http://www.sccu.ndo.co.uk/grad.htm
It's known that a few results were missing and it can be supposed that a number of duplicated players had been identified. Thus the conclusion that (to quote the site)
is a logical consequence of universal membership.Halfgames up, graded players down!
If you match the names on consecutive grading lists or grading files, I believe it an established result that there's around a 15 % to 20 % turnover. The standing observation is that the leavers and joiners are broadly balanced. I'm fairly sure there was a one-off loss of players when the non-member Yorkshire leagues were evicted from the grading list, but I don't know if there's ever been an exact measurement of how many were removed.
There was a "white is black" argument that membership encouraged marginal players to stay within the system. I would have thought it an obvious deduction that would apply only until their membership expired, at which point they failed to renew.
Still it's a policy decision that the ECF would rather take the money by demanding membership when it can, than promote participation.
(That's an admission on the official site from a Director by the way)
FIDE require that all players appearing on the rating list are registered through a national chess federation (Rating regulations 13.1). It is surely implicit in this that such Federation may set reasonable conditions (such as membership) on the players they register.