Europe’s weekly-contact rate came close to its pre-pandemic level
ESS round 1 (2002–03) to round 11 (fieldwork 2023–24) · adults meeting socially weekly
↗Data story · Europe and the United States
Europe’s weekly-contact rate nearly recovered. The US is 8 points below 2018.
Derek Thompson's “You are no longer invited to dinner” argues that Americans have stopped having people over. Surveys that have asked the same question for decades, to a fresh random sample each time, let us check a close cousin of that claim on both sides of the Atlantic: how often adults see friends, relatives or neighbours, and how much time they spend with them.
By João Tomé · September 30, 2026 · European Social Survey 2002–2024 · US General Social Survey 1974–2024 · American Time Use Survey 2010–2024 · Eurostat time use and EU-SILC
One change deserves a closer look: Americans aged 15–29 spent about 70 fewer minutes a day with friends in 2024 than in 2010. Around 77% of that decline had already happened by 2019. Read the short story and see the two share cards.
TL;DR · Four signals in the data
How social contact is changing
The surveys track different things: how often adults meet, how that varies by age, and how much time people spend with friends.
1–2 / 4
The US weekly-contact measure remains below its 2018 level
GSS · weekly socialising · 1974–2024
↗Weekly contact among Europeans aged 15–24 fell between 2002 and ESS round 11
ESS round 1 (2002–03) to round 11 (fieldwork 2023–24) · ages 15–24
↗Young Americans spent about 70 fewer minutes a day with friends
ATUS · average minutes per day · ages 15–29 · 2010–2024
↗Each line shows change within its own survey. The levels cannot be compared across the ESS, GSS and ATUS.
Both drifted down slowly for decades, then split after 2020
In 15 European countries surveyed in every round, the share meeting people socially every week went from 65.7% in 2002 to 62.7% in 2018. It dipped to 58.0% in the pandemic round. ESS round 11, fielded in 2023–24, recorded 61.7%. In the US it slid from 61.6% in 1974 to 55.3% in 2018, and was still at 47.0% in 2024.
Europe · 15-country ESS mean
United States · GSS
Share of adults meeting socially at least once a week. Each panel uses its own vertical scale, showing change within that survey. The European line is an unweighted mean of 15 countries present in all 11 ESS rounds, with each country weighted by the ESS post-stratification weight (Atlas of Social Attitudes export, pulled September 30, 2026). The US GSS is weighted by wtssps; weekly means “almost daily” or “once or twice a week” for any of relatives, neighbours or friends. US single-year margin about ±3 points (weights only).
Read each panel on its own scaleThe ESS asks how often people meet friends, relatives or colleagues. The GSS asks about social evenings with relatives, neighbours or friends. Their levels are not comparable. Each panel shows change within its own survey.
Both interview groups report less contact than in 2018
The GSS moved from in-person interviews to the web in 2021 and to a mix after that. In 2022 and 2024 it randomly assigned every household to be offered either an in-person interview or the web form first. The in-person-first group reported 2.3 to 2.5 percentage points more weekly socialising. The difference is small beside the gap from 2018, and this comparison alone cannot explain the full change.
GSS 2022 and 2024, split by randomly assigned first-mode offer (modesequence), not by the mode people ended up using. The larger gap among people grouped by completed interview mode (57.5% in-person vs 38.8% web in 2024) is descriptive and can reflect differences between respondents as well as mode effects, so it is not treated as a causal estimate.
Europe’s 15–24 group meets socially less often; US declines span generations
Europe's 15–24s fell 8.4 points from ESS round 1 (2002–03) to round 11 (2023–24), more than any other age group. The over-45s ended close to where they started. In the US every age group is lower than in the 1970s. The young fell 13 points. The 45–64s fell 15 points, 6 of them since 2014–18. US years are pooled so each point has 350 to 2,600 interviews.
Europe, 15 countries, by ESS round
United States, pooled GSS years
Same measures as the first chart. Europe's youngest group starts at 15 and the US at 18. The US 2022–24 margin is ±7 points for 18–24s and ±4 for the other groups. In the randomized arms, 18–24s offered in-person first reported 68.9% and those offered the web first 71.2%.
