Protecting local voters from insufficient testing by Secretaries of State

In spite of the comforting recent message from the Colorado Secretary Of State that our election quality relies on individuals with integrity, there is a desperate need for oversight for Clerks and the Secretary herself at every point in the election process from the making of procedures to the training of election officials to the performance of the pre and post election tests.

To achieve a quality election in Colorado we must have both oversight and a repository for election problems as well as constructive improvements in the procedures. We have been making such improvements in Eagle County.

As an example, and as a first minimal step, forms for reporting election events such as the "post election random audit" must provide space for exceptions and criticisms to be described by the county canvass boards when reporting to the Secretary of State.

There is no question that as a result of the recent Conroy trial in Colorado, additional security now required will be used as an excuse to avoid necessary oversight and testing of our voting systems and procedures. I am very concerned about the overall effects of the recent Colorado court case and hope that the energy is redirected towards better testing of all kinds and real oversight and attention to the critical details of counting votes before securing the machines with additional layers of technology.

I am also concerned that Clerks are limiting access to the very important Public Logic and Accuracy Tests in some counties. I am the Canvass Board member and party Co-chair and testing board representative for the Democrats in Eagle County. Any county Logic and Accuracy Test in this election needs extremely close and careful attention by a team from each party. These tests need people who are critical and questioning both in conducting and observing. There are still unasked and unanswered questions about election procedures and systems in every election jurisdiction. It is not wise to have only those who are sure of their knowledge and ability conducting our election tests- or our elections.

This year the Eagle County LAT (Logic and Accuracy Test) has found positive evidence that Diebold Accuvote 2000 (optical scanner) can not be depended on to accurately count all races on a folded ballot. This machine is very popularly used throughout Colorado for absentee voting and in some cases still used for precinct voting (e.g. Pitkin County). In case of absentee (voting by mail) ballots are probably always folded.

We found problems in the vote counting by acting on the simple guess that the Accuvote had ever been properly tested. We tested with folded ballots, first by testing the special "test" ballots which were provided flat, but then hand folded, and then by testing the actual machine folded absentee ballots from inventory.

Spurious votes and overvotes which cancel out the voters intention are appearing at or near the folds. We now do have proof using simple tests of 4 or 5 ballots passed through the machines four times each. These folds are near but not crossing over the ballot oval itself. Tests of flat ballots do not show these problems.

Because of our recent tests, the Eagle County Clerk has decided and I agree that we must hand count problematic races in the absentee counting of the 2006 General Election just as we were forced to do in 2005 when transferred toner or ink voted extra votes for at least one of the candidates on our Diebold manufactured and pre-folded ballot. This time we are no longer using Diebold as a ballot manufacturer, and so far we have no evidence of mirrored text appearing on our ballots.

As a result of our experience this year we started to test folded ballots to get closer to the conditions of real absentee ballots. I folded 25 test ballots myself, stuffed them and ran them through an automated postage machine to simulate mailing. Folding the ballots leads to votes appearing where no marks are found on the ballot. On machines which were returned to Diebold for servicing after the 2005 election because of a similar problem, we found shocking differences in the performance of each machine tested in the extent to which they detected non existent votes.

Every election jurisdiction in the world should be made aware of the potential for the Diebold Accuvote 2000 to count in error. But who is responsible for accomplishing this: how will this information be shared? Sorry, there is no answer for this question. Our election process is optimized to maximize voter trust. Evidence of failure does not promote trust and this bad news is not wanted or often shared.

Colorado election procedures require the word "test" to be put on test ballots. Printing the word "test" damages the validity of the Logic and Accuracy Test. What this means is that we almost never test the real absentee ballots... and they are the ones which are folded and don't say "TEST" on them. This year Eagle County test ballots are voting "no" on one of the Court of Appeals judges and one of the local initiatives because the screen print of the word "TEST" actually goes across two of the ovals on the ballot, invalidating both the logic and the accuracy of the test.

The LAT is no longer a perfunctory matter. It is now crucial to carefully perform the LAT at the very least to be sure that the ballot on a DRE (LCD screen voting device) is programmed correctly.

