They looked at the mental health (as measured by self-reported surveys) among schools over time and cross-referenced that with the rollout of Facebook over time. So they could compare the change in mental health at schools the received Facebook access and compare it to the change in mental health at schools that did not receive Facebook access at the same time.
The methodology appears to be fairly novel and does isolate them from several reverse-causation biases, as it is difficult to imagine that the rollout of Facebook was influenced by factors that led to the decline of mental health in student bodies.
I assume to do that you have to establish the complete pathway and mechanism from someone using facebook to an increase in depression, like showing observations of changes in neurotransmitters or brain structure that have been proven to cause changes in mental health, and then proving that facebook caused the changes in those levels. (FWIW I assume this could be done and that we may see those kinds of results if it were done, but I haven't actually seen a study like that. I also assume the hypothesis in general.)
For instance, using the example of smoking from another commenter, from the CDC website [0]:
> - Poisons in cigarette smoke can weaken the body’s immune system, making it harder to kill cancer cells. When this happens, cancer cells keep growing without being stopped.
> - Poisons in tobacco smoke can damage or change a cell’s DNA. DNA is the cell’s “instruction manual” that controls a cell’s normal growth and function. When DNA is damaged, a cell can begin growing out of control and create a cancer tumor.
These seem more like things that can be tested in laboratory settings that are easily reproducible and rely on more objective observations than self-reporting.
I'm neither a neuroscientist or social scientist so I'm just trying to understand, not saying they're wrong or that the research is even flawed.
[0]: https://www.cdc.gov/tobacco/campaign/tips/diseases/cancer.ht...