As far as I can tell, it's mostly to do with:
1. How people choose their subjects, and consequently, how consistent these people are in their study scores amongst their subjects
And to some (but more minor) extent:
2. How much certain subjects scale - which is essentially, a consequence of how consistent the cohort of a subject was amongst their other subjects anyway (this esp. applies to spesh from what I can see - the amount this scales often has quite a large impact on the required aggregates at the high end, most evident when it comes to getting 99.95s and such)
Essentially, in the years where aggregates are higher, the higher study scores are essentially being pooled by the people higher up as they've performed more consistently. But simply put, no it's not because of how 'smart' the cohort is as most would assume. It seems to me to be more about how consistent people were amongst their subjects, and how many people in that year essentially tried to abuse scaling (which ironically results in a decrease in the amount that subjects get scaled).