Re: LARGE Human factor problems in florida e-voting??

From: charlie strauss <cems_at_earthlink_dot_net>
Date: Thu Nov 04 2004 - 19:15:21 CST

Kathy, I just updated the figures (for legibility) and changed the bottom axis label to avoid a confusion that we appeared to have. The bottom axis in the first two figures is then number of expected voters based on party registration percentage multiplied by turn out. My original caption was confusing and suggested this was the number of registered democrats or republicans. I meant this mean the expected number who voted.

I answer your questions below

-----Original Message-----
From: Kathy Dopp <kathy_at_directell_dot_com>
Sent: Nov 4, 2004 3:41 PM
To: voting-project@lists.sonic.net
Cc: cems@earthlink.net
Subject: Re: LARGE Human factor problems in florida e-voting??

charlie strauss said:
> the results can be found here for now:
> http://vvnm.org/resources/florida2004/florida_vote_patterns.htm

>I've had two other mathematicians do regression analysis whose conclusions
>disagree with yours, and I am not sure on what data you used to plot some
>of those charts yet.

the data is off your web site.

>In most of your charts, all the Red REP voters on the top of the 45 degree
>line, and all the Blue DEM votes underneath which shows a clear pattern
>for optical scan votes that is completely at odds with the touchscreen
>data and what one would normally expect.

If you look at the blow up graph you see that most of the touchscreen (solid points) lie amongst the scatter of the paper ballots. So I dont understand the comments. The means trend of these sets seems to be close. One could argue they are not close enough statistically. But that argument is not neccessary because the fact that the paper trails APPEAR to have a larger scatter alone indicates that people dont vote the same way on touch screens as on paper ballots. I dont entirely understand why some counties are outliers. I would have expected these to be damped out when averaged over an entire county. That remains curious, and I cant explain it.

>Another one of your charts shows clear unmistakable divergence too with
>most of the Repub on the top of the 45% line and most Dems below it for op
>scan systems, so that for some mysterious reason Dems voting on opscan
>machines seemed to abandon their party and everyone seemed to abandon
>their party, and join the Republicans

the region of strongest divergence is in the smaller counties and those are only optical scan. there is no point to compare with e-voting. thus I dont see how you conclude that e-voting is not following optical scan in this region. perhpas you did not realize the e-voting is the solid dots?

The place on my figures that I think is the most controversial is the lower lefthand corner of the blow up. are the optical scan democratic counties that lie well off the trend line of the e-voting disagreeing with the e-voting? I think you (kathy) would say yes. I'm less certain. Perhaps this is edge of the onset of the second trend I noticed that peeled the republican and democrat lines off this trend. It is hard to say. That's why I blew it up so the reader could decide. It's a tough call. I lean towards agreeing with what I think you are saying but I' reluctant to go the whole way.

>>
>> the e-votes appears to follow party line (more predictable by the people
>> registered in a county to a given party) much much more tightly than
>> the optical scan process.

>Yes. Agreed. Op-scan is showing a very suspicious pattern, but it might
>yet be explained by other factors or demographics which people are working
>on putting together.

yes there could be many factors here. its a good guess that opscan districts are poorer though not neccessarily so. One person (edward) suggested that the media market might be correlated with the county size and thus they might have seen different advertising sets. Perhaps the media market could be correlated with voting machine type as well (via affluence perhaps).

>
>> Another hypothesis would be that optical scan machines are wildly
>> inaccurate or have incidences massive ( perhaps >30% !!) undervotes that
>> tend to favor Republicans a bit more often than Democrats. (Most

>It doesn't look like there were many undervotes at all in this race, but
>I'll look at it as soon as I get a chance.

agreed undervotes explaining that are a long shot. That second hypothesis was a long shot just placed there to offer a counter point to the previous hypothesis that serial voting distorts the voting process when compared to paper.

>I don't particularly like your chart that maps actual votes for party pres
>vs. number of voters registered to party because it doesn't seem to take
>into account voter turnout in each county, or does it?

I clarifed the figure labels. It does take into account turnout.

==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Tue Nov 30 23:17:10 2004

This archive was generated by hypermail 2.1.8 : Tue Nov 30 2004 - 23:17:43 CST