What Loneliness Policy Cannot See Without a Longitudinal Study

The U.S. Surgeon General’s 2023 advisory reports that about half of American adults experience loneliness. Harvard’s Making Caring Common survey, published in 2021, found 61% of young adults aged 18 to 25 reporting serious loneliness, against 36% of Americans overall. Gallup’s 2023 global survey put the youngest cohort, 19 to 29 year-olds, at the highest loneliness rate of any age group worldwide, 27%. These figures get cited constantly in policy documents as evidence that young adults are a distinct risk group requiring targeted intervention.

They are almost all cross-sectional. Each is a single measurement, at a single moment, across many different people. That design tells a policymaker how loneliness is distributed right now. It cannot tell them how any individual’s loneliness changes over time, what precedes a rise, or whether the young adults who report high loneliness at 22 are still lonely at 27. Those are different questions, and the policy response to each would look different.

What a snapshot actually licenses

A cross-sectional survey answers one question well: at this moment, what fraction of this population reports this state. The Harvard survey’s 61% figure among 18-25 year-olds, and the AEI Survey Center’s 2021 finding that 12% of all Americans report no close friends at all, up from 3% in 1990, are both legitimate answers to that question. The AEI comparison across two points in time — 1990 and 2021 — is closer to longitudinal, but it compares two different samples of Americans at two moments, not the same people tracked through their lives. It shows the aggregate has shifted. It does not show any individual’s trajectory.

This distinction matters more for young adults than for most groups, because early adulthood is a period when social ties are unusually unstable. People finish education, move cities, change jobs, form and dissolve romantic partnerships, at a pace that later decades of life do not repeat. A single cross-sectional reading of a 24 year-old’s loneliness could be catching a transient state tied to a recent move, or a stable condition that will persist for a decade. The Harvard figures cannot distinguish these. Neither can Gallup’s.

The three things a longitudinal design would show

Persistence versus turnover. If the same cohort of young adults were surveyed annually across, say, ages 20 to 30, a longitudinal design would separate people who are lonely once from people who are lonely repeatedly. These are plausibly different populations with different needs. A person lonely for one year following a specific disruption — a graduation, a breakup, a move — may resolve without intervention as their circumstances change. A person lonely across five consecutive annual waves is a different case, and probably the one health systems should prioritize, per the general logic in the National Academies’ framing of isolation as a health-system concern for older adults. No comparable youth-focused tracking exists at the scale of the youth loneliness surveys that get cited in policy.

Direction of the arrow. Cross-sectional data can show that loneliness correlates with, say, remote work, low social network diversity, or heavy phone use, but it cannot establish which came first. Holt-Lunstad’s 2015 meta-analysis in Perspectives on Psychological Science found isolation, loneliness, and living alone each independently predictive of early mortality, with effects that held after adjusting for baseline health status — a design feature that partly addresses reverse causation for that specific outcome. But most youth loneliness surveys do not have an analogous adjustment, because they only have one wave. Whether social withdrawal produces loneliness, or loneliness produces withdrawal, or a third factor drives both, is not answerable from a single reading. Policy that assumes the first causal direction — for instance, funding programs to increase social contact on the theory that contact reduces loneliness — is making an assumption the data does not test.

Whether declared national strategies work. The UK’s 2018 loneliness strategy, the first of its kind from any national government, committed to routine measurement through the Office for National Statistics specifically because a repeated, comparable measure was recognized as necessary to know whether the strategy was having an effect. That is the correct instinct. A repeated measure on the same population, or at minimum the same cohort definition, over years is what would let a government say whether youth loneliness fell after a given policy rather than merely differed from one year’s snapshot to the next. Absent that, a government can report that youth loneliness was 61% in 2021 and some different figure later, but cannot attribute the change to anything it did.

The nearest thing to an exception, and its limits

The JACSIS study, tracking Japanese respondents across 2020 and 2021, is longitudinal rather than cross-sectional, and for that reason it can speak to change within people rather than change in an aggregate. That is a meaningfully different and stronger kind of evidence. Its limits are narrower ones: two waves across a pandemic period, a specific national context, and an interval too short to say much about a normal, non-pandemic young-adult trajectory. It shows what a longitudinal design can do. It does not, by itself, substitute for one built around the specific question of youth transition.

BMC Public Health’s 2023 review of the state of loneliness research names inconsistent measurement as a central barrier to comparing findings across studies — a separate problem from the cross-sectional/longitudinal gap, but one that compounds it. A repeated-measures cohort study using a fixed instrument, tracked through the specific transitions of early adulthood — leaving education, first job, relocation, first serious partnership — would let researchers separate transient distress from chronic loneliness, test which precedes which, and give policymakers something closer to a dose-response answer about what interventions, if any, shorten the duration of a lonely period rather than simply correlating with a lower prevalence at one point in time.

Until that exists for young adults specifically, the honest summary is narrower than the headline figures suggest: a large fraction of young adults report loneliness at any given moment, the fraction is high relative to other age groups, and beyond that, the standard evidence base does not currently support claims about who stays lonely, for how long, or why.