If you do your intervention and 2 out of 127 children enroll in a college when you would have expected 0.7, you've tripled the expected rate, but that was easy to do by chance.
In this case, the sample size was presumably half that, around 64.
Say... Kids get 10.5 years schooling on average in neighborhood A. 12 years in neighborhood B. That is a 1.5 years or 15% difference. In a 12 grade system, this is the best possible result implying a 100% completion rate.
Neighborhood A sends 10% of all kids to college. Neighborhood B sends 50% of all kids to college. That's a 500% difference.
This example seems pretty plausible. A neighborhood where an average kid makes it halfway through 10th grade probably sends a lot fewer kids to college than a neighborhood with a 100% completion rate for grade 12.
I haven't seen the data, so I don't know if this is the case, but one explanation is that this intervention works really well for some people, and not at all for others. Also, I think "3x more likely" means 200% more people.
So the numbers could be something like: 100 participants, 5 would have gone to college. The social stimulation really benefited 10 kids. They, in turn, did years more schooling and went to college. 5% -> 15% college rate is 3x more likely, but the 10 kids' large increase in attendance was averaged out over the group of 100, so the overall increase was just 5%.
Not saying this is the case, just showing a way those numbers could be plausible.
The 5% more and the 300% more just seem wildly out of kilter
(It kind of makes sense - if all your peers stop school at 16 and you stay for 7 months or more, suddenly getting to college is a much easier step.) I think i am just surprised by how surprised i am