The 35-minute socializing figure: fewer people each day, more time among those who do

The American Time Use Survey puts socializing in 2025 at 0.58 hours a day, about 35 minutes. That figure averages over everyone age 15 and over and over every day of the week, including the roughly seven in ten people who did no socializing on the day they described. Only 30.4% socialized on an average day in 2025. Those who did spent 1.92 hours on it, about 115 minutes. [1]

The Bureau of Labor Statistics (BLS) reports 35 minutes in 2025 against 41 in 2015, and 30% of people against 38%. [2] Over the same decade, time among the people who socialized went up, from 107 minutes to 115. The average fell because fewer people socialized on a given day.

Below, that split year by year, what the category leaves out, and one weekend figure that needs correcting. We recomputed every number from BLS tables downloaded on 30 September 2026; the series is in the CSV linked at the end.

Two numbers inside one average

Minutes for everyone equal the share who did the activity times the minutes among those who did. BLS publishes all three in Table 1 of each annual release, so the average can be taken apart without modelling. [1] [3] [4] [5]

Minutes are the published decimal hours × 60, rounded. People age 15 and over, primary activities, all days of the week.

Most of the change is one step. Between 2019 and 2021 the share who socialized fell from 34.8% to 29.1%, while minutes among those who did rose from 110 to 118. From 2021 to 2025 both series barely move. The survey records what people did, so it cannot say what caused the step. We did not retrieve standard errors, and no year-to-year move in this table has been tested for significance.

Why 2020 is blank

BLS suspended collection in spring 2020 and has no data about 18 March to 9 May of that year. The ATUS User's Guide says 2020 data "are not representative of the full year". [6] BLS's own 2020 release compared May–December 2019 with May–December 2020 instead of publishing ordinary annual averages. [7] A trend line drawn through 2020 would treat ten months as a year, so the table and the CSV leave it empty.

What the 35 minutes counts

BLS defines the category as "face-to-face social communication and hosting or attending social functions". [8] Only the main activity counts: respondents doing two things at once name the main one, and apart from childcare nothing done at the same time is collected. [8] Each activity then gets one code from the ATUS lexicon, and a coding manual settles borderline cases. [9] [10]

Sources for the codes: the 2025 lexicon (pages 13, 16, 44, 45, 47 and 65) and the 2025 coding rules (pages 36, 37 and 48). [9] [10]

Two rows need explaining. BLS files texting with telephone calls: the Technical Note says the telephone category "also includes texting and internet voice and video calling", and the coding rules say to code text messaging "the same way you would code telephone calls". [8] [10]

Meals are the second, and here the answer is our reading of the BLS definitions; BLS publishes no statement about meals with friends as such. The Technical Note assigns eating and drinking "whether alone, with others, at home, at a place of purchase, or somewhere else" to its own category (Major activity category definitions, "Eating and drinking"). [8] The 2025 coding rules add, under category 11 (page 36, rule 2), that "talking with people who are present at a restaurant meal" is coded as socializing. [10] So a dinner with friends counts toward the 35 minutes only if the respondent reports the talking as the main activity. If they describe it as eating, it goes to Eating and drinking.

Calls, texts and video calls, one row down

The category that holds calls and texts shows the same pattern. In 2025, 19.2% of people spent time on "Telephone calls, mail, and e-mail" on an average day: 0.19 hours (11 minutes) averaged over everyone, and 1.01 hours (61 minutes) among those who did. In 2015 the figures were 21.4%, 10 minutes and 45 minutes. [1] [3] Slightly fewer people did it on a given day, and those who did spent longer.

Social media sits in a third place: BLS says its games and leisure computer category, up from 25 to 37 minutes a day since 2015, includes "using social media". [2] Adding these rows does not give a total for "contact time". They are separate main activities, and the tables do not say who was on the other end.

The weekend figure is 55 minutes

Table 2 of the 2025 release splits the day types. On weekdays, 28.4% socialized and the average was 0.44 hours, about 26 minutes. On weekends and holidays, 35.2% socialized and the average was 0.92 hours: 0.92 × 60 = 55.2, so 55 minutes. [11]

You may also see 56 minutes. That figure is real, but it belongs to 2024. BLS's 2024 release says people spent "twice as much time socializing and communicating on weekend days (56 minutes) than on weekdays (26 minutes)", and the 2024 Table 2 gives 0.94 hours. [4] Both releases also give 35 minutes a day and 30% overall, which makes it easy for a 2024 sentence to end up in a 2025 summary.

The split by day type holds on both kinds of day:

Source: Table 2 of the 2015 and 2025 releases; hours × 60, rounded. [3] [11]

How much weight 2015 against 2025 can bear

The definitions did not move. BLS's record of lexicon changes says all activity codes were identical from 2013 through 2025, and the category wording quoted above appears in the Technical Note of every release from 2014 to 2025 data. [12] [5] Collection did change:

• Sample. About 10,900 people were interviewed for 2015 and about 6,100 for 2025. [3] [8]
• Response. The response rate before final processing was 48.5% in 2015 and 25.7% in 2025. BLS weights the data to adjust for nonresponse, but the Technical Note warns that errors "could occur if nonresponse is correlated with time use". [6] [8]
• Shutdown. There are no diaries for 30 September to 11 November 2025. BLS weighted the later fourth-quarter diaries to stand in for the missing weeks and says the effect "is not possible to quantify". [2] [13]

BLS publishes the 2015-to-2025 change in socializing as a finding, and its analyses use a 90% confidence level. [2] [8] The participation-and-duration split is our arithmetic on its estimates and is descriptive only.

What the number means for keeping in touch

The 35-minute figure measures one thing well: time spent with people in person as the main thing you were doing. It says nothing about the call on the drive home, the group thread, or the conversation over dinner that was logged as eating. If you are weighing your own habits against it, compare like with like.

On a given day, fewer people now spend time with others in person, and those who do spend longer at it. Much of the contact in between is filed under calls and texts, and its record sits in your message history. Kluro reads the Messages history stored on your Mac once you grant the Full Disk Access that macOS requires for it. [14] Our look at that permission covers what granting it opens. For the habit itself, the guide to keeping in touch starts with a short list of people.

Method

We downloaded Table 1 of each ATUS annual release for data years 2014–2019 and 2021–2025, and Table 2 for 2015, 2024 and 2025, on 30 September 2026, and parsed them with a Python 3.11 standard-library script. Dividing minutes for everyone by the participation share reproduces each published per-participant figure within 1.7 minutes, the margin expected from rounded inputs. Response rates come from Table 3.3 of the ATUS User's Guide (September 2026). We did not use microdata, age breakdowns or standard errors.