Whose Loneliness Gets Counted: Sampling Frames and the Young Adult Numbers

Two numbers circulate constantly in reporting on young-adult loneliness. One is 61%, the share of American 18-to-25-year-olds that the Harvard Graduate School of Education’s Making Caring Common project described as experiencing serious loneliness, in a report published in February 2021. The other is 27%, the share of 19-to-29-year-olds worldwide that Gallup found feel very or fairly lonely, in a survey released this past October. Both are described in coverage as “the loneliness rate for young adults.” They differ by 34 percentage points. That gap is not noise. It is the sampling frame.

Different populations, different years

The Harvard figure comes from an online survey of 950 U.S. respondents, fielded in October 2020, with young adults 18-25 drawn as one age band among several. The data collection sits squarely inside the acute phase of the COVID-19 pandemic in the United States, a period of closed campuses, remote work, and suspended in-person socializing for a population disproportionately built around universities and shared housing. The report itself is explicit about this: 43% of young adults said their loneliness had increased since the pandemic began, and about half of lonely young adults said no one had spent more than a few minutes recently asking how they were doing in a way that felt genuine. This is a snapshot of a specific cohort at a specific and unusual moment, not a baseline reading of youth loneliness in general.

The Gallup figure comes from a different instrument entirely: roughly 1,000 respondents per country across 142 countries, fielded between June 2022 and February 2023, as part of the Gallup-Meta State of Social Connections project. It is a much later moment, well past the acute pandemic disruptions in most of the countries surveyed, and it is global rather than American. It also uses a broader and shifted age band, 19-29 rather than 18-25, which changes who is included at both ends – Gallup’s youngest bracket excludes 18-year-olds still in the transition out of adolescence, and its oldest bracket reaches into the late twenties, a life stage with typically higher rates of partnership and settled routine than the very early twenties.

Put those two facts together – a pandemic-year U.S. sample versus a post-acute global sample with a different age cut – and a 34-point gap stops looking like a contradiction and starts looking like the predictable output of two different measurement designs pointed at two different things.

The instrument itself is not identical

Sampling frame is only part of the mismatch. The two surveys also do not ask the same question in the same way. Making Caring Common’s report describes “serious loneliness” without publishing the item wording as a single validated scale score in the same way, say, the AARP Foundation’s 2018 survey used the full 20-item UCLA Loneliness Scale. Gallup’s item asks directly how lonely a respondent felt “yesterday,” a single self-report question embedded in a broader well-being survey rather than a dedicated loneliness instrument. Neither approach is illegitimate. But a direct one-day recall question and a broader retrospective framing of “serious loneliness” are not interchangeable measurements, and treating one figure as an update to the other invites exactly the kind of false trend line that gets built into slide decks and op-eds.

Cigna’s 2020 workplace loneliness survey adds a third data point that is instructive precisely because it does not fit neatly next to either: 73% of workers aged 18-22, and over 80% of employed Gen Z overall, reported loneliness in that report, using yet another set of questions and a sample restricted to the employed. A young adult who is unemployed, in school, or out of the labor force is excluded from that frame by design. The figure is often cited as if it describes “young adults” broadly, when it describes young adults who hold jobs – a population that skews toward those with at least some structured daily social contact, which if anything should suppress isolation figures relative to the full young-adult population, not inflate them. That the figure is nonetheless among the highest in circulation is worth sitting with.

What a mode difference can do to a number

Online panels, phone surveys, and in-person interviews do not draw identical respondents even when the target population is the same. A person willing to complete a 20-minute online loneliness survey during a pandemic lockdown may not resemble a person reached by a global multi-mode survey operation with in-person interviewing available in lower-income countries. Gallup’s global methodology, run in partnership with Meta, explicitly excludes China and therefore covers roughly 77% of the world’s adult population rather than all of it – a limitation the organization states plainly rather than papers over. Making Caring Common’s sample, at 950 respondents split across multiple age bands, yields a young-adult subsample in the low hundreds, wide enough for a topline percentage but with a confidence interval rarely reported alongside the headline figure when it is quoted elsewhere.

None of this means either survey is wrong. It means the two numbers answer different questions: what did a specific, pandemic-affected American cohort report in late 2020, versus what did a broad international sample of a differently-bounded age group report roughly two years later. Reporting that swaps one for the other as evidence of an interim trend – loneliness falling from 61% to 27% between 2020 and 2023 – is manufacturing a time series out of two cross-sections that were never designed to be compared.

The American Enterprise Institute’s 2021 survey on friendship offers a useful corrective, because it measures something more stable than a felt state on a given day: the count of close friendships people report having. That figure – 12% of Americans reporting no close friends, up from 3% in 1990 – moves slowly and is less sensitive to the mood of the survey week, though it carries its own limitation, since “close friend” is left to each respondent to define.

What would actually let these numbers be compared

A defensible trend line on young-adult loneliness would require the same age brackets, the same instrument wording, and the same mode of administration, repeated at intervals, ideally with a published subsample size and confidence interval for the youngest cohort specifically rather than a topline figure that submerges it. The BMC Public Health review of the loneliness research base, published earlier this year, names inconsistent measurement across studies as a structural barrier to exactly this kind of comparison, not a minor footnote to it. Until a single instrument is applied longitudinally to a fixed age definition, the honest response to “is young-adult loneliness rising or falling” is that the current data cannot answer it, because no two widely cited estimates were built to be placed next to each other in the first place.