I saw a link on Facebook to this story that claims that Mitt Romney is stealing the election by siphoning votes off other candidates. Intrigued, I dug in further. The thrust of the article seems to be that if you take the election results from Pima county for the 2012 Republican primary, and order them a certain way, you see curves appear in the resulting numbers. The ordering is from the precinct with the fewest votes cast to precinct with the most. The numbers are the cumulative total percent of total for each candidate. You can see the resulting graph here.
It's then asserted that the upward curve for Romney, mirroring the downward curve for Paul represents some sort of fraud where votes are 'transferred' from Paul to Romeny, the size of the transfer being dependent on the size of the precinct. This is a really convoluted way of showing that the more people voted in a district, the greater the percentage of them voted for Romney, therefore, somehow, 'fraud'. It's further asserted that an unmanipulated election would see all the lines being roughly horizontal.
Take a moment and put on your critical thinking hat. What would it mean if the lines were horizontal? It would mean that the distribution of votes between the republican candidates was roughly constant, regardless of the voting district. So for instance, Romney should have gotten the same number of votes in downtown Tuscon as he did in Diamond Bell Ranch off Route 286. That doesn't seem very likely. So 'horizontal' is a non-starter.
So why are we supposed to believe that the curves are suspicious? Well, because the votes have been ordered from smallest to largest precinct (as measured by total votes cast). For some reason we're supposed to think that this measure is orthogonal to the voting preferences of the people in that district. Nothing could be further from the truth.
First off, voting districts are the prime target of gerrymandering. There's nothing random about them, their borders or the political demographics of the people therein. Take a look at the actual voting districts in Pima county. Those shapes aren't random. The two leftmost districts are 1 and 2. Note that 1 (the square) has a tiny section out of it on the lower right so that district 2 can continue around it. Note the district in the upper right that is entirely surrounded by another district. All these things are done specifically to make sure incumbents stay incumbent.
Second, this is the Republican primary, so only Republicans are voting. That means the more liberal a voting district is, the fewer people there are voting. It's not unreasonable to expect that the more liberal a district is, the Republicans within that district would lean to the left of the spectrum within their party.
Finally, if you look at the actual vote count details per precinct you'll notice that the more votes were cast in a precinct, the higher the turnout was. In every case where it's over 4 digits, the turnout is over 50%. In some cases it's much higher. That's not random either. That's the result of campaigning. If I were Romney's campaign manager for AZ, I'd have been spending my money trying to get out the vote in the districts where it would do the most good.
Ultimately, manipulating data and seeing curves doesn't prove fraud. You have to spend the bulk of your time explaining why you'd expect to see something other than curves. The article doesn't do that. It just asserts that 'curves = fraud', presumably hoping that those who link it and pass it along won't think too hard about it.
Work on re-electing Obama. Don't get suckered into wearing the tin-foil hat.