What evidence would distinguish these two possibilities?
Select arbitrary productivity metric. Select people in a controlled environment (like cubicles). Randomly assign their IP addresses to a correctly configured DNS server or a DNS server with a faulty entry for 'facebook.com'. See if there is a difference.
Distraction is a Symptom of a Deeper Problem: The Convenience Principle and the Destruction of American Productivity is a good article on distractions versus getting things done. With extra emphasis on how many of our distractions are the result of a desire for convenience rather than something more substantial.