Yes, the study design is able to tell which way the causality points. Not all colleges got access to Facebook at the same time (back before it was open to the general public) so this is sort of a natural experiment: you can look at the colleges that had Facebook access and compare them to the ones that didn't, assuming that they're probably pretty similar in all other confounding variables, and that people don't choose their college based on whether or not it has Facebook access. For more information on this type of design, the phrase to google is "difference in differences".
https://en.wikipedia.org/wiki/Difference_in_differences