[EM] PR in student government

[Abd ul-Rahman Lomax]

At 01:14 PM 4/17/2007, James Gilmour wrote:
>You may be interested to know that in our elections on 3 May, electors
>will complete conventional ballot papers (real paper) with an old
>fashioned stubby pencil (chained to the polling booth) or vote by post.
>The ballot papers for the Local Government STV elections (and for the
>Scottish Parliament elections that are being held on the same day  -
>MMP voting system!!) will be scanned and data files created by
>intelligent optical character recognition software.  The ballot data
>will then be counted electronically.

Yes. The insanity of using "voting machines" here, given that 
equipment for scanning paper ballots has existed since I was a child, 
is ... manifest. I remember using a number two pencil to fill in 
ovals on paper, and these were machine-counted, for standardized 
tests. Before 1960.

Now, any fax machine could be used as the scanner, and any computer 
could analyze that data.

>So the full ballot data will be readily available in an electronic
>format and COULD be published in a suitably anonymised form, i.e.
>removing all references to the ballot paper number from the file of
>ballot data.

Actually, the ballot paper number does not identify the voter, or it 
shouldn't. However, if there is some way that it *would* identify the 
voter, some other means of serializing the ballots should be used, so 
that people can independently count them and compare results ballot-by-ballot.

(Actually, of course, they would produce lists of ballots with the 
votes counted for each identified ballot, and these lists can then be 
compared for discrepancies. It's an old technique used for keypunched 
data, to have two people keypunch it and then compare the results, 
which should match precisely. If they don't, then one looks at the 
discrepancies. When Florida 2000 was on the TV here, I was screaming 
at the box, because it was allegedly difficult to count ballots, 
every ballot had to be examined carefully, etc. Not. The care is only 
necessary where independent counters don't come up with the same 
totals for a batch, and the batch size is kept small enough that the 
majority of batches are in agreement. And if a counter is biased, it 
will show up quickly!)

>   With thousands of voters in every one of the 353 elections
>and only small numbers of candidates in each of the 353 wards (local
>government electoral district), the chances of identifying any voter
>from a "unique" sequence of preferences are probably so small that they
>can be safely set aside in the greater public interest of making the
>full ballot data available at ward level.  This was done with the full
>STV-PR ballot data for three constituencies in Ireland in 2002.

I agree. On the Range Voting list, where we consider all kinds of 
election reforms besides Range Voting, I've written extensively about 
the publication of ballot scans. It is indeed interesting to know 
that it is being done, and I'd consider the data extremely valuable 
for more than one reason. Understanding voting patterns can reveal 
far more about the preferences of the public than simple aggregated 
totals. And where election fraud is suspected, allowing the public to 
validate the counts is actually crucial. It's devastating here that 
we have, too often, a lack of confidence in our election processes.

>Unfortunately, our government (= "Scottish Executive") has followed the
>conventional approach to paper records relating to elections and so the
>Election Rules specifically prohibit the release or publication of any
>of the electronic information although all the electronic information
>has to be retained for four years (until the next election).  However,
>following some agitation on this issue, the government has agreed to
>carry out a consultation on this after the elections are over.  So we
>may yet see the full STV ballot data from all our 2007 local government
>elections.  Then we might have independent validation of the results and
>lots of interesting (and amusing) political and sociological analysis of
>the preference patterns.

It also becomes possible, then, to study the effects of reforming 
election methods. There really are many possible benefits to the data 
being made public, and the down side is very small. It does not add 
costs; indeed, it might reduce them. (It's expensive to challenge an 
election! If the ballot images could be seen, any challenger would 
know whether or not they have a basis, and a full formal recount 
isn't necessary at all. A court would require a challenger to present 
*specific* evidence of error. Now, necessity is the mother of 
invention, and stubborn desire is the mother of necessity, so there 
might still be improper challenges, but it would become easier to 
resolve them if they involve the actual counting.)

There were very good reasons to sequester paper records, the original 
ballots, and those reasons remain. But images of the ballots are not 
subject to the same risks. Indeed, they become safer if released, 
because they no longer exist in only one or a few places, under 
unified control.

I've argued that if ballots are properly imaged, it would normally be 
the case that, even for manual counting (i.e, not OCR assisted) 
ballots, it would be the images which are counted, not the original 
ballots. One of the avenues for election fraud is ballot alteration 
during the counting process. Imaging would be much faster and could 
be more closely supervised, with the actual ballots then being placed 
under seal.

(It's also true that imaging improves ballot security from the other 
direction as well: if the original ballots are lost, the images 
generally would not be. Obviously, a loss of original ballots would 
be a serious matter, but accidents do happen, armored cars crash and 
burn, and so do buildings.)