Example with half-votes

Hugh R. Tobin htobin at ccom.net
Sat Jul 13 13:17:34 PDT 1996

Mike Ossipoff wrote:
> > >
> > > Say we were using Condorcet's method, with the added provision,
> > > recently proposed, that if a ballot ranks X & Y equally, below
> > > everything else in the Smith set, then that ballot is counted
> > > as casting half a vote for X over Y, & half a vote for Y over
> > > X.
> > >
> > >
> And what if they _are_ ignorant of false-preference strategy? That's
> a good thing. Most likely  not many voters would be into
> strategy enough to attempt it on their own, and a publicity
> campaign, or even people talking in bars, etc, could never be
> kept secret, and that guarantees that it would backfire reliably.
> If you're going to say that some voters are strategically-inclined
> enough to attempt false-preference strategy, than you can't
> say that their intended victims aren't strategically-inclined
> enough to thwart it.

If voters are not strategically inclined, then my half-votes proposal 
accurately reflects their preferences and non-preferences when they 
truncate, and no opt-out is necessary.  But if voters fail to choose 
half-preferences out of ignorance, when doing so would elect their most 
preferred candidate, I count this as a failure of the system.

If some voters are strategically inclined, then even though other voters 
may recognize the potential adverse effect on themselves, these "victims" 
may have no way to thwart the strategy; e.g., in a prior example, Clinton 
cannot win if enough Dole voters participate in order-reversal, just as 
he cannot win if too few Dole voters actually prefer him to Dole. In that 
case the strategy of "zero-option" truncation can only help Nader win 
over Dole in that case -- a result that they consider as bad as a Dole 
victory (indeed, I think order-reversal is more likely to be attempted 
when the middle voters are believed to prefer the plurality over the 
other extreme, and can be counted on to give Dole, in our example, 
substantial support).  But if the numbers are a little different than the 
Dole plotters are hoping for (or if the circular tie is natural), then 
for the truncating Clinton voters, choosing the half-option may elect 
Clinton, while failing to do so elects Nader.

> In public elections, the voting of false preferences, either
> by order-reversal, or intentionally voting false preferences
> via the contrary half-vote option, would be most unlikely
> to happen on a scale sufficient to change the election result.

First, choosing the half-vote is not a false preference if one is
sincerely indifferent.  Failing to choose it when truncating, which
sacrifices some chance for one's favorite to win a circular tie, can
better be regarded as a false preference.
Second, if massive order-reversal is unlikely in the nature of things for
public elections, then we don't need the zero option available as
part of an attempted deterrent for the middle voters.

> Because people aren't into strategy that much, or, if theyk
> are, then their victims are into it too, and will use
> defensive deterrent strategy. If you knew that I was going
> to falsify preferences to take the election from your candidate,
> would you still vote for my candidate, which is what would
> make it possible for me to get away with it?? Not likely.

Yes, I would, in a pairwise contest against another candidate that I
liked less, at least unless I thought order-reversal on my part would 
help elect my own favorite.  We all vote for lesser evils, and for 
candidates whose campaigns use dubious tactics, all the time.  I take for 
granted that major campaigns are run by cynical manipulators.  The 
encouragement of order-reversal would not be high on my list of 
objections to Dole or his party.  Voting out of pique is irrational.  
Whether voting to "deter" some voters from trying a similar tactic in a 
future election might be rational belongs to the realm of game theory, 
but please note: the plurality voters may reverse order again in order to 
deter the middle from trying deterrence again -- this is the only way for 
them to get their candidate into office and escape what they regard as 
the tyranny of the middle.  Or the plurality voters may figure next time 
that order-reversal will work because the middle voters think they 
established their credibility in the last election and can safely vote 
their true interests this time, because "order-reversal on a scale that 
would change the election result" is now so unlikely.  Any benefits of a 
deterrent strategy are distant and uncertain.  I doubt that I would vote 
differently in the face of expected attempt at order-reversal" than I 
would if I thought a large portion of the plurality voters sincerely 
preferred my last choice to my first (middle) choice, and were thus 
likely to create a circular tie.  I would be extremely reluctant to use 
any "deterrent" strategy that actually reduces my candidate's chance of 

> An important thing about preference-falsification in Condorcet
> is that it only works against people who trust you & who
> vote for your candidate. It's a big betrayal of friends &
> allies. And they wouldn't be inclined to co-operate with it.

It seems to work against a middle that fears the opposite extreme,
regardless of the level of trust between plurality and middle.  If the
Clinton voters think Nader is a nut, they don't have to like or trust
Dole or his supporters in order to balk at a "deterrent" strategy that
elects Nader, and the Dole voters won't find that strategy credible.

> >
> > Consider a slight modification of this scenario: Dole's strategists get a
> > minority of his supporters, 18 per cent of the total, to order-reverse
> > (alternatively, they really prefer Nader), the rest sincerely vote Dole,
> > Clinton.
> >
> > 28 Dole, Clinton
> > 18 Dole, Nader
> > 20 Clinton
> > 34 Nader, Clinton
> >
> > Now we have a circular tie, and if no 1/2 votes are counted between Nader
> > and Dole for the 20 Clinton voters, Nader is least-beaten with only 46
> > against, though he lost by 12 to Dole, while Clinton lost by only 4 to
> But to use "margins-of-defeat" as a standard calls for justifying
> margins as a standard. Though margins can sound convincing,
> applying that standard doesn't get rid of the lesser-of-2-evils
> problem. Maybe I sound like a 1-issue advocate, but I consider
> that to be the important goal. As I said, the LO2E problem has
> progressive voters so cowed & dominated that we may never have the
> opportunity to find out how many Democrat voters are really
> progressives who want something better. I've written here about
> why GMC follows from a basic fundamental majority-rule principle,
> and I've talked (in "Advantages of Condorcet's Method") of
> how Condorcet's method's "votes-against" count transparently
> ensures that your vote for Clinton will do what you're trying to
> do when you vote for him: vote against victory for someone you
> like less. Margins doesn't do that.

