One in Three or Six in Ten: Why Loneliness Surveys Disagree

The AARP Foundation’s 2018 survey of adults 45 and older found that one in three respondents were lonely. Cigna’s 2020 workplace survey found that 61% of U.S. adults were lonely, sometimes or always. Both are large, professionally fielded surveys. Both claim to measure the same thing. The two figures differ by roughly a factor of two, and the difference has almost nothing to do with how lonely Americans actually were at the time each was conducted.

It comes down to what the surveys asked, and how they counted an answer as “lonely.”

Two different instruments, two different questions

AARP’s survey of 3,020 adults used the UCLA Loneliness Scale, a 20-item instrument developed in the academic literature and used in decades of published research. It asks about specific experiences — feeling left out, feeling isolated from others, having people to turn to — and produces a continuous score that researchers then bin into categories. AARP counted respondents above a validated clinical threshold as lonely. That threshold is conservative by design; it is meant to identify a state that recurs, not a passing mood, and it is comparable across studies precisely because it is not bespoke.

Cigna’s figure comes from a different kind of question. Its loneliness index asked whether respondents felt lonely “sometimes or always” — a much lower bar than a clinical cutoff, and one that captures anyone who has had a bad stretch of weeks alongside anyone who is chronically isolated. A person can answer “sometimes” to a single-item question and mean something quite different from what the UCLA scale’s threshold is built to detect. The Cigna figure is not wrong. It answers a different question than AARP’s, and treating the two as measuring the same underlying rate is the error, not either survey individually.

This is not a minor technical point. It means the two most-cited loneliness prevalence figures in circulation cannot be placed on the same axis. A journalist writing “loneliness has roughly doubled” between the two surveys would be describing an artifact of instrument design, not a trend.

What the Harvard survey adds, and complicates

The Harvard Graduate School of Education’s Making Caring Common project published a third data point in February 2021: 36% of Americans reporting “serious loneliness,” rising to 61% among young adults aged 18 to 25. That 36% figure sits closer to AARP’s one-in-three than to Cigna’s six-in-ten, which might suggest AARP’s instrument is the more reliable anchor. But the Harvard survey was fielded during the pandemic, its youngest respondents skew far lonelier than its older ones, and its own definition of “serious loneliness” is again constructed differently from AARP’s UCLA-scale threshold. The apparent convergence between AARP and Harvard around a third of respondents may be coincidence rather than confirmation. Three surveys using three different operationalizations of loneliness do not triangulate on a true rate; they demonstrate that there isn’t yet a single agreed rate to triangulate on.

What the Harvard data usefully add is a life-stage pattern that recurs across other work: young adults report loneliness at much higher rates than older adults do, even though public attention and policy have historically concentrated on the old. Cigna’s survey found the same skew in a workplace population — 73% of workers aged 18 to 22 reported loneliness, and over 80% of employed Gen Z respondents did. Whatever the absolute prevalence number, the relative ordering by age holds up across instruments that disagree wildly on the baseline. That consistency is more trustworthy than any single top-line figure, precisely because it survives the instrument-mismatch problem.

Why “lonely” resists a single number

Part of the difficulty is that loneliness is not a fact about the world in the way that, say, household size is. Cacioppo and Patrick’s account of loneliness frames it as an aversive internal signal, evolved to motivate reconnection — closer to hunger than to a demographic category. Two people can be objectively isolated to the same degree and report very differently on how they feel about it, and a single-item “do you feel lonely sometimes or always” question will pick up both the acute and the chronic case without distinguishing them. A validated multi-item scale, by forcing respondents through several angles on the same underlying state, filters out some of that noise. It does not eliminate the deeper problem: loneliness is subjective by definition, and any instrument is measuring a self-report of an internal state, not a directly observable quantity.

This is also why loneliness prevalence figures should not be confused with isolation prevalence figures, even within the same body of work. AARP’s own survey found that the size and diversity of a person’s social network and their degree of physical isolation were the strongest predictors of loneliness — network structure predicting a subjective state, not the same thing as it. Holt-Lunstad’s 2015 meta-analysis in Perspectives on Psychological Science treated loneliness and social isolation as related but statistically distinct mortality risk factors, with different odds ratios (1.26 for loneliness, 1.29 for isolation, 1.32 for living alone) that did not collapse into one number even after adjustment for health status. If isolation and loneliness do not collapse into each other in a mortality meta-analysis with 148 cohort studies behind it, there is little reason to expect two prevalence surveys using different single instruments to collapse into each other either.

What would actually resolve the disagreement

The AARP-versus-Cigna gap will not be closed by picking a winner. It would be closed by a survey research programme that fields the same validated instrument — the UCLA scale or an equivalent — across a representative sample at repeated intervals, so that any change in the resulting score reflects a change in the population rather than a change in question wording. The UK’s national loneliness measurement, embedded in the Office for National Statistics as part of its 2018 strategy, is closer to this model than most American survey efforts, precisely because it standardizes the instrument across waves rather than relying on whichever organization happens to commission a poll that year.

Absent that, the honest response to “is loneliness at one in three or six in ten” is that both figures are real answers to different questions, and neither should be quoted without naming the instrument behind it. The AARP figure describes something closer to a clinical threshold; the Cigna figure describes something closer to a mood. A policy brief that averages them, or treats the higher one as more alarming and therefore more citable, is doing something the underlying data do not support.