In the conferences I've experienced, each paper gets a serial number that can tell its authors how many submissions were made prior to theirs.
Many conferences have a fixed number of tracks (and days), and thus the number of accepted papers cannot vary too much. (The organizers do have some flexibility, but it is rather limited.)
Thus, knowing the number of submissions can help one assess his chance of being accepted to the conference, which might tempt him to submit at the very last moment just to figure out the submission count.
Obviously, not everyone can do this. Why don't the organizers simply post it on the website after the submission date is due?