eventually i'll be posting or publishing the paper, but let me give you a slice:
If the initial probability that your position on a single factual issue is responsive to the evidence is at most .5 (50%), then that probability multiplies when the issues are multiplied. That is, if your position on whether New York will drown because of global warming correlates with that of your friends or your demographic, and so does your belief that Palin influenced Loughner, then initially we should say that the chances that your belief that the earth is warming and that Palin influenced Loughner is evidence-based are at most .25. Add, for example, the view that higher gun ownership correlates with higher levels of violence, for example, and you're at .125 and so on. At ten shared beliefs, the probability that your belief is based on evidence is about .0005.
Hence it follows from the 50% principle that the most partisan voices - the most consistent advocates of consensus factual positions of their political group - ought to be regarded as the voices with the least actual credibility, short of actively delusional schizophrenics and infants.