[epidemiology]

Wow this is an excellent point and really undermines the article conclusions. We should always be looking at unadjusted scores as well as a whole series of adjusted scores with a variety of methods.

Just grabbing one highly adjusted score and drawing conclusions solely off of that is not enough. It's really only giving you one piece of a very complex puzzle in the case of something like education scores.

Observational stats in social sciences turns out to be a lot like epidemiology and strongly held conclusions are hard won.