self-care and independent-living difficulties.
- Employment status: CPS PEMLR employment status among internet users aged 15+. Not a workplace-use measure.
- Metropolitan status: CPS metropolitan statistical area classification, not literal urban/rural density. Unknown stays a separate group.
- School-aged child in household: NTIA household classification from age 6–17 child/relative records; this remains an internet-user 15+ denominator, not a count of households or children.
- Veteran status: CPS PEAFEVER categories among internet users aged 15+; missing status is not assigned to non-veteran.
No cross-classification is present. Selecting age and income together, or creating a local estimate from national data, would require evidence not in this release. The interface offers neither. Broad activities may overlap and cannot be added to estimate unique users.
Historical snapshots retain changes in the questions
The activity pages show three national snapshots: November 2019, November 2021 and November 2023. They are dated observations, not a harmonized growth series. Question changes can affect responses, and this collection does not estimate their effects.
- Video, by 2021: the introductory wording changed from “streaming or watching” to “streaming or downloading.”
- Calls: examples changed from Skype and FaceTime in 2019, to Skype, FaceTime and Zoom in 2021, to FaceTime and Zoom in 2023. A gaming question appears immediately before calls by 2021.
- Audio, by 2021: the optional probe changed to listening to music, radio programs or podcasts using the internet.
- January 2022: CPS population controls changed. This is our comparability caution; no NTIA-specific effect is quantified here.
Sources: 2019 questionnaire, 2021 questionnaire, 2023 questionnaire and BLS 2022 population-control adjustments.
Each activity page reproduces the relevant dated wording and places caution markers beside its history chart. No interpolation, current-year extrapolation or cross-wave percentage-point change is generated. A future wave enters the site only after actual source observations and its questionnaire are available.
Data Commons: useful discovery, checked against the source
A real Data Commons v2 observation query on 2026-10-08 returned NTIA activity counts and World Development Indicators internet-use observations. Actual dates and values were examined; a mapped variable or future date label alone was not treated as available data.
A person/household mapping that we do not carry forward
The Data Commons importer maps NTIA’s adultInternetUser activities to Household and householder properties. For example, Count_Household_SocialNetworksUsage_IsInternetUser returns counts matching NTIA’s person-based cells. Matching numbers do not make the household label correct.
We quarantine those imported NTIA variables. Published activity percentages and sampling errors come directly from the original NTIA table and its person-based eligibility rules. We do not divide the imported count by a different population, relabel it as households, or silently turn it into a platform audience.
Evidence: official Data Commons import mapping, age-variable node, and matching original NTIA counts and shares. The mismatch is an audit finding about this import, not a claim that all Data Commons data have this issue.
Explicit facets for country internet adoption
The queried variable Count_Person_IsInternetUser_PerCapita returned the World Development Indicators facet 15880917103117484836, provenance dc/base/WorldDevelopmentIndicators, annual period P1Y and scaling factor 100. The returned values are already percentages; multiplying them by 100 again would be an error.
The returned U.S. and UK 2024 values matched the original World Bank API. Country/year survey notes are then taken from that original source, because the Data Commons observations do not provide all the age and recall-window exceptions needed for these profiles.
Data Commons variable · original indicator definition and methods. The integration selects facets explicitly and rejects incompatible or changed facet metadata; it does not choose whichever source happens to be first.
Country internet use is a separate denominator
World Bank / ITU indicator IT.NET.USER.ZS targets individuals using the internet from any location in the past three months, as a percentage of the in-scope population. Historical cells may use different ages or windows. For example, Canada’s 2010 footnote explicitly warns against comparison with earlier years after a redesign; a U.S. 2015 cell uses a six-month window.
Country profiles preserve every supplied footnote, including source-labelled estimates. Empty notes mean the details were not supplied. They do not establish a uniform age range, an unchanged survey or a direct measurement. Exact questions, fieldwork dates and uncertainty remain unknown where absent.
Our four selected country histories have actual values through 2024. The generic metadata range ending in 2025 does not change their missing 2025 cells. The charts visibly separate missing values from zero and mark documented scope notes and source-note changes.
Sources: selected original World Bank API records, source 2 with footnotes; WDI methodology. Twenty-eight sampled country/year values were checked against the separate official CSV distribution; its July 2026 revision date remains distinct from the October API revision.
Rights, versions and updates
NTIA’s selected federal aggregates support our attributed tables, original charts and downloads. World Bank labels the selected ITU-derived indicator CC BY 4.0; downloads retain both World Bank and ITU credit, source terms and a notice of our selection and formatting.
The Data Commons FAQ distinguishes its compilation licence from underlying-source terms. It is not a blanket licence for every source document, agency graphic or third-party artwork. We use no institutional logos, respondent records or bulk copies of publisher pages.
How a release becomes public
Batch imports go into a private source cache. Validation checks identities, facets, populations, missingness and numerical derivations against retained source files. Assisted review binds the candidate, original evidence and previous snapshot by SHA-256. A publication lock and immutable version archive protect the last good release.
A network error, missing credential, changed definition or invalid review stops that import. It does not turn missing evidence into zero or replace the current published snapshot. Source corrections receive a new version and a documented revision.
NTIA refreshes follow actual survey-table releases; WDI refreshes follow source releases or deliberate availability checks. Research pages do not call upstream APIs on each visit. Browser controls use the same reviewed static records as their HTML tables.
Review here is assisted factual and technical checking, not independent human editorial review. The release date never substitutes for the data period.