The argument against explanation 3 (roughly, "NCLB created incentives for schools to teach kids to be good at math tests and not good at real-world problem solving") is that no one teaches to the NAEP, and that the NAEP has stayed mostly constant in its material coverage since '78.

This would be sound if the NAEP tested real-world problem solving. However, judging by their sample question booklets ( http://nces.ed.gov/nationsreportcard/about/booklets.asp ), the NAEP consists of the sorts of questions one would assign for homework in a high school algebra or geometry class: e.g., "train" problems, proofs of properties of triangles, and so on -- not exactly things I would call "real-world problem solving."

My own personal experience as an undergraduate math instructor hints very strongly to me that #3 is the most significant issue.