In a recent and updated post, I reviewed the evidence clearly showing that there was and is not a so-called “loneliness epidemic.” (Surprise! Despite my post, the epidemic of epidemic alarms continues.) Many media stories about loneliness (e.g., here, here, here) also assert that American adults have fewer friends or weaker friendships than they did some time before. While similar to the panic about loneliness, this claim is about relationships, not about emotions. So, what is happening to friendship?
In Still Connected (2011), I assessed the data from 1970 to 2010 and concluded that not much happened to friendship in those years; a 2014 post reaffirmed that conclusion. However, later reports seem to describe a more recent decline in American friendships. And, unlike with loneliness, there may indeed be something happening here, though what it is ain’t exactly clear (apologies to Buffalo Springfield).
Friendship is a messy topic in part because of confusion about what “friends” mean and how to count them. So, this post also becomes one about meaning, measurement, and method. In the end, the 2010s do look a bit different. The reasons why range from the social media (probably not) to economics to housing arrangements (maybe).
What is a Friend?
Friendship as we know it is a modern development. Only in the last few centuries in the West have personal, close relationships independent of other social bonds such as lineage, patronage, village, alliance, or commerce emerged. Having friends of this sort became increasingly important to middle-class westerners, especially women, in the nineteenth century and to yet more people in the twentieth. “A wider range of people” in the first half of the last century, writes historian Mark Peel, “came to regard a particular form of intimate and emotional friendship as a crucial component of a good life. . . . [It] was the social glue of modernity.” The later twentieth century “saw the triumph” of this ideal as yet more people came to rely on friends for support and as cultural authorities increasingly “prescribed” such social bonds.[1]
Among westerners, Americans have used the word friend especially loosely compared to, say, the French. We typically apply the label to just about anyone with whom we are friend-ly and who is not a relative. And even then, we vary a lot;[2] some people, for example, do include relatives when they count their friends.
A 1990 Gallup Poll illustrates this looseness. It asked Americans, “Approximately how many friends, if any, do you have?” The median answer was 15, but estimates ranged from three at the 10th percentile to 91, or thirty times as many, at the 90th percentile.[3] This tells us something about how Americans vary in friends, but it may tell us more about how Americans vary in what they mean by friend.
Given such fuzziness, how people answer questions about friends may change over time even if their actual relationships do not. Who knows, for example, what extra confusion was created when Facebook decided about 15 years ago to call users’ online contacts “friends,” instead of calling them, say, “contacts,” “chums,” or whatever?
So, it’s muddled, American friendship, but it is nonetheless highly important to contemporary Americans. And they read articles online that worry over its condition. What do the data, however ambiguous, tell us about American friendship in recent years?
Counting Friends
Not much–because we have few surveys with representative samples that use the same question over a range of years. (Many surveys ask about friends, but few are true replications.) And sometimes the same question is not really the same. Let’s try, anyway.
First, consider the percentage of Americans who told surveys that they had zero or only one or two “close friends.” I found two series with very similar wording. One only covered three years in the 2000s–shown in orange in the graph below. The wording of the question about friends suggested a high (and modern) standard for who respondents should count: “people you feel at ease with, can talk to about private matters or call on for help.” About 23% said that they had two or fewer such friends.
The blue dots represent results from three different organizations using what seem to be the same questions. And there we see a change: Gallup’s tally of the percentage of Americans with zero to two “close friends” (not counting relatives) dropped slightly between 1990 and the 2000s. But closely similar versions of the question asked in 2021 and 2023 by other organizations found much higher proportions similarly short of friends. And, thus the headlines: “Americans Have Fewer Friends than Ever Before,” said the New York Post in 2021–though “ever” is a long, long time.
Why? What happened between 2003 and 2021? The Great Recession happened; Donald Trump and extreme political polarization happened; Covid happened; and, many point out, social media happened. Facebook launched in 2004 and had 235 million American users by 2020. We can look for explanations in these places—and also, I will suggest, in changing living arrangements.
Unfortunately, what also happened between 2003 and 2021 was a technical change in the polling. When is the same question not the same question? The short answer here–the long answer is in this endnote[4]–is that the ASC in 2021, unlike Gallup in 1990, prefaced the friends question with almost three dozen items that focused attention on respondents’ feelings and friendships, asking, for example, “When was the last time you … shared personal feelings” with a friend?). That long introduction probably raised the bar for who counts as a “close friend” and discouraged the 2021 respondents, unlike the 1990 ones, from claiming many friends.
Maybe so, but what about the 2023 Pew Survey, when 29% said that they had two or fewer close friends? That survey does not have the same issue with location of the question.[5] Pew did use an online panel rather than telephone polling like Gallup–and people are less willing to admit an embarrassment such has having having few friends when talking to an interviewer (1990) than when tapping on a computer (2023)–but the 20+ point change in claiming no or few friends is probably too large to explain by method. Something probably happened between 2003 and 2023. I’ll return to that later.
