The supposed decline of copyleft

I personally think any study of this data is going to be wrong. The data sets will be polluted with all kinds of stuff that you probably wouldn't want to include to get a reasonable call on whether there is a trend away from Copyleft. As an example there are thousands of little scripts and such posted to github and other places that are typically freely licensed precisely because there is literally no reason to even copyright them.

To effectively derive any kind of meaning (IMO) you'd need not only lines of code, but number of contributors and downloads, activity and such to try to exclude these one shot code releases that in all likelyhood there is no significant use and IMO aren't of any value in making this evaluation. There's been a dramatic increase in these one shot postings of scripts and other code that before would have just been posted to pastebin or some forum before (along with no copyright statement) and are now posted to github because of the simplicity and ease of use. You include these things in there and it skews the data towards permissive but no one would license this stuff copyleft simply because there isn't a point because the author has no intention of maintaining anything or making a project out of it. Without excluding these kinds of data you simply can't get an accurate representation.