It is an interesting question, but I think it would be hard to quantify. I know you've been keeping statistics on arena populations, the only question I would have is what times were those samples taken? It's entirely possible that numbers online dropped during some times and increased in others, I would also imagine that arena numbers could be a decent analog for subscribers the two numbers are technically unrelated. I could keep an account and not log at all during a given tour (I know I've done it.
) There also could be multiple accounts for one household that get consolidated down to one. Plus how to quantify lower numbers and arena caps and not other external factors like a down economy?
But is it an interesting question. 
First, let me claryify that I never said the caps are the (sole/main) reason for dropping numbers. It's my personal belief (<-important) that the caps did not have any considerable negative influence (on player numbers online) for a long time. HT stated that the quote of players keeping a subscription after their 2weeks did improve.
However I do have the opinion that now they possibly do contribute to the problem, as it takes more and more time to get to "critical mass" during off peak times.
In other words: For one's individual gameplay, it doesn'T make that much difference if there are 500 or 250 players on a large map. But having to join a 40 players arena instead of playing with 180 makes a much bigger difference.
But all in all, the phenomenon of decreasing numbers is a different, (and quite complex) discussion.
On the numbers:
I have three sets of numbers I'm currently basing my argumentation on.
The first, and weakest data source is my memory. While I always took big interest in all numbers (as you all know

), I did not care to keep records during the "golden days". While I used to fly a lot both at Euro as well as US times, I'm reluctant to compare US numbers now and then. But I belive I have a very good memeory for offpeak numbers.
But as memory is fallible, it try to get different data to compensate for that. One crutch to measure some kind of "activity" is simply to count the number of kills made in the arenas over time - this number is available. While there can & will be some minor variation due to holidays, tour length and maps, it's quite comparable as the basic distances & gameplay design didn't change during the past 5 years:

(Black column= Arena Split)
And the third set is data actually recorded by me. I started this last month, to see how player numbers and caps work out during the day. By chance I found one set of numbers showing arena population during Euro prime from early 2009, which allowed me a comparison based on hard data.