Fifty Years of American Evenings

The "friendship recession" is usually told with time-use diaries, and the diaries only go back so far: the American Time Use Survey began in 2003, and the alarming numbers — face-to-face socializing down by a third, companionship at record lows — cover barely two decades. Robert Putnam, writing Bowling Alone in 2000, had to stitch his story together from club rosters and league memberships, and his data ran out a quarter-century ago.

But one instrument has been pointed at the American evening, without interruption and without rewording, since Gerald Ford was president. In 1974 the General Social Survey began asking how often respondents "spend a social evening" with relatives, with friends who live outside the neighborhood, with neighbors, and at a bar or tavern — the same seven answer choices, from never to almost daily, in twenty-nine survey rounds across fifty years. Add the survey's television question (asked since 1975) and its internet-hours question (since 2000), and you can watch, in one dataset, the American evening empty of other people and fill with screens.

This article does that — and then asks the question the decline curves don't answer on their own. As evenings with other people became scarcer, did they come to matter more for happiness? The second half of the data holds a quiet surprise, and the barroom holds a bigger one.

Read the dusk-colored stream first. In 1974–77 the average American did 2.2 of these four things at least monthly. Through 2016–18 — the last rounds conducted face-to-face — the figure had only slipped to 2.1. Then it dropped to 1.8 in 2021–24. That last cliff comes with an asterisk the size of a methods section: in 2021 the GSS moved from in-person interviewing to a mostly web-based design, and answers given to a screen are not strictly comparable to answers given to a human being in your living room. Every figure in this article marks that boundary. Treat the post-2020 level as a new series, not a verdict.

The honest fifty-year story, then, is not "Americans stopped seeing people." It is stranger and more specific than that — and you can see it by unstacking the stream.

What actually died: the near evening

Evenings with relatives never declined at all before the mode change: 72.3 percent monthly-plus in 1974–77, 74.1 percent in 2016–18. Evenings with friends look almost as durable at the monthly threshold — 63.7 percent then, 64.3 percent in 2016–18 — but the weekly habit was quietly compressing underneath: 23.1 percent of Americans saw friends weekly in the mid-1970s, 18.8 percent by 2016–18, 15.6 percent now. Friendship wasn't abandoned; it was rescheduled, from a weekly rhythm to a monthly one.

The unambiguous casualty is the neighbor. A social evening with neighbors was a majority practice in 1974–77 (57.1 percent monthly-plus). It fell in nearly every decade since — 45.7 percent by 2016–18, 36.4 percent in 2021–24 — and the share who say they never spend an evening with a neighbor rose from 23.4 percent to 37.7 percent. This is the steady, mode-proof, fifty-year decline; it was well underway when Putnam wrote, and it has continued for twenty-five years past his data.

The bar tells a subtler story. The monthly-plus share barely moved before 2021 (28.3 percent in 1974–77, 28.2 percent in 2016–18). But the regulars vanished: weekly-plus bar-going fell from 11.1 percent to 7.0 percent to 4.3 percent — a 61 percent collapse across the half-century. Meanwhile the share who never set foot in one actually fell, from 52.2 to 44.5 percent. The tavern shifted from a habitat to an occasion: fewer barflies, more birthday parties.

And the screens? Television, the villain of Bowling Alone, turns out to have been remarkably constant: weighted mean viewing has stayed within a narrow band — between 2.76 and 3.03 hours a day — across every pre-2021 era of the survey (2.91 in 1975–78, 2.76 in 2016–18, with a pandemic-era 3.17 since). The screen that grew is the other one. Self-reported internet time outside email went from 6.4 hours a week in 2000–04 to 12.9 in 2014–18 to 15.7 in 2021–22 — an increase of roughly nine hours a week, which is to say, several evenings.

The chart Putnam never got to draw

A cross-section can't tell you whether a decline is a story about eras or about generations. For that you need to follow birth cohorts through the age ladder — and with fifty years of identical questions, the GSS can finally do it.

The ribbons are strikingly parallel. Every cohort enters adulthood gregarious — among 18-to-24-year-olds, between 82 and 88 percent see friends at least monthly, whether they were born in the 1950s or the 1990s — and every cohort slides down the same hill as marriages, mortgages, and children arrive. Boomers at 32–38 (the 1950s cohort): 69.5 percent. Gen X at the same age: 69.5 percent. The life course, not the calendar, does most of the work.

