Welcome to European Tribune. It's gone a bit quiet around here these days, but it's still going.
Display:
Given that the simplest way of tampering with the results is to simply reverse the counts for Clinton and Obama, this county-level chart would appear to be consistent with fraud in Hillsborough county only. This, in fact was the suggestion of the EDA people when it still looked like their statewide total "uncanny coincidence" held.

So, I did a similar ward-level chart for Hillsborough county. Is there anything in it that jumps out at you?



We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Wed Jan 16th, 2008 at 07:28:07 PM EST
[ Parent ]
So let's test the hypothesis that the Hillsborough machine counts were not reversed. We start by redoing the above plot, with machine counts reversed. I added warnings to the plot in three different places so people can't say they are being misled.



We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 07:26:47 AM EST
[ Parent ]
Please be patient with the slower amongst us...

I read the above two charts as saying in chart 1) that clinton's lead is positively correlated both with ward size and count method (machine).

Flipping count method virtually eliminates any correlation between Clinton's lead and ward size.

Ergo - Clinton's lead is actually correlated with ward size, but because that also correlates with count method, it has been confused with the correlation with count method.

Therefore there is no count method fraud.

But why is there such a strong correlation between ward size and clintons's lead?  Can it be explained by demographics, or is it easier to stuff (and hide the stuffing) of a ballot box with extra ballots in a larger ward?

However, for the correlation between Clinton's lead and count size to be so "smooth", the amount of stuffing would also have to be proportionate to ward size.  Do we really think that a fraudster would be that resourceful and clever?

I vote demographic factors...as the more likely explanation, but which ones?  Gender, education, income, race, class.....???  Do we have enough demographic data to come to a conclusion?

Index of Frank's Diaries

by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Thu Jan 17th, 2008 at 08:09:51 AM EST
[ Parent ]
It appears just about every democraphic factor that has been tried, including the size of the precinct both as a continuous variable and as a large vs. small categorical variable, has failed to remove the Diebold effect. I am not sure gender has been tried, though. I am also not sure that the regressions involving both size and method have been carried out properly.
However, for the correlation between Clinton's lead and count size to be so "smooth", the amount of stuffing would also have to be proportionate to ward size.  Do we really think that a fraudster would be that resourceful and clever?
No, in this case the suggestion is a simple reversal of the vote counts between Clinton and Obama.

Flipping count method virtually eliminates any correlation between Clinton's lead and ward size.

Ergo - Clinton's lead is actually correlated with ward size, but because that also correlates with count method, it has been confused with the correlation with count method.

No, this is not "flipping count method". It is flipping the sign of the lead between Clinton and Obama for one of the two methods.

We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 11:19:57 AM EST
[ Parent ]
As a first test, the distribution of dots is:

      Above Below Total
Black   3    11    14
Red     6    30    36
Total   9    41    50

We can do a two-way chi-square test for the independence of red/black and above/below (unfortunately the 3 is a little too low, but we'll go ahead anyway). The expected values are

      Above Below Total
Black  2.52 11.48  14
Red    6.48 29.52  36
Total   9    41    50

The chi-square contributions are

      Above Below Total
Black  0.09  0.02  0.11
Red    0.04  0.01  0.05
Total  0.13  0.03  0.16

This has a p-value of 69%. That is, a pretty good fit.

We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 08:23:39 AM EST
[ Parent ]
For fun, rather than vote differential vs. ward size, I plot vote differential vs. democratic/republican ratio in total votes. Top plot, numbers 'as they are'. Bottom plot, machine vote flipped.
Numbers as they are, opposite trends for machines/hand-count with ratio of dem. votes. Visually quite striking, top vs. bottom.
by someone (s0me1smail(a)gmail(d)com) on Thu Jan 17th, 2008 at 09:23:08 AM EST
[ Parent ]
Top chart shows clinton doing better in predominantly Democratic wards - as one would expect - given apparent independent preference for Obama.

