What Sampling Frame Sits Behind 'Half of Americans Are Lonely'

The Surgeon General’s advisory, published earlier this month, states that approximately half of American adults report experiencing loneliness. That figure has already begun circulating as a fixed national fact, the way “one in five” figures for mental illness once did. It is worth asking what survey, administered to whom, and with what instrument, sits behind it.

The advisory itself is a synthesis document. It does not run a new survey; it aggregates existing prevalence estimates and cites the range that other studies have produced. That is a legitimate thing for a Surgeon General’s advisory to do. But it means the “half of adults” figure is not one number from one sampling frame. It is a rounded midpoint sitting somewhere near several different surveys that used different questions, different populations, and different modes of administration, and that produced measurably different results.

Three surveys, three numbers

Consider three sources that plausibly feed into the general public understanding of “how lonely is America.”

Cigna’s 2020 loneliness index reported that 61% of U.S. adults said they sometimes or always feel lonely, a figure the report noted was up seven points from the year before. That survey found loneliness heavily concentrated among younger workers: 73% of employees aged 18 to 22 reported loneliness, and over 80% of employed Gen Z respondents did. Cigna’s sample was explicitly a working population, recruited to speak to workplace loneliness. A national estimate drawn from an employed sample is not the same as one drawn from all adults, including those who are retired, unemployed, or out of the labor force for health reasons — groups whose loneliness rates and social patterns differ substantially from working adults.

The Harvard Making Caring Common survey, published in February 2021, found 36% of Americans reporting serious loneliness, with the figure rising to 61% among young adults aged 18 to 25 and 51% among mothers with young children. That is a general population sample, not restricted to employment status, and it used a different threshold — “serious” loneliness rather than any loneliness — which produces a lower headline number even though the underlying instrument may be picking up a similar construct at a stricter cutoff.

The AARP Foundation’s 2018 survey of adults 45 and older found that one in three reported being lonely, using the 20-item UCLA Loneliness Scale, a validated instrument with a long history in the academic literature going back to Holt-Lunstad and colleagues’ 2015 meta-analysis. That sampling frame excludes everyone under 45 by design — a substantial limitation given that younger adults consistently report the highest loneliness rates across nearly every other survey in this list. An estimate built on adults 45 and older will understate loneliness relative to the full adult population, not because the measurement is wrong, but because the frame excludes the age group most likely to report the outcome.

None of these three numbers — 61%, 36%, one in three — is wrong on its own terms. Each is a defensible estimate of loneliness within its stated population, using its stated instrument, at its stated moment. The problem is treating any of them as a stand-in for “American adults” without specifying which slice of American adults, measured how, and when.

What the instrument is actually asking

The UCLA Loneliness Scale asks respondents to rate agreement with statements like “I feel left out” and “I lack companionship” across twenty items, producing a continuous score that researchers then bin into categories. Cigna’s index used its own composite measure built around similar constructs but not identical wording. A single-item “do you sometimes or always feel lonely” question, of the kind embedded in some workplace surveys, is a coarser instrument than a twenty-item scale, and coarser instruments tend to produce higher affirmative rates because respondents self-select into a binary state rather than being pushed toward a threshold by an aggregated score. This is not a criticism of any single survey. It is a reason that headline percentages from different instruments should not be placed side by side as if they were repeated measurements of the same quantity.

Mode and timing compound the problem

Beyond the instrument, there is the question of when data were collected relative to when a report is published. The Harvard survey, published in February 2021, was fielded during a period of acute pandemic disruption to social contact; 43% of young adults in that sample reported increased loneliness specifically since the pandemic began, a context-bound figure that says as much about early 2021 conditions as about a stable underlying rate. Cigna’s 2020 report describes a rise from the prior year that likely reflects the same disruption. Neither figure should be read as a baseline for what American loneliness looks like in a non-pandemic year, though both are sometimes cited that way in secondary summaries.

Mode of administration matters too, though the sources reviewed here do not report enough detail to fully disaggregate it. Online panels, phone surveys, and workplace-distributed surveys draw from different pools of people who differ systematically in internet access, employment status, and willingness to disclose an experience some respondents still regard as stigmatizing. A survey fielded to employees through an employer-sponsored benefits platform will not reach the same population as a probability sample of U.S. households, even if both call themselves national.

What the advisory’s “half” actually represents

Given this, the Surgeon General’s approximately-half figure is best read as a rounded synthesis across a literature that itself spans instruments producing results from roughly one-third to roughly two-thirds, depending on age group, employment status, survey year, and question wording. That is not a flaw in the advisory’s arithmetic. It is an accurate reflection of a genuinely heterogeneous evidence base, and the advisory’s own framing acknowledges it draws on a range of existing estimates rather than asserting a single precise national rate.

The practical consequence for anyone using these numbers is straightforward. A claim that “half of Americans are lonely” should specify, at minimum, the age range, the survey year, and whether the underlying instrument was a validated multi-item scale or a single self-report item. Absent that, the number functions more as a rhetorical anchor than a measurement.

A stronger evidence base here would come from a single, repeated, nationally representative probability sample using one validated instrument administered on a fixed schedule — closer to how the Census Bureau tracks other social indicators — rather than the current patchwork of employer surveys, foundation polls, and age-restricted panels that each capture a true but partial picture. Until that exists, any single-number summary of American loneliness is really a summary of whichever sampling frame happened to generate it.