Why 'One in Three' and 'One in Four' Are Not the Same Number

Two figures circulate constantly in writing about older adults and social connection. One in three. One in four. The first comes from AARP Foundation’s 2018 national survey of 3,020 adults aged 45 and older, which found that 33% met the threshold for loneliness on the 20-item UCLA Loneliness Scale. The second comes from the National Academies of Sciences, Engineering, and Medicine’s 2020 consensus report, which estimated that roughly one quarter of adults aged 65 and older are socially isolated. Both numbers get dropped into the same sentence routinely, as if they describe the same problem at slightly different rates. They do not.

Different populations, different constructs

Start with what looks like the smaller discrepancy: the age bands do not match. AARP surveyed adults 45 and older; the National Academies figure applies specifically to those 65 and older. Loneliness in the AARP data was in fact highest among the youngest cohort in the sample, adults in their late 40s and 50s, not the oldest. So the two studies are not simply reporting on overlapping populations at different moments — one includes a large block of late-middle-age adults who are, by the AARP data itself, more likely to report loneliness than people over 65.

The more consequential mismatch is conceptual. AARP measured loneliness: a subjective state, captured by asking people to rate agreement with statements like “I feel left out” or “I feel isolated from others.” The UCLA scale is the most widely used instrument in the academic loneliness literature, which makes the AARP survey unusually comparable to other research, but it is still measuring how a person feels about their social life, not how many people are actually in it.

The National Academies figure measures something structurally different: social isolation, meaning the objective size, frequency, and diversity of a person’s social contacts. A person can have a small network and feel entirely content with it. A person can have a large network and feel cut off from all of it. Cacioppo and Patrick’s 2008 work on the biology of loneliness makes exactly this argument — that loneliness functions as a subjective alarm signal, evolved to prompt reconnection, and that the alarm does not correlate cleanly with the actual number of people around a person. Isolation is a property of the network. Loneliness is a property of the person’s experience of it. They are related but not interchangeable, and a substantial body of research treats them as separate variables precisely because they behave differently.

Why this is not just semantics

The distinction would be academic if isolation and loneliness predicted the same outcomes to the same degree. They do not. Holt-Lunstad’s 2015 meta-analysis in Perspectives on Psychological Science, covering studies with mortality as the outcome, found an odds ratio of 1.29 for social isolation and 1.26 for loneliness — close in magnitude, but derived from measuring two different things, and the paper treats them as independent risk factors rather than proxies for one another. A separate figure in the same analysis, an odds ratio of 1.32 for simply living alone, adds a third, purely structural measure that overlaps with isolation but is not identical to it either.

This means a health system, or a local program, deciding where to intervene is not choosing between two estimates of the same underlying problem. It is choosing between two different targets. A program built around expanding contact — a phone-based outreach service, a transportation benefit, a senior center shuttle — addresses isolation as the National Academies defines it. It may do very little for someone whose loneliness stems from the quality or meaning of the contact they already have, rather than its quantity. The AARP survey’s own data point toward this: 33% of respondents who reported ever speaking with their neighbors were lonely, versus 61% of those who never had — a gap that shows contact matters, but the presence of some contact clearly does not prevent loneliness for a third of people who have it.

The instrument problem underneath the construct problem

There is a second layer worth separating out. Even within loneliness research, instruments differ. AARP used the UCLA scale, a validated, multi-item measure developed for research settings. Other surveys — Cigna’s workplace loneliness index, for instance — use their own single-item or composite questions calibrated to different thresholds. Two surveys asking “are you lonely” in different words, to different populations, at different times, will not produce comparable numbers even when both call the result “loneliness.” The AARP figure is comparatively strong methodologically because the UCLA scale is the same instrument used across decades of the academic literature, which is precisely why researchers can compare AARP’s 2018 result to other UCLA-scale studies but should be far more cautious comparing it to isolation estimates built on entirely different measurement logic, such as network-size counts or frequency-of-contact questions used in the National Academies’ underlying sources.

What this means for reading the two reports together

None of this means one study is right and the other wrong. Both are credible within what they measured. The error is treating “one in three” and “one in four” as competing estimates of a single phenomenon that a reader should try to average or reconcile. They are answers to different questions: how many older and midlife adults feel lonely, and how many older adults have objectively thin social networks. A population could plausibly have low measured isolation and high measured loneliness, or the reverse, and both conditions would be real and both would carry independent health risk according to the mortality literature.

The practical implication for anyone using these figures in a policy or program context is to name which construct is being cited and why. A grant application or briefing document that cites “one in three older adults are lonely” to justify a network-expansion intervention is citing the wrong number for that intervention’s theory of change; the AARP figure supports interventions aimed at subjective experience and meaning, not contact frequency. The National Academies report is explicit that it is calling for the health care system to screen for isolation and loneliness as distinct items, not a combined score, which is itself an acknowledgment that the two do not move together predictably enough to be treated as one measurement.

What would resolve the apparent disagreement is not a better survey of either construct alone, but a study that measures both isolation and loneliness in the same older-adult sample, using validated instruments for each, and tracks how far one predicts the other over time. Absent that, the honest reading of these two widely cited reports is that they are not in tension. They were never measuring the same thing.