Bottom chart shows clinton doing worse in predominantly Democratic wards - as one would not expect

therefore the vote switch hypothesis is not supported?

Index of Frank's Diaries

by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Thu Jan 17th, 2008 at 09:35:34 AM EST
[ Parent ]
Top chart shows Clinton doing better in machine counted dem. wards, and Obama doing better in hand counted dem. wards. Note, this is not strictly dem. wards, it is the ones with more people voting in the democratic primary. Those people could be either registered democrats or independents:
New Hampshire primary - Wikipedia, the free encyclopedia

Undeclared voters -- those not registered with any party -- can vote in either party primary.
...
Additionally, as of 2002, 25.6% of New Hampshire residents are registered Democrats and 36.7% are Republicans, with 37.7% of New Hampshire voters registered as "undeclared" independents. This plurality of independents is a major reason why New Hampshire is considered a swing state in general U.S. presidential elections.
by someone (s0me1smail(a)gmail(d)com) on Thu Jan 17th, 2008 at 09:52:08 AM EST
[ Parent ]
Now let's apply the same nonparametric test to the data with the machine counts flipped. The result is quite different:

        Wilcoxon rank sum test

data:  (Clinton..d - Obama..d)[machines] and (Obama..d - Clinton..d)[!machines]
W = 234, p-value = 0.7086
alternative hypothesis: true mu is not equal to 0
95 percent confidence interval:
 -0.10751315  0.07106908
sample estimates:
difference in location
            -0.0175131

This means that one cannot reject the hypothesis that the hand counts and the reversed machine counts have the same distribution.

We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 11:46:27 AM EST
[ Parent ]
This is a nonparametric test of the hypothesis that the two subpopulations, machine and hand counts, come from the same distribution.

        Wilcoxon rank sum test

data:  (Clinton..d - Obama..d)[machines] and (Clinton..d - Obama..d)[!machines]
W = 422, p-value = 0.0001150
alternative hypothesis: true mu is not equal to 0
95 percent confidence interval:
 0.1186801 0.2903706
sample estimates:
difference in location
             0.2027052

The test is astonishingly simple: you rank the combined population and add up the ranks of the elements of one of the two subpopulations, and then look up in a table. This is telling us that the Hillsborough hand counts and the machine counts come from different populations at 99.99% confidence level.

We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 11:42:29 AM EST
[ Parent ]
do there is a 99.99% chance the machine counts were tampered with rather than be explained by random variation.  But since the distribution of machine counts WAS not random, what does it tell us?

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Thu Jan 17th, 2008 at 11:49:16 AM EST
[ Parent ]
That there is a very strong effect, amounting to a 20% swing in Hillsborough county alone.

It doesn't tell us where the effect came from.

We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 11:58:02 AM EST
[ Parent ]
Flip 42.7% of 80542 = 34391 votes for called for HRC with 34% of 80542 = 27384 for BO

80542 : total votes in Hillsborough

Adjust for the 7007 difference in the total vote tally - and they flip as well 112000 - 105000 ! And the undoctored handcount percentages - 39.3% for BO and 36.6% for HRC - reappear for the whole state!

Hillsborough has 3 handcounted precincts (Atrim, New Boston, Wilton) reported in the last yellow table, Appendix E of the Paired Precinct study.

Those 3 handcounted precincts have 30% for HRC and 42% for BO. It must be highly unlikely that Diebold precincts within the same county diametrically differ from handcounted precincts.  

Hillsboroug as an urban area would tilt even more towards BO looking at demographics. Rural areas would be less black, less young and more conservative. A female bias in urban areas cannot be that pronounced to override this. Indeed BO's advance is even more pronounced in urban areas.

I bet that the recount will show that the Hillsborough county total was flipped.

One single operation performed by a remote hacker on the biggest county. Couldn't resist the Freudian temptation to leave a clue by taking on Hillsborough for Hillary.. A simple and workable method that could be started at the beginning of the poll evening.
The tactic would survive a first hand recount check that the individual Diebold precincts were correctly reported to the total tally before the additions.

