In the special school, 100% of the kids/parents actively took an interest in education. In the alternatives, less than 100% did.
School quality is largely determined by the students in the school. This doesn't just mean that the performance of student X is determined by student X caring about school. The performance of student X is also determined by the fact that student Y cares or doesn't care about school.
So in all the non-special schools, there exist the sort of uncaring disruptive students who will drag down everybody. It isn't that the special school is itself somehow wonderful, but that the bad influences are excluded.
Simply put, one can not scale this solution. Providing the supposed quality school to everybody is not possible because the quality of a school is largely determined by the students.
No. They are simply stating the truth that there is no evidence it will work.
It is basic selection bias.
I think "confounding variable" would the describe the problem. Basically there are two variables that seem to affect the outcome, but since they only controlled for one of them, they can't tell if the second variable is affecting the experiment or not.