Skip to content
All tools

Tools

Close friends, 1990 → 2024

Drag to turn · scroll or pinch to zoom · click a light to follow one person

Close friends · more friends, more light
none123456–910+

Building the town…

Sample City · a town of 500

1990Gallup
by phone
no survey in this series
How this was built
SurveySaid they have no close friends
1%
of everyone asked
SurveySaid they have one close friend or none
6%
of everyone asked
ModelLive in a small world
–
under 10 people within 2 steps
ModelConnected to the main web
–
can reach the big group

Follow one person

Click any light in the town, or pick someone in the middle of the pack, to see their circle in every year.

The town in 1990

Model

What people said in 1990

Survey

“Not counting your relatives, about how many close friends would you say you have?”

10%20%30%40%1None5110211384165146–93410+
Survey, share of those who answeredThis town

Gallup, Jan 18–21, 1990, phone. Mean 9.9, median 5. 2% didn't answer. Source

How this was built

  • The surveys set how many close friends each person has, year by year. The town matches them to within a tenth of a point.
  • The model decides who is friends with whom. Reach, small worlds and the main web come from that, so they're modeled, not measured.
  • How tight friend circles are was tuned to what people said in 2021 about whether their friends know each other.
  • 1990–2003 were phone surveys and 2021–2024 online, so part of the jump may be the method.
Full method, tables and sources

What you're looking at

Each light is a person and each warm line is a close friendship. Every dot is the same size; what changes is its color and how much light it gives off, which grows with close friends. People with 10+ glow the most, and a neighborhood full of friendships lights up. Cyan rings are people with no close friends; they give off no light. A line takes the color of its less-connected person, and long lines and lines to people with 10+ friends are drawn dimmer, so they don't drown out everyone else's. The town has 500 people in 18 neighborhoods made of small friend circles, plus some people who live between them. On screen each friend circle is spread out a little so its friendships are visible, and neighborhoods are drawn closer together so their light overlaps; who is friends with whom was decided before either.

The surveys

All of them asked a version of “Not counting your relatives, about how many close friends would you say you have?” and grouped the answers the same way.

YearWhoHowNone123456–910+No answer
1990GallupJan 18–21, 1990, phone1%5%10%11%8%16%14%33%2%
2003GallupDec 11–14, 2003, phone, n=1,0112%6%8%11%11%17%18%27%1%
2021Survey Center on American LifeMay 14–23, 2021, online (Ipsos KnowledgePanel), n=2,01912%7%13%17%11%13%12%13%1%
2024Survey Center on American LifeMar 29–Apr 15, 2024, online (Ipsos KnowledgePanel), n=6,59717%5%13%12%8%14%11%15%6%

How the town is wired

  1. Each person gets a fixed place in the town and a fixed sociability rank. Whoever is the most social person in 1990 is the most social in every year.
  2. For each year, close-friend counts are handed out by that rank so the town's shares match the survey. Counts inside 6–9 are spread evenly. Inside 10+ they follow a long tail averaging about 20, which is what Gallup's printed averages imply (1990 21.1, 2001 18.2, 2003 20.0; 2021 and 2024 printed no average and use 19.8). Nobody gets more than 50.
  3. Every pair of people gets an affinity: higher when they live closer, plus a fixed random spark. Friendships form from the strongest affinity down until everyone has their count. The sparks stay the same every year, so most friendships in a sparse year also exist in the denser years.
  4. How tight the friend circles are was tuned so the 2021 town matches what people said in 2021 when asked “How many of your close friends know each other?”

How satisfied people are

In 2021 the Survey Center on American Life asked how satisfied people are with the number of friends they have, and published the share who were very or completely satisfied at each friend count: none 29%, one 39%, two or three 43%, four or five 56%, six to nine 67%, ten or more 75%. Applied to each year's answers, that gives the estimates below. For 2021 it gives 51%, the same as the 51% the survey reported. It assumes people at each friend count felt the same way in every year, which no survey checked.

The same survey found no clear line between one close friend and none. After accounting for age, gender, race, education and marital status, 57% of people with no close friends and 57% with one said they had felt lonely in the past week, against 54% with two or three and 35% with ten or more. Averages can't say whether one friend is enough for a particular person, introverts included.

Very satisfied with number of friends (est.)1990200320212024
Share of adults60%59%51%51%

Nearby friends only (2024)

The surveys count close friends anywhere, including friends who've moved away. In 2024 the Survey Center also asked how many of those close friends live in your neighborhood or close by. The answers are groups, not counts, so this view assumes “all” means all of someone's close friends, “most” 75%, “some” 50%, “a few” 25% (at least one) and “none” zero. Each person with a close friend gets an answer in the survey's proportions. Answers run together by neighborhood, so people in the same neighborhood tend to be local together, the way places work; how strongly was tuned so the town's answers match the survey's. A 2024 friendship counts as nearby only if both people have room for it, so nearby friends are always some of a person's real friends. Earlier years didn't ask, so there's no nearby view for them.

Which numbers are measured and which are modeled

The surveys measured how many close friends people have. Reach, small worlds and the main web come from the model, because they depend on who is friends with whom, and no survey measured that. Read them as what a town shaped like this one looks like, not as findings.

Across 20 random towns, the typical person's reach within 2 steps went from 49 [range 42–51] in 1990 to 25 [21–28] in 2024, the bottom quarter's from 25 [21–28] to 6 [5–8], and the share living in a small world from 8% [5–9%] to 31% [27–33%]. A town three times the size gives nearly the same numbers.

Limits

  • Gallup also asked in December 2001. Its answers sit almost on top of 1990's (2% none, 35% with 10 or more, median 6), so it isn't drawn here. It still counts toward the 10+ average above.
  • 1990, 2001 and 2003 were phone interviews. 2021 and 2024 were online. People may admit to having no close friends more readily on a screen than to an interviewer, so part of the change could come from the method. The Survey Center on American Life says this about its own comparison.
  • The 2021 and 2024 question starts “Now, thinking only about the friends you are close to…” and ends “Just your best estimate”.
  • In 2024, 6% didn't answer (1–2% in the other years). Shares here are of people who answered.
  • The Survey Center reports 1990's “no close friends” as 3%. Gallup's own published table shows 1% none and 2% no answer. This page uses Gallup's table.
  • Survey answers aren't mutual: you can count someone who doesn't count you. In the town, every friendship counts for both people.
  • 500 people is a small town, so reach within 3 steps runs into its edges.

Sources

  • Gallup, “Americans Satisfied With Number of Friends, Closeness of Friendships,” Joseph Carroll, March 5, 2004. Table for 1990, 2001 and 2003. news.gallup.com
  • Survey Center on American Life, May 2021 American Perspectives Survey, topline Q.2, Q.6, Q.13 and Q.14; and “Yes, Having More Friends is Better,” August 9, 2021. topline PDF · report · loneliness by friend count
  • Survey Center on American Life, 2024 American Social Capital Survey, topline Q.22 and Q.23. topline PDF · report