What the Headline Loneliness Rate Leaves Out

The Harvard Graduate School of Education’s Making Caring Common project reported in February 2021 that 36% of Americans experience serious loneliness. That figure has circulated widely, usually stripped of the number sitting right next to it in the same report: 61% of young adults aged 18 to 25 report serious loneliness, more than 60 percentage points higher than the rate reported by some other groups in the same survey. A single national rate that averages across an age gap that large is not describing one population. It is describing at least two, and possibly obscuring a third.

This is a general property of prevalence statistics, not a flaw specific to the Harvard survey. Any topline number is a weighted average across whatever subgroups the sample contains, and averaging is a form of information loss. The question worth asking of any loneliness headline is what subgroup the average is quietly absorbing.

The age gradient the topline hides

The Harvard survey’s breakdown shows loneliness is not smoothly distributed by age. Young adults report it at nearly twice the rate of the overall sample; mothers with young children report it at 51%. Both figures sit well above the 36% headline, which means some other segment of the sample must sit correspondingly below it to pull the average down. The report does not offer a single clean number for, say, adults in stable mid-career circumstances without young children, but the arithmetic makes clear that such a group exists and is doing a lot of work to keep the topline moderate.

Compare this with the AARP Foundation’s 2018 survey of adults 45 and older, which found one in three report loneliness using the 20-item UCLA Loneliness Scale — a validated instrument, not a bespoke question, which makes it more comparable to the academic literature than most consumer surveys. That AARP figure, roughly 33%, sits close to the Harvard topline of 36%. On its face this looks like convergence: two different surveys, two different years, roughly the same overall rate. But the AARP survey only samples adults 45 and up. It cannot report anything about the 61% figure among 18-to-25-year-olds because that group is not in its sample at all. The near-agreement between the two topline numbers is coincidental, not corroborating; they are averages over different populations, and treating them as validating each other misreads what each instrument actually asked and of whom.

Where the AARP data itself does the same thing

The AARP survey, to its credit, breaks its own topline down further, and the breakdown is instructive precisely because it shows the same averaging problem recurring one level down. The strongest predictor of loneliness in that data was not age itself but network size and diversity, and a specific behavioral marker: 33% of respondents who had spoken with a neighbor in the past week reported loneliness, against 61% of those who never had. That gap — 28 points — is comparable in size to the age gap in the Harvard data. A “45 and older” prevalence rate is itself an average across people with active neighborly contact and people with none, and the rate obscures which side of that divide any given respondent falls on. Subgroup analysis does not stop being useful once you have broken a sample down by age; it is useful again on whatever axis you break it down by next.

Gender inside the age gradient

The American Enterprise Institute’s Survey Center on American Life reported in June 2021 that 15% of men report having no close friends, a fivefold increase since 1990, versus 12% of Americans overall reporting the same. The share of men reporting six or more close friends fell from 55% in 1990 to 27% in 2021. Overlaying this onto the Harvard age data is not something the sources permit directly — they are different surveys with different sampling frames and different questions, one asking about loneliness as a felt state, the other asking about friendship counts as a structural fact. But the two lines of evidence point in a similar direction: young men are plausibly a subgroup within a subgroup, sitting at an intersection of the highest-loneliness age band and a friendship-network decline concentrated in men.

The Cigna 2020 workplace survey offers a third angle on the same age slice: 73% of workers aged 18 to 22 reported loneliness, and over 80% of employed Gen Z workers did. That is a workplace-specific sample, not a general population one, and the instrument differs again from the UCLA scale and from Harvard’s questions. But three separate surveys, using three different questions on three different samples, each single out people in their late teens and early twenties as reporting loneliness at rates far above whatever number gets quoted as the national average. That convergence across instruments is more informative than any one topline figure, because it does not depend on the wording of a single question.

What the National Academies report gets right by not averaging

The National Academies’ 2020 consensus report on older adults deliberately avoided producing one national loneliness number. Instead it reported that roughly one quarter of adults 65 and older are considered socially isolated, and treated isolation as a structural property of a person’s network distinct from the subjective loneliness measured in the Harvard and AARP surveys. This distinction matters here: Holt-Lunstad’s 2015 meta-analysis found isolation, loneliness, and living alone carried separate and only partly overlapping mortality risks, each roughly comparable in size. A subgroup that scores low on loneliness questions can still be structurally isolated, and vice versa. Any single prevalence figure, even a well-disaggregated one, is typically reporting on one of these two constructs, not both, and readers often assume it captures more than it does.

What would settle this

A survey designed to report age-by-instrument-by-construct breakdowns as its primary output, rather than as an appendix to a topline number, would let researchers compare structural isolation and subjective loneliness within the same age bands using the same instrument. None of the sources reviewed here do that consistently; each optimizes for a different headline. Until such a design exists, the honest use of any single loneliness statistic is as an average whose value depends entirely on knowing which subgroup it is quietly submerging.