[RK]

As a grad student in the 2000's I was told by a professor that the department would take into consideration the expected attrition rate (typically ~50%) when admitting grad students. The idea being that they would have enough TA's for intro classes. Attrition would make the number of advanced students more inline with the money available to fund PhD students till they graduated. In other words, they purposefully over-admitted to get TA's, most of whom would leave with an MS as best.

This was in science departments. In departments that had few or no "general ed" courses, such as engineering, and thus little need for TA's, the graduation rate was much higher.