[EM] Hand counting election methods

[Gervase Lam]

> Date: Sun, 16 Nov 2003 20:59:39 +0100
> From: David GLAUDE <dglaude at gmx.net>
> Subject: Re: [EM] Electronic Voting Bill of Rights?

> I would like to know wich one are possible to handle with manual
> counting on large scale.

I am sure there are lots that can be hand counted.  It depends on how much 
time and effort you are willing to expend.  This leads me to some fleeting 
thoughts that I have had for the past several months.

Many methods have been discussed here with the main examples being IRV, 
Approval and Condorcet.  However, what are the most optimal and error-free 
ways of hand counting each of them?  What about other methods like Borda 
or MCA?

I'll start off with Approval as finding a good way to hand count Condorcet 
makes my brain hurt!  I may get on to IRV.  I don't know whether the "IRV 
is summable (sort of)" thread may help or not.

Suppose there are candidates A, B, C, D, etc... in an Approval vote.  The 
most obvious way to count them is to count the number of ballots for and 
against A.  The count is then redone for candidate B, and so on.

This is the slowest way.  Nevertheless, I think this is the most error 
free way.

An idea I thought of for counting Approval votes involves using a 2x2 grid 
on a piece of paper.  The columns of the grid are labeled "Yes" and "No" 
respectively for candidate A.  The rows of the grid are labeled "Yes" and 
"No" respectively for candidate B.  The ballots are put into the 
appropriate quadrant of the grid.  Repeat the process but with candidates 
C and D etc...

It is probably best to count the candidates in order of appearance on the 
ballot.  This enables the counters to see easily what has been voted 
without having to look up and down the ballot for the correct pair of 
candidates.

The idea could be extended by having two 2x2 grids in order to simulate a 
2x2x2 grid.  The top 2x2 grid is for the ballots who have voted Yes for 
candidate A while the bottom 2x2 grid is for the ballots who have voted No 
for candidate A.  The Yes/No columns/rows on each of the grids are for 
candidates B and C.

The 2x2x2 grid can of course be extended yet again to a 2x2x2x2 grid by 
using two sheets of paper.  The first sheet is for those who voted Yes for 
candidate A while the other is for those who voted No for candidate A.  
The top grid on each page is for those who voted Yes for candidate B while 
the bottom grid is for those who voted No for candidate B.  The Yes/No 
columns/rows on each of the grids are for candidates C and D.

I think the way the grids are physically laid out, with one above the 
other and on different pages, is very important as they act as 
visual/psychological cues for the counters.  Therefore, I don't think you 
can go practically any further than a 2x2x2x2 grid.

Thanks,
Gervase.