He constructs a scenario that gives a similar linear relationship for the Obama/McCain results. Of course, the US results were reported by time zone, not alphabetically, so it has an artificial ring to it. But his point is well taken - a high R^2 does not *automatically* prove fraud.

http://www.fivethirt...es-not-prove.html

John Cole discusses some reason why the results are suspect, here: http://www.juancole....ian-election.html

Cheers,
Scott.