Where once it was possible for a couple of people including a Clerk to examine the paper ballot for accuracy, now it is necessary to operate the DRE (Direct Recording Electronic) machines sometimes for days to find out if the ballot is programmed correctly on all ballot styles. And there is always a potential for lack of care and resulting error during these time consuming tests.

Testing is the hidden cost of the DRE, and it is a heavy cost, both in time, money, and in the integrity of the test where practical considerations invariably require sacrifices in testing quality. Meaning? Paper ballots can be easily tested,but LCD screen DREs take too long to test.

Because ballots are programmed specifically for each election, we must check if the voting machine counts correctly in each election. Sometimes, if done very methodically and with more than the required minimum number of ballots, the (LAT) test can find an underlying problem with a voting machine.

Most counties have the voting machine manufacturer program their ballot for them. Eagle County is one of the few in Colorado with access to the ballot programming software for our machines. Other counties are additionally depending on the manufacturer and in those counties there is even more reason to test with great scrutiny.

Eagle County's Logic and Accuracy Test ran to 7 days. It revealed that voters verifying the Voter Verified Paper Audit Trail would not find the judge names listed on the printed ballot and instead only a list of 5 identical races ("Court of Appeals") to be verified.

We Democrats (Republicans fail to participate in the test) requested a change to add the judge name to the title of each of the five races. After checking with John Gardner at the SOS office, our Clerk Teak Simonton took time to reprogram our DRE verifiable paper ballot format so that voters would recognize the Court of Appeals races when verifying on the paper ballot printout (V VPAT). Our Clerk , Teak Simonton is doing the right thing now whenever she is given permission to do so.

Now, we need a similar solution for all of the referenda and initiatives.

How will initiatives be verifiable by voters who can not remember when looking at the V-VPAT printout which letter and number goes with which initiative?

The identification of initiatives on the small paper records kept in the voting machine will need to be addressed in the future in the Secretary of State rules if not by statute. We must agree on a way to make an official short title to characterize each of these lengthy issues so that the paper record portion of our new LCD screen voting machines can actually be understood by humans.

A close examination of our voting system will reveal many improvements to be made. The public must hold our soon to be elected Secretary of State responsible to take the steps necessary to bring our systems and procedures up to par. I will strongly suggest that the newly elected officials in Denver including a new Secretary of State should take a look at what we have been doing in Eagle County to protect our election.

Harvie Branscomb
Canvass Board Member, Eagle County
Co-Chair, Eagle County Democratic Party

(the following list can be used as a guide to asking questions of officials in the right places)

Items which need to change to protect our Colorado elections in the future:

Improved LAT procedures- e.g. no use of "test mode", no use of “test” printed ballots;

Recommend improved post election audit- e.g. never recount during the audit, always hand count a selection of ballots which produces a real election tally subtotal;

Recommend improved hand count procedures to be recommended by SOS- e.g. sorting before counting method; (the cost and reliability of these different methods should be measured and contrasted);

Recommend examining the ½ percent recount rule… set the percentage to fit the quality of testing (probably greater than 1/2 percent threshold is required), require hand count for all recounts and specify procedures;

Recommend new rules for identifying all lines on V VPAT (voter verified paper audit trail) so voter verification can be done including a procedure for creating a 2-3 word description of all ballot initiatives to make this possible;

Recommend requiring voting machine manufacturers to strongly motivate actual V VPAT verification, by hiding the screen display at the time of verification of the paper record;

Revisit certification and equipment acceptance procedures to remove influence of manufacturers;

Provide clerks with motivation to support public interest ahead of their own personal interest in being re-elected;

Stop putting “restore voter confidence” as the top priority over substantive election reform… this produces more deception and poor or nonexistent testing and oversight;

Include the possibility of returning to all hand counting on paper as one of the possibilities in discussions of potential voting systems;

It is important to me to somehow address the filtering effect of media coverage of elections and election problems and officials' defensive and preventative reaction to it. We need to address the inherent need of Clerks and SOS to “put a happy face on elections”.

