[EM] STV ballot question

[Rob Lanphier]

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 
tool, since Java applets can be seamlessly downloaded today.

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 
Perl 5.001 scripts, you can get the latest version at:
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