Not sure what you are asking here, are you trying to get a handle on average regression rates across the industry?
I am always dubious about any such stats related to "lines of code".
My doubts can best be expressed by referring to estimation techniques based on "lines of code".
Most apps are not built from the ground up any more, so how do you calculate lines of code?
e.g. A few lines on an Oracle form can in fact be drawing on hundreds or thousands of lines of "hidden code".
The lines of code approach also suffers from ignoring complexity, ignoring design approaches, ignores code standards (e.g. a few generic classes as opposed to thousands of specific classes in OO languages).
You end up not comparing like with like!
All such estimation techniques demand tuning - calculating factors to apply that bring the results closer to how your company works.
When it comes to collecting generic stats, how do you collate these "tuning factors"?
Sorry to be negative.....