Other Approaches to Assessing Friendship
Some surveys try to measure people’s activities with friends.[6] The longest such series is the General Social Survey’s (GSS) question asking respondents how often they “spend a social evening with friends who live outside the neighborhood,” as shown in the graph below. Over about 40 years more respondents said they “often” spent such an evening and fewer said they rarely did so (and recently-born generations reported this social activity more than did earlier generations). But that trend of greater sociability reversed sharply, dropping from 2010 to 2018. (I am setting aside the dramatic shift that coincided with Covid.[7]) Notably, the proportion who said that they saw friends at least weekly–the blue line–dropped, eventually by 6 points; that decline was particularly acute among the unmarried, young respondents, 13-points.[8]
Some researchers have used time budget surveys in which respondents describe in great detail what they were doing and with whom in each hour of the previous day. One analysis showed no change over the 20 years running up to Covid in the proportion of Americans who went without any social contact throughout a whole day–about one in 14. Another study, though, focused on how much time respondents reported being in the company of “friends” and found a drop of about half between 2003 and 2019, friend minutes dropping especially during the 2010s. A closer look reveals that between 2003 and 2015 only 15-to-24-year-olds reported a decline in time with friends, but between 2015 and 2019 all age groups reported some decline.[9]
Yet another analysis of the same data examined how much time Americans spent in “solitary leisure.” Between 1965 and 2018 Americans increasingly reported time relaxing alone. The major reason by far was that more Americans were watching more television. After 2003, the trend was fueled less by tv and more by young and unmarried Americans spending time on computers and playing video games. Notably, married Americans reported showed no further increase in solitary leisure in the 21st century. (A study of time use by young Swedes yielded similar results.)
Let’s focus on that gaming. Most Americans these days say that they play video games. Their gaming time seems to have come largely out of television time; total screen engagement has been pretty flat over the last couple of decades. We can think of gaming as friendless but most players say that their gaming is a social activity done with others, in person or virtually, and is even a source of new friends. Some research backs up the claim that gaming can foster friendship. Does such joint gaming fill in the drop we saw in in-person friend time? I don’t know (and I don’t game).
Tentative Conclusions
Friendships seem to have changed some for some Americans specifically in the last several years before Covid (and, of course, changed more during Covid). If Americans, especially young adults, were reporting fewer friends and/or less activity with friends in the late 2010s than before, why?
* I am skeptical about a smartphone or social media explanation. The SMS, IM, DM, and even the phone functions of smartphones promote more sociability, not less. That’s what Americans overwhelmingly believe about mobile technology. Moreover, the purported injuries of social media–say, distraction and invidious comparisons–are about too much friend activity, not too little.
* The economic injuries of the Great Recession and its slow, unequal recovery may have crimped young people’s social lives (and spirits, as well).
* The rise of video gaming as a substitute for in-person leisure is a plausible explanation, especially given how the trends I’ve reviewed are concentrated among the young and unmarried. But gaming seems to have mainly replaced watching television and it is probably a more social technology than television.
* The centrality of young adults in this story leads to another intriguing explanation: In the last thirty years, especially during the 2010s, many more young adults, especially men, have been living with their parents rather than with a spouse or partner (doing so perhaps because of low pay, high housing costs, and extended schooling).[10] Living with Mom and on a limited budget is more conducive to gaming alone than to having friends over.
The travails of young adult men, notably those without a college degree, is a big story on its own, of course. And for another post.
END NOTES ————————-
[1] Peel, in Caine (ed.), Friendship: A History, 2009, pp. 279, 317.
[3] Later in this 1990 poll, Gallup asked respondents how many “close” friends who were not relatives that they had. The median answer then was five, with range from 2 at the 10th percentile to 20 at the 90th percentile. I downloaded the data from iPoll.
[4] The most persuasive presentation of a change in the number of friends Americans claim to have is from a report of the American Survey Center. It juxtaposes the answers Gallup Poll respondents gave in 1990 to the answers its own respondents gave in 2021 to a very similar question. The differences are dramatic: In the 1990 Gallup Poll, 7% said they had no or only one friend (33% said 10 or more) and in the 2021 ASC poll 19% said no or onloy one friend (13 percent said 10 or more), almost tripling the relatively friendless. My rule when I see such large historical changes is, first, check the method; is it an artifact? In 2022 I had a congenial talk with ASC’s excellent director, Daniel Cox. We mused about a methods difference in the mode of interview: Gallup used random telephone interviews, the Center a panel surveyed online. That difference is important—people are less likely to give socially desirable answers, like reporting many friends, online—but is not likely to explain such a big difference in results. I think I have since found larger methodological concerns (thanks to the Roper Center’s iPoll archive and to ASC’s online materials).