Americans' time with friends halved in 14 years
For Americans aged 15–29, that means about 70 fewer minutes a day with friends: 122 minutes in 2010, 68 in 2019 and 52 in 2024. About 77% of the net decline had occurred by 2019. The displayed annual averages are rounded to whole minutes; the change and share use the unrounded source values. These are population averages, including days with no time together. They do not measure loneliness or establish what caused the change.
Frequency surveys ask how often. Time diaries record each day, minute by minute, and who was in the room. Among Americans aged 15 and older, average time with friends fell from 1.03 hours a day in 2010 to 0.51 in 2024. Most of the decline happened before the pandemic. People may still meet weekly while spending less time together each time.
American Time Use Survey (US Bureau of Labor Statistics), via Our World in Data “Who Americans spend their time with” (data refreshed 2 October 2026). ATUS covers the civilian, noninstitutional population aged 15 and older. Time with anyone counted as a friend, including roommates. Sleep and personal care excluded. No interviews March 18 to May 9, 2020.
Europe's time diaries are too sparse to repeat this comparison consistently. Across three Eurostat waves, reported time in social-life activities fell from 104 to 77 minutes a day in Norway and from 50 to 38 in Poland. The wave categories changed, so smaller movements elsewhere are difficult to interpret.
Eurostat tus_00age and tus_20age, all adults, minutes a day of “social life”. For 2000 and 2010 this includes visiting and feasts plus other social life. For the 2020 wave it includes socialising with family, with others, and virtual social life (phone and online). Fieldwork ran between 2019 and 2023 depending on the country. Categories changed between waves, which limits direct interpretation of small differences. Eurostat's “time spent alone” series also changed definition and is not used.
Near-constant Internet use rose while time with friends fell
Pew has asked Americans how often they go online since 2015. In its phone polls the share saying “almost constantly” rose from 21% in 2015 to 31% in 2021. For 18–29s it rose from 39% in 2018 to 48% in 2021. Pew then switched to web and mail questionnaires, and the next reading was 41%. Pew warns that the switch can change answers, so the jump across the break is not a measured increase. Since 2023 the figure has been flat, at 41% overall and 62–63% for 18–29s.
Pew Research Center, US adults. Phone polls 2015–2021, web and mail in 2023, web, mail and phone in 2025. The 2018 figures come from AP's report of the January 2018 Pew survey (n = 2,002). The 2019 overall figure comes from secondary reports of Pew's July 2019 post, which is no longer online. Pew said 2024 was “on par” but did not publish the number. Values are in data/manual/pew_almost_constantly_us.csv, with a link for each.
Europe has no matching trend, but Pew asked the same question in 10 European countries in spring 2025. Only Spain matched the US level, at 41%. Everywhere else it was between 14% and 32%. Young adults were far above older ones in every country.
Pew Research Center, spring 2025 Global Attitudes Survey. The US figure comes from Pew's separate 2025 National Public Opinion Reference Survey, which used a different method, so treat the US vs Europe gap with care. Values were read from Pew's published charts (September 8, 2025).
The longest European series only counts daily use, and for young people it maxed out years ago. Among EU 16–24s, 90% were online every day in 2015 and 98% in 2025. The growth since then has been among older people, from 40% to 78% of 55–74s. On intensity, the UK's Ofcom measures time online at 4 hours 30 minutes a day in May 2025, 31 minutes more than in 2021, and 6 hours 20 minutes for 18–24s. The WHO's survey of schoolchildren aged 11, 13 and 15 (HBSC) found problematic social media use rising from 7% in 2018 to 11% in 2022. In 2022, 36% said they were in near-constant online contact with friends and others.
Eurostat isoc_ci_ifp_fu, indicator I_IDAY (Internet used every day or almost every day, as a share of all individuals), EU-27 aggregate. Ofcom “Online Nation 2025” (December 10, 2025): personal devices, excluding work. WHO Europe / HBSC, September 25, 2024: 44 countries and regions, about 280,000 children.
Americans lost evenings with neighbours first; relatives held steady until 2020
Weekly evenings with neighbours fell from 30.7% in 1974 to 18.3% in 2018. With friends outside the neighbourhood they went from 24.1% to 18.7%. With relatives they moved between 32% and 41%, with no clear trend. All three were lower in the mixed-mode years after 2020.
