Reading and designing questions
What social-media survey questions actually measure
“Do you use social media?” leaves several decisions unstated.
A social-media survey question needs a defined population, activity, recall period and set of response options. The result also needs a stated denominator. Visiting, posting and owning an account answer different questions; a percentage without those boundaries is easy to misread.
U.S. source examples · NTIA November 2023 and NCI HINTS 2024 · Original teaching examples are not validated instruments
Examples: NTIA questionnaire, PDF p148 defines occasional activity in the prior six months; HINTS 7 B12a, PDF p5 asks frequency of visits in the prior 12 months. Their populations and measures differ.
Read the question together with its instructions
A question fragment may omit the recall period or eligibility rule supplied immediately before it. Read that context before naming the measure.
| Survey | Population and activity | Window and interpretation |
|---|---|---|
| NTIA Internet Use Survey | Internet users aged 15 and older in the civilian noninstitutionalized population; within the stated group or geography Reported use of social networks at least occasionally; not platform-specific adoption. | At least occasional activity during the past six months Internet-user eligibility; not all U.S. adults or a daily-use measure. |
| NCI HINTS 7 | U.S. civilian noninstitutionalized adults aged 18 and older Frequency of visiting a social-media site; not posting or account ownership. | Past 12 months All survey participants in the codebook denominator, including missing/invalid item responses. |
NTIA gives Facebook, Twitter and Instagram as examples. HINTS introduces Facebook, TikTok, YouTube and Instagram. Neither item produces a separate percentage for every named platform. NTIA also permits proxy answers; HINTS uses a self-administered paper survey with a push-to-web option. These differences are part of the evidence, not labels that can be edited away.
HINTS response options and the meaning of “weekly”
- Almost everyday
- At least once a week
- A few times a month
- Less than once a month
- Never
These are the source's mutually exclusive choices. “Almost everyday” is not an exact every-day count. “At least once a week” is one response category here, not the combined share who visit at least weekly. Preserve the source labels. A cumulative statistic would require its own defined inputs, calculation and uncertainty; do not silently relabel the published category.
Source: NCI HINTS 7 B12a and Public Codebook, PDF p80 / printed p79. No NCI endorsement implied.
For the actual published results, use the NTIA participation dataset and HINTS frequency guide. This guide does not compare their percentages or turn either into a platform audience count.
The same answers can produce different percentages
Imagine 100 fictional, equally weighted eligible respondents: 60 answer Yes, 30 answer No, and 10 leave an item unanswered. This is an invented binary question, not NTIA or HINTS data.
Same Yes responses; different populations in the calculation
Source: Social Usage illustration and methods. Invented counts, equal weights, no population inference.
Read the fictional counts and calculations
| Denominator rule | Yes | Denominator | Percent |
|---|---|---|---|
| All eligible respondents | 60 | 100 | 60.0% |
| Valid Yes/No answers | 60 | 90 | 66.7% |
All eligible: 100 × Yes ÷ (Yes + No + missing). Valid answers only: 100 × Yes ÷ (Yes + No). Display percentages are rounded to one decimal; calculations retain full precision.
The first calculation includes unanswered items in its denominator; it does not recode them as No. The second excludes them and answers a different question. Neither tells us how the missing respondents would have answered. Keep the rule explicit rather than choosing the larger result.
Real weighted surveys are more complex than this equal-weight illustration. The retained HINTS national codebook percentages include missing and invalid item responses in their denominator. That rule must remain attached to its results; this example is not a replacement HINTS analysis.
Source for the HINTS rule: HINTS 7 public package, Public Codebook p80, universe “All Survey Participants”.
Three original drafts for three different aims
These are Social Usage teaching drafts, not copied source questions, tested survey instruments or ready-made national benchmarks. Define the eligible group and what counts as social media in the introduction. Test interpretation with that intended group before fielding a real survey.
| Aim and draft | Response plan | What it cannot answer |
|---|---|---|
| Days with visits During the last 7 days, on how many days did you visit a social-media app or website? | Choose a whole number from 0 through 7; record unable-to-recall and skipped answers separately. | Does not measure minutes, posts, or account ownership. |
| Posting During the last 7 days, did you publish a post on a social-media app or website? | Yes / No, with unable-to-recall and skipped answers retained separately. Define whether comments, reposts and private messages count. | Does not capture viewing-only activity or number of posts. |
| Account ownership Do you currently have an account on any of the social-media services listed here? | Yes / No, with unsure and skipped answers recorded separately. Supply the exact service list before fielding. | Does not establish recent visits or active use; covers only the listed services. |
Keep posting separate from viewing, and distinguish a respondent who reports zero activity from a skipped question or someone not eligible to receive it. If asking about named platforms, collect a response for each and report each platform separately; people can select more than one, so the platform percentages cannot be added into a unique-person total.
A convenience poll of your followers describes its respondents. The wording alone does not make it representative of U.S. adults, and neither a large response count nor a familiar question turns it into the NTIA or HINTS sample.
Before comparing two survey results
- Match the eligible population, age range, geography and how respondents were selected.
- Check the activity, recall window, response choices and any skip instructions.
- Keep the denominator, missing-answer rule and weighting method with the percentage.
- Record fieldwork dates separately from publication dates; disclose changed wording or methods.
- Retain uncertainty and source limitations. Matching labels alone do not justify a trend, causal claim or significance claim.
For duration questions, the screen-time measurement guide explains activity coverage and overlapping device time. For a published finding, build a citation that preserves its scope.
Sources, dates and methods
Retained evidence and original arithmetic
NTIA Internet Use Survey: November 2023 technical documentation, PDF148 / Attachment8 questionnaire p6, USEINT and SOCIAL; PDF7 for proxy responses.. Data period: November 2023. Source publication/update: June 6, 2024. Original retrieval: 2026-10-01T22:17:47.5845721Z. Original question source.
NCI HINTS 7: Annotated English instrument B12a PDF5 / printed4; Public Codebook PDF80 / printed79, SocMed_Visited.. Data period: March 25–September 16, 2024. Source publication/update: August 2025 package update; exact day unknown. Original retrieval: 2026-10-01T23:47:51.356625+00:00. Original question source.
Question definitions and category labels come from previously reviewed releases, reused without fetching them again or advancing their source dates. This guide adds no population observations. Its fictional counts and two formulas are stored and validated separately from source statistics. They have no sampling uncertainty because they are arithmetic examples, not estimates.
Reuse and review scope
Selected attributed federal definition facts and original Social Usage prose follow the retained NTIA/Census and NCI reuse decisions. No source questionnaire pages, logos, respondent records or source graphics are distributed. Original question drafts are unvalidated; no publisher endorsement is implied. No new dataset, download or embed is offered. Editorial and technical review was assisted, not independent human review.