Many studies and discussions about social media use the time spent on the platforms as evidence of various theories and conclusions – but it turns out that according to research by Facebook, people are actually very poor at telling how much time they spend on it.
If you were to conduct a study that looked at how social media use might affect or be affected by mood, you would likely rely on self-reported statistics for either measure.
Of course, there is no objective way to measure mood. So you have to rely on what the participant is saying. And you'd think these people would have a pretty good idea of how long they spent scrolling through Facebook. Instagram or Twitter. Not as much!
Everyone understands that these self-reported numbers will be in error, and some studies have shown it, but this meta-study by Facebook comparing self-reporting to actual server logs shows that the connection may not be reliable enough to be used for serious scientific purposes to become work.
The answers to some questions compared to internal data showed that the time on site was overestimated by hours on average. But they also underestimated how often they opened the app or website. Check out this exciting table:
As you can see, very few people spent more than three hours a day on the website, with the vast majority spending around an hour on the website. And the opposite was the case with logins: Comparatively few people thought they had opened the app 10 times a day or more, but that was extremely common. Younger people in particular were prone to error, which, given that these studies tend to show more of this population, only accentuates the problem.
None of this means Facebook isn't a website that people spend a lot – maybe too much – time on, even by their own estimation. However, it is important that studies on these phenomena are based on reliable data, and it appears that self-reported data are not.
As Facebook says, "We suggest that researchers don't use these values directly, but rather interpret people's self-reported time as a noisy estimate of where they fall within a distribution compared to other respondents."
In other words, instead of saying that "teenagers who spend two hours or more on Facebook are more likely to …", you could say that "users in the top 10% of self-reported Facebook usage are more likely to …" or something similar. If exact online times are needed, a tracking app or working with Facebook is probably a good idea.
The full article on Facebook's research can be found here.