What Actually Changed: Reading the AARP Loneliness Surveys Against Each Other

AARP’s 2018 survey of adults aged 45 and older found that one in three respondents were lonely, measured with the 20-item UCLA Loneliness Scale across 3,020 participants. Its 2025 successor, “Disconnected,” reports rising loneliness in the same population using the same instrument. That sentence should not need saying, and the fact that it does is the point of this note. Comparability across waves of the same survey series is rarer in this literature than most readers assume, and when it is absent, headline comparisons over time are close to meaningless.

Why the AARP pair is unusual

Most loneliness surveys are one-offs, or they change their question wording between rounds without flagging it prominently. AARP’s two surveys, seven years apart, used the same 20-item scale on a broadly comparable population of midlife and older U.S. adults. That means a change in the topline number can plausibly be attributed to a change in the underlying experience of loneliness among that age group, rather than to a change in how the question was asked. This is close to the ideal condition for tracking prevalence over time: same instrument, same target population, different point in time. It is worth naming because the 2023 review of the field in BMC Public Health singled out inconsistent measurement as one of the central barriers to comparing findings across studies at all. AARP’s repeat use of the UCLA scale is a partial answer to that problem, at least within its own series.

The 2018 survey also did something else useful: it identified network size and diversity, and physical isolation, as the strongest predictors of loneliness within that population, and it showed a striking gradient by neighbourly contact — 33% of respondents who had spoken with neighbours were lonely, against 61% of those who never had. Whether that gradient held, widened, or narrowed by 2025 is a testable question precisely because the instrument did not change. A methods note cannot report that comparison without the underlying tables in hand, but the point stands: this is the kind of question the AARP series is built to answer, and most series are not.

What a broken comparison looks like

Contrast this with Cigna’s two loneliness reports. The 2020 U.S. report used a loneliness index and found 61% of American adults sometimes or always felt lonely, with the sharpest concentration among the youngest workers — 73% of those aged 18 to 22. The 2025 report, “Loneliness in America,” reports 57% of Americans experiencing loneliness, drawn from fieldwork conducted in mid-2024 with more than 7,500 U.S. adults. On its face, that reads as a four-point decline. Whether it actually is one depends on whether the two reports asked the same question of comparably sampled people, and the Center’s remit does not extend to auditing Cigna’s underlying instrumentation across the two waves. The safer conclusion is that a five-point shift in a self-report loneliness figure, on its own, tells a reader very little unless the survey design is held constant. Absent that assurance, the two Cigna figures are two separate snapshots that happen to share a sponsor, not a trend line.

A second, more subtle version of the same problem shows up in publication timing. The CDC’s Morbidity and Mortality Weekly Report on loneliness was published in June 2024, but the survey data behind it were collected in 2022. The American Journal of Preventive Medicine’s HINTS-6 analysis, reporting that over 37% of U.S. adults experience moderate-to-severe loneliness, was published in November 2025 but also draws on 2022 data. Both are legitimate, recent-looking citations. Neither describes 2024 or 2025. A reader assembling a timeline of American loneliness who treats publication date as data date will place two years’ worth of change in the wrong place, and will be tempted to read a gap between the CDC figure and a genuinely 2024-fielded survey as movement over time, when it may simply reflect different instruments applied to the same underlying year.

What comparability actually requires

Three conditions have to hold for a repeated survey to license a claim about change: the same instrument, administered to a comparably defined population, with the fielding dates clearly separated from the publication date. The AARP pair satisfies the first two explicitly, because the 2025 report describes itself as a follow-up using the earlier instrument. It satisfies the third by virtue of being close enough in publication-to-fielding lag that the distinction matters less. The Cigna pair is more ambiguous on the first condition, and the CDC and HINTS-6 examples show why the third condition cannot be waved away — a report published this month can describe a country as it stood two or three years ago, and readers who conflate the two dates will misread the pace of change in either direction.

None of this is a case for distrusting survey data on loneliness generally. Gallup’s cross-national work, the Surgeon General’s 2023 advisory, and the National Academies’ 2020 consensus report all rest on sound methodology within their own scope. The caution is narrower: prevalence figures answer the question “how much loneliness did this instrument detect in this population at this time,” and stacking two such answers into a trend requires that only the third term changed. When a series is designed to hold the first two constant, as AARP’s is, a rise or fall in the number is informative. When it is not, the honest response to “loneliness went up” or “loneliness went down” is to ask which of the three terms moved, and to treat the headline figure as provisional until that question has an answer.

A better version of this evidence base would standardise on a small number of validated instruments — the UCLA scale is the obvious candidate given how widely it is already used — and require survey sponsors to report fielding dates alongside publication dates as a matter of course. Short of that, the practical rule for anyone reading two waves of the same-sounding survey is to check, before comparing the numbers, whether anything besides time was actually held fixed.