NTIA · United States · November 2023
Online classes and job training in the U.S.
Who uses the internet for classes or job training, and how do age and employment groups differ?
Internet users aged 15+ · Prior six months · Assisted analysis checked 2026-10-09
of U.S. internet users aged 15+ were reported as using the internet for educational classes or job training in NTIA’s November 2023 survey.
The activity sequence asks about at least occasional use during the prior six months. Classes and job training share one Yes/No question: the estimate does not separate students taking classes from people undertaking job training.
People · U.S. civilian noninstitutionalized internet users aged 15+ · November 2023 · Not all residents, school pupils or all adults aged 18+
Source: NTIA / U.S. Census Bureau Analyze Table, Nov 2023 / onlineClassUser / adultInternetUser, usProp; EDTRAI, questionnaire page 7, and USEINT six-month context, page 6. Social Usage converts the source proportion to a percentage.
The highest age-group estimate is for ages 15–24
Participation was 37.7% among internet users aged 15–24, compared with 29.8% among ages 25–44, 21.9% among ages 45–64 and 8.5% among ages 65+.
U.S. internet users within each age band · November 2023 · Prior six months
Source: NTIA aggregate CSV, row 765, age1524 / age2544 / age4564 / age65p Prop columns. Exact values and source standard errors.
The descriptive difference between ages 15–24 and 65+ is 29.1 percentage points, calculated before rounding. These estimates compare age groups in one survey; they do not follow the same people as they age or establish why participation differs.
Ages 15–24 include teenagers and adults, whether or not they attend school. That bar is not an online-school enrollment rate, a teen-only statistic or a result for all students. The age groups also cannot tell us how much participation was educational coursework and how much was job training.
Employment groups answer the same combined question
The estimate was 30.3% for internet users classified as employed, 28.7% for those classified as unemployed and 13.2% for those not in the labor force.
U.S. internet users aged 15+ in each source employment-status group · November 2023
Source: NTIA aggregate CSV, row 765, workEmployed / workUnemployed / workNILF Prop columns; source PEMLR recode in the official tabulation code, line 40.
The employed and not-in-the-labor-force estimates differ by 17.2 percentage points. “Unemployed” and “not in the labor force” are separate source classifications; the latter cannot simply be relabelled as unemployed, retired or a student group.
The unemployed estimate has a source standard error of 1.8611 percentage points, compared with 0.4164 for the employed estimate. The apparent ordering of point estimates is not a significance finding. We do not calculate a significance test or a confidence interval for a gap.
Employment status does not establish a person’s status when the class or training took place during the prior six months. Nor does the combined question identify training supplied by an employer, learning at work or a course that led to a job. These marginals cannot support an age-by-employment comparison.
One question covers educational classes and job training
“What about online classes or job training? (If needed) [Do you/Does NAME] use the Internet for educational classes or job training?”
Source: EDTRAI, November 2023 questionnaire, PDF page 149 / Attachment 8, page 7. USEINT, PDF page 148, gives the at-least-occasional, past-six-month activity context and selection of an internet user aged 15+ per household. Household proxy answers are permitted.
A Yes response can cover classes, job training or both. Neither component has its own percentage in these cells. The question applies to the selected internet user, including people who are not employed; the nearby employed-only routing in the questionnaire belongs to TELEWK, the separate telework question.
The measure records participation, not how many courses were taken, hours spent, fees paid, qualifications earned, completion or skills gained. It does not identify a provider, subject, degree program or whether instruction was entirely online. A school-enrollment dataset would answer a different question with a different population.
Informal learning is not separately identified. The survey cannot tell us which online videos were educational or how many YouTube users took classes. The separate activities can overlap; adding their percentages would not produce a unique learning audience.
All estimates and source uncertainty
Read the national, age and employment-status figures
| Group | Share (%) | Source SE (pp) | Source column |
|---|---|---|---|
| United States | 24.5529 | 0.2925 | usProp |
| 15–24 | 37.6503 | 1.0525 | age1524Prop |
| 25–44 | 29.8218 | 0.5165 | age2544Prop |
| 45–64 | 21.8632 | 0.4469 | age4564Prop |
| 65+ | 8.5294 | 0.3404 | age65pProp |
| Employed | 30.3222 | 0.4164 | workEmployedProp |
| Unemployed | 28.7050 | 1.8611 | workUnemployedProp |
| Not in the labor force | 13.1599 | 0.4580 | workNILFProp |
NTIA supplies proportions and proportion standard errors to six decimal places. Both are multiplied by 100 here, preserving the published precision as four-decimal percentages and percentage-point SEs. An SE is not a confidence interval or margin of error. Item sample sizes and confidence intervals are not supplied in these selected cells.