I don't argue for margins; for the reasons indicated in my original
posting, I think my proposal better reflects voter intent.  But where 
equal ranking results only from sincere truncation, "margins" is 
equivalent to my proposal and accurately reflects the voters' preferences 
and non-preferences.

> > Nader.  But if the half-votes are counted, Clinton is least beaten, with
> > 52 against to 54 and 56 against Dole and Nader.  No rational Clinton
> > voter who understood the system would refrain from choosing the 1/2 vote
> > option, but I suggest that few voters will ever understand the definition
> Again, not at all. What you're saying is that the Clinton voters
> will help Dole beat Nader, because they trust the Dole voters to
> not make Clinton more beaten than Dole.

Sorry, that is not at all what I was trying to say.  I was assuming the 
Clinton voters were sincerely indifferent between Dole and Nader; I am 
sorry if I did not make this clear.  The Clinton voters don't have to 
trust anybody; they only have to believe the probability that Clinton 
will be less beaten than Dole is greater than zero.  If the Clinton 
voters are indifferent between Dole and Nader, I
am saying they will vote half-preferences if they are rational because it
might elect Clinton and can't hurt Clinton.  If Clinton is not
least-beaten the half-vote may help Dole beat Nader, but again, by
hypothesis they are indifferent to this effect, so even if the likelihood
that the results would be as given in my example is perceived as small,
it would be irrational not to cast the half-votes.

>If the Dole voters don't make Clinton more beaten than Dole than why should they bother
> preference-falsifying? No, any serious attempt to steal the
> election requires making Clinton more beaten than Dole.

Sorry, I was assuming the Clinton voters were sincerely indifferent, not 
that they were falsifying anything.  The fact that Dole strategists try 
order-reversal does not mean they will succeed in what you have called a 
risky endeavor! Remember, nobody knows how close the Dole voters are to a 
majority, much less how many will follow the strategy.  As a Clinton 
voter, how can I be certain they can pull it off well enough to make Dole 
less beaten than Clinton?

>The Clinton voters therefore certainly can't assume that the Dole
> voters, while using preference-falsification against Clinton
> , won't make Clinton more beaten than Dole.
> And if the Dole voters do make Clinton more beaten than Dole,
> which has to be their goal when they falsify, and the Clinton
> voters oblige & co-operate with them by making Nader more beaten
> than Dole, then the Clinton voters are being a little too
> co-operative with their victimization.
Again, they don't have to assume anything.  The Clinton voters are merely
voting rationally, given their uncertainty and sincere indifference
between Nader and Dole.  The Clinton voters are sincerely trying to elect 
Clinton by casting half-votes between Nader and Dole.

> No, the Clinton voters won't choose the contrary half-votes option
> because it would be co-operating with their victimization, under
> conditions requiring strategy, and would otherwise serve no purpose,
> voting contrary insincere preferences. And the Dole voters
> won't do it because they know the Clinton voters aren't going
> to help them steal the election from Clinton.

Again, the purpose of Clinton voters in casting the half-votes when they
truncate is so that Clinton has a better chance to win a circular tie,
however it may be caused (conveniently, in my example this worked).
There is nothing insincere about choosing half-votes for this purpose.
> And will there be lots of voters who don't even know about
> these strategies? Absolutely. That's one reason why falsification
> won't happen on a scale sufficient to change the election result.
> But even if voters found out about it, it still wouldn't happen,
> for the reasons I've given. In that kind of an atmosphere,
> everyone would know what the result would be.

By the same token, I fear that voters will be ignorant of the fact that 
they need to choose to cast half-votes in order to express fully their 
sincere desire to maximize the chance that their higher-ranked candidate 
should prevail over either of the equally-ranked candidates.  Therefore, 
at a minimum the half-vote option, as in my proposal, should be the 
default.  If we have a zero option then perhaps we also should have four 
choices: (1) my proposal, (2) zero in all cases, (3) zero only if all 
other Smith set candidates are ranked below the voter's equally ranked 
candidates on his ballot, otherwise half-votes; (4) half-vote in all 
cases.  But this gets very complicated.  As a sincere voter, if I can 
have only two options, I would rather have (1) and (3) than any other 
pair.  If my equal rankings will usually be at or near the bottom, then 
(2) is the option I would miss the least were it not offered.
> > think it should be made difficult to elect the "zero" option if its is
> > allowed.  Perhaps the voter should have to give a rational explanation of
> > why he would not want the 1/2 vote to count in case of a tiebreak, in
> > order to opt out of it.
> Oh no: The voter should be in charge of what preference votes
> he casts, and shouldn't be accountable for it. In particular, he
> certainly shouldn't have to explain why he didn't cast a false
> preference vote, or contrary halves of false preference votes.
I do not seriously propose any Condorcet literacy test.  But again, I do 
not see where any false preference is involved.  I will always vote the 
half-votes when I truncate out of sincere indifference between  
candidates, for the reasons stated above and in my original posting.

   -- Hugh Tobin
> > .-
> >
> Mike
> --

More information about the Election-Methods mailing list