This argument becomes quite similar to the strategic argument of “electability” which is often used to supercede the importance of candidates positions on issues. There is a similar effect in the dispersion of information about election problems… election officials believe that the public must not hear bad things about elections overall, or the public will not participate. I do not have a solution to this problem, but we need to find one. A non-partisan election quality group would have the best chance of creating such a solution (such as a seal of conditional approval on voting procedures and equipment)… something like underwriters labs… to set a threshold of acceptability while listing the deficiencies and motivating efforts to correct them.

Not everywhere! In NY we have 20,000 lever machines ready to go and hand counted paper ballots for the disabled with accessible ballot markers.

FOREWARNED IS FOREARMED
10/28 Update 2006 Polling Analysis

Generic Polls
-------------
Data source:
http://www.pollingreport.com/2006.htm

THE DEMOCRATS WON EVERY ONE OF 110 GENERIC POLLS SINCE SEPT. 2005.

The following trend analysis is based on a 5-poll moving average. The margin of error (MoE) for the moving average is 1.41%, assuming a combined sample size of 5000 (1000 per poll)

The current Democratic moving average share is 57.8%
Undecided vote allocation (UVA) to Democrats: 60%.

Based on the MoE, UVA, the 57.8% share, and assuming ZERO FRAUD, there is a 97.5% probability that the Generic Democratic Vote will EXCEED 56.4%.

These are the probabilities that the Democratic Generic vote will exceed the given vote shares:

Vote: 55% 56% 57% 58% 59%
Prob: 100% 99% 87% 39% 5%

http://www.geocities.com...n2006_16921_image001.png
http://www.geocities.com...n2006_19660_image001.png