GSS socommun, socfrend, socrel; share answering “almost daily” or “once or twice a week”; weighted by wtssps. The shaded years were not interviewed fully in person.
Across Europe, Ireland lost the most; five countries gained
Of the 18 countries with both a 2002 and a 2023 round, Ireland fell 26.9 points, Hungary 14.3 and Poland 12.6. France, Sweden, the Netherlands, Portugal and Slovenia ended higher than they started. The European average hides a map that moved in both directions.
ESS sclmeet, answers “once a week”, “several times a week” or “every day”, as a share of all answers including don't know. Round 1 fieldwork 2002–03, round 11 2023–24. Greece has no 2018 round.
Religious attendance fell faster than socialising on both sides
Thompson contrasts the “modest” fall in church-going with the collapse in hosting. In these surveys it runs the other way. With 2002 set to 100, Europe's monthly attendance ended at 72 and its weekly socialising at 94. In the US, attendance ended at 77 and socialising at 83. The randomized GSS arms show no in-person bump for attendance (34.7% in person first vs 37.7% web first in 2024).
Attendance: at least monthly, excluding weddings and funerals. ESS rlgatnd, same 15 countries: 27.5% in 2002, 19.9% in 2023. GSS attend (“about once a month” or more): 47.9% in 2002, 36.9% in 2024. Socialising as in the first chart. Index = value ÷ 2002 value × 100.
Two European surveys disagree about 2015–2022
The OECD's 2025 report finds that Europeans meeting friends daily fell from 21% in 2006 to 16% in 2015 and 12% in 2022. That series comes from EU-SILC. In the 17 countries both surveys cover, EU-SILC's weekly-or-more meetings with friends fell by an average of 7.4 points between 2015 and 2022. The ESS measure fell by 1.0 point between 2016 and 2023. Sweden, Spain and Finland show falls of 17 to 20 points in EU-SILC and almost none in the ESS. I could not find why. Neither question asks about a specific recent period.
Change in points, by country
Can't afford a monthly drink or meal with friends or family, EU-27
EU-SILC ilc_scp09: friends only, daily or weekly, ages 16+. ESS sclmeet: friends, relatives or colleagues, weekly or more. Below the diagonal means EU-SILC shows the bigger fall. Affordability: EU-SILC ilc_mdes10a, ages 16+, “cannot afford” to get together with friends or family for a drink or meal at least once a month, at home or out.
While the hosting survey halved, monthly evenings with friends stayed flat
Thompson's long series comes from the DDB Needham Life Style survey. There, the share who entertained at home 12 or more times a year fell from 42% in 1975 to 19% in 1998. Over the same years the GSS share spending a social evening with friends at least monthly went from 62.3% to 64.0%. These are different activities, so both can be true: people may have kept seeing friends while hosting them less. It does suggest the change was in where people met, not only in whether they met.
DDB Needham Life Style survey, weighted, as charted in Thompson's post (endpoints only, read from his chart). GSS socfrend “about once a month” or more often, weighted by wtssps.
Why the 2026 hosting number needs careThompson's 12% for 2026 comes from an online panel run by Data For Progress, compared with a mail panel from the 1970s–90s. Data For Progress says the comparison is “suggestive of trends, not a precise estimate of change”. The GSS test above shows a web-vs-in-person gap of about 2 points on these questions. A mail-to-web gap on the hosting question has not been measured.
Europe
European Social Survey sclmeet, rounds 1–11, country shares from the Atlas of Social Attitudes CSV export (ESS post-stratification weight). Panel: .
United States
General Social Survey cumulative file 1972–2024, release 3a (July 2026), gss7224_r3a.dta. Asked in 29 years between 1974 and 2024, about 900–2,700 respondents per year.
Not comparable
Levels across the two surveys, and the pandemic rounds (ESS round 10, 2020–22; GSS 2021) with anything else. Some ESS countries used web or paper questionnaires in round 10. Round 11 was mostly face to face again.
The chart notes link to the source surveys and explain their scope. Download the processed chart series. Prompted by Derek Thompson, “You are no longer invited to dinner”, and Data For Progress, September 29, 2026.