Gallup’s 1990 survey opened with three personal questions (“Who do you usually talk to first when you have a personal problem?;” “When something good happens to you, who do you usually tell first?” “How often do you ever feel lonely?”) and then asked, “Approximately how many friends, if any, do you have? Just your best estimate please. (Include anyone considered a friend.)” Note: This is not the question used in the graph above nor the one the American Survey Center used; that question came later. Only three percent said none or one to the first friend counting variant. The Gallup script for interviewers then asked six questions about friends, from “Are you satisfied with the number of friends you have…?” to “Is your best friend a member of your family…?” and who that would be. Then came the key question, the one approximately later copied by the Center, “Not counting you relatives, about how many close friends would you say you have?” (7% said zero or none in 1990). The 1990 Gallup poll followed with many detailed questions about friendship.
In 2021, ASC did not ask the first version of the friends question at all. But it did ask 34 detailed questions about friendship before coming to the key question (“Not counting your relatives, … ”). These 34 preceding questions asked about feelings, personality, and mood, and also asked “How satisfied … are you with … the number of friends you have?,” “When was the last time you … shared personal feelings…,” “… received emotional support from a friend?,” and “How long ago did you most recently make a new friend?” Then came its variation on the Gallup “close friends” question, with a new preface: “Now, thinking only about friends you are close to [italics added], not counting your relatives, about how many close friends would you say you have?”
I submit that (a) the 1990’s general friend question served as a first estimate, a marker, for 1990 respondents when they later answered the close, non-relative friends question. (Thirty percent of respondents estimated having as many or even more friends who were “close and not relatives” as they estimated who were friends of all kinds.) This marker effect was absent in 2021. (b) The ASC 2021 version raises the standard for “close friends” with that introduction, “Now, thinking only about friends you are close to…” (c) And probably most important, many of the 34 questions asked in 2021 before the count-your-close-friends question raised the standard for who is a close friend in profound ways. That is, a close friend is someone who shares personal feelings, provides emotional support, and so on. It would make sense, then, for the 2021 respondents to low-ball their number of number of close friends after getting that primer on close friendship.
[5] It appears after a battery of questions concerning work and another battery concerning gender issues.
[6] One might also look at satisfaction with friendships but that opens up another box of complexities, notably, whether there are historical changes in friendships or instead historical changes in expectations of friendships.
[7] And setting aside methodological issues.
[8] This compares unmarried respondents aged 18 to 34 (dropping from 47% to 34%) to all respondents. And the change was especially great for those unmarried, young respondents living in large center cities–30 points, from 58% to 28%. Why, I don’t know.
[9] Figure 4 and the online supplement for Kanan and Viaze (2023) shows the following complex story: (1) Between 2003 and 2015, 15-24 year-olds’ reports of time with friends dropped from 153 to 105 minutes, minus 48. Other age groups started at about 40 minutes with friends in 2003 and stayed there through 2015. (2) Between 2015 and 2019, the youngest’s minutes dropped an additional 38 and other groups’ minutes by about 10. So something changed for everyone in 2016 through 2019, but again much more for the youngest.
[10] U.S. Census data shows the following living arrangements over the last 30 years for Americans aged 25 to 34. (The numbers shown are three-year averages around the dates.) Note that “Parents, kin” includes–mostly for women–living just with one’s children. For men, the percentage living with parents (and no children) increased from 14.1 to 20.8 percent, almost a 50 percent jump in relative terms.
PERCENTAGE OF 25-to-34 YEAR-OLDS, BY LIVING ARRANGEMENTS AND GENDER
MEN | WOMEN | ||||||||
ALONE |
SPOUSE/PARTNER |
PARENTS, KIN |
NONKIN |
ALONE |
SPOUSE/PARTNER |
PARENTS, KIN |
NONKIN |
||
1990 | 11.5 | 60.5 | 20.2 | 7.9 | 7.4 | 66.2 | 22.7 | 3.6 | |
2000 | 11.7 | 57.1 | 20.4 | 10.8 | 8.1 | 63.3 | 23.4 | 5.1 | |
2010 | 11.3 | 55.1 | 24.7 | 8.9 | 8.1 | 63.1 | 24.9 | 4.0 | |
2020 | 11.9 | 49.4 | 29.9 | 8.8 | 9.1 | 59.2 | 26.9 | 4.8 | |
2020- 1990 |
0.4 | -11.1 | 9.7 | 0.9 | 1.7 | -7.0 | 4.2 | 1.2 |