The Loneliness Scare is Back

Maybe you noticed the well-publicized, bathetic Atlantic Monthly cover story on whether Facebook is making us lonely — wait, let me check, uh, no— although economic distress may be.

http://hoeg.deviantart.com/art/alone-7296534

Or perhaps you saw Sherry Turkle’s N.Y. Times essay publicizing her book, Alone Together, which argues that our mini-screens are stopping us from really talking to each other anymore (no systematic evidence on this either).

It looks like we are having a small resurgence this month of the old meme that communications technology — indeed, modern life — is making us lonely. It’s back to ’80s, back to ’50s, back to ….  Anyway, the folks at Boston Review asked me to discuss the loneliness scare. And I do, at this link.

If you find that essay interesting, come back to the blog for a bonus: a special, short report on new findings that make us rethink the claim that more Americans became friendless in the last couple of decades.

Friendless Americans? Not.

A trio of sociologists published a study in 2006 suggesting a dramatic jump in Americans’ isolation. In the 1985 General Social Survey, fewer than 10% of respondents reported that they had no one with whom they discussed “important matters.”  In the 2004 General Social Survey, about 25% of the respondents said no one. Virtually within moments, the media had published scores of stories that Americans had become isolated, friendless, and lonely. In 2009, I published a comment suggesting that the 2004 findings were anomalous at best. I argued, first, that other data contradicted this apparent near-tripling of isolation. (My book, Still Connected, reviews these data and I discussed them in three previous posts on  this blog — see “American Ties.”) I argued, second, that there seemed to be oddities in the data suggesting some kind of systematic error. The authors of the original study replied, of course, that I was wrong.

The folks who run the General Social Survey were concerned about the hyper-dramatic 2004 results and conducted an experiment to see what might have happened — was the 25%  real or an artifact? In the 2010 running of the GSS, they asked the same question about discussing important matters in three different ways to three different random subsets of interviewees. One procedure tried to repeat the context in which the question was asked in 1985 and another to repeat the context in which it was asked in 2004. In 2004, the  critical question had followed a long, complicated, and nosy set of questions about respondents’ organizational memberships. (There is also a 1987 variant, but I’ll set that aside here. It only reinforces the point.)

GSS co-investigator Peter Marsden will present a thorough analysis of data from this experiment. Here, I report one basic finding. (All GSS data are available to the public; visit the GSS website and run the numbers yourself.) The graph below shows the percentage of respondents in 1985 and 2004 who said they had no one they discussed important matters with — and then the percentages who said the same in 2010 depending on whether they were asked the question the way the 1985 survey did or the way the 2004 survey did. It seems that if you ask the question in the survey context used in 1985, one gets the 1985 results; if you ask it in the survey context used in 2004, you get the 2004 results. It looks like the apparent 2004 increase in isolation is an artifact.

One lesson not to draw from this experiment is that surveys cannot be trusted. Surveys are often the only way we can learn about what people think, feel, and do. The lesson is that all studies have to be closely cross-checked and confirmed. The GSS has systematically cross-examined its methods and results over the years to establish solid findings. This experiment is another demonstration of how confidence can be built by identifying where there are errors and why. It’s way better than the impressionistic generalizing from one’s and one’s friends’ experiences that underlies much of the loneliness scare.

Update, June 2013:

Anthony Paik and Kenneth Sanchagrin provide a fuller analysis of the networks experiment in the American Sociological Review, June 2013 (here). They conclude, as their article is titled, “Social Isolation in America: An Artifact.” They find, drawing on the 2010 experiment and other surveys as well as the 2004 survey, that the high isolation rate in 2004 resulted from interviewer (not respondent) fatigue. A small number of interviewers account for a great number of “isolates,” seemingly because these interviewers choose to avoid the difficult and tiring effort of presenting the network questions. “Our research,” they conclude, “thus suggests that there is no evidence of increasing social isolation among Americans.” They also warn that much of our data on this issue may be tainted by interviewer effects.

Update, June 23, 2017:

I have just come across a solid study, published in a health journal, that shows–at least at first glance–that heavy use of Facebook being associated with decreases in self-reported well-being: by Shakya and Christakis, 2017.