Vote counts for Clinton/Obama exchanged in New Hampshire

It's funny how things happen sometimes. After the New Hampshire primary last Tuesday there were some suggestions that the Democratic primary might have been tampered with. I have taken part in some discussions of this on and off the blog, and my basic take was that

1. the implications are important enough for the allegations to be taken seriously;
   
2. "taking the allegations seriously" means carrying out some further tests before jumping to conclusions, especially if you're going to call for an official recall like Dennis Kucinich has done!
   My initial suggestion was to compare the actual vote counts by hand vs. by machine against the exit polls, if an exit poll could be found that aggregated the data according to the vote counting method used in the precinct where the voter was interviewed. Of course this variable was not in the published exit polls. In my enthusiasm I imagined someone in the blogosphere knowing someone in one of the companies that do exit polls, so they could try to get the raw data from election night re-analysed. But see below.
   
3. if a statistically significant discrepancy between the vote percentages and the exit polls, aggregated separated by vote counting method, were found, one would have to remember that correlation doesn't imply causation. One could imagine socioeconomic variables correlating independently with both the Clinton/Obama swing and the use of voting machines in a precinct. For instance, rural vs. urban precincts, the size of the town, the average income of the town, whether the local government is democrat or republican controlled, etc. All of these plausible explanations would have to be controlled for before one could claim to have evidence of election fraud.
   

   A very interesting discussion, promoted by In Wales - some intro moved below the fold

---

Given the impossibility of getting my hands on disaggregated exit poll data and the observation that the exit polls matched the election results (but see below), I decided not to give the issue any more thought. That is, I ceased to take it seriously.

However, supporters of both Ron Paul and Dennis Kucinich among other people didn't lay this to rest and started posting fragmentary statistical analyses of actual voting data. These got more and more elaborate until, earlier today, Drew, who had nagging suspicions and so had been tracking what the blogosphere was saying about this issue, pointed me to a couple of websites that changed my mind, because they contain serious statistical analysis of the kind I suggest in 3) above, albeit performed only on the election data, not on exit polls. But that is actually better as I'll also discuss below.

Now I think there is a high likelihood that the vote counts for Obama and Clinton were exchanged by Diebold voting machines.

The first indication that something might be amiss came in this comment by ThatBritGuy:

From a comment by "soros" at the big orange monster:

read more | digg story