Convened by the American Association for the
Advancement of Science
Main | Participants
In September, 2004, our predecessor workshop posed a
long list of questions and research topics important to the future of
elections. There was a particular emphasis on the lack of adequate “baseline”
information on the election process as it existed at that time.
Since that time, the “baseline” has shifted radically.
New voting systems have been introduced across much of the nation, centralized,
statewide voter registration databases have been implemented in most states
and new voting procedures have been developed in response to new requirements
of law and new technologies.
I seem to recall noting (I think out loud) that the legislative
bodies around the country were not likely to wait for our research results
before forging ahead with new legislative “remedies” to perceived problems.
This has certainly proven to be the case, and in the process whole new
sets of problems have been generated.
Election administrators and
policy makers share a common objective. Both are primarily interested
in (and in need of) methods that will improve the process of casting and
counting ballots in the next election. We cannot, or will not,
wait for the research results to come in. (By “casting” I mean more than
the voting device. I include the whole vote enabling process.)
If researchers want to make a significant contribution
to improving the election process, they must target their research toward
providing sound practical responses to the most recently perceived problems
in the election process. (And, by
the way, if you have any engineering skills your value will increase exponentially.)
I do not want to imply that the proposed “baseline” research is not important
or that academic understanding of the election environment is of little
value. I consider both highly valuable and have spent most of the past
twenty years as an election administrator building my own knowledge of
that “baseline” especially for my jurisdiction. Recently, however, that
baseline has been a rapidly moving target.
Among the most pressing problems
created by the latest round of legislative initiatives around the country
relate to the “paper trail” requirement for DRE voting systems. Examples
of research needs suggested by our recent experience with these “voter
verifiable paper trails” include:
1. Voting
system reliability--By far the most frequent source of voting
machine “breakdown” with the iVotronic touchscreen voting machine with
its “Real Time Audit Log” (paper trail) was the printer itself. A preliminary
review of 92 reported voting machine problems leading to voting stoppages
on November 7, 2006, in Guilford County, revealed that 59 were directly
related to the paper trail printer. This incidence of voting machine
malfunction, whether due to mechanical or human causes, is unacceptable
high. (In reality the incidence was higher. We are finding that not
all printer failures were documented.)
How frequent and how widespread are such
printer related disfunctionalities on DRE systems with “paper trail” add
ons?
What alternatives exist to paper audit
or backup systems for DRE voting machines? How do the usability and reliability
of such systems compare to paper trail systems?
2. Auditability--Related
to the printer malfunction problem is a tabulation issue. Of the more
than 20 states that have required some form of “paper trail” associated
with DRE voting systems, presumably all have some form of audit requirement
and many have deemed the paper ballot the “controlling” ballot in the
event of a recount. Missing paper records, as described in item 1, appear
to substantially undermine the usefulness of such “paper trail” dependent
systems for audit or recount purposes. Auditability requires a much higher
degree of reliability than appears to be likely with paper audit systems.
3. Audit
and Recount Procedures--Manual recounts of paper “ballots” raise
a number of issues. Existing research on manual recounts is sparse and,
to date, has addressed only manual vs. machine recounts of paper based
voting systems (1) or hand recounts of previously hand counted ballots.(2)
(As of early 2005, these were the only two research efforts I could locate
actually attempting to measure the accuracy of vote tabulations.) The
widespread manual audit requirements recently implemented in many states
should provide a tremendous breadth and depth of readily accessible data
for analysis of manual vs. machine tabulation of a variety of voting systems.
Significant questions arising from North Carolina’s brief experience with
the manual audit include:
Where paper trails are required on DRE
voting machines, how should an audit sample be designed given that some
paper records are likely to be lost or destroyed?
Can any audit discrepancies between DRE
and manual paper trail tabulations be attributed to factors other than
missing paper records or errors in manual tabulation?
What factors account for discrepancies
between machine and manual tabulation of optical scan ballots?
State and local election jurisdictions
need prompt answers to these questions. If such issues are too pedestrian
to merit academic research efforts, then the administrators and politicians
will continue to devise “solutions” based on best guess or popular appeal
criterion.
The 2008 presidential election
is less than 2 years away and solutions that are not ready to implement
within the next six to ten months will not be implemented by November,
2008. After that, the landscape may well change again.
One final research topic:
How do you get policy makers to pay attention to substantive research
results?
1. Alvarez, R. Michael; Katz,
Jonathan N & Hill, Sarah A.; MACHINES VERSUS HUMANS: THE COUNTING
AND RECOUNTING OF PRE-SCORED PUNCHCARD BALLOTS; VTP WORKING PAPER #32,
September 2005
2. Ansolabehere, Stephen & Reeves, Andrew, “USING
RECOUNTS TO MEASURE THE ACCURACY OF VOTE TABULATIONS: Evidence from New
Hampshire Elections 1946-2002,; VTP Working paper #11, January 2004.
| 
