Thursday, June 01, 2006

Lies, Damn Lies, and Statistics

As a trained political scientist with more experience than I care to recall at creating and analyzing polls & voting behavior, the story in the new Rolling Stone by RFK Jr. on the results of the analysis of the 2004 exit polls is, in a word, shocking.

The compelling piece of data is the difference between the margin of error on the exit polling (approximately 1%) and the swing between the exit poll and the reported votes. A slight difference could be explained, but consistently between 7% and 10% gap between the exit poll and the vote, and overwhelmingly in Bush's favor? Sorry, no.

There needs to be some aggressive fact checking here, particularly in those precincts with unnaturally high/low votes. One simple thing to do is check the voter list. Every precinct must maintain a check list of who came in and voted there. Who do they say voted? In the super high turn-out precincts, are those people actually there? In the low ones, did more people turn out than that?

What militates against the vote-corruption (messing with voting in whatever manner), is the many kinds of messing around that would have to have been done. Coordination and secrecy issues are difficult to keep straight under these circumstances. However, what militates for it is a consistent pattern of extra votes for George in safe Republican precincts where the exit polling data did not show him winning by such margins.

In particular, the claim that Bush voters were underrepresented in the exit polls - the only phenomenon that can save the data - is proved wrong. Voters in Kerry strongholds were less likely to participate, while voters in Bush strongholds were more likely to, thus providing inflation of Bush's numbers in the exit polls - which still showed him behind.

The discrepancy in margin of error points to the distortion of numbers more than anything else. Now the question is to identify the patterns of vote distortion and how they can be matched to specific and plausible kinds of vote fraud.

Me, I say most of it is old fashioned ballot box stuffing. That's why it is important to scrutinize precinct numbers and names. The precincts that showed a disproportionate number of Bush votes in comparison to the exit polls need to have their voter rolls closely examined, and see if the people who "voted" A) were actually alive (the cemetary wards being a long-standing US tradition) and B) were equal to the number of votes cast.

Fund raising is always important, but the real challenge of the coming elections is to get a really good look at voter registration lists. The fraud is happening at the precinct level, a few dozen here, a half-hundred there. It is happening in people being turned away (be on the lookout for that in November as the wingnuts target "Hispanic" voters) or being denied adequate voting conditions. GOTV efforts have to start now, and have to focus on ensuring that the rolls are clean.


No comments: