What the Surgeon General's Loneliness Figure Actually Measured

“Approximately half of U.S. adults report experiencing loneliness.” That single line, from the 2023 U.S. Surgeon General advisory on social connection, has become the most repeated statistic in American public health writing about isolation. It appears in op-eds, in state legislative testimony, in corporate wellness slide decks. It is treated as a settled fact about the country’s emotional state, on the order of an unemployment rate.

It is worth asking what that figure is actually built from, because the advisory itself is a synthesis document, not a single survey. It does not report a single study with a single sample and a single question. It aggregates findings across a body of prior research, and the “approximately half” figure is a summary claim drawn from that aggregation rather than the output of one measurement instrument applied at one point in time.

That distinction matters more than it might seem.

Different surveys, different questions, different numbers

Consider what else exists in the literature the advisory draws on. The AARP Foundation’s 2018 survey of adults 45 and older, using the 20-item UCLA Loneliness Scale on 3,020 respondents, found one in three reporting loneliness. Cigna’s 2020 workplace survey, using its own proprietary loneliness index, found 61% of U.S. adults sometimes or always lonely. The Harvard Graduate School of Education’s Making Caring Common project, surveying in early 2021, found 36% of Americans reporting serious loneliness, rising to 61% among adults 18-25. Gallup’s 2023 cross-national work, covering 142 countries, found 24% of people worldwide lonely “very” or “fairly” often, with the same age skew toward young adults.

These are not small discrepancies. One in three, 36%, 61%, and roughly half are four different numbers describing, ostensibly, the same underlying phenomenon in overlapping time periods. The gap between AARP’s one-third and Cigna’s 61% is not explained by any plausible change in the population’s actual emotional state between 2018 and 2020. It is explained by instrument mismatch. AARP used a validated multi-item clinical scale designed to capture a durable trait. Cigna used a company-built index designed, at least in part, to generate a publishable and shareable number. Harvard’s team asked about “serious loneliness” specifically, a narrower category than “loneliness” as a general frequency question. Gallup’s cross-national instrument asked about feeling lonely “a lot of the day yesterday,” which measures a recent affective state rather than a chronic condition.

None of these is wrong. They are measuring different things and should not be expected to agree.

What “approximately half” is doing in the advisory

Read against that backdrop, the Surgeon General’s “approximately half” reads less like a precise national estimate and more like a rounded midpoint chosen to be defensible against a range of instruments that disagree with each other by a factor of two. That is a reasonable thing for a policy document to do. An advisory intended to mobilize a “National Strategy to Advance Social Connection” needs a number simple enough to state once and durable enough not to be undercut by the next survey release. It is not, however, the same rigor as a peer-reviewed prevalence estimate from a single validated instrument, and it should not be cited as though it were.

The federal government’s own surveillance data illustrate the same problem. The CDC’s Morbidity and Mortality Weekly Report on loneliness, published in June 2024, drew on 2022 survey data — a two-year gap between collection and publication that is itself worth flagging whenever the report is cited as current. The MMWR figures were generated through a different survey architecture than the advisory’s underlying sources, and they were not designed to produce a single top-line “half of adults are lonely” statistic; they broke findings down by state, demographic group, and specific measures of social and emotional support, which is a more granular and arguably more useful contribution than a single national percentage.

The number obscures who is actually affected

The bigger problem with treating “half of U.S. adults” as the headline fact is that it flattens a distribution that is not remotely flat. Every survey in this literature that breaks results down by age finds the same pattern: young adults report loneliness at roughly double the rate of older adults. Harvard’s Making Caring Common survey found 61% of 18- to 25-year-olds seriously lonely against much lower rates among older cohorts. Gallup’s global data found the 19-29 age group at 27% against 17% for adults 65 and older. Cigna similarly found the highest rates concentrated in Gen Z workers.

A single population-wide average obscures this. If “half of adults” is lonely, but the real distribution is closer to two-thirds of young adults and roughly one-third of older adults, then a policy response calibrated to the average — a general public-awareness campaign, say — will be poorly matched to where the burden actually concentrates. The National Academies’ 2020 consensus report on older adults, working from a different literature focused specifically on isolation rather than loneliness, put the isolation figure for adults 65 and older at roughly one in four, a number that is not directly comparable to any of the loneliness percentages above because isolation and loneliness are conceptually distinct: isolation describes the structure of a person’s network, loneliness describes a subjective state, and the two are only moderately correlated. Aggregating everything into one national loneliness percentage erases that distinction along with the age distinction.

What the figure can and cannot support

None of this means the advisory’s underlying argument is wrong. The mortality evidence it draws on, particularly Holt-Lunstad’s 2015 meta-analysis finding elevated mortality risk associated with social isolation, loneliness, and living alone even after adjusting for health status, is methodologically stronger than any single prevalence survey, because it aggregates many studies using a formal meta-analytic procedure rather than picking a rounded figure from a range of incompatible ones. The claim that social disconnection carries a mortality risk comparable to smoking is a defensible synthesis of that body of work. The claim that “approximately half” of adults are lonely, cited as a precise national fact, is not carrying the same evidentiary weight, even though both claims appear in the same nine-page section of the same document.

A 2023 review in BMC Public Health mapping the state of loneliness research made the underlying problem explicit: inconsistent measurement across studies is one of the field’s central methodological barriers, and it makes cross-study comparison unreliable in exactly the way the prevalence figures above demonstrate.

What would resolve this is not more surveys asking slightly different loneliness questions of national samples. It is a longitudinal study using one validated instrument, applied consistently across age cohorts and repeated at fixed intervals, so that any change in prevalence over time reflects an actual change in the population rather than a change in who happened to run the survey and what they called the phenomenon they were measuring. Until that exists, “half of Americans are lonely” is best read as a rough, era-specific synthesis of a fractured literature, not a number with the precision its constant repetition implies.