[ExI] FW: elections again (with new analysis)

[spike]

> Damien Broderick

> Subject: Re: [ExI] elections again
> 
> a comment from a pal on Spike's post, and my reply:
> 
> =========

http://neggie.net/vote2008/nh_primary.cgi?min_votes=300&max_votes=900

> 
> >the stealth code hypothesis also looks like a loony
> conspiracy theory -
...
> 
> Indeed. I assume it's what some commentator explains--rural voters are 
> different from city voters, and as it chanced each predominantly used 
> different voting systems. It's just surprising that the skews should 
> be so symmetrical... Damien Broderick


A rocket scientist and his numbers.  It's a beautiful thing.  {8-]

Damien's comment about the difference between rural and city voters made me
go back once again and match even more closely the size of the community
(thus filtering out the city/country voting delta.)  This time I filtered by
town ranges of 200 voters:  I took the vote counts of all NH towns of voter
population 200 to 400, then 300 to 500, then 400 to 600, then 500 to 700 and
so on up to 1000 to 1200.  So there is some intentional overlap in this
analysis.  These are all rural people.  The largest communities I compared
are those with 1000 to 1200 voters, so there are no city people involved in
this analysis at all. 

I took each group of voters and calculated the average deviation of the
machine counted votes from the hand counted votes.  The top line shows that
the average deviation for Clinton is 10.2% higher for machine-counts than
for hand-counts.  For Obama it is 10.8% lower for machine-counts than for
hand counts.  (The Edwards and Richardson percent deviations are high, but
they are based on a lot fewer votes.  If I use wider ranges, such as
yesterday's 300 to 900 voters, their percentage deviations are closer to
zero.)

Then I started swapping the machine-counts for pairs of candidates.  With
four candidates there are six possible pairs to swap.  The first line is no
swaps, the second line, I swapped the number of machine counts for Sen.
Clinton with the number of machine counts for Sen. Obama.  The other two
percent deviations (of Edwards and Richardson) didn't change of course.  

You can see the Clinton deviation went from 10.2 percent to 2.3 percent, and
the Obama deviation went from -10.8 percent to -4.1 percent.  So in that
case only did the percentage deviation between machine-count and hand-count
actually go down by swapping the machine counts between candidates (the
agreement between methods improved with that swap).  In the other five
possible swaps, the swaped percent deviation went way up, in some cases
waaay up (so the methods disagreed way more after the swap).  But in the
Clinton-Obama machine count swap, the agreement between hand and machine
became waay better after the swap:


				clinton	obama		edwards	richardson

no swaps			0.102		-0.108	0.105	-0.134

clinton-obama swap	0.023		-0.041	0.105		-0.134
clinton-edwards swap	-0.396	-0.108	1.096		-0.134
clinton-richardson	-0.814	-0.108	0.105		5.599
obama-edwards swap	0.102		-0.503	0.979		-0.134
obama-richardson swap	0.102		-0.877	0.105	5.244
edwards-richardson	0.102		-0.108	-0.726	2.466

This analysis has convinced me that the Diebold machines somehow counted
Sen.
Clinton's votes as Obama's and Obama's votes as Clinton's.

Is there any other internet chatter to this effect?  Seems this should
become a SMIR, or Snowballing Massive Internet Rumor: that the voting
machines miscounted the NH primary votes.  The implications are stunning.

spike