I think this pretty much sums it up:
1 The decline represents a change in the rate of retention of desirable, good-faith newcomers.
- The proportion of newcomers that edit in good-faith has not changed since 2006.
- These desirable newcomers are more likely to have their work rejected since 2007.
2 This increased rejection predicts the observed decline in retention.
- Semi-autonomous vandal fighting tools (like Huggle) are partially at fault.
- An increasing proportion of desirable newcomers are having their work rejected by automated tools.
- These automated reverts exacerbate the predicted negative effects of rejection on retention.
- Users of Huggle tend to not engage in the best practices for discussing the reverts they perform.
3 New users are being pushed out of policy articulation.
- The formalized process for vetting new policies and changes to policies ensures that newcomers' edits do not survive.
- Both newcomers and experienced editors are moving increasingly toward less formal spaces.