Re: Questions from election officials

From: Douglas W. Jones <jones_at_cs_dot_uiowa_dot_edu>
Date: Mon May 03 2004 - 13:37:46 CDT

On May 3, 2004, at 1:07 PM, charlie strauss wrote:

> 1) records retention.
> One of the reasons states currently using DREs are averse to returning
> to paper based systems is that new standards make paper more of a
> hassle than it once was. In particular places like Los Angeles would
> have to retain millions of paper ballots for 22 months. They did not
> have this issue to contend with in the past.

The 22 month record retention requirement has been in place for a while.
It applies to any election for federal office. Indeed, it's a hassle.
However, a sound economic analysis of the lifecycle cost of DRE versus
paper versus VVPT has yet to be done, and there are all kinds of
unwarranted assumptions all over the place.

> 2) paper quality.
> If non-card stock is used then recounts become problematic. The claim
> is that paper ballots on ordinary paper jam more frequently on each
> pass through a recounting scan. In new mexico substandard paper was
> a contributing factor in a problem that was only resolved when 60,000
> ballots were literally hand tallied.

I believe this. I've seen paper ballots on every kind of paper from
stock that was barely better than tissue paper to semi-rigid cardstock.
Clearly, you want ballot stock that can stand up to handling. The
VVPT systems that use cash-register receipt tape are going to be
real problems because that kind of paper handles very badly.

And, of course, the OVC model, since we're designing in terms of
commodity printers, can be counted with commodity auto-feed scanners,
but they'll certainly work better if the jurisdiction uses decent
paper and doesn't simply go out and buy whatever Office Max happens
to have on sale.

> 3) recounting
> if OVC's answer to the above issue is simply that counts are done by
> hand anyhow (currently with a wand), this requires a lot of people:
> that'll work on election day when the election judges are plentiful
> but then how are recounts to be done after election day?

State law is, of course, the guide here. In many states, machine
countable paper ballots may be recounted by hand or by machine,
at the discretion of the commissioner of elections. If the recount
was called because of an allegation that the machines miscounted,
a hand count is appropriate. If the recount was called for some
other reason, for example, because of an allegation that the polls
were improperly closed at some precincts, resulting in a failure to
count all the ballots from that precinct, then a machine recount
makes good sense.

In any case, California's law mandating a hand recount of paper
ballots from precincts representing 1% of the voters after every
election is good policy! This law guarantees that, when there is
a call for a hand recount, there will be a corps of workers with
experience conducting such counts, and it provides a useful audit
on the integrity of the vote tabulating software.

> 4) ballot size
> this has already been a point of discussion on OVC recently, but she
> wanted to know how her typical 70 position ballot would be come out on
> paper and if it would be readable.

The OVC model is aiming us at standard office-paper sized ballots.
This gives us a page-size comparable to that used on standard
optical mark-sense ballots, and lets us try to be at least as
readable as such ballots.

The OVC model is not aiming to print on cash-register tape, as many
VVPT vendors propose to do. Ted Selker has produced some mockups
of what a VVPT ballot would look like in cash-register tape format!
They're readable, but a general election comes out to be over a
yard long.

> 5) precint/county level manipulation
> paper is more susceptible to manipulation by the same low tech means
> they have problems with now.

California election law requires ballot reconciliation as part of
the canvassing that will catch this, whether it involves DRE systems,
lever voting machines, or hand-counted paper. I recommend this
auditing model. The canvass requires reconciliation of the number
of signatures in the pollbook with the number of ballots counted, with
a serious attempt to explain every discrepancy.

> 6) machine sealing
> Apparently machines get sealed for some period before and after
> elections. What parts of the OVC machine would need to be sealed?
> Just the live CD? The ballot image data output? It turns out that
> machine sealing is a major hassle because it means machines cannot be
> used for multiple purposes during an election period. FOr example you
> have to have different machines to count absentee ballots because the
> precint level ones are sealed during the same election period.

We need to do a security analysis of the system to determine what needs
sealing when. Also, some states don't have sane laws in this arena,
so even if the machine doesn't need sealing, they may require it anyway.

The absentee ballot problem is different, because in many states,
it is illegal to begin tabulation of absentee ballots until the polls
close. This restriction is there because if you could tabulate the
absentee ballots in advance, any leak of that tabulation could be
of incredible value to the recipient. So, absentee ballot tabulation
will be required to be done on different machines from the ones in the
polling place.

                Doug Jones
                jones@cs.uiowa.edu
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Mon May 31 23:17:04 2004

This archive was generated by hypermail 2.1.8 : Mon May 31 2004 - 23:18:15 CDT