If, as many articles seem to be suggesting, we decrease the number of grad students, then that just means that the ones left will teach more, which distracts them from the research and (presumably) lengthens the PhDs. It's not a completely linear tradeoff, but certainly something to bear in mind.