Poll Survey Dates DEM GOP Other Diff Dem5 GOP5 DemUVA GopUVA
Number Average All 49.3 38.2 12.5 11.0 49.3 38.3 56.8 43.2
2005
1 Newsweek RV. 905 50 38 12 12 50.0 38.0 57.2 42.8
2 Pew RV...... 911 52 40 8 12 49.3 39.0 56.8 43.2
3 DemCorp LV.. 921 48 39 13 9 50.0 39.0 55.8 44.2
4 Newsweek RV. 930 47 42 11 5 49.3 39.8 53.6 46.4
5 DemCorp LV.. 1010 46 41 14 5 48.6 40.0 54.4 45.6
_____
6 GWU LV...... 1012 47 41 13 6 48.0 40.6 54.8 45.2
7 Hotline RV.. 1016 40 31 29 9 45.6 38.8 57.4 42.6
8 DemCorp LV.. 1023 48 39 12 9 45.6 38.8 55.2 44.8
9 Gallup RV... 1023 50 43 7 7 46.2 39.0 54.2 45.8
10 ABC/WP RV... 1102 52 37 12 15 47.4 38.2 59.2 40.8
_____
11 DemCorp LV.. 1106 48 40 12 8 47.6 38.0 55.2 44.8
12 Newsweek RV. 1105 53 36 11 17 50.2 39.0 59.6 40.4
13 Hotline RV.. 1115 41 35 24 6 48.8 38.2 55.4 44.6
14 DemCorp LV.. 1120 48 41 11 7 48.4 37.8 54.6 45.4
15 Time RV.... 1201 48 36 15 12 47.6 37.6 57.0 43.0
_____
16 DemCorp LV.. 1204 49 39 12 10 47.8 37.4 56.2 43.8
17 CBS/NYT RV. 1206 42 33 25 9 45.6 36.8 57.0 43.0
18 DemCorp LV.. 1212 49 41 9 8 47.2 38.0 54.4 45.6
19 Hotline RV….. 1213 43 33 25 10 46.2 36.4 58.0 42.0
20 NPR LV........ 1218 45 37 17 8 45.6 36.6 55.2 44.8
_____
21 ABC/WP RV... 1218 51 41 9 10 46.0 37.0 56.4 43.6
22 Gallup RV... 108 49 43 8 6 47.4 39.0 53.8 46.2
23 CBS/NYT RV. 125 43 34 23 9 46.2 37.6 56.8 43.2
24 DemCorp LV.. 125 49 41 10 8 47.4 39.2 55.0 45.0
25 ABC/WP RV... 126 54 38 9 16 49.2 39.4 59.4 40.6
_____
26 Pew RV...... 205 50 41 9 9 49.0 39.4 55.4 44.6
27 Gallup RV... 212 50 43 8 7 49.2 39.4 54.8 45.2
28 GWU LV...... 215 46 41 14 5 49.8 40.8 54.4 45.6
29 Hotline RV.. 219 46 31 23 15 49.2 38.8 59.8 40.2
30 DemCorp LV.. 227 48 40 12 8 48.0 39.2 55.2 44.8
_____
31 Gallup RV... 301 53 39 7 14 48.6 38.8 57.2 42.8
32 FOX LV..... 301 48 34 18 14 48.2 37.0 58.8 41.2
33 Gallup RV... 312 55 39 7 16 50.0 36.6 59.2 40.8
34 NPR LV........ 314 52 37 11 15 51.2 37.8 58.6 41.4
35 Newsweek RV. 317 50 39 11 11 51.6 37.6 56.6 43.4
_____
36 Time RV.... 323 50 41 9 9 51.0 38.0 55.4 44.6
37 CBS RV...........409 44 34 22 10 50.2 38.0 57.2 42.8
38 ABC/WP RV... 409 55 40 5 15 50.2 38.2 58.0 42.0
39 Gallup RV... 409 52 42 6 10 50.2 39.2 55.6 44.4
40 Pew RV...... 416 51 41 8 10 50.4 39.6 55.8 44.2
_____
41 CNN RV ...... 423 50 40 9 10 50.4 39.4 55.4 44.6
42 Cook.......... 430 44 32 24 12 50.4 39.0 58.4 41.6
43 Gallup RV... 430 54 39 7 15 50.2 38.8 58.2 41.8
44 FOX LV..... 503 41 38 21 3 48.0 38.0 53.6 46.4
45 CNN RV ..... 507 52 38 10 14 48.2 37.4 58.0 42.0
_____
46 CBS/NYT RV. 508 44 33 23 11 47.0 36.0 57.8 42.2
47 Newsweek RV. 512 50 39 11 11 48.2 37.4 56.6 43.4
48 ABC/WP RV... 515 52 40 9 12 47.8 37.6 57.4 42.6
49 Fabrizio LV. 517 39 36 25 3 47.4 37.2 54.0 46.0
50 Hotline RV.. 521 42 36 22 6 45.4 36.8 55.2 44.8
_____
51 Gallup RV... 604 51 42 7 9 46.8 38.6 55.2 44.8
52 Gallup RV... 611 51 39 10 12 47.0 38.6 57.0 43.0
53 FOX LV..... 614 46 33 20 13 45.8 37.2 58.0 42.0
54 CNN RV ..... 615 45 38 16 7 47.0 37.6 54.6 45.4
