I'd be interested in how reading more about how they're controlling for non-air pollution factors, particularly since they say the effects are most pronounced among the non educated.
That'd mean if you live in a city and have not had much education, it might be because you are less intelligent (on average). While if you're in a rural area, a lack of education might not mean anything, because you have less opportunity to get educated in general, regardless of intelligence.
If you did a simple regression in this case with years of education and air pollution (more common in cities) in the model, I think you'd find that -- controlling for education -- air pollution would be correlated with a reduction of intelligence, even if the air pollution itself wasn't causal.
There are ways of controlling for this, and I'm not at all saying the study didn't take this into account, but it'd be useful to know more. There are a lot of ways (as in any model) the relationship could be correlated without being causal.