The neighbors left first. In 1974, 61 percent of American adults said they spent a social evening with someone in their neighborhood at least once a month. By 2018 it was 46 percent. In 2024 it was 38 — a figure dimmed further by a change in how the survey is taken, but the descent was forty years old before that asterisk arrived. Evenings with relatives and evenings with friends, by contrast, held almost eerily steady for four straight decades — roughly three in four adults saw family at least monthly, just under two in three saw friends — before both slipped in the 2020s. Convert each answer to a rough count using the survey’s own frequency categories and the average American’s calendar of company shrank from about 12.2 expected social evenings a month in 1974 to 10.9 by 2018, and 9.5 in 2024.
What filled the vacated hours is less novel than the usual story suggests. Television never really rose: 3.0 hours a day in 1975, 3.0 hours a day in 2024. It was already there, the wallpaper of the American evening, fifty years running. The new tenant is the internet — from 5.1 hours a week in 2000 (the survey excludes email) to 16.5 hours by 2022, a tripling that sits on top of the old TV habit rather than replacing it.
That is the familiar act of this story, the one Robert Putnam was telling in 2000 with data that ended in the late 1990s. The GSS lets you run the tape 25 years past Bowling Alone. But it also lets you ask a second, stranger question: as evenings in company became scarcer, did their association with happiness fade — or sharpen? The answer, in these data, is that it sharpened. In the 1970s, Americans who often spent evenings with relatives or neighbors were about as likely to say they were “very happy” as those who rarely did. In the 2010s and 2020s, the gap between frequent and rare socializers runs six to nine percentage points. And one activity — the bar — flipped its sign entirely.
The shape of an evening, 1974–2024
Expected social evenings per month, from stated frequencies. The four activities are asked separately — they are not parts of a whole, and one evening can count in two bands — so read the stack as an additive index of company, not a diary.
Meanwhile, the screen
Hours per day: television (asked since 1975) with internet use stacked on top (asked 2000–2022, hours per week divided by seven; excludes email).
The chart’s long middle is the surprise. The friendship recession, as told from time diaries, is usually a story of the 2010s. Seen from the GSS’s longer window, evenings with friends and family were remarkably durable; what eroded for half a century was the proximate tie — the neighbor, the person you see because they are simply there. That is Putnam’s decline, still running a quarter century after he described it. The 2020s drop in friends and relatives is real in the data but arrives exactly when the survey changed modes, so some unknown share of it is methodology, not loneliness. We flag it rather than lean on it.
Toggle the chart to birth cohorts and the age–period–cohort picture is laid out for inspection rather than modeled. Each ribbon is a birth decade flowing through the ages of life; every generation socializes most in its twenties and settles toward a floor. But the ribbons stack lower with each decade of birth: Americans born in the 1950s averaged 12.4 expected social evenings a month at ages 25–31; those born in the 1990s, observed at the same ages, averaged 10.5. The evening didn’t empty because old people aged; it emptied because each new cohort arrived a little less social-by-evening than the one before.
Scarcer company, bigger gap
Now the second act. For each activity and each pooled bin of surveys, take the people who report that activity at least monthly (“frequent”) and those who report it less than monthly (“rare”), and compare the share saying they are very happy. The result is a gradient — an association, emphatically not a causal effect; happy people may seek company at least as much as company makes people happy.
Evenings with relatives carried essentially no happiness signal in the mid-1970s — a gap under one percentage point, statistically indistinguishable from zero. By 2010–14 the gap was 9 points; in the 2021–24 surveys it stands at 6.5. Neighbors trace the same arc: nothing in the 1970s, 6.4 points in the most recent bin — the very activity that collapsed is the one whose remaining practitioners look happiest relative to abstainers. Friends are the honest exception: the friends gradient is small and noisy across the whole half century, flickering in and out of significance, never the dependable signal the other two became. If you expected the friendship recession to make friendship the most precious currency, the GSS only half obliges: in a 2018–2021 snapshot, 51 percent of frequent friends-socializers rated their social life very good or excellent against 38 percent of rare ones — but on the long happiness series, family and neighbors are where scarcity and signal moved together.
In the 1970s, the regulars at the bar were measurably less happy than everyone else. In the 2020s, they are measurably happier.
The bar flipped
The wildcard was always the tavern question. Bar-going is the only one of the four that was already a minority habit in 1974 (29 percent monthly, drifting to 25 by 2024), the only one with a strong age skew, and the only one whose happiness association ran negative at the start. Age-adjusted, frequent bar-goers in 1974–77 were 7.2 percentage points less likely to be very happy than their same-age peers who rarely went — the tavern as refuge, the lonely-regular reading. That deficit shrank through the 1980s and 1990s, crossed zero in the 2010s, and in 2021–24 sits at positive 3.9 points, an interval that excludes zero. The institution didn’t change its drinks; its clientele and its meaning changed. As the casual tavern faded, an evening at a bar stopped marking the person with nowhere else to be and started marking the person who still has somewhere to go — a scheduled, social, increasingly optional “third place.” That is one reading; selection is surely doing work in both directions, and the data cannot split it. But the sign flip itself is not subtle, and it is the cleanest single emblem of the half-century: the same behavior, in an emptier evening, now keeps different company.
What this shows — and what it can’t
- Association, not causation. Every gradient here compares people who do and don’t socialize; it cannot say which way the arrow runs. Happiness recruits company at least as plausibly as company produces happiness, and third factors (health, marriage, money) feed both.
- Behavior vs. evaluation. The four evening items measure the frequency of a behavior. “Very happy” is a global self-evaluation. We juxtapose them; we never treat one as the other.
- The 2021 asterisk. The GSS moved from in-person interviewing to push-to-web in 2021–22. Post-2021 levels are partly mode-shifted, which is why every trend wears a dashed line there and why we anchor the long declines to 2018 wherever possible. A mode change can impersonate a social collapse; we decline the impersonation.
- Ordinal scales, stated thresholds. “At least monthly” means a code of “about once a month” or more often; the evenings-per-month index uses stated midpoints and is an approximation, not a diary count.
- The bar item is the trickiest. Most Americans never go; bar-goers are young. The bar gradient is therefore shown age-adjusted, and one early bin is withheld for thin cells. The flip survives the adjustment; that is precisely why it is reported.
- Loneliness is a snapshot, not a trend. The GSS loneliness items exist only in a 2018 module (and the social-life rating in 2018–2021). They are reported once, never trended — and in that one snapshot, frequent socializers actually score slightly worse on lack-of-companionship (1.97 vs 1.77 on a 1–5 scale, higher = lonelier), a reminder that the loneliest socializers are often the youngest, and that single-year module cuts are confounded by age.
What the half-century portrait does establish is the pairing: the long emptying of the American evening was led by its most casual ties, and over the same decades the remaining company came to travel with happiness more tightly — family steadily, neighbors at the end, the bar by way of a full reversal. Whether connection became more precious because it became scarce, or the people left gathering are simply the flourishing kind, the GSS cannot say. It can say that in 1974 an evening among others told you almost nothing about whether an American was very happy, and that today it tells you quite a lot.