About that particular piece of research, the wiki article states (I don't know where to confirm it) "However, in the analysis the datasets from universities were discarded as they were considered possibly biased and insufficient. Instead only data from schools was used for the analysis."

This unfortunately is quite a political topic and, as such, research here can be politically motivated (for and against) so it is tough to break through the noise.