Do Americans even make friends in high school anymore?

No, Americans aren't getting lonelier in general. But teenagers might be.

This post features original analysis. As always, you can review my code and replicate my work via my GitHub.

This week, NPR's On Point ran an hourlong episode based on the premise that "Americans are spending less time with friends." It seemed to take its cue from a piece last month in the Post, which cited data from the American Time Use Survey. In fact, as a cursory look at the actual ATUS data makes clear, the premise of this episode is mostly false.

Americans in general are not spending less time with their friends. For most age groups, the amount of time that Americans spend with their friends has remained remarkably stable over the past twenty years. It is the case, however, that American teenagers, in particular, are spending less time with their friends — dramatically less so.

This makes it something of a disappointment that NPR dedicated a whole hour to answering a fundamentally misguided question, including the obligatory Bowling Alone reference and interviews with a "friendship coach" and a "friendship expert" with a new book to sell. There is a real story here — it's just not the one that NPR ran with.

In order to figure out what that story is, we have to dig in a bit.

Lies, damn lies, and statistics

The ATUS is an annual survey that basically takes a random sample of Americans aged 14 and up on a random set of days and asks them to respond to a bunch of detailed questions about how they spent those days. For a big group of activities, they also ask people to explain who they were with when they were doing the activity—studying, working, water polo, whatever. This enables the ATUS to tally up those "who-were-you-with" questions in order to calculate how much time Americans spend with people to whom they have different sorts of relationships.

For people identified as "friends", the ATUS data records a variable called TRTFRIEND. It's defined as "Total nonwork-related time respondent spent with friends (in minutes)."

When you look at the time series , there does appear to be something going on here. When you look at ATUS data, it does indeed show a precipitous decline in the value of TRTFRIEND starting in about 2015.

When you see a graph with a like this, though, you should get curious. America is a diverse place, and the nature of friendship differs dramatically across geographies, religions, social classes, education levels and, perhaps most obviously (to some), age.

When you disaggregate by age, this graph tells a very different story.

Instead of a topline story in which Americans across the board have experienced a roughly 30-minute-per-day friendship decline, we see one in which most Americans' social lives (but for a small pandemic interlude) have remained the same. Today's teenagers, however, seem to be spending about two hours less per day with their buddies than their predecessors did twenty years ago.

Dubious explanations

Like police who arrest a burglar carrying a comically large sack with a dollar sign on it, NPR can be forgiven for chasing down an obvious-seeming culprit: social media. Social displacement theory suggests that people replace face-to-face friendship with social media use. This theory is not borne out by the ATUS, which asks respondents to detail how much time they spend using a computer for leisure:

The ATUS is a bit behind the times: there is no activity code for smartphone use, and since it's up to respondents to enter their own data, we don't really know whether teenagers consider their smartphone a computer. But I still don't think the social media explanation is credible, for reasons I'll get into in a moment.

There are a number of activities that seem to be negatively correlated, either at the individual level or in time series, with TRTFRIEND:

• Working

• Traveling (to or from school or work)

• Watching tv or movies

• Sleeping

• Playing games

• Being in class

• Doing homework

Looking at the data, I thought—I really thought—surely one of these is the culprit. Surely the decline in TRTFRIEND will be mirrored by a telltale increase in some other variable. As someone who always hated school, homework was my favored explanation—our overworked kids have no time for friendship!

No such luck for my preexisting bias:

Here's the problem. The change in the amount of time Americans spend with friends, as estimated by the ATUS, is bigger by a factor of five than almost any other change in the period 2003-2021. So there be no single explanation, no single activity greedily sucking time from our children's friendships. If what were happening were displacement of some kind — kids prioritizing or being forced to prioritize some other activity over friendship — we'd expect to see some concomitant increase in those activities. But there don't seem to be any.

That's weird. And it should make us stop and think this whole thing over.

Good explanations?

Let's start at the top. The way the ATUS defines the TRTFRIEND variable is as a byproduct of asking people what they're doing more generally. "Spending time with friends" isn't an activity in itself; rather, it's a characterization of who is around when a respondent is doing *whatever.*

Scroll up and look at the TRTFRIEND time series disaggregated by age. In 2003, according to the ATUS, the average 14- to 18-year-old was spending around three hours per day, on average, with their friends. Does this strike you as strange?

Were you alive in 2003? Did you know anybody who spent three hours per day with their friends?

This may become a bit clearer if we break it up by days of the week:

There's some weird things about this chart. First, in 2003, teenagers reported spending almost half as much time with their friends on Sundays as they did on Mondays. And second, weekdays seemed like a weirdly social time. I don't know about you, but in high school I usually didn't spend four hours at a friend's house after school on a Wednesday — especially not on an average Wednesday.

What's going on here is that time with friends in school is counted as time with friends. In the ATUS, class time (or time at work, or playing sports, or acting in a play) is included in the TRTFRIEND variable, just so long as your friends are there.

The puzzle

The puzzle, then, is what changed. It seems to me that there's two possible explanations:

1. The meaning of the word "friend" has changed, substantively and steadily over time, for American teens.

2. Americans are simply not making friends in school and, consequently, are spending less time with their (non) friends outside of school.

I am inclined to think that explanation (2) is more likely. If the meaning of the word "friend" were changing, then we'd expect American teens to be spending Saturdays and Sundays with their actual friends, and we wouldn't see the 2003/2021 gap that we see in the chart above for those days. Instead, teenagers seem to be spending less time on the weekend with people they consider their friends, and filling in that time with an inconsistent mismash of other activities.

Why wouldn't American teenagers be making friends in school? I don't have a good explanation for this, but I find it difficult to believe that social media or technology in general are major causes. This decline began in in earnest well before Facebook or even MySpace had significantly penetrated the high school market.

I hasten to suggest that this decline may well have been going on for decades before the American Time Use Survey began in 2003. The typical American school day lasts about seven hours, and it's not difficult to imagine an idealized 1950s teenager reporting that he'd spent all seven with his very best friends in the world