The national estimate is separately weighted, not an unweighted average of the age or employment percentages. Each bar describes participation within its own group, not that group’s share of all learners. Age and employment are separate classifications, not eight parts of a whole.
Missing under-15 activity cells stay missing. Missing employment classifications are not added to the unemployed group. Nonusers of the internet are outside the activity denominator.
Source: NTIA Analyze Table, Nov 2023 / onlineClassUser / adultInternetUser; named Prop columns and their corresponding PropSE columns.
Methods and reproducibility
Selection, employment categories and weighting
Fieldwork ran 12–18 November 2023. The Census Current Population Survey supplement covers the civilian noninstitutionalized population in the 50 states and District of Columbia. It randomly selects one internet user aged 15+ per household for this activity sequence, skipping it when all internet users are under 15. A household member may answer for the selected person.
NTIA assigns onlineClassUser to adultInternetUser and defines it as PEEDTRAI = 1 for the selected respondent. The internal name does not change the age minimum to 18. Estimates use the random-respondent weight PWPRMWGT; source sampling errors use 160 respondent replicate weights with successive difference replication.
The source age recode uses PRTAGE. Employment status uses PEMLR: codes 1–2 become Employed, 3–4 Unemployed and 5–7 Not in the Labor Force; −1 is missing. The employment rows cover internet users classified in the stated group. We retain NTIA’s activity eligibility and do not substitute a general labor-force statistic’s population or denominator.
Social Usage selects published aggregates and processes no individual responses. Census editing and hot-deck imputation, nonresponse, reporting and coverage error can affect the estimates beyond the reported sampling SEs. No adjustment for age or other differences between employment groups is performed.
Sources: technical documentation, PDF pages 7, 142, 148 and 149; November 2023 code, lines 16/26, 33/36, 39/40 and 97; official aggregate tabulator, lines 330–368.
Conversions and descriptive calculations
Find Nov 2023 / onlineClassUser / adultInternetUser in the original CSV. Row 765 includes the header in the retained version; the stable variable, wave and universe identify the record if later row positions change.
National proportion: 0.245529 × 100 = 24.5529%. Source SE: 0.002925 × 100 = 0.2925 percentage points.
- Ages 15–24 minus 65+: 37.6503 − 8.5294 = 29.1209 percentage points.
- Employed minus not in the labor force: 30.3222 − 13.1599 = 17.1623 percentage points.
These Social Usage calculations subtract full-precision percentages before rounding. They are descriptive differences, not percent growth or causal effects. The source SEs do not provide the covariance needed for uncertainty around the gaps.
Retained CSV SHA-256: c9b744849847775cf9419c8565553830b9fa374f62bbd674d4f22888521fd10a. Analysis: ntia-learning-2026-10-09-v1. Source, questionnaire, tabulation code, selected cells, formulas and this page are bound to an assisted review.
Dates, history and permitted reuse
- Data period
- November 2023 · Fieldwork 12–18 November
- Source publication
- 2024-06-06 · 2023 Data Explorer update
- Retained statistical retrieval
- 2026-10-01 (UTC)
- Assisted analysis review
- 2026-10-09
November 2023 is the latest onlineClassUser wave in the retained table. This page selects one wave and does not extrapolate a current-year figure. Earlier rows need a separate review of questions, populations and weighting before a historical comparison can be made.
Selected federal aggregate facts and original attributed charts/tables are used in this ad-supported publication under the retained rights basis. This page offers no new data download, image export or embed. Classes and job training are outside the existing digital-behavior explorer’s five-activity selection and downloads.
U.S. federal statistical work; 17 U.S.C. § 105; original rights verification 2026-10-01. Source: National Telecommunications and Information Administration (NTIA), Internet Use Survey, November 2023; U.S. Census Bureau Current Population Survey. Selection and presentation by SocialUsage; no agency endorsement implied.
This is assisted source and calculation checking, not independent human editorial approval. No agency artwork, logo or individual response is reproduced.