GOP spreadsheet: a proof of tampering?
Submitted by erikderzweite on Wed, 02/15/2012 - 18:37I have downloaded Official Vote Breakdown from Maine GOP - http://www.mainegop.com/wp-content/uploads/2012/02/me_gop_ca...
Then I have converted it to an editable format (LibreOffice Calc in my case) and did the math. The result is available here:
http://dl.dropbox.com/u/20804013/MaineCaucus2012.ods
First I have seen that if you sum all candidate's totals, you get 5585 votes, the official number. But if you sum the "Totals" column (column "H"), you get 5493 votes. Hmm... I've made my own "Totals" columns then (just the sum of votes cast for all candidates, column "I") and have compared both columns. At first, I did not see anything spectacular, Bangor Totals was apparently missed with 77 votes cast and a 0 in "Totals", a +1 here, a -4 there... Several honest mistakes at a first glance. True, some of the entries with irregularities shown 1 vote lead of Romney, but I am not that paranoid.
But then I did a sum of all the differences (absolute values, -4 votes counts as 4, for example).
The result: 100 votes exactly. Now, what is the chance? Romney "wins" with a margin of 196 votes and the official spreadsheet's totals are off by exactly 100?
It is fishy at best.
Relevant?
I saw something that reflected that the spreadsheet was originally Excel XLS sheet.Believe it was excel 2006 if I remember right. It was changed to a PDF later and posted on the web.
Not sure that will help with anything, but figured I'd mention it.
Sorry, I love numbers. I
Sorry, I love numbers. I just had a thought.
The distribution of Paul-Romney is normal and centered at 0 over all samples.
If we assume that the errors in the entries were normal, then shouldn't the Paul-Romney differences within only these error totals also be centered at 0? Well, it isn't, it is shifted to Romney.
Ok, now let's look at that distribution of Paul-Romney differences within the error rows some more:
The Wilcoxin-Rank sum test I ran showed that the distributions of Paul-Romney are significantly different between the error rows and the correct rows (p = 0.0173)
There are only 12 samples we have that are errors, but that's what the stats are for...
It's great that you did this.
It's great that you did this. I'm looknig at the numbers now. It would be interesting to see if there was a pattern to the votes in the precincts that were different.
I noticed that if you sum the vote differences between Romney and Paul in only the towns that had messed up totals, then you get -40 (i.e. 40 votes more towards romney in precincts that had different totals than the reported value). That's weird, but I don't know what it means...
I think there could be a way to check the "randomness" of the situation. I've started looking at the distributions of the votes in all counties for all candidates.
I had an idea to find the best linear function between the columns and the totals to see if there was some sort of pattern to the "incorrect totals" or just in general (i.e. is there a consistent formula that gives the total based on the number of votes in those "messed up precincts?", how about in general?)
There are only 12 messed up totals, but I was thinking there might be a way to determine if they were just random errors of if there was some consistent pattern...
Here are the numbers for
Here are the numbers for those "messed up totals"
The last columns are as you had them, but I added a "Paul-Romney" diff column at the far right.
24 23 8 4 0 56 59 3 -1
29 28 19 9 0 81 85 4 -1
18 17 3 0 0 40 38 2 -1
19 6 2 5 0 28 32 4 -13
2 2 3 1 0 7 8 1 0
0 2 0 0 1 0 3 3 2
26 36 9 4 2 0 77 77 10
1 0 1 0 1 4 3 1 -1
9 6 6 1 0 21 22 1 -3
0 0 5 3 1 8 9 1 0
20 12 3 6 2 44 43 1 -8
29 5 11 2 0 45 47 2 -24
went through your post and
went through your post and spreadsheet but I am not sure about what you are trying to say here.
Yes, the 'total' column seems to be off, but what does that mean? Is it relevant?
No idea, that's why I'm
No idea, that's why I'm asking.
I just find it odd that the end number is so weird for a mistake. There are many small mistakes which could be just that -- mistakes and inability to use a spreadsheet program. And by a remote chance these honest mistakes make exact 100 votes. I am not saying here that 100 votes were moved from Paul to Romney, that is not the case as the bulk of this mistakes is one missed total of 77 votes.
Still, it's a somewhat big irregularity given that the winner is determined by 196 votes. I know that some folks investigate deeper on the issue, maybe it could be useful to them.