55 Pew RV...... 619 51 39 10 12 48.8 38.2 57.0 43.0
_____
56 Hotline RV.. 625 41 36 24 5 46.8 37.0 55.4 44.6
57 ABC/WP RV... 625 52 39 9 13 47.0 37.0 57.4 42.6
58 Gallup RV... 625 54 38 7 16 48.6 38.0 58.2 41.8
59 TIME LV.... 629 47 35 18 12 49.0 37.4 57.8 42.2
60 Gallup RV... 709 51 41 9 10 49.0 37.8 56.4 43.6
_____
61 AP-Ipsos RV. 712 51 40 9 11 51.0 38.6 56.4 43.6
62 FOX LV..... 712 42 34 25 8 49.0 37.6 57.0 43.0
63 Hotline RV.. 723 48 32 20 16 47.8 36.4 60.0 40.0
64 CBS/NYT RV. 725 45 35 20 10 47.4 36.4 57.0 43.0
65 Gallup RV... 770 51 40 8 11 47.4 36.2 55.8 44.2
_____
66 CNN RV ..... 803 53 40 7 13 47.8 36.2 57.2 42.8
67 ABC/WP RV... 806 52 39 8 13 49.8 37.2 56.8 43.2
68 AP-Ipsos RV. 809 55 37 8 18 51.2 38.2 59.8 40.2
69 FOX LV..... 809 48 30 22 18 51.8 37.2 61.2 38.8
70 Gallup RV... 810 50 41 9 9 51.6 37.4 55.4 44.6
_____
71 Newsweek RV. 811 51 39 10 12 51.2 37.2 57.0 43.0
72 Pew RV...... 813 50 41 9 9 50.8 37.6 55.4 44.6
73 Hotline RV.. 820 40 33 27 7 47.8 36.8 56.2 43.8
74 Gallup RV... 820 47 45 7 2 47.6 39.8 51.2 48.8
75 CNN RV ..... 820 52 43 6 9 48.0 40.2 55.6 44.4
_____
76 CBS/NYT RV. 821 47 32 21 15 47.2 38.8 59.6 40.4
77 TIME LV.... 824 51 40 9 11 47.4 38.6 56.4 43.6
78 Newsweek RV. 825 50 38 12 12 49.4 39.6 57.2 42.8
79 FOX LV..... 830 48 32 21 16 49.6 37.0 60.6 39.4
80 CNN LV ..... 902 53 43 4 10 49.8 37.0 55.4 44.6
_____
81 ABC RV.......... 907 50 42 9 8 50.4 39.0 55.4 44.6
82 Pew RV...... 910 50 39 11 11 50.2 38.8 56.6 43.4
83 Gallup RV... 910 53 41 7 12 50.8 39.4 57.2 42.8
84 FOX LV..... 913 41 38 21 3 49.4 40.6 53.6 46.4
85 Gallup LV…….. 917 48 48 4 0 48.4 41.6 50.4 49.6
_____
86 CBS/NYT RV. 919 50 35 15 15 48.4 40.2 59.0 41.0
87 CNN LV ..... 924 55 42 3 13 49.4 40.8 56.8 43.2
88 FOX LV..... 927 49 38 14 11 48.6 40.2 57.4 42.6
89 Hotline RV.. 927 43 33 24 10 49.0 39.2 57.4 42.6
90 Zogby LV….. 928 42 33 25 9 47.8 36.2 57.0 43.0
_____
91 CNN LV ..... 1002 53 42 5 11 48.4 37.6 56.0 44.0
92 AP-Ipsos RV. 1004 51 38 11 13 47.6 36.8 57.6 42.4
93 Pew RV...... 1004 51 41 8 10 48.0 37.4 55.8 44.2
94 TIME LV.... 1005 54 39 7 15 50.2 38.6 58.2 41.8
95 Newsweek RV. 1006 51 39 7 12 52.0 39.8 55.2 44.8
_____
96 ABC RV....... 1008 54 41 5 13 52.2 39.6 57.0 43.0
97 CNN LV ..... 1008 58 37 5 21 53.6 39.4 61.0 39.0
98 Gallup LV…….. 1008 59 36 4 23 55.2 38.4 61.4 38.6
99 Harris LV…. 1009 49 36 15 13 54.2 37.8 58.0 42.0
100 FOX LV..... 1011 50 41 9 9 54.0 38.2 55.4 44.6
_____
101 CNN LV ..... 1015 56 40 4 16 54.4 38.0 58.4 41.6
102 NBC RV..........1016 52 37 11 15 53.2 38.0 58.6 41.4
103 Newsweek LV. 1021 55 37 8 18 52.4 38.2 59.8 40.2
104 Gallup LV.. 1023 54 41 5 13 53.4 39.2 57.0 43.0
105 ABC RV..........1023 54 41 5 13 54.2 39.2 57.0 43.0
_____
106 CNN LV ..... 1022 57 40 3 17 54.4 39.2 58.8 41.2
107 Hotline RV.. 1023 52 34 13 18 54.4 38.6 59.8 40.2
108 Zogby LV….. 1025 44 33 23 11 52.2 37.8 57.8 42.2
109 FOX LV..... 1025 49 38 13 11 51.2 37.2 56.8 43.2
110 Newsweek LV. 1027 53 39 8 14 51.0 36.8 57.8 42.2

