The change is large whichever end of the friendship scale you examine. In 1990, 55 percent of American men said they had at least six close friends outside their family. By 2021, that share was 27 percent. Over the same period, the proportion reporting no close friends rose from 3 percent to 15 percent.
Those numbers come from the Survey Center on American Life’s State of American Friendship report, which compared its May 2021 American Perspectives Survey with a Gallup poll conducted 31 years earlier.
The percentages support the description of a dramatic contraction. They do not amount to a continuous measurement of the same people, however, and the two surveys were administered in different ways. That distinction matters when a fivefold increase is built from two endpoints.
Where the headline numbers came from
The 1990 benchmark was a Gallup telephone survey of 1,226 American adults. Its archived record is available through the Roper Center for Public Opinion Research. Respondents were asked not to count relatives and to estimate the number of close friends they had.
The 2021 comparison surveyed 2,019 adults between May 14 and May 23. Participants belonged to Ipsos’s KnowledgePanel, a probability-based online panel recruited through random selections of residential addresses and telephone numbers. People without internet access were provided with a device and connection, helping the panel represent more than the already-online population.
The 2021 results were weighted for age and gender, race, education, region, metropolitan status and household income. The stated margin of error for the whole sample was plus or minus 2.4 percentage points at the 95 percent confidence level. Estimates for men and women separately have wider uncertainty because each is based on a subgroup.
“Close friend” was left to respondents to interpret. That subjectivity is not necessarily a defect. Closeness is personal. But it means the count records whom each person believed qualified, not a standardized number of relationships that researchers independently inspected.
A phone conversation and an online form can produce different answers
The most important limitation is the change in survey mode. In 1990, a person gave the answer aloud to a telephone interviewer. In 2021, a person entered it privately online.
The report’s authors explicitly raise social desirability bias. Someone speaking to another human may be reluctant to say they have no friends, or may round their estimate upward because the socially comfortable answer sounds better. A private questionnaire may reduce that pressure.
This creates uncertainty in both directions. The newer survey may have captured isolation more honestly. Or part of the apparent historical decline may reflect a greater willingness to report it. The comparison cannot calculate how much each explanation contributed.
Question context can matter too. Respondents may set a stricter threshold after answering several questions about support, personal disclosure and contact. The word “close” supplies no fixed rule about duration, frequency or emotional intimacy.
None of these caveats makes the 2021 result disappear. They mean that 55-to-27 and 3-to-15 should be read as survey estimates with a method break, not a precise census of friendships lost.
Women’s networks contracted as well
The gender contrast is real within the reported comparison, but the decline was not exclusive to men. In 1990, 41 percent of women reported at least six close friends. In 2021, 24 percent did. Ten percent of women in the later survey reported no close friends.
Men therefore showed a 28-percentage-point drop in the six-or-more category, compared with 17 points among women. That is what the title means by a sharper contraction in large friendship networks. It does not mean most men had no friends: 85 percent still named at least one in 2021.
The same survey found differences in how friendships were used. Forty-eight percent of women, compared with 30 percent of men, said they had shared personal feelings in a private conversation with a friend during the previous week. Forty-one percent of women and 21 percent of men said they had received emotional support from a friend during that period.
Those are separate outcomes. The number of friends, the frequency of communication, emotional disclosure and support can move differently. A person may have several activity friends but no confidant, or one close friend who provides substantial support.
The pandemic was immediate context, but not the whole timeline
The 2021 interviews took place after more than a year of pandemic disruption. In the same survey, 47 percent of Americans said they had lost touch with at least a few friends during the previous 12 months, while 9 percent said they had lost touch with most of them.
Yet the results were not simply a record of friendships collapsing. Forty-six percent also said they had made a new friend within the preceding year. The survey captures both disruption and renewal during an unusual period.
Earlier Gallup results help locate the timing. When Gallup repeated its telephone question, the overall share of adults reporting six or more close friends was 47 percent in 1990 and 45 percent in 2003. On that measure and using the same survey mode, little changed through the early 2000s. The largest apparent contraction seems to have occurred later.
Nor did the low count obviously rebound after the acute pandemic period. The 2024 American Social Capital Survey of 6,597 adults found that 17 percent had no close friend outside their family, while 25 percent reported six or more. This was a different survey rather than a perfectly continuous series, but it points away from treating 2021 as a momentary anomaly.
Friend count is not the same as friendship quality
Six is a useful survey category, not a biological requirement for a good social life. Two reliable friends can be more meaningful than a much larger network of occasional contacts. Reporting zero also does not automatically mean someone is lonely, because family, partners and other relationships may provide closeness.
ScienceBlog has previously covered a 10-year study showing that older adults often lost peripheral ties while preserving close relationships. Shrinking networks can reflect selectivity as well as deprivation. The 1990-to-2021 survey comparison cannot determine which process applied to each respondent.
A 2024 Pew survey adds another complication. Men and women were about equally likely to report at least one close friend, but women were more likely to say they would turn to a friend for emotional support, 54 percent versus 38 percent. Among adults with friends, women communicated more frequently by text, social media, phone and video, while men were slightly more likely to see a friend in person often. The Pew findings separate access from use.
This is why “male friendship crisis” can be too blunt. Friendlessness, network size, contact and emotional reliance are related, but none can substitute for all the others.
The contraction is credible; its cause remains open
The survey report discusses later marriage, geographic mobility, parents devoting more time to children and changes in work as possible structural pressures on friendship. It did not experimentally test those explanations. Smartphones and social media are also frequently blamed, but the 1990-to-2021 comparison cannot isolate their effect from changes in family life, employment, housing, health or survey practice.
It also cannot show whether individual men lost friends. These were two cross-sections of the population, not a panel that followed the same respondents for three decades. Cohort replacement, changing interpretations of closeness and demographic shifts can all affect the endpoints.
The careful conclusion is substantial enough. American adults reported smaller friendship circles in 2021 than Gallup respondents did in 1990, and the fall in large networks was steeper among men. Later surveys continue to find many adults without a close non-relative friend. The size of the change is approximate, and the reason is unresolved, but the social contraction is difficult to explain away as one stray percentage.