Of course KIPP selects for motivated students. But it also selects for severely disadvantaged students. So while it's true that we can't assume KIPP's results can be generalized across the entire population, we know that additional school resources works in a subset.