Which Loneliness Are Governments Measuring

The UK’s 2018 loneliness strategy cites a figure derived from a specific choice: a single self-report item, asked with a specific set of response options, embedded into the Office for National Statistics’ ongoing surveys. The U.S. Surgeon General’s 2023 advisory cites a figure derived from a different tradition: multi-item psychometric scales developed for clinical and social-psychological research, adapted for population surveys. Both numbers get reported in the press as “the loneliness rate.” They are not measuring quite the same thing, and the gap between them is not noise. It is a policy choice made decades before anyone wrote a national strategy.

Two lineages, one word

The dominant multi-item instrument in this field is the UCLA Loneliness Scale, developed in the late 1970s and revised into a widely used 20-item version. It asks respondents to rate agreement with statements like “I have nobody to talk to” or “I feel part of a group of friends,” and sums them into a score. AARP’s 2018 survey of adults 45 and older used the full 20-item version, on a sample of 3,020 respondents, and reported that one in three fell into the lonely range. AARP’s 2025 follow-up, “Disconnected,” used the same instrument on the same age band, making it one of the few genuinely comparable before-and-after readings in this literature — most survey series change their wording between waves and lose that comparability entirely.

The other lineage is the single-item self-report: “How often do you feel lonely?” with a handful of frequency options. This is what Gallup uses in its global surveys, what Cigna has used since its 2020 workplace report and carried into “Loneliness in America 2025,” and what much of the CDC’s surveillance work relies on. Gallup’s 2023 and 2024 global figures — roughly a quarter of the world’s population, higher among young adults — come from asking people to report a feeling on one axis. Cigna’s 57% figure for 2025, drawn from fieldwork run in mid-2024, comes from Cigna’s own item set, not the UCLA scale.

These are legitimate instruments for different purposes. The UCLA scale was built to capture a construct — trait-like loneliness, embedded in a person’s relationship to their social world — with enough items to average out noise from any single question. The single-item measures were built for surveillance: cheap to field, fast to administer, easy to repeat across waves of a survey that already has thirty other things to ask about. What they are not is interchangeable. A population can show a stable UCLA score while single-item “felt lonely yesterday” figures swing sharply, because the single item captures state as much as trait.

What this does to a national strategy

The UK’s 2018 strategy chose to build its own primary measure rather than adopt the UCLA scale wholesale, embedding a loneliness question into the Community Life Survey administered by the ONS. This was a deliberate methodological decision with a policy rationale: a government tracking progress against a strategy needs a repeatable, low-burden item it can run every year on a large sample, not a 20-item clinical scale that raises fielding costs and respondent burden across a national statistical infrastructure. The tradeoff is that the UK figure is not directly comparable to AARP’s UCLA-based figure, or to Gallup’s global item, even where the underlying phenomenon overlaps.

The U.S. federal surveillance system took a related but not identical path. The CDC’s 2024 Morbidity and Mortality Weekly Report on loneliness, using 2022 survey data, and the 2025 analysis of the 2022 Health Information National Trends Survey (HINTS-6) both rely on short-form items rather than the full UCLA instrument, and they report different structures of the same underlying idea: HINTS-6 breaks loneliness into moderate and severe categories, finding roughly one in seven adults at the severe end and over 37% at moderate-to-severe combined. That is a genuinely different reporting convention from AARP’s binary lonely/not-lonely cutoff on the UCLA scale, and neither is wrong. They answer different policy questions. A health system deciding whether to screen for loneliness at all needs a binary threshold. A health system trying to allocate intervention resources by severity needs the graded version.

The National Academies’ 2020 consensus report on older adults, and the clinical commentary that followed it, pushed toward routine assessment in health care settings — a recommendation that itself assumes a short, validated instrument clinicians can administer during a visit, not a research-grade psychometric battery. The Surgeon General’s 2023 advisory then had to synthesize across all of this: UCLA-scale findings, single-item surveillance data, and clinical short forms, arriving at “about half of U.S. adults” as a working figure. That number is best read as an order-of-magnitude synthesis across instruments rather than a single measured quantity, and the advisory does not claim otherwise.

An assumption built into every version

Every one of these instruments, multi-item or single-item, asks about the subjective feeling of loneliness. None of them, on their own, measure the structural fact of social isolation — how many people someone actually sees, how large their network is, how often it activates. The World Health Organization’s 2025 Commission report keeps these separate for good reason: its headline claim, that loneliness is linked to an estimated 871,000 deaths annually, sits alongside distinct measures of network size and contact frequency, because a person can score low on isolation while still reporting high loneliness, and the reverse. A national strategy built entirely on a subjective-feeling instrument will miss people who are structurally cut off but not consciously distressed about it — a population health-relevant, but invisible to any of these surveys.

The 2023 review in BMC Public Health names this directly: inconsistent measurement across the field is a barrier to comparing findings, not a footnote to them. That inconsistency is not a sign of a young or sloppy field. The UCLA scale is nearly fifty years old. The instrument choices have been stable for decades; what has changed is which governments decided the construct was worth measuring at all, and each one inherited whichever tradition happened to be dominant in its own statistical or academic infrastructure at the moment it acted.

What would resolve this

A national loneliness strategy that wanted comparability across countries, or across its own waves over time, would need to report both a validated multi-item score and a single frequency item on the same sample, alongside a separate structural measure of network size and contact — and hold that combination constant across every wave. No government surveyed here does all three consistently. Until one does, cross-national and cross-report comparisons of “the loneliness rate” should be read as comparisons of instruments as much as comparisons of populations. The AARP 2018-to-2025 pairing is the closest thing in this literature to a genuine trend line, precisely because it changed nothing about the question being asked.