_______________________________________________________________________

Senate Polling Analysis
-----------------------
Latest Update: Oct. 28
Current: GOP 55; Dems 44; Ind 1

The Dems need to win 6 of 8 GOP seats to gain control of the senate.

Data Source: Real Clear Politics (RCP)
Note that average poll shares are calculated by RCP based on the latest polls (usually 3-5).

http://www.realclearpoli....com/epolls/2006/senate/
Now, lets simplify things. This is not rocket science. The races to watch closely for fraud are VA, TN, MO. The Dems need to win TWO of them.

The probability is 78% that the Dems will WIN if they get 2/3 (67%) of the undecided vote (UVA), assuming ZERO FRAUD. But it's only 30% if the UVA is 60%.

Sensitivity of Democratic Win Probability to UVA%


UVA 50 55 60 67 75
N Win Probability
6 1 9 30 78 99
7 0 1 4 30 77
-----------------------------

MT Tester;Burns...... 50 44 48.3 44.3 52.7 47.3 100%
MO McCaskill;Talent.. 51 42 46.0 47.3 50.0 50.0 50%
NJ Menendez;Keane.... 44 40 43.8 41.8 52.4 47.6 100%
OH Brown;DeWine...... 54 42 51.5 39.8 56.7 43.3 100%
PA Casey;Santorum.... 52 39 50.0 39.8 56.1 43.9 100%
RI Whitehouse;Chafee. 51 42 47.3 40.5 54.6 45.4 100%
TN Ford;Coker.........48 46 45.5 47.8 49.5 50.5 18%
VA Webb;Allen.........47 49 46.0 47.5 49.9 50.1 43%

_________________________________________________________________

GOP House Polling Analysis
--------------------------

There are 60 GOP seats in jeopardy as of 10/24.

The Dems need 15 GOP seats (net) to gain control of the House. They currently lead in 31 congressional GOP districts.

Monte Carlo Simulation:
If they (conservatively) capture just 50% (UVA) of the undecided vote, there is a 100% probability the Democrats will win the House and gain 25 or more seats, assuming ZERO FRAUD.

If the Democrats capture 60% UVA, there is a 99% probability they will gain 30 or more seats, assuming ZERO FRAUD.

Sensitivity Analysis
-------------------

UVA 50 55 60 67 75
N Probability (Dems win at least N races)
25 100 100 100 100 100
30 39 86 99 100 100
35 0 4 39 96 85
40 0 0 0 15 70
45 0 0 0 0 3
--------------------------------------------------------------------
Assuming the Dems get 60% UVA, these are the probabilities they will win the following number of seats, assuming ZERO FRAUD.
------------
Dems win....25 30 31 32 33 34 35 36 37 38 39
Probability 100 99 96 91 77 59 38 21 9 3 1

http://www.geocities.com...n2006_22243_image001.png

Assuming the Democrats win the 10 GOP seats in which they lead beyond the margin of error, then the next 25 races in the following chart (FL-16 to PA-4) should be monitored closely for fraud.

http://www.geocities.com...n2006_17065_image001.png


DemU, GOPU are vote shares assuming UVA: 60% Dem/40% GOP.
Prob is the probability of a Democratic win based on UVA.

It's going to get hot. Very hot.
November elections will be the start of something much bigger.

www.wintersteel.com/RobertDuncanOFinioan.html

I'm educated, knowledgable, highly literate, learn fast and willing to work all day on election day in whatever capacity they needed me, pollworker or technician (I am very comfortable with technology). I filled out an application to work the election in Cleveland over a month ago. When I didn't hear back, I called, waited on hold for 45 minutes and finally talked to someone who said they had all the pollworkers and election-day technicians they needed. We'll see.