[EM] the simplest election reform

[Abd ul-Rahman Lomax]

At 08:25 AM 6/16/2005, Andrew Myers wrote:
> > Further, in a few days, so far, of searching, there seems to be a distinct
> > lack of cogent arguments for the rule in the first place. As near as I can
> > tell, the reason for it is a variant on "It seemed like the thing to do at
> > the time." The rule is clearly very old.
>
>I can think of one important reason. The rule ensures that if there are N 
>valid ballots, summing the number of votes for all candidates yields N. 
>This is a simple check on the integrity of the election process.

This argument has been raised by others, and I claim that, on examination, 
it is not cogent, for the following reasons:

(1) Because there are abstentions, i.e., undervotes, simple vote totals do 
not check for fraud. To check for fraud, more sophisticated means are 
necessary. Among other things, overvotes and undervotes should *always* be 
reported, as well as polled in exit polls, whether they are discarded or 
not. Significant variation between exit polls (cautious ones, conducted in 
an impartial way to the extent possible) and reported results, coupled with 
a suspicious pattern of overvotes (as, for example, overvotes for 
candidates widely separated in views, i.e. Buchanan/Gore in Florida 2000), 
would indicate either fraud or systematic error.

Indeed, consider this scenario: the criminal defacer of ballots wants Bush 
to win in Florida 2000, but does not want to leave an obvious trace. 
Noticing that there is a butterfly ballot which could be expected to 
produce an increase in overvotes for Buchanan and Gore, the criminal 
punches out the Buchanan holes on a number of Gore ballots. (No accusation 
by me against the Bush campaign should be taken from this example.) So:

(2) Discarding the overvotes may accomplish the goal of the one who has 
altered the ballot. Leaving them counted partially reduces the harm over 
discarding them; this is not difficult to show.

(3) There are better ways to detect and guard against fraud. For starters, 
protecting the ballots from access under any conditions where there are 
insufficient observers to guarantee the absence of fraud is necessary, and 
I think this is done, including the use of seals on containers of ballots. 
Which does not mean that fraudulent alteration of ballots does not occur, 
just that it can and should be kept to a minimum.

(4) The most common cause of overvotes is voter error, not fraudulent 
overvote. In creating the overvote, the voter is attempting to correct the 
error. If overvotes are not discarded, the effect is that the voter has 
abstained from voting in the pairwise contest between the two marked 
candidates, and yet the voter's desired vote does count in the pairwise 
contests between those candidates and all other candidates. Often this 
extra undesired vote would be harmless (as with Buchanan in Florida 2000), 
and allowing the overvote preserves the true intention of the voter.

It should be realized that discarding overvotes is not necessarily 
intuitively obvious. I think that a voter could go for a long time, being 
reasonably aware of public affairs, and not know that the ballot will not 
be considered if it is overvoted. Rather, the voter might well think that 
the ballot will be counted as a vote for both. As I've noted elsewhere, 
this is what happens, almost certainly, when show-of-hands votes are taken 
in open meetings. It is what a voter is very likely to do in such an open 
meeting if they change their mind or realize that they erred in voting for 
an earlier candidate in the announced list. They partially undo the damage 
of the error by voting again. Because some people generally abstain, these 
open overvotes are not very detectable, unless someone is specifically 
looking for it. Even then it might be missed.

If a voter considers that the original vote were truly harmful, the voter 
can cancel all votes by voting for all candidates. This is exactly 
equivalent to abstention.

But why doesn't the voter just go get another ballot? It's their right! 
Unfortunately, the reality is that not only are some voters not aware of 
this, but they may not have time. Suppose a single mother is rushing to 
pick up her child at day-care, and stops to vote. It's a serious problem if 
she is late.... She may well decide to minimize the damage by overvoting 
instead of going back and getting another ballot.

And then there would be deliberate overvotes where a voter is actually 
doing Approval voting, in the mistaken belief that the votes will count.

IN SUMMARY, the reason given, that overvotes should be disallowed because 
it is possible to check the "integrity of the election process" by summing 
the votes for all candidates and comparing it with the total valid ballots, 
is technically true, but what would be detected in this case is only simple 
counting or math error. Counting errors that cause a ballot to be rejected 
as overvoted or having no valid vote are not detected. Fraud remains quite 
possible and is not detected.

Any cogent counterarguments (i.e., refuting or claiming to refute what has 
been written about the validation argument)? Or other arguments against 
allowing overvotes?

It would also be interested to find if there is any history of the rule. I 
couldn't find a discussion of it on the net. I did find at least one state 
statute stating the rule, and presumably there would be legislative history 
somewhere. But I think the rule is very old and there might not be much....