[EM] Re: Voting Systems Study of the League of Women Voters of Minnesota

[Abd ulRahman Lomax]

At 03:38 PM 6/10/2005, Chris Benham wrote:
>Abd,
>You wrote:
>
>>Now, I wonder if there is anyone on this list who thinks that overvotes 
>>should not be counted?
>
>Let me first make it clear that I agree that FPP is awful and that 
>Approval would be a huge improvement. However, simply counting all 
>overvotes at full value is an attempt to sneak
>Approval in through the back door which is arguably not appropriate in a 
>civilised and democratic country. (Maybe not relevant.)

It actually is an attempt to note that, were it not for rules prohibiting 
overvotes, we would not even need to discuss the question of Approval, it 
would already be there, as it is in informal elections. As far as I know, 
the matter was never debated specifically (but it might have been, the 
world is vast). It would be quite interesting to know the history of the 
rules that prohibit overvotes.

I found this in a search:
http://www.sptimes.com/News/111201/news_pf/Lostvotes/Without_overvotes_Gor.shtml

Nowhere, so far, have I been able to find a cogent reason for discarding 
true overvotes. (The above URL points to a claimed disqualification of 
enough Gore ballots from technical overvotes that were *not* unclear as to 
voter intent, which was supposed to be the Florida rule, such as a voter 
writing down the name of the candidate *in addition* to punching the hole 
for the candidate, to throw the election to Bush in 2000.) Everything I've 
found simply assumes that overvotes are voter error and that *therefore* 
the votes quite properly are discarded. I haven't found a single 
justification for this.

I find that quite strange. I assume that if I continue to search, I'll find 
something somewhere.... but the question is notably invisible.

>While the method is FPP and not Approval, I think the fairest way to 
>handle overvotes is to count them at fractional value summing to one on 
>each ballot.

That, however, causes the split vote to be weakened. That is, by approving 
an *additional* candidate, the voter is lessening the chance that the 
preferred candidate will be elected, but, instead, some third candidate. I 
don't think that is fair at all. It is, however, superior to discarding the 
votes.

I think the crucial argument is that simply counting the overvotes 
correctly produces desired results in all pairwise considerations of the 
contest. In the simplest case, an approval of two candidates, the double 
vote only counts as one vote in each paired race with non-approved 
candidates, and it is equivalent to an abstention in effect for the pair of 
approved candidates.

>I put forward two arguments:
>(1) If  overvotes are regularly counted at full value, and this is widely 
>known, then overvoting will become a regular way of voting, and so if 
>votes are tampered with by having them
>vote for an extra candidate this could be done on a much bigger scale 
>without looking so suspicious.

I think I already answered this one. It's quite weak. If the criminals have 
sufficient access to the ballots to create overvotes on a "bigger scale," 
they can similarly alter the outcome by other means. As with most election 
fraud, serious manipulation can become visible through comparison of 
results with exit polls. Not a proof. But neither is a suspicious increase 
in overvotes.

>(2) What is the ballot instruction? If it is to check one candidate, then 
>its absurd that very many (maybe all) the voters have an incentive to do 
>otherwise.

The ballot instruction is a separate matter. However, a ballot instruction 
can be informative rather than prescriptive. It could merely note what the 
law prescribes for the election, i.e., the number of candidates to be 
elected. Or it could be kept the same. Or it could be modified to "Vote for 
one, or more if you approve of more than one." The point, for me, is to 
*allow* voters to overvote, not to tell them to.

>Also its unfair that voters who obey the ballot instruction (which I 
>assume is a "law") should be penalised in comparison with overvoters who 
>have the luxury of  fully voting both
>their favourites and compromises.

Yes, the ballot should contain correct instructions. If overvotes are 
allowed, the voters should be clearly informed.

>   That is like having a speeding limit, not enforcing it, and creating 
> incentives to speed.
>Another possible source of unfairness and absurdity is that some voting 
>equipment might not enable voters to overvote. That would mean that the 
>system is unofficial Approval
>for some voters, but enforced FPP for others.

That would truly be unfortunate. However, that's a specious argument, 
because election machines are designed to allow multi-candidate elections. 
In counting the votes, you simply set the machine for a number equal to the 
number of candidates, perhaps even that number plus one. It will then count 
all the votes, even if rejecting overvotes was hard-wired in. The totals 
can then be used to determine the winner.