When 'Up 7 Points' Doesn't Mean What It Sounds Like

Cigna’s 2020 loneliness report states that 61% of U.S. adults report sometimes or always feeling lonely, and that this figure is up 7 points from the year before. That single sentence carries a lot of unstated weight. A trend claim of this kind is only as good as the guarantee that the 2020 wave asked the same questions, of a comparably drawn sample, in the same way as the 2019 wave. The 2020 report available for review does not reproduce the prior year’s questionnaire or sampling frame, so the “up 7 points” figure has to be taken on trust rather than verified. That is not a reason to dismiss it. It is a reason to treat it as a claim about a survey series, not a claim about the population, until the methodology behind both waves is visible side by side.

This distinction matters more than it might seem, because most of what gets reported as “loneliness is rising” is actually a comparison between two cross-sectional samples that happen to have been fielded a year apart under the same organizational banner. That is a weaker design than it looks. A cross-sectional wave draws a fresh sample each time; differences between waves can reflect sampling variation, changes in who responded, or small wording shifts, not just a change in the underlying rate.

A cleaner comparison: Japan’s JACSIS panel

The contrast worth drawing is with the JACSIS study of isolation and loneliness in Japan, covering 2020 to 2021. That study is longitudinal rather than cross-sectional: it follows the same cohort across two time points rather than drawing two independent samples and comparing their averages. This is the design that actually supports a claim about change, because it holds the “who” constant and varies only the “when.” A rise or fall in reported loneliness in a panel study reflects something happening to real people over time. A rise or fall between two cross-sectional Cigna waves might reflect that too, or might reflect nothing more than ordinary sampling noise combined with an unreported tweak to the screening question. Absent the full methodology for both years, there is no way to distinguish the two.

Instrument mismatch within a single organization’s own output

A related problem shows up even when there is no time gap at all. Gallup published two loneliness-adjacent figures on the same day in October 2023: its World Poll estimate that 24% of people worldwide feel very or fairly lonely, and, jointly with Meta, an estimate that 72% feel very or fairly socially connected. A later Gallup analysis explicitly warns against treating these as opposite sides of the same coin, noting that connectedness and loneliness are not simple inverses of one another. That is worth taking seriously as a methods point, not just a rhetorical caveat. Two figures from the same fieldwork period, released by the same organization, measuring related but distinct constructs, can produce numbers that look reconcilable and are not. If that mismatch exists within a single release, it is a reasonable bet that it exists across releases separated by a year or five, where sampling, question order, and mode of administration are all liable to drift without anyone flagging it.

AARP’s 2018 survey of adults 45 and older offers a useful point of contrast on instrument choice, if not on time-series design. It used the 20-item UCLA Loneliness Scale rather than a proprietary index, which means its 33% figure for that age group is directly comparable to the wider academic literature that also uses the UCLA scale. That is a different kind of comparability than the year-over-year kind Cigna is claiming, but it illustrates the same underlying requirement: a number is only comparable to another number if both were produced by the same instrument, or by instruments with known equivalence. A proprietary index built for a corporate report has no such track record, which is one reason the BMC review of the loneliness and social isolation literature flags inconsistent measurement as a persistent barrier to comparing findings across the field, not just across studies from different authors, but sometimes across successive reports from the same source.

What would settle it

None of this means the direction of Cigna’s finding is wrong. Loneliness may well have risen among U.S. adults between 2019 and 2020; there are good substantive reasons, unrelated to survey design, to expect that it did during that period. The point is narrower: a single sentence citing a point change between two annual waves does not, on its own, establish a trend, and readers who want to cite that number as evidence of a rising rate should look for the underlying instrument documentation for both waves before treating the comparison as settled.

A stronger version of this kind of series would publish the exact wording and response options for every wave, alongside the sample size and sampling method, so that a reader could check whether “up 7 points” survives scrutiny or dissolves into methodology. Better still would be a genuine panel, in the JACSIS style, tracking the same respondents rather than two independent snapshots. Until survey series covering loneliness routinely do one or the other, year-over-year percentage changes reported in press materials should be read as suggestive at best, and never cited as if they carried the same evidentiary weight as a study designed from the outset to measure change.