[E-voting] Report on the 1% Manual Recount for Special Election
Nov. 2005 Los Angeles County
cansbro at eircom.net
Fri Jul 28 16:23:43 IST 2006
This is an amazing short paper. The small table at the end displays
properly if you go to the link. The implications are enormous--it shows
the weakness of relying on properly carried out audits, since the
results can be ignored, as happened here.
It's also interesting to note just how inaccurate the computer counts were.
Posted on Friday, July 28, 2006 - 07:42 am: Edit Post
View Post/Check IP
Move Post (Moderator/Admin Only)
Report by Judy Alter. Shows that computers sometimes add votes, and miss
votes, and that the difference between computer and hand count may
change the results. Also reveals that:
1) When there is a discrepancy they do NOT do a larger audit
2) When there is a discrepancy they do NOT notify the candidate
3) When there is a discrepancy they do NOT inform the citizenry.
Here is the report, also provided in an attachment at the end.
July 18, 2006
*REPORT ON THE 1% MANUAL RECOUNT FOR SPECIAL ELECTION*
NOV. 2005 LA COUNTY
By Judy Alter
This report analyzes the manual 1% recount for the Special Election of
Nov. 2005 in Los Angeles County. The central question I asked is: how
often does the computer count match the hand count? I found in the two
parts of the Special Election of Nov. 2005, the 8 initiatives and the129
local races, that the hand count and computer count matched an average
of 22%: 28 % of the time in the 496 initiatives (62 precincts times 8)
and 16% of the time in the 129 local races on that ballot.
*Added and subtracted votes*
After watching the actual recounting process in Norwalk after the
Primary Elec¬tion on June 6, 2006, I further refined my question about
what I can learn about the accuracy of the computer vote-counting. When
people examine a ballot for a mark that indicates a vote, people can
discern even faint marks. Thus, a logical result from the manual recount
will be that the people find more votes than when the computer counted
them. In the recount tallies of the initiatives, I found that the
computer missed a total of 436 yes votes, 1.3% of 33,863 ballots and 479
no votes, 1.4% of 33,863 ballots.
A more serious finding is when the 1% manual recount shows votes that
need to be subtracted from the computer totals. Somehow, in the
mechanical electronic counting process of the initiatives, the computer
added 110 yes votes, and 99 no votes. Although these numbers, 110 and
99, constitute a very small percentage of the votes recounted, the fact
that computers can add votes not written by a voter is very disturbing.
In the local races, the people recounting found 643 votes that the
computer did not count, and 354 votes that the computer added and, thus,
had to be subtract from the totals.
Changed winners and losers?
By looking only at the raw numbers, what I saw appeared to suggest
different winners and losers in a few local races. I asked Michael
Milroy to analyze the 1% manual recount results using tools found on
Excel. He found 4 contests (out of 129), like the one below, where even
the 1% manual recount found discrepancies that showed different winners.
That result should have triggered a larger manual recount of more
precincts in each contest (but apparently did not). The one percent
manual recount of precincts that is required by CA election law is too
small a number to produce a statistically meaningful audit of our
I asked Deborah Wright, information liaison for the LA County Registrar,
about what the election officials do with the findings of this 1% manual
recount. She said they sometimes recount some batches of ballots on
another scanner when the recount shows high discrepancies. They do not
notify the candidates nor do they publicize the findings of the recount.
I list, below, the serious discrepancies from the 129 local elections
for school board, community college board, or water district board, in
November 2005 election.
-In three contests between 9 to 11 more votes were found by the hand
count for an individual candidate;
-in 3 contests between 12 to 15 more votes were found by the hand count
for an individual candidate;
in two contests between 15 to 20 more votes were found by the hand count
for an individual candidate;
and in one contest 23 more votes were found by the hand count for two
Said another way, in only 16% of the contests people found no
differences between the computer count and the hand recount. The hand
count seems repeatedly to show that the automatic scanner/computer
counts produce highly inaccurate results.
In his analysis Michael Milroy addressed this question: do the 1% hand
tally results suggest different winners than do the final election
returns? (based on the difference between the 1% hand tally votes and
the 1% computer-count votes).
Computer Count Tally
R J Buonocore 110 110
Paul Helzer 111 113
L. Sanchez-Ramirez 131 131
Sonny Santa Ines 76 78
Bill Ste Marie 109 111
The winners in this contest suggested by the 1% tally hand count were
Sanchez-Ramirez, Helzer and Ste. Marie.
The winners suggested by the computer count were
Buonocore, Helzer and Sanchez-Ramirez.
The actual winners, announced in the final returns, were Buonocore,
Helzer and Sanchez-Ramirez.
I focus on the difference between the 1% hand tally and the 1% computer
count because I don’t think that the 1% hand tally count has much
predictive value as to who would have won the final results (it’s a
small sample, and it’s not representative of the entire jurisdiction.
To rectify this situation Michael suggests: “But I agree with you that
the differences between the 1% hand tally votes and 1% computer count
votes still warrant further investigation, and a higher percentage
recount. (And by the way, that’s why I included the number of precincts
in each contest.) In a contest that includes 15 precincts, a 25% hand
recount would have involved 4 precincts).”
Michael found three other contests where the 1% hand recount showed
different winners than the computer count. While the effort to manually
recount 1% of LA County’s precincts is worthwhile and carried out with
great care, the election officials do not use the results to examine the
contests further nor do they publish the results for public scrutiny. I
only could study and report on these results because I requested them.
Director of Study California Ballots (Protect CA Ballots)
application/mswordLos Angeles 1% manual count - report (36.4 k)
More information about the E-voting