One Question, Five Numbers: What Loneliness Surveys Actually Ask

Gallup published a figure this week that 24% of people worldwide feel very or fairly lonely, drawn from surveys of roughly 1,000 people aged 15 and older in each of 142 countries between June 2022 and February 2023. Three years earlier, Cigna reported that 61% of U.S. adults sometimes or always feel lonely. Both numbers are described as measuring loneliness. They differ by a factor of 2.5, and the difference is not evidence that loneliness fell, or that Americans are unusually afflicted. It is evidence that the two organizations asked different questions.

This is worth taking seriously rather than resolving. The instinct when confronted with divergent prevalence figures is to pick the one that seems most credible and treat the rest as noise. A better approach is to line up what each survey actually asked, because the wording differences are themselves informative about what “loneliness” is being asked to mean.

The direct-question surveys

Gallup’s global figure comes from asking respondents directly whether they feel very lonely, fairly lonely, not very lonely, or not lonely at all — a single self-rated item, translated across 142 countries and administered by phone or in person depending on the market. It produces a clean, comparable global number: 24% lonely, 49% not lonely at all, with young adults aged 19-29 reporting the highest rate at 27% and adults 65 and older the lowest at 17%. The companion Gallup-Meta report frames the same dataset the other way around, as social connection rather than its absence: 72% of respondents, representing roughly 3.2 billion people in the countries surveyed, felt very or fairly socially connected. Notably, that survey excludes China, so it covers about 77% of the world’s adult population rather than all of it — a caveat that matters when the topline is presented as a global figure.

Cigna’s 61% comes from a different kind of instrument. Rather than asking respondents to self-label as lonely, Cigna’s index is built from a battery of items — feeling left out, feeling isolated from others, lacking companionship — that are then aggregated into a loneliness classification. This is closer to how the academic literature typically measures the construct, using multi-item scales rather than a single self-rating, and it tends to produce higher prevalence estimates than direct self-report. People are often more willing to endorse a specific statement like “I feel left out” than to apply the label “lonely” to themselves, which many treat as a more freighted admission.

The AARP Foundation’s 2018 survey of adults 45 and older split the difference in an instructive way: one in three respondents were classified as lonely, using the 20-item UCLA Loneliness Scale, the most established instrument in the academic literature. That scale also does not ask “are you lonely” directly; it asks about specific experiences and derives a score. The result — 33%, versus Cigna’s population-wide 61% — cannot be explained by age alone, since older adults tend to report less loneliness than younger ones in most datasets, including Gallup’s. Some of the gap is plausibly the difference between a validated multi-item scale and a proprietary index; some of it may be sample and timing differences that neither report allows a reader to fully untangle.

The Harvard Graduate School of Education’s Making Caring Common project, surveying in early 2021, found 36% of American adults reporting “serious loneliness,” with young adults aged 18-25 at 61%. That survey used its own operational definition tied to frequency and intensity thresholds, collected during a period of pandemic disruption to in-person contact that makes direct comparison to pre-pandemic or non-pandemic surveys difficult regardless of instrument.

What the instrument choice is doing

None of these five figures — 24%, 61%, 33%, 72% felt connected, 36% — is wrong. Each is an accurate report of what its respondents said in answer to the question they were actually asked. The mistake is treating them as five measurements of one stable quantity that a reader can average, rank, or plot on a single trend line, as though “loneliness” were a blood pressure reading with an agreed cuff and threshold.

Three design choices seem to matter most. First, direct self-labeling versus multi-item scoring: asking someone to say “I am lonely” screens out people who feel the underlying experiences without accepting the label, and generally produces lower numbers than scale-based batteries like the UCLA instrument or Cigna’s index. Second, the reference period: some instruments ask about current or recent feelings, others about general life experience, and the two are not asking the same question even when the words overlap. Third, whether the survey frames the question around connection or its absence — Gallup’s own paired reports show how differently a topline reads depending on which side of the ledger gets the headline, even when it is the identical underlying dataset.

The Surgeon General’s 2023 advisory tried to cut through this by settling on a single widely cited figure — approximately half of U.S. adults report experiencing loneliness — but that figure itself is a synthesis drawn across a literature that uses inconsistent instruments, which the advisory’s own framing acknowledges by pairing it with the more robust claim that the mortality risk associated with social disconnection is comparable to smoking up to 15 cigarettes a day. That comparative claim, built on the meta-analytic mortality literature, rests on more stable methodological ground than any single prevalence percentage, because relative risk within a study is less sensitive to how “lonely” was defined than a cross-national prevalence comparison is.

This is not a minor technical quibble. A June 2023 review in BMC Public Health, surveying the current state of loneliness and social isolation research, identifies inconsistent measurement as a structural barrier to comparing findings across studies — not an incidental flaw in a few surveys but a feature of the field as it stands. The National Academies’ 2020 consensus report on older adults made a related point in a different register: it estimated that roughly one quarter of adults 65 and older are socially isolated, a figure describing network structure — size, contact frequency, isolation from others — rather than subjective feeling, and it is not directly comparable to any of the loneliness percentages above precisely because isolation and loneliness are different constructs measured by different instruments, one structural and one experiential.

What a comparable measure would look like

A cross-national or cross-survey comparison that meant something would need at minimum a shared instrument, a shared reference period, and ideally both a self-rated item and a validated multi-item scale administered to the same respondents, so the gap between “calls themselves lonely” and “scores as lonely” could be measured directly rather than inferred across separate surveys. The UK’s 2018 national loneliness strategy took a step in this direction by embedding a standard loneliness question into the Office for National Statistics’ regular data collection, which at least allows within-country tracking over time using a fixed instrument — a lower bar than cross-national comparability but a meaningfully more useful one than a single snapshot survey.

Until something like that exists at scale, the honest reading of the current numbers is not that loneliness is rising or falling by some precise margin, or that any one country is more or less lonely than another by some exact percentage. It is that researchers, survey firms, and public health bodies are asking meaningfully different questions and reporting the answers under the same word. The word is doing more work than any of the surveys can support on its own.