What the National Loneliness Average Hides About Men

The U.S. Surgeon General’s 2023 advisory put the headline figure at roughly half of American adults reporting loneliness. Gallup’s 2023 global survey found 24% of people worldwide report feeling very or fairly lonely. Cigna’s 2025 report puts American loneliness at 57%. These are useful numbers, and they are also averages, which means they describe a distribution while concealing where the mass sits within it.

One subgroup the average consistently flattens is men, and specifically men’s friendship networks, which several sources track with unusual precision and which show a trend distinct from the general loneliness figures reported alongside them.

The number that does not show up in most loneliness surveys

Most loneliness instruments ask about subjective feeling: how often does a person feel isolated, or lack companionship. The Survey Center on American Life’s 2021 survey asked a different, more structural question: how many close friends does a person actually have. The answer, drawn from a nationally representative sample, is not subtle. In 1990, 3% of Americans reported having no close friends. By 2021, that figure had risen to 12%. Among men specifically, the share reporting no close friends rose fivefold over the same period, to 15%. The share of men with six or more close friends fell from 55% in 1990 to 27% in 2021 — roughly half.

This is a structural measure, not a feeling. It describes a property of a person’s network, not their emotional response to it. That distinction matters because isolation and loneliness are independently predictive of health outcomes and do not move in lockstep; a person can have a small network and not feel lonely, or a large one and feel lonely anyway. But a fivefold rise in the share of men reporting zero close friends is not a fact any topline loneliness percentage captures, because most loneliness surveys do not ask the network-size question at all. This is an instrument mismatch worth naming directly: a 57% loneliness figure and a 15%-of-men-have-no-friends figure are not competing estimates of the same thing. They are answers to different questions, both real, both consequential, and not reducible to each other.

Where the age breakdown goes the other way

The subgroup story gets more specific once age is added. AARP’s 2018 survey of adults 45 and older, using the 20-item UCLA Loneliness Scale, found one in three reporting loneliness, with network size and diversity as the strongest predictors — consistent with the friendship-network data. AARP’s 2025 follow-up, using the same instrument and therefore genuinely comparable to the 2018 wave, reports loneliness rising further in this age group. That is a rare thing in this literature: a repeated measure with a stable instrument, which means the change can be attributed to something other than a change in what was asked.

Set against that, Gallup’s global data finds loneliness highest among the youngest adults (27% among 19-to-29-year-olds) and lowest among adults 65 and older (17%). Read naively, an older reader might conclude loneliness is a young person’s problem and largely resolves with age. The AARP data complicates that: among adults 45 and older specifically, loneliness is common and, on a stable instrument, increasing. The apparent contradiction is not really a contradiction — it likely reflects the fact that “older adults” is a wide and heterogeneous category, and that the youngest end of it looks different from the oldest. But it is exactly the kind of thing an average by decade of birth will erase, and neither survey alone resolves it, because they use different populations, different instruments, and different reference periods.

What the intervention evidence says about matching help to the right group

The subgroup problem recurs in what has been tried. Social prescribing — referring lonely or isolated people into community activities, often through a health provider — has been evaluated mostly through qualitative and small-sample work, and a 2022 meta-synthesis found that participants describe benefit mainly when the activity restores a sense of purpose, not simply contact for its own sake. That is a finding about mechanism, and it implies something uncomfortable for the “no close friends” subgroup identified above: an intervention built around low-stakes social contact may not be built for people whose deficit is structural network size rather than a lack of casual encounters.

The two randomised trials in this literature offer a sharper test, because unlike most social prescribing evaluations, they have a control arm. The 2024 HEAL-HOA trial tested volunteering and prosocial engagement against a control among lonely older adults in Hong Kong. A 2025 trial of befriending in residential aged care found reductions in UCLA Loneliness Scale scores of 2.39 points at eight weeks and 2.71 points at sixteen weeks relative to control — a real, measured effect, on the same instrument AARP uses, which makes it unusually easy to compare against prevalence data rather than against other interventions’ own bespoke measures. But a 2026 follow-up trial deliberately used befriending as the comparison condition against a telephone-delivered behavioral activation program for older adults living in poverty, alone, and digitally excluded, and found the more structured intervention outperformed the more common one. Two takeaways follow. First, befriending works, on average, for a general older-adult population. Second, “on average” is doing a lot of work in that sentence, and for the most isolated and most disadvantaged subgroup within that population, a lighter-touch social contact model was outperformed by something more structured.

None of the randomised evidence above was designed around the specific subgroup this piece has been describing — middle-aged and older men with structurally small friendship networks. That is the gap. The friendship-network data comes from a general survey; the intervention trials were run on lonely older adults broadly, on aged-care residents, or on people in poverty, not on men selected for having zero close friends. There is no trial in this list that tests a social connection intervention against that specific population, and it is a reasonable guess that the mechanism failure described in the 2022 meta-synthesis — contact without restored purpose — would apply with particular force to a group whose baseline network size has fallen by half in three decades.

What would strengthen this picture is a trial stratified by baseline network size rather than by age or housing status alone, testing whether structural isolation and subjective loneliness respond to different interventions, in the population where the friendship-network decline is sharpest and least studied.