It's in the studies, but the gist is that they try to compare 'matched students'. So, they try to remove the selection effects by comparing comaprable pupils in

It's in the studies, but the gist is that they try to compare 'matched students'. So, they try to remove the selection effects by comparing comaprable pupils in each school type. I'm sure it's possible to quibble with exactly how they do this, of course.