Not elaborate and subtle doctoring precinct by precinct which would have meant a bigger conspiracy, a considerably higher risk and an uncertain outcome. Creating votes out of thin air could probably be detected by uncorrelated turnout documented elsewhere. The flip can also be denied by to be the whim of a computer or a software.
Unfortunately, the Hillsborough flip also flipped the statewise total ! Now, I can't imagine that the flipper would have the nerve to swap the candidates at the TOP LEVEL presentation after the additions, but why not ?

Elling

by Elling (elling@torium.se) on Thu Jan 17th, 2008 at 03:56:19 PM EST
[ Parent ]
Yes, what we know (in particular the charts I have posted in this subthread) is consistent with flipping of the vote counts at the ward level in Hillsborough county only. But it doesn't prove that is what happened.
Those 3 handcounted precincts have 30% for HRC and 42% for BO. It must be highly unlikely that Diebold precincts within the same county diametrically differ from handcounted precincts.  
Yes, that is really unlikely, but if you have 10 counties that one will be nearly diametrically flipped is 10 times more likely, so the bar is higher: to have 95% confidence that the effect is there in one county out of 10 you need 99.5% confidence for the result in that county only.

However, that the result in the largest county is flipped is equally significant regardless of the number of counties. Unless, of course, you have formulated the hypothesis after looking at the data, in which case the bar rises again.

We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 04:21:31 PM EST
[ Parent ]
Does flipping Hillsborough machine counted votes change the overall result of the Primary - i.e. does Obama then win?

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Thu Jan 17th, 2008 at 06:25:34 PM EST
[ Parent ]
The original results are
Clinton  Obama
 112610 105007
with
Clinton  Obama
  31928  25525
in the Hillsborough county machine-counted wards. Reversing those gives
Clinton  Obama
 106207 111410


We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Thu Jan 17th, 2008 at 07:28:13 PM EST
[ Parent ]
Thank for the corrected exact numbers.

Now , maybe the irregularitites extnds to wards outside Hillsborough. The table contains other 40-34 pairs seeminglynext to each other that are reversed from  hand to Diebold.

Also, the exit polls estimated a double digit Obama win. So the fraudster may have have had to flip more than Hillsborough : Changing 40-34 for Obama into 39-36 for  Clinton would rather mean flipping the WHOLE Diebold count. Is that feasible ?  


Elling

by Elling (elling@torium.se) on Fri Jan 18th, 2008 at 04:48:54 AM EST
[ Parent ]
The county-level data do not support the hypothesis that the machine counts were flipped in all counties. Only Hillsborough presents a near match.

The total number of Democratic votes was 288055. The preliminary counts gave Clinton 39.1% to Obama's 36.5%. Flipping the Hillsborough machine counts gives Obama 38.7% to Clinton's 36.9%. This also looks like a near reversal, just because of the size of Hillsborough relative to the whole state.

It has been argued that the order of the candidates on the ballot can explain up to a 3% difference between pre-election polls and actual results. Going from 40:34 to 39:37 for Obama due to this ballot placement effect and then to 36:39 due to a Hillsborough flip is not so far-fetched.

We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Fri Jan 18th, 2008 at 05:02:57 AM EST
[ Parent ]
But if it can be verfied that the ballot layout and its ordering of candidates has this dramatic effect, isn't that reason for a replay of elections ?

Who gets to design the ballots anyway ?

Elling

by Elling (elling@torium.se) on Fri Jan 18th, 2008 at 05:19:49 AM EST
[ Parent ]
Well, you have to have the candidates in some order or another on the ballot, don't you? In this case it was alphabetical. Have you seen the ballot, by the way? [PDF]

Though I have read somewhere (no link, sorry) that it used to be that each precinct had a different randomly generated ordering so this effect was minimised, except that procedure wasn't used this time around.

