It's a probability thing. For instance, suppose 90% of great people go to great schools. Suppose only 20% of non-great people go to great schools. Suppose only 1% of people are great. (Just making up numbers, as the exact numbers don't matter to illustrate the point. All that matters is that great people have a significant tendency to go to great schools, and that there are a lot more non-great people than great people).
Then a random candidate from a great school will have a 4.3% chance of being great, whereas a random candidate from a non-great school will only have a 0.13% chance of being great.
If you have a ton of resumes, and only time to examine, say, 1000 of them, pulling at random from a pool that mixes those from great and non-great schools will give you a pile with about 9 great candidates (8 from great schools, 1 from a non-great school).
If you tell HR to filter out the non-great schools, and then look at 1000 resumes from just the great schools, you'll have around 43 great candidates.
Then a random candidate from a great school will have a 4.3% chance of being great, whereas a random candidate from a non-great school will only have a 0.13% chance of being great.
If you have a ton of resumes, and only time to examine, say, 1000 of them, pulling at random from a pool that mixes those from great and non-great schools will give you a pile with about 9 great candidates (8 from great schools, 1 from a non-great school).
If you tell HR to filter out the non-great schools, and then look at 1000 resumes from just the great schools, you'll have around 43 great candidates.