Same Word, Different Question: How Loneliness Surveys Diverge Across Countries

The U.S. Surgeon General’s 2023 advisory reported that approximately half of American adults experience loneliness. Gallup’s 2023 global survey, covering 142 countries, put the worldwide figure at 24%. Both numbers are defensible. Neither is wrong. They are answers to different questions, asked in different ways, over different reference periods, and treating them as comparable is the most common error in loneliness reporting.

The gap is not a matter of one survey being more rigorous than another. It is a matter of what “lonely” was operationalized to mean before anyone picked up a phone or opened a browser tab.

Three design choices that move the number

Three features of a survey instrument do most of the work in determining where a country’s estimate lands.

The reference period. Gallup’s global surveys, including the 2023 Gallup-Meta State of Social Connections report and the follow-up analysis on the world’s social connections, ask whether a respondent felt lonely “a lot of the day yesterday” — a single day, close to the interview. That framing captures a mood state, not a standing condition, and it tends to produce lower prevalence than instruments asking about loneliness “in general” or “over the past year.” The Cigna 2020 workplace report, by contrast, asked U.S. adults whether they “sometimes or always” feel lonely with no fixed window, and found 61% — nearly triple Gallup’s global figure from a comparable period.

The instrument itself. The AARP Foundation’s 2018 survey of adults 45 and older used the 20-item UCLA Loneliness Scale, the closest thing the field has to a shared academic standard, and found that one in three respondents qualified as lonely. Harvard’s Making Caring Common project, surveying a broader age range in 2021, used its own items and arrived at 36% reporting “serious loneliness,” with 61% among adults 18 to 25. The CDC’s 2024 surveillance report, drawing on 2022 BRFSS-adjacent data, used yet another framing tied to social and emotional support rather than loneliness as a felt state. None of these are interchangeable, even when the resulting percentages look similar.

Whether the survey measures loneliness at all, as opposed to isolation. Loneliness is subjective — a felt gap between the connection a person wants and the connection they have. Isolation is structural — the size, density, and contact frequency of an actual network, independent of how a person feels about it. Gallup’s connectedness surveys largely measure the latter; the UCLA scale and its derivatives measure the former. A 2023 BMC Public Health review of the loneliness and isolation literature names this conflation as one of the field’s persistent methodological gaps, and the difficulty is not academic. A country can report high isolation and modest loneliness, or the reverse, and a survey that measures only one will misdescribe the other.

Where national strategies get their numbers

Government responses to loneliness have generally been built on whichever instrument was available domestically at the time, which means cross-country comparison of policy is comparing responses to non-comparable inputs.

The UK’s 2018 loneliness strategy, the first national strategy of its kind, rested on measurement embedded into the Office for National Statistics’ existing household surveys — a long-run domestic instrument with its own item wording, not the UCLA scale and not Gallup’s daily-affect question. The U.S. Surgeon General’s advisory synthesized multiple U.S. sources, including Cigna’s workplace data and academic meta-analyses, arriving at an aggregate “approximately half” framing that is itself a judgment call about which underlying surveys to weight.

Japan offers a useful case because it has both a dedicated government structure — a cabinet-level post on loneliness and isolation, created in 2021 — and a longitudinal academic study, JACSIS, that tracked isolation and loneliness across 2020 and 2021 using consistent items on the same panel of respondents. That longitudinal design is rarer than it should be. Most national estimates are cross-sectional snapshots, so a single number like “36% of Americans” or “24% of the world” cannot by itself tell readers whether loneliness is rising, falling, or flat in that country. The JACSIS design can speak to change over time within Japan; it cannot be laid alongside the UK’s ONS trend line or the Harvard 2021 U.S. figure and treated as a global time series, because the instruments and populations differ.

What the Gallup global data can and cannot do

The Gallup-Meta 2023 survey is the most methodologically consistent attempt at cross-national comparison available, precisely because it asked the same items, in the same reference period, to roughly 1,000 respondents aged 15 and older in each of 142 countries between June 2022 and February 2023. That consistency is real and valuable: it is the reason Gallup can report that young adults aged 19–29 show the highest loneliness rate globally (27%) against adults 65 and older the lowest (17%), a pattern that would be much harder to trust if it were assembled from a patchwork of national instruments.

But consistency in method does not mean the underlying construct translates evenly. A single-day affect question may carry different cultural weight in a country where discussing loneliness is stigmatized than in one where it is not, and Gallup’s own follow-up analysis notes that connectedness and loneliness are not simple inverses of each other across countries — a population can report high social connectedness and still show a meaningful minority reporting frequent loneliness. The survey also excludes China, meaning its “world” figure of 24% describes roughly 77% of the world’s adult population, not the whole of it. That is a caveat worth stating plainly rather than folding into the headline number.

What a genuinely comparable estimate would require

A cross-national loneliness estimate that supported real comparison would need three things current data mostly lacks: the same instrument, translated and validated rather than merely rendered into another language; the same reference period, stated explicitly rather than left as “generally” or “yesterday”; and a paired measure of structural isolation alongside the subjective loneliness item, so that a low loneliness score could be distinguished from a well-connected population rather than a stoic one.

Until that exists, the honest way to read the current landscape of numbers is not to average them or rank countries against each other, but to ask, for any single figure, exactly what question was on the screen when the respondent answered — and over what stretch of time they were being asked to look back.