They used census data to estimate the number of users under 18.

This is just BS PR from probably the least reliable kind of researcher, the social science crowd.