MCAT numbers go up because more people are applying.

Let's say school X has 10 spots available, and it wants to pick those that are the most competitive.

The MCAT is a preset curve (or at least the % change very little between tests). So lets say 1997 100 people took the test. Based on aamc %

http://www.aamc.org/students/mcat/admissionsadvisors/examstatistics/scaledscores/combined08.pdf
50 people < 26

25 people 26-29

10 people 30-32

10 people 33-35

5 people >35

So school X will take the top 5 (with MCAT >35) and 5 of the next 10 (33-35)

In 2006 because of increase interest in the medical field, 200 people take the MCAT. Using sampe %

100 people < 26

50 people 26-29

20 people 30-32

20 people 33-35

10 people >35

School X hasn't changed it's class size since medical admissions were capped, so it still only has 10 spots. So it takes the highest 10. Those 10 all have MCAT >35.

So in 1997 matriculates at school X had an average MCAT of [5*(>35)+ 5*(33-35)]/10. In 2004, school X's average MCAT was [10*(>35)]/10, which would be greater than that of 1997.

This is an exageration, as I doubt medical school applicants (or MCAT takers for that matter) has doubled between 1997 and 2004.