Why Small Experimental Effects of Social Media Use Are Compatible With Large Real-World Effects

The social media hypothesis posits that the widespread adoption of social media is largely responsible for the decline in mental health among US teenagers. Critics of this view argue that it is incompatible with estimated treatment effects from randomized controlled trials (RCTs), which tend to be small. I argue that these RCTs provide little evidence against the social media hypothesis: the causal effect identified by RCTs fundamentally differs from the causal effect invoked by the social media hypothesis in at least three ways. First, the social media hypothesis concerns the effects of joining social media, whereas RCTs typically identify the effects of quitting social media. Second, the social media hypothesis concerns the effects of using social media for many years, whereas RCTs typically identify the effect of quitting social media for several weeks. Third, and most overlooked, the social media hypothesis concerns the effects of altering the behavior of large swaths of the population—a global intervention—whereas RCTs typically identify the effects of local interventions. In such local interventions, a person may be encouraged to quit social media, but both her immediate peers and the general public continue to use social media in large numbers. In this commentary, I discuss the local–global distinction at length and explain why the global intervention plausibly had much larger effects than experimental local interventions.