Subject: RE: California Turnout Update and Undervote Analysis
From: "Steven Hertzberg" <steven@votewatch.us>
Date: 10/10/2003, 7:57 PM
To: "'Larry Levine'" <larrylevine@earthlink.net>, mmcdon@gmu.edu, election-law@majordomo.lls.edu, election@csd.uwm.edu

Larry,

I am not sure I completely understand your question, but I will attempt
to answer it to the best of my ability.

The undervote takes into consideration a cast ballot that has at least
one unanswered question.  It could include any one of the four
questions, or combination of the questions.  Each unanswered question is
counted as an undervote.

Votewatch placed poll monitors inside a polling location in Santa Clara
county.  Santa Clara and Los Angeles counties utilized the same voting
equipment and ballot design for this election.  We saw, before Tuesday
morning was over, that there was an issue with voters who were trying to
answer Question 1 using this ballot type.  For example, our Santa Clara
poll monitor regularly witnessed poll workers announcing, in a loud
voice, their answer to the question "how do I vote for Gray Davis?"
every time it was asked by a voter.

We also saw, from our own survey work, that approximately 15% of voters
realized that they left a question blank, either unintentionally or
intentionally. 

Furthermore, we also received anecdotal evidence that voters were
missing Question 1, with several comments from voters on Election Day.

Even with the above, we saw from our online survey that 76% of voters
were either very confident or confident that their vote will be counted
correctly.  This number jumped to 91% in our surveys administered at
polling locations on Election Day.


________________________________
Steven Hertzberg  
Votewatch, a nonprofit Corporation
San Francisco, California 94123

http://www.votewatch.us
Your Eye on Elections  



-----Original Message-----
From: owner-election-law_gl@majordomo.lls.edu
[mailto:owner-election-law_gl@majordomo.lls.edu] On Behalf Of Larry
Levine
Sent: Friday, October 10, 2003 4:19 PM
To: mmcdon@gmu.edu; election-law@majordomo.lls.edu; election@csd.uwm.edu
Subject: Re: California Turnout Update and Undervote Analysis


Does the "undervote" on the replacement matter take into consideration
the possibility that some voters may have made a decision to vote
(probably
against) on the recall and not vote for a replacement candidate? Also,
is there, or will there be, any demographic data developed (particularly
level of education) to differentiate between the various voting systems?
Larry Levine



----- Original Message ----- 
From: "Michael McDonald" <mmcdon@gmu.edu>
To: <election-law@majordomo.lls.edu>; <election@csd.uwm.edu>
Sent: Friday, October 10, 2003 4:35 PM
Subject: California Turnout Update and Undervote Analysis


The California Secretary of State now reports 8,374,804 ballots have 
been cast, for a turnout rate among eligible voters of 39.6%.  The 
final
turnout
rate will probably creep up in the coming days as the remaining 
absentee, provisional, and write-in votes are tallied.  From talking 
with my friends on the ground in California, I am expecting a final 
tally somewhere in the neighborhood of 8.5 million, or a turnout rate 
among eligible voters of approximately 40.5%.  This turnout rate will 
be higher than the 37.2% rate in 2002 and lower than the 45.4% rate in

1998.  It is an unusual special election, but is not a record vote for

a governor's election.

As for the undervotes (those persons casting a ballot, but not voting 
for
an
item on the ballot) and voting systems, using the current numbers
available
at the Secretary of State's web site, I have calculated the following 
undervoting rates:

Punchcards
Recall undervotes: 7.1%
Replacement undervotes: 9.0%

Optical scan
Recall undervotes: 5.0%
Replacement undervotes: 9.4%

Touchscreen
Recall undervotes: 1.4%
Replacement undervotes: 6.7%

Exit Poll (statewide)
Recall undervotes: 2.6%
Replacement undervotes: 7.0%

One other bit of information of interest:

Vote "Yes" Recall
Punchcards 50.0%
Optical scan 55.0%
Touchscreen 50.8%

Statewide, there were 7,989,828 ballots cast on question 1, the 
recall,
for
a turnout rate of 37.8%.  This was 95.4% of all ballots cast, for an 
undervote of 384,976.  If we assume the exit poll is correct, then 
4.6% minus 2.6% or 2.0% of the persons who cast ballots intended to 
vote on the recall, but their vote was not recorded.  The number of 
undervote errors would therefore be approximately 160,000 (a little 
less than the 178,170 vote differential between the "no on recall" and

vote for Schwarzenegger). Interestingly, we see that even though 
voters who voted by touchscreen and punchcard were very similar in 
their preference for recalling Davis, 1.4%
of
touchscreen voters did not vote in the recall while 7.1% of punchcard 
(and 5.0% of optical scan) voters did not.

On question 2, the replacement, 7,698,407 persons voted for a turnout 
rate of 36.4%.  This was 91.9% of all ballots cast, for an undervote 
of
676,397.
Again, if we assume the exit poll was correct, 8.1% minus 7.0% or 1.9%

of the persons who cast ballots intended to vote for a replacement 
candidate, but their vote was not recorded.  The number of undervote 
errors would be approximately, 150,000.  We see the same pattern as 
with the recall
question: 6.7% of touchscreen voters did not vote for a replacement
candidate, while 9.0% of punchcard and 9.4% of optical scan voters did
not.

In both cases, touchscreen voters had smaller undervote rates than the

punchcard and optical scan voters, and also had a rate smaller than 
the
exit
poll indicated.  Either of two conclusions may be reached: touchscreen

recorded fewer errors, or more people were prompted to vote by having 
a computer screen tell them to do so.  The difference of undervote 
rates between systems provides evidence for the first conclusion, 
while the difference between the exit poll and the touchscreen 
undervote rate
provides
evidence for the second conclusion (though sampling error on the exit 
poll may account for this evidence, too).

If the prompting of the touchscreens induced more touchscreen voters 
to completely fill out the ballot, this may have some profound 
implications
for
future elections.  If voters feel obliged to complete the touchscreen 
prompts, I image that they will be more likely to vote for downballot
races,
and thus there will be less rolloff.  (The proposition voting is a 
clear test of the hypothesis.)  If voters do behave in this manner, it

will be curious to see what information will govern their decision 
making process...will they use partisanship as a guide when it is 
available?

Finally, as a fun aside consider...

Schwartzman vote (as percent of question 2):
Punchcards 0.17%
Optical scan 0.12%
Touchscreen 0.11%

I will post the data used to make these calculations on my website 
later this evening (http://elections.gmu.edu/voter_turnout.htm).

==================================
Dr. Michael P. McDonald
Assistant Professor
Dept of Public and International Affairs
George Mason University
4400 University Drive - 3F4
Fairfax, VA 22030-4444

Office: 703-993-4191
Fax: 703-993-1399
Efax: 561-431-3190

mmcdon@gmu.edu
http://elections.gmu.edu/