Friendship circles shrink by almost one person a decade, research finds
A 2013 meta-analysis finds friendship networks shrink by almost one person a decade, and what that means after 40.
By The Graze · November 17, 2026 · 6 min read

Midlife
The idea that, at middle age, you cannot afford unsuccessful friends sounds like a rule of the club. The research on how friendships change after 40 is more precise, and less useful as a status test. Networks shrink. Close ties tend to stay. Americans say they have fewer close friends than a generation ago, but that comparison is imperfect. And the best large study of who is friends with whom does not support pruning for lack of worldly success. If anything, ties across class lines are among the strongest predictors of upward mobility for poorer children.
What shrinks, and what does not
In 2013, Cornelia Wrzus and colleagues published a meta-analysis in *Psychological Bulletin* of 277 studies and 177,635 people. The global social network, they found, increases up until young adulthood and then decreases steadily. Both the personal network and the friendship network decrease throughout adulthood. The family network is stable in size from adolescence to old age.
The shrinkage is modest in any single year, but it adds up. Personal and friendship networks "became smaller by almost one person per age decade." The unstandardised annual coefficients were −0.09 for personal networks and −0.08 for friendship networks. Studies of life events that occur at particular ages, including the transition to parenthood, job entry and widowhood, found network changes similar to the age-related pattern.
There is a period effect as well, not just an ageing one. Friendship networks shrank by 0.9 people with every five-year period. Studies from 1980 to 1985 recorded, on average, about 3.6 more friends than studies from 2000 to 2005.
Socioemotional selectivity theory, associated with Laura Carstensen, says the cuts should fall mainly on peripheral ties, as older adults prefer emotionally meaningful relationships. The meta-analysis fits that picture: friendship networks fall, family networks do not. It is not a finding about the worldly success of the people who remain.
The American survey, and its own caveat
The sharpest popular claim is that close friendship itself has thinned in the United States. The usual source is the Survey Center on American Life's May 2021 American Perspectives Survey, directed by Daniel A. Cox at the American Enterprise Institute. Interviews were conducted online, between 14 and 23 May 2021, among 2,019 adults on Ipsos's KnowledgePanel, a probability-based panel designed to represent the US general population.
Close to half of those adults, 49 per cent, said they had three or fewer close friends. A further 36 per cent had between four and nine, 13 per cent had 10 or more, and 12 per cent had none. Nearly six in 10, 59 per cent, said they had one person they considered their best friend; 40 per cent said they did not. Two-thirds, 67 per cent, said they had a friend known since childhood.
The report compares those figures with a Gallup poll from January 1990, in which 27 per cent of adults had three or fewer close friends, 33 per cent had 10 or more, and only 3 per cent had none. Three-quarters, 75 per cent, had a best friend. The comparison is not a clean before-and-after. The 2021 report's own footnote is plain: the Gallup poll was administered over the telephone, with a live interviewer; the 2021 study was conducted entirely online. Telephone respondents, the report says, "might have been less willing to report having no friends or inflate the number of close friends they had" because of a live interviewer — social desirability bias, in the report's phrase. The direction of the change can still be real. The jump from 27 per cent to 49 per cent is not a like-for-like measurement.
The study that is not a midlife pruning rule
If the question is whether successful people should drop unsuccessful friends, the most ambitious recent work on friendship and status points the other way, and at a different stage of life. In 2022, Raj Chetty and colleagues measured social capital with data on 21 billion Facebook friendships. The first paper in the series, published in *Nature* and circulated as NBER working paper 30313, uses 72.2 million users aged 25 to 44 — 84 per cent of that US age group.
The measure that matters is economic connectedness: the fraction of high-socioeconomic-status friends among low-socioeconomic-status people. It is, the authors write, "among the strongest predictors of upward income mobility identified to date." If children with low-SES parents "were to grow up in counties with economic connectedness comparable to that of the average child with high-SES parents, their incomes in adulthood would increase by 20% on average."
That is a finding about children's later mobility, and about ties that cross class lines. It does not test whether a successful 45-year-old does better by keeping only successful friends. If it is borrowed as a rule of the club, it is being borrowed backwards.
| Measure | Finding | Source |
|---|---|---|
| Friendship network, adulthood | Falls by almost one person per decade (annual coefficient −0.08) | Wrzus et al., 2013, 277 studies |
| Close friends, US, 2021 | 49% have three or fewer; 12% have none | American Perspectives Survey, May 2021 |
| Close friends, US, 1990 | 27% had three or fewer; 3% had none (telephone Gallup; not like-for-like) | Gallup, January 1990, compared in the 2021 report |
| Economic connectedness | Among the strongest predictors of upward mobility; 20% income lift in the county thought experiment | Chetty et al., 2022 |
Compiled by The Graze from Wrzus et al., 2013; the May 2021 American Perspectives Survey, including its comparison with Gallup, January 1990; and Chetty et al., 2022, with figures checked against NBER working paper 30313. The 1990 and 2021 surveys used different modes. The 20 per cent figure is a thought experiment about children, not a midlife rule.
Our take
The question "can you afford unsuccessful friends?" treats midlife friendship as a portfolio. The evidence does not. What reliably falls after young adulthood is the outer ring: weaker friends, acquaintances, the ties that a child, a job or a death makes harder to keep. Family networks stay the same size. A childhood friend is still a common fact of American life. That is a story about time and close ties, not about other people's careers.
The comparison with 1990 is the one that looks most like a friendship recession, and it should be told with the report's caveat. A telephone poll with a live interviewer is not the same instrument as an online panel. People may have claimed more friends in 1990 than they would put on a screen. The direction of the change can still be real. The size of the jump, from 27 per cent to 49 per cent with three or fewer close friends, is not a clean measurement.
Chetty's work is the study most often borrowed as a status argument. It is the opposite kind of finding. The paper is about poor children's later incomes, not about successful people curating a table. If the share of high-status friends among low-status people is among the strongest predictors of upward mobility, a rule that the successful should drop the unsuccessful is not a research result. It is a prejudice looking for a citation.
Sources
- Cornelia Wrzus, Martha Hänel, Jenny Wagner and Franz J. Neyer, Social network changes and life events across the life span: a meta-analysis, *Psychological Bulletin*, 2013, Vol. 139, No. 1, pp. 53–80 (277 studies, 177,635 participants; friendship and personal networks; family-network stability; life events; period effects)
- Raj Chetty, Matthew O. Jackson, Theresa Kuchler, Johannes Stroebel et al., Social capital I: measurement and associations with economic mobility, *Nature*, 608, 108–121 (2022) (economic connectedness and upward mobility)
- Raj Chetty et al., NBER working paper 30313 (the working-paper version of the same study, from which the sample size, 21 billion friendships, 72.2 million users aged 25–44, and the 20 per cent county thought experiment were checked)
- Survey Center on American Life / American Enterprise Institute, Daniel A. Cox, The State of American Friendship: Change, Challenges, and Loss (American Perspectives Survey, 14–23 May 2021, N = 2,019, Ipsos KnowledgePanel; close-friend and best-friend figures, and the comparison with Gallup, January 1990, including the telephone-versus-online caveat)
- Survey Center on American Life / American Enterprise Institute, PDF of the same report (tables and footnote 7 on social desirability bias)