The exception sits at the newest edge. Americans born in the 1990s fell to 66.8 percent at ages 25–31 — a step their elders took a decade of life later (the 1980s cohort was still at 80.7 percent at the same age). Some of that cell is post-2020 interviewing, so judge it gently; but it is the one place in fifty years of ribbons where a cohort visibly leaves the path early. If the friendship recession has a generational signature in this data, it is Millennials and Gen Z beginning the midlife social descent in their twenties.

The second act: scarcity and value

Here is the under-asked question. The GSS measures happiness in every round ("taken all together… very happy, pretty happy, or not too happy"). So for each decade we can compute a simple gradient: among people who do an activity frequently (several times a month or more), how much likelier are they to say very happy than people who rarely do it (never, or about once a year)? If shared evenings became scarcer and the happy increasingly cluster among those who kept them, the two curves together say something the decline alone cannot.

For three of the four activities, the association strengthened as the activity thinned. The clearest case is friends: in the 1970s, frequent friend-seers were no happier than people who almost never saw friends — a gap of 0.3 percentage points, statistically indistinguishable from zero. The gap turned positive around the turn of the century and has been remarkably stable since: +6.3 points in 2000–08, +6.3 in 2010–18, +6.3 in 2021–24, each with confidence bands clear of zero. The very decades in which the weekly friend-evening eroded are the decades in which having one began to separate the very happy from everyone else.

Neighbors follow the same script on a smaller scale: nothing in the 1970s (+0.3 points), a significant +3 to +4 points from the 1990s, +6.4 in the 2020s — the gradient peaking precisely when the practice hit bottom. Even relatives, the most stable activity, saw their gradient roughly double from +6.6 points in the 1970s to a peak of +12.7 in 2000–08 (in that decade, 35.6 percent of frequent family-socializers were very happy versus 22.9 percent of rare ones).

One reading is Sandstrom and Dunn's: social contact is a nutrient, and gradients steepen when a nutrient grows scarce. But honesty requires the other reading too. These are cross-sectional associations, not effects. Happy people may seek company rather than company producing happiness; depression both isolates and saddens; and a society in which socializing requires more initiative may simply sort the flourishing into friendship more efficiently than one where the neighbors drop by unbidden. The data cannot pick among these. What it can say is that the link between a peopled evening and reported happiness is two to six points stronger now than when the questions were first asked — whichever way the causal arrow runs.

The barroom flip

Then there is the bar — the one activity whose happiness association didn't strengthen but reversed.

In 1974–78, people in a bar several times a month were 11.7 points less likely to be very happy than people who never went. By 2010–18 the raw gap was +2.0 points — a sign flip of nearly fourteen percentage points across four decades, and it stayed positive (+1.7) in the 2020s.

Age explains part of the starting point but not the flip. Frequent bar-goers have always been young (mean age 34.7 in the 1970s) and rare ones older (47.8), and the young are less likely to say "very happy" in any era; adjusting for age trims the 1970s deficit to −8.4 points. But the adjusted series flips sign on the same schedule, reaching +2.1 by 2010–18. Something about who is in the barroom changed. The regulars thinned out — weekly-plus bar-going fell by more than half — while the never-share also fell; and the frequent bar-goer aged thirteen years, from 34.7 to 43.2, as the activity drifted from a young person's habitat toward a sociable adult's occasion. The bar evening of 1975 kept company with the unhappy; the rarer, older, more deliberate bar evening of the 2010s sits on the happy side of the ledger — though its confidence band still grazes zero. The wildcard delivered.

Coda: what the new questions say

In 2018 the GSS began asking directly about social satisfaction and loneliness, and the new items land on both sides of this article's argument. Satisfaction first: among 2018–2021 respondents who see friends frequently, 54.8 percent rate their "social activities and relationships" very good or excellent, versus 35.1 percent of those who rarely do — a 20-point gradient that echoes the happiness results. But the 2018 loneliness item (n = 381 across the three friend-frequency groups) refuses to cooperate: frequent friend-seers were more likely to report lacking companionship at least sometimes (34.1 percent) than rare ones (23.4 percent). The likeliest culprit is age again — the frequent group skews young, and the young are the loneliest Americans — plus the possibility that the lonely seek company out. It is a useful closing caution: an evening with friends is associated with a happier life, not a guaranteed cure for an empty one.

Fifty years ago the American evening held two and a quarter kinds of company a month and three hours of television a night. Today it holds slightly less company than that — almost all of the loss borne by the people nearest by — the same television, and two extra hours a day of internet. And the evenings with other people that remain now carry a happiness premium they did not carry when they were abundant. Whether that premium is scarcity pricing or self-selection, the long series leaves one fact uncontested: the Americans who kept their evenings peopled are increasingly the happy ones.