IMO a disservice is done by the current practice of reporting admission and matriculation statistics of multi-college universitites such as Cornell and CMU as an aggregate, when admissions criteria and applicant pools differ by college. This was not always the case; when I applied to colleges the college guides generally reported statistics for each of the individual colleges separately.
Traditionally Cornell’s engineering college had an acceptance rate somewhat higher than the university’s composite average. But the profile of its incoming freshman class on purely objective academic criteria (eg SAT scores) was higher than most of the university’s colleges; similar overall to CAS but with higher math scores. More recently engineering has become an increasingly popular choice, and now the college may actually be somewhat more selective that the university’s aggregate.
But my point is one has no necessary relation to the other, even if they happen to be the same in a particular year you happen to look. Engineering college students constitute only about 1/5 of all undergrads there, the applicant pools are different, and the relationship varies. Moreover, even if the acceptance rates happen to match, at some point that you happen to look at them, that does not mean the academic profiles of the student bodies at each of the colleges match identically.
The most recent breakout for engineering alone I’ve seen is here:
http://profiles.asee.org/profiles/6574/screen/19?school_name=Cornell+University