I was once told directly by some faculty when I was in grad school that programs in the natural sciences will often intend for a 30-50% dropout rate for grad st

I was once told directly by some faculty when I was in grad school that programs in the natural sciences will often intend for a 30-50% dropout rate for grad students, because they need lots of TA's for large first year undergrad classes, but don't have funds/means to graduate all of those students as PhD's. This wasn't along the lines of plan for such a dropout rate, but actively desire it. It was a pretty disheartening thing to hear. I think most faculty would like to be able to graduate all of the students, but they also see it as an additional way of weeding out the bottom half of the class. I also discussed this with some other professors in chemical engineering (a much smaller program, with very few TA's) and they told me that their dropout rate was essentially zero.