We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Fri Jan 18th, 2008 at 05:54:38 AM EST
[ Parent ]
I had a quick look at the NH secretary of state website and the recount doesn't seem to be showing up any significant anomalies.  Are they hand counting the paper ballots that were previously machine counted?

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Fri Jan 18th, 2008 at 06:47:44 AM EST
[ Parent ]
Forgot to say - congrats Migeru - you got another 200+ comment thread to your name - thanks in part to all my stupid questions....

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Fri Jan 18th, 2008 at 06:49:37 AM EST
[ Parent ]
Here is the recount page.

Yes, they are recounting the machine ballots in Hillsborough and Rockingham counties, I believe.

We have met the enemy, and he is us — Pogo

by Carrie (migeru at eurotrib dot com) on Fri Jan 18th, 2008 at 07:21:21 AM EST
[ Parent ]
Yep, that's the page I was looking at - which doesn't seem to be showing any anomolies - which either means there is no fraud, or they are using the same methodology, or their methodology is flawed

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Fri Jan 18th, 2008 at 01:30:36 PM EST
[ Parent ]
The other hypothesis is that the actual ballots were tampered with.

Black Box Voting : 1-17-08: Ballot boxes found slit; NH stops putting ballots in vault;

No worries, say New Hampshire officials when cuts up to eight inches long are spotted in newly delivered ballot boxes. "The only seal that counts is the one on top."

Except the seal on top can be peeled off without leaving a trace, then reaffixed.

Black Box Voting has been doing a chain of custody exam for the New Hampshire Primary's recount. On Wednesday night, Election Defense Alliance's Sally Castleman mentioned a troubling observation: After following the ballots back to the ballot vault following Wednesday's recount, she had the opportunity to enter the ballot vault, and noticed what looked like cuts, or slits, in the side of many ballot boxes. New Hampshire officials assured us that these cuts, which slice through the tape and seals do not permit access to the uncounted ballots, pointing to a label on the boxtop which they call a seal.
Drew pointed me to this page.

We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Fri Jan 18th, 2008 at 01:49:56 PM EST
[ Parent ]
One remote programmer can make a change to a program to flip results for two candidates and then erase the change afterwards to cover his tracks.  It takes a conspiracy of an entirely different order to enter vaults, slit boxes, and stuff them with exactly the right number of ballot papers per box or ward to replicate the effects of the counting switch - which would require adding c. 7000 Clinton ballots and removing c. 7000 Obama ballots. Is there ANY evidence that could support this hypothesis - other than the amazing fact that NH procedures seem to be so lax as to make it possible?

Index of Frank's Diaries
by Frank Schnittger (mail Frankschnittger at hot male dotty communists) on Fri Jan 18th, 2008 at 02:01:35 PM EST
[ Parent ]
No, there is no evidence of this except that NH's chain of custody is so lax as to give LHS the means and opportunity to do it after the recount was called (which would provide the motive). However, this is highly speculative as far as I'm concerned. There is a pile of circumstantial evidence but no hard evidence any crime was actually committed.

Black Box Voting : 1-17-08: Ballot boxes found slit; NH stops putting ballots in vault;

I confirmed this morning that many if not most of the boxes scheduled to be counted today had slits in them. I went out when a vanload of ballots arrived, and saw that they were slit at the time they arrived by van. Susan Pynchon and I drove to two nearby towns and watched as they handed over their ballot boxes to "Butch and Hoppy", the two men who drive around in the state in a van picking the ballots up. We observed as they loaded boxes of ballots into the van with no slits at all in them. We videotaped each of these up close. They arrived at the destination without slits. The label on the top was affixed, but in some cases was crumpled, or also damaged.


We have met the enemy, and he is us — Pogo
by Carrie (migeru at eurotrib dot com) on Fri Jan 18th, 2008 at 02:11:01 PM EST
[ Parent ]

Display:

Top Diaries

Occasional Series