Yes, but the number of students in my program is higher than the other two (wealthier) schools in my district, and a lot higher than schools with comparable demographics.
Even if Joe Blub teaches nearly as well as I do (and I'm willing to accept that he may), I recruit a whole lot better than he.
But if I teach my students Java (as the AP exam requires), does that make me Joe Blub?!? :)
Regarding teacher quality, he found the best fit was a failure model. Most teachers perform adequately, and these teachers are statistically indistinguishable. Some teachers fail, with varying degrees of failure.
Basically, think hard drives. Most work, and these are all the same. Some fail, with varying degrees of data loss.
If the sample in the study you mention comes from the partially failing region, they could easily find an increasing relation. Or possibly this program was special, that I don't know.
However, I do stand by my statement to leave teaching to those who aren't great at other things. That's just basic comparative advantage.
You are comparing your students with you as a teacher to your students with no teacher. The proper comparison is "you program, joe blub teaches" vs "you teach, joe blub programs." If joe blub teaches almost as well as you, but you program much better than him, then it's far better if you leave the teaching to him.
Of course, if you enjoy teaching, you certainly are not obligated to do the most economically efficient job. I'm certainly not (postdoc in math here).