What the Time Diaries Say That the Surveys Don't

Cigna’s 2020 survey found 61% of U.S. adults sometimes or always felt lonely. Gallup’s global survey, fielded 2022 to 2023, found 24% felt very or fairly lonely. The 2022 HINTS-6 data, published in 2025, put moderate-to-severe loneliness at just over 37%. These are not small discrepancies, and they did not arise because American loneliness tripled and then halved in three years. They arose because three different instruments asked three different questions of three different populations at three different moments, and each produced a number that behaves, on the page, as if it were the same fact.

This is the recurring problem with self-report loneliness data: it is sensitive to wording, response scale, and framing in ways that make year-over-year and survey-over-survey comparison close to meaningless without matching instruments. The BMC Public Health review of the field’s state as of 2023 named inconsistent measurement as a central barrier to comparing findings across studies, and the numbers above are exactly why. AARP’s advantage, by contrast, is that its 2025 follow-up survey used the same 20-item UCLA Loneliness Scale as its 2018 baseline, which is the rare case in this literature where a trend line can actually be trusted rather than merely plotted.

What happens when the measure is behavioral instead

Set the self-report figures aside and look at what people actually do, as recorded by time diaries rather than recalled feeling. Reporting on American Time Use Survey data found that face-to-face socializing among Americans has been declining for years, a decline that began before the pandemic and continued after it. That is a materially different shape from the self-report record, which lurches: a spike attributed to pandemic disruption, partial recovery, renewed concern. The time-diary trend does not lurch. It slopes downward steadily, because it is recording something structural — hours actually spent with other people — rather than a subjective judgment about how connected someone feels on the day they answer a phone survey.

The distinction matters because isolation and loneliness are not the same construct, a point a 2024 Scientific Reports study on the pandemic period made explicit: the relationship between the two varies by age and does not move in lockstep. A person can report feeling less lonely while spending measurably less time with others, and vice versa. Time-use data captures the former category — isolation as a structural fact — and self-report surveys capture the latter. Comparing headline “loneliness” percentages across a self-report survey and a diary study is not really a comparison at all; it is two different variables wearing the same label.

Administrative data adds a third, harder-edged check

The most useful recent evidence on this point is not a survey at all. A 2026 Science study used administrative records — actual use of mental health services and actual prescriptions filled — as its outcome, rather than asking people how connected they felt. It found that the shift to remote work increased time spent alone and worsened measurable well-being outcomes, and that remote work’s rise explains roughly a third of the total increase in isolation and mental distress between the 2011–2019 period and 2022–2024. It further found that remote workers who live alone saw their days-spent-alone increase roughly tenfold relative to workers who live with others.

That is a number self-report cannot easily generate, because it does not depend on anyone rating their own loneliness on a scale. It depends on whether a person filled a prescription or used a mental health service, which is a coarse proxy for distress but not one subject to the framing effects that make survey percentages jump around. The administrative measure and the time-diary measure agree with each other in direction — more time alone, worse outcomes, and a trend that moves steadily rather than in survey-driven spikes — even though neither one asked a single person how lonely they felt.

The mismatch is informative, not just annoying

None of this means self-report should be discarded. Subjective loneliness predicts outcomes — mortality risk, health service use — independently of objective isolation, which is why the Surgeon General’s 2023 advisory and the mortality meta-analyses treat it as a distinct risk factor rather than a stand-in for isolation. The point is narrower: when a self-report percentage and a behavioral or administrative trend diverge, the divergence is data, not noise to be explained away. It says something about which construct is moving and which is not.

A better study on this question would follow the same individuals over years with all three measures at once — a repeated self-report instrument held constant, time-use diaries, and linked administrative health records — so that the gap between what people say and what they do and what shows up in a clinical record could be measured directly, in the same people, rather than inferred by lining up three separate literatures and hoping they rhyme.