Goldilocks is online
A year or so ago, there was a startling article by Jean Twenge in The Atlantic, asking, Have Smartphones Destroyed a Generation?
There is a cynical rule of thumb which says, any question in a headline can be answered in the negative, but Twenge's article was based on her recent book, which was wide, deep, and scholarly, and associated the rise of smartphones with an increase in depression among young women, and I know from bitter experience that writers don't get the last say on the headline.
Nonetheless, it's very hard to tease out cause and effect, and to disentangle the often modest effects of social media use on wellbeing from everything else, like poverty or austerity.
The same year, Przybylski and Weinstein published a large-scale analysis of the What About Youth? data set, finding that the highest wellbeing scores were found for modest amounts of time spent online. Lots of online time was associated with poorer wellbeing, but those who spent no time online also had lower wellbeing scores than those spending a bit of time online. They suggest that a middling amount of involvement with social media might be most desirable -- neither too little, nor too much, but just right -- as Goldilocks might have concluded.
Przybylski AK & Weinstein N (2017). A Large-Scale Test of the Goldilocks Hypothesis: Quantifying the Relations Between Digital-Screen Use and the Mental Well-Being of Adolescents. Psychol Sci, 28:204–215.
The WAY? sample was a postal self-report study of 15-year-olds, and produced oddly low figures for smoking, but it was large, nation-wide, carefully designed, and the authors did their best to control for confounding effects, so I don't see any reasons there for rejecting the idea. In fact, I was interested recently to explore it with a large sample of young people from a SHEU study in one local authority, and found that Goldilocks is online there too: the highest average wellbeing scores are seen for those spending modest amounts of time online.
Mean wellbeing scores (SWEMWBS) of Year 10 females in one local authority, 2018, by time spent the previous evening on the Internet (on any device) or using a 'phone (for talking/texting).
The same pattern is seen if we select from the sample just those students who live with both parents at home. We also see the same pattern in most deprivation quintiles, although some of the samples become rather small and thereby wobbly.