This paper shows that a family of methods for attributing market value gains to effects shares a key dilemma that does not arise in an idealized model.  Specializing the idea to the Brinson attribution model reveals certain refinements in practical algorithms.

Refinements in Multi-Period Attribution