One Word, Many Instruments: How Countries Actually Measure Loneliness

Nineteen countries were asked, in one form or another, whether their populations are lonely, and the answers range from 17% to 61%. That spread is not primarily a story about which nation has a loneliness crisis and which does not. It is a story about instruments.

The same word, four different questions

The World Health Organization’s Commission on Social Connection, in its 2025 report, put the global figure at roughly one in six people affected by loneliness, drawing on the Gallup-Meta survey of 142 countries conducted between June 2022 and February 2023. That survey used a direct single-item question — asking respondents whether they felt lonely, on a scale from “very” to “not at all” — and found 24% reporting they felt very or fairly lonely, per Gallup’s own 2023 release of the same fieldwork. Gallup’s later 2024 update, using a slightly different framing focused on the previous day’s feelings, found just over one in five adults worldwide reporting loneliness “a lot” of that day.

Compare that to the U.S. Surgeon General’s 2023 advisory, which cited approximately half of American adults reporting loneliness — a figure built from a different measurement tradition entirely. Cigna’s Loneliness in America 2025 report, based on fieldwork from May and June 2024, found 57%. The Harvard Making Caring Common survey found 36% reporting “serious loneliness,” but 61% among adults aged 18 to 25. The CDC’s 2024 surveillance report, using 2022 data, produced yet another number. None of these figures is wrong. They are measuring related but distinct things, with instruments that ask different questions, use different reference periods, and set different thresholds for what counts as “lonely” at all.

This is the instrument mismatch that recurs constantly in this literature, and nowhere is it clearer than across borders.

Three families of instrument

Loosely, three approaches dominate.

The first is the direct single-item question: “How often do you feel lonely?” This is what Gallup uses globally, and it is cheap to field across 142 countries in a single wave. Its weakness is well documented — self-labeling as “lonely” carries stigma, and cross-cultural comparability is uncertain. A respondent in a culture where admitting loneliness is more socially acceptable may answer differently than one in a culture where it signals personal failure, even at identical underlying levels of subjective isolation. Gallup’s own analysis of the results noted that connectedness and loneliness are not simple inverses of each other across countries, which is itself a clue that the single-item measure is capturing something narrower than the full construct.

The second is the UCLA Loneliness Scale, a 20-item multi-question instrument developed for academic research and used by, among others, AARP’s 2018 and 2025 surveys of U.S. adults 45 and older. Because AARP used the identical 20-item instrument in both waves, the two AARP surveys are among the very few loneliness datasets in the world that are genuinely comparable over time, methodologically speaking. That comparability is rarer than it should be. Most national tracking exercises either change their question wording between waves or never had a fixed instrument to begin with.

The third is the composite or de Jong Gierveld-style scale, which distinguishes emotional loneliness from social loneliness and often separates subjective loneliness from objective network size. The 2025 study using HINTS-6 data, drawing on 2022 fielding, applied this kind of layered approach and reported that roughly one in seven U.S. adults experience severe loneliness and one in five moderate loneliness — a finer gradation than a single “are you lonely, yes or no” question permits, and one that produces a combined moderate-to-severe figure above 37%, a number that sits closer to Cigna’s than to Gallup’s, despite measuring a U.S.-only population at a different point in time.

Japan’s case: a national instrument built for policy, not research

Japan offers an unusually clean illustration of how national measurement infrastructure shapes the numbers a government sees. Following the 2021 appointment of a Minister of Loneliness and Isolation — the second such post globally after the UK’s — Japan built its own tracking apparatus rather than adopting an existing academic scale wholesale. The JACSIS cohort study, covering 2020 to 2021 and published in 2023, is longitudinal rather than cross-sectional, which lets it speak to change during the pandemic period rather than offering a single snapshot the way most national loneliness surveys do. That longitudinal design is a genuine methodological advantage over most of the comparative literature, which relies on repeated cross-sections that cannot track the same individuals over time.

The UK took a different institutional route. Its 2018 loneliness strategy, the first national strategy of its kind, embedded loneliness measurement directly into the Office for National Statistics rather than relying on a standalone survey vendor. That decision means UK loneliness estimates are produced with the same infrastructure and continuity as unemployment or inflation statistics — a different kind of comparability than Japan’s cohort design or AARP’s fixed instrument, but a comparability nonetheless.

Why “one in six” and “one in seven” can both be defensible

It is tempting to treat the WHO’s global one-in-six figure and HINTS-6’s American one-in-seven-severe figure as roughly convergent evidence. They are not really testing the same claim. The WHO number is a 142-country average built from a single-item question fielded in one wave; the HINTS-6 number is a U.S.-only estimate built from a graded severity scale applied to 2022 data and published three years later. That gap between when data is collected and when a report is published is itself a recurring source of confusion — the CDC’s 2024 MMWR report on loneliness is likewise built on 2022 survey data, so a reader citing it as “2024 loneliness levels” is misdating the underlying reality by two years.

A recent eight-country study, published in early 2026, tried to hold the comparison more constant by administering the same instrument across all eight countries simultaneously and reporting that nearly half of young adults report loneliness, with the loneliness-depression-anxiety association holding after adjusting for demographics and prior diagnoses. That design — one instrument, multiple countries, one wave — is closer to what genuine cross-national comparison requires than stitching together Gallup’s global figure, AARP’s American figure, and Japan’s cohort figure and treating them as a single trend line.

The 2023 review in BMC Public Health made the general point plainly: inconsistent measurement is a structural barrier to comparing findings across studies, not an incidental inconvenience. It is worth taking that seriously rather than as academic throat-clearing. When the Center for Social Connection is asked whether loneliness is worse in the United States than in Japan, or worse now than in 2018, the honest answer is usually that the question cannot be answered from existing data without first establishing whether the same instrument, the same reference period, and the same population were used in both measurements. Often they were not.

What better cross-national measurement would look like

A stronger comparative literature would field one validated instrument — ideally the UCLA scale or an equivalent multi-item measure, rather than a single self-label question — across a fixed set of countries, at a fixed interval, with published item-level data rather than only summary percentages. It would also separate isolation, a structural count of network size and contact frequency, from loneliness, a subjective state, since the AARP data show these can diverge sharply: 33% loneliness among people who talk to neighbors, against 61% among those who never do, a gap that a single-item “are you lonely” question would flatten. The 2024 Household Pulse Survey work moves in this direction domestically, breaking loneliness and isolation out separately by demographic and socioeconomic factors, but no equivalent exists yet at global scale. Until it does, cross-country loneliness comparisons will keep producing numbers that look like disagreement about the world when they are often just disagreement about the question.