If we suppose that this theoretical child is a problem that is disrupting their classroom, then we are not saving any children from that disruption, merely spreading it around. They will be a problem for one classroom this year, a different one next year, and so on. Assuming class sizes are the same, we will have on average the same number of disrupted-pupil-hours with the expulsions as without them, just a different distribution.

If you want to argue it's a better distribution, or that there is some reason to believe expulsions will push "problem children" to smaller classrooms on average, then fine, but absent those arguments it's very clearly worse for the student and their parents but not better for anyone else (or rather, it's maybe better for specific other people, but also worse for yet other ones, and on average no better for anyone).