Re: Election in Boulder County nullified on constitutional grounds?

[Ralph Shnelvar <ralphs@xxxxxxxxx>]

Mr. Tiger:

On Mon, 9 Aug 2004 18:10:02 -0600, you wrote:

>There's nothing about the barcode on the ballot that identifies the voter.

Wrong.

>
>"He then went on with a bunch of irrelevancies."
>
>My 'irrelevancies' point to the issues of voter and ballot fraud. I might
>assume from your insistence that ballot not be uniquely identified that you
>would like to encourage ballot fraud. The barcode identifies the ballot, not
>the voter.

You can serialize the ballot inside the computer after it is cast.

Or the ballot could be printed with a holographic image to make it difficult
to duplicate ... or any of a number of other ways to may it difficult to
stuff ballots into a ballot box.

>
>Please describe a method to all that have been addressed here how this
>barcode can be used to identify the voter that cast the ballot. Inquiring
>minds want to know.

God, Paul, if you'd think for a couple of minutes then maybe you'd not make
a fool of yourself.  I provide the answer, below.  But I suggest you think
about what you wrote.  I'm quoting you:

> [|>] The bar code on the sides of the ballots identify the ballot  
> style, not the individual ballot. Ballot style is about what issues a  
> person can vote on in a particular geographical location. My wife and  
> I would have the same ballot style in a general election, but  
> different in a primary. We live in the same house, but we are members  
> of different parties. I get a Libertarian ballot, she gets a  
> Democratic ballot. If we were members of the same party the barcodes  
> would be identical.

And, by the way, at least in my house my wife is a registered Libertarian.

>
>Paul Tiger


I hope you gave it some thought, Paul.

Here's the answer.  It's called "vote selling".

Part of the purpose of a secret ballot is to make it difficult or impossible
to do vote selling.

In vote selling what happens is that you sell your vote to someone for money
(duh!).

The problem is "How does the buyer know that the seller has kept his
promise"?  It is a multi-step process.

Step 0: A valid ballot is mailed from the County Clerk to the proper voter.

Step 1: Voter X scans/faxes the ballot to seller.

Step 2: Seller Y stores the ballot image on his computer.

Step 3: Voter X votes.

Step 4: Voter sends in ballot.

Step 5: Ballot with identification is scanned it at the County.

Step 6: All scanned ballots are stored in County computer.

Step 7: Let's say that there are 100,000 votes cast.  Each vote is stored as
a .jpg  and is, say,  600K (two 300K b&w images) in size.  That's a total of
60 gigs of data.  Today that 60 gig drive can be bought for a lot less than
$100 and easily fits in a coat pocket.

Step 8: Unscrupulous out-sourced temporary employee Z take drive with
scanned images to Seller Y.  Seller Y now scans his 60 gigs of data at 25
megabytes/sec for a total time to match a particular vote seller in 40
minutes.  It wouldn't take much more than 40 minutes to match a couple of
thousand purchased/sold votes against that database.  Looking for barcodes
at fixed positions in scanned images is fast ... as anyone who has purchased
a bottle of milk at the supermarket knows.

Note that all of this matching becomes (nearly) impossible if there is no
serialization (barcode) number.



So let me put some objections into Paul's mouth.



(1) No one buys or sells votes.

The history of voting is full of vote buying and selling.  One of the
objectives of a secret ballot is to minimize this possibility.

See, for instance, http://slate.msn.com/id/91418/ in which 1,131 Illinois
residents put up there votes for sale in 2000.


(2) You've got to have a dishonest employee to copy an entire database.
Security at the County Clerk's is so tight ...

If entire computer disks can be stolen out of Sandia National Labs where
national security concerns are so high ...



(3) Hey! This only applies to absentee ballots!

What proportion of ballots need to be cast absentee in order to swing an
election?  It isn't much and there is a bigger and bigger push to have
mail-in ballots.



Can I identify Paul Tiger's vote just given a barcode?  I don't know.  Part
of my problem is that I don't know the internals of the computer system that
is counting the vote.

Assume that Paul votes at a polling place.  I do know that somewhere a
computer spit out a ballot and barcode for a Libertarian Party ticket that
Paul Tiger eventually filled out and was eventually scanned in.

I don't know if the Hart machinery time-stamps the scanned-in image.  (How
would I find out?  Ask the vendor?  Trust the vendor?)  Thus it wouldn't be
too hard to guess that a particular ballot matched with a particular voter
just based on the intersection of time, district, and (third) party
affiliation.  There are seven-hundred-or-so Greens in Boulder County.  If
they all voted in person I bet you I could make a really good guess as to
which ballot went with which voter.

But I'm only guessing.



Thirty years ago I made my living as a research assistant doing econometrics
on a variety of projects.

There was a famous (and now forgotten) story back then about the "bionic
data set".

There was a TV show that I never watched called "The Six Million Dollar
Man". In the show he was known as "The Bionic Man."

Anyway, there was this longitudinal and time-seies dataset of thousands of
people. The dataset was so constructed that the anonymity of the
participants was supposedly guaranteed.

For reasons I don't remember, they desperately needed to add another year of
data to the time series.  But the dataset was something like the census
tract data in which it is supposedly impossible to get individual census
data from the summary data.

Well, they spent an additional $6,000,000 (hence the "bionic dataset") going
through the data looking for clues about where people lived and how many
children they had, their general age, etc.

They eventually managed to reconstruct the underlying respondents even
though this was in breach of contract.

But, hey, this was a government agency so contracts with individuals really
didn't mean much because there was a government need.

Anyway, this lesson proved to me that confidentiality can often be broken by
effort.  In this case it was just a bunch of government bureaucrats
interested in some time-series data.

Imagine what resources unscrupulous people would throw into an important
election.



The bar-coding of ballots is not just a theoretical threat.  It's a real
encouragement of vote buying/selling.