# [EM] STV ballot question

Rob Lanphier robla at eskimo.com
Mon Feb 19 13:15:25 PST 1996

```On 18 Feb 1996, Ed Still wrote:
> Let's say we are trying to use existing optical scanner voting machines for
> STV (Preference Voting).  This machine will not recognize characters like
> "1", "A", etc., but senses marks inside small circles or bubbles on the
> ballot.
>
> What we plan to do is list the candidates down the left side of the ballot
> and have columns for marking candidate-choices:
>
>           --------choices------------
> candidate 1st 2nd 3rd 4th 5th 6th 7th
> Anderson   ()  ()  ()  ()  ()  ()  ()
> Bush       ()  ()  ()  ()  ()  ()  ()
> Clinton    ()  ()  ()  ()  ()  ()  ()
> Dukakis    ()  ()  ()  ()  ()  ()  ()
> Eisenhower ()  ()  ()  ()  ()  ()  ()
> Forbes     ()  ()  ()  ()  ()  ()  ()
>
> Ideally, we would like to have the same number of columns as we have
> candidates, but there is a finite width and length to the ballot, so we may
> have to truncate the number of columns.

Here is a ballot format that is decidedly more complex, but offers the
flexibility you would need in elections with large numbers of
candidates.  This is based on the way most standardized tests (CAT, PSAT,
SAT, etc) ask for personal information (birthdate, etc).  It has the
advantage of being more manageable when there are a large number of
candidates (>30).  You write the rank underneath the candidate (purely for
verification purposes) and fill in the bubbles that correspond to the
numbers you assigned:

Anderson    Bush     Clinton   Dukakis Eisenhower  Forbes
Rank      Rank      Rank      Rank      Rank      Rank
|0||1|    |0||2|    |0||3|    |0||4|    |0||5|    |0||6|

( )(0)    ( )(0)    ( )(0)    ( )(0)    ( )(0)    ( )(0)
(1)( )    (1)(1)    (1)(1)    (1)(1)    (1)(1)    (1)(1)
(2)(2)    (2)( )    (2)(2)    (2)(2)    (2)(2)    (2)(2)
(3)(3)    (3)(3)    (3)( )    (3)(3)    (3)(3)    (3)(3)
(4)(4)    (4)(4)    (4)(4)    (4)( )    (4)(4)    (4)(4)
(5)(5)    (5)(5)    (5)(5)    (5)(5)    (5)( )    (5)(5)
(6)(6)    (6)(6)    (6)(6)    (6)(6)    (6)(6)    (6)( )
(7)(7)    (7)(7)    (7)(7)    (7)(7)    (7)(7)    (7)(7)
(8)(8)    (8)(8)    (8)(8)    (8)(8)    (8)(8)    (8)(8)
(9)(9)    (9)(9)    (9)(9)    (9)(9)    (9)(9)    (9)(9)

Thank God we have lotteries to get people used to the idea of filling in
little bubble forms.  Perhaps lottery forms will provide yet another model.

Ultimately, though, Lucien may be right in thinking a computer is
necessary.  With several manufacturers rumbling about the possibility of
cheap Java terminals, maybe we should work on a rewrite of Lucien's tool to
Java.  If it turns out that we do end up with cheap Java machines, then
great, we have a head start.  Regardless, we'll have a voter education

I'm working on a CGI front-end to my Perl script for tallying preference
ballots Condorcet-style.  In fact, I would have it installed for the
world to see and use, but the sysadmins here need to install the latest
version of Perl, first, and then need to review my script, all of which
could take a while.  If someone out there has the ability to install CGI
http://www.eskimo.com/~robla/politics/condorcet.cgi

It requires Perl 5.001 or later, and CGI.pm, available from
http://www.perl.com.

> When I met George Hallett in 1974, he told me that he had once voted in the
> New York City council election and something like his 21st choice helped
> elect a candidate.  His first choice never made the quota, but he remained an
> active candidate until Hallett's choices 2-20 had either been elected or
> eliminated.  So I realize there are some practical consequences to limiting
> the number of columns.
>
> Since (as someone said recently on the elections-reform list) we need to be
> the guys in the white lab coats, I need to be able to tell folks the real
> world consequences of the limitation of the choices to less than the number
> of candidates.
>
> Are there any empirical studies or theoretical ones that I can cite?

I would imagine this would be a difficult thing to study.  However, I'm
betting that a formula can be determined for a given flavor of STV,
number of candidates and level of truncation that will give you the
maximum error caused by the truncation.  What nasty math though.

Rob Lanphier
robla at eskimo.com
http://www.eskimo.com/~robla

```