[EM] Winning-votes intuitive?

Adam Tarr atarr at ecn.purdue.edu
Thu Feb 21 18:46:11 PST 2002


Blake wrote:

>So, is the point of your example that the Bush voter's are dishonest 
>then?  

Dishonest?  Is all strategic voting tantamount to dishonesty?  If so, 
then I agree that the Bush voters are dishonest.  If not, I see no 
reason to slander the (imaginary) Bush voters like that.  Their choice 
to truncate their ballots strikes me as no more dishonest than a 
plurality voter casting a vote for the "lesser of two evils" rather 
than their true first choice.

>Because if you just judge by the votes you give, without any 
>preconceptions about what Bush voters would really prefer, I think it
>is obvious that Bush really does deserve to win in your example.

My assumption, which seems reasonable given the left-right spectrum 
that most voters are comfortable placing themselves on, is that nearly 
all Bush voters would prefer Gore to Nader.  If it pleases you, we can 
include this stipulation as part of my example.

They choose not to rank Gore or Nader (in my example) because it cannot 
help them, but it can hurt them.  I am arguing that, in this example, 
Ranked Pairs has created an incentive for truncation.

>But as for the strategy argument, my response is as follows.  Let's 
>imagine that some voters aren't aware that they are allowed to vote a 
>partial ranking.  Then, they will have to randomly rank candidates 
>instead of placing them as equal.  On average, these random votes will 
>tend to cancel each other out, with an effect as if they had voted in 
>margins.  So in margins, knowing that they are allowed to vote partial 
>rankings doesn't give a new opportunity for strategy that did not
>exist before.

This is true... would-be truncators voting randomly will have the same 
effect as Ranked Pairs would predict otherwise.  But is this a good 
rationale for using Ranked Pairs?  The idea that one can truncate is 
not even remotely hard to grasp.  It's no harder than the idea you can 
only vote for one candidate in plurality.  Very few real-world voters 
will mistakenly fill out full preferences.

Now, I think you're driving at a more subtle point here, that 
truncation actually makes a strategic difference in SSD, and blunting 
that difference makes the voting strategies simpler.  My counter to 
this is that truncation also has a strategic impact in Ranked Pairs, as 
my example (with the attendant assumption about the true preference of 
the Bush voters) shows.  It's true that the Bush voters could achieve 
the same effect by ranking their ballots randomly after Bush, but the 
coordination required to pull that off is dramatically harder to 
achieve.  For all practical purposes, truncation has a strategic impact 
in both methods.

>However, if you count ballots in a way that tends to penalize those
>who vote partial rankings, which SSD does, 

I still haven't seen the example that shows this.  I don't doubt that 
one exists, although I do doubt that the example will seem as likely as 
the one I provided.  I could be wrong of course.

>the voters would actually be better off believing that they aren't
>allowed to vote partial rankings, since this will cause them to
>randomly rank, and thereby avoid the penalty.  In fact, an argument
>can be made that this is true even counting with margins, but the
>effect is minimized.

I've shown an example here where truncation produced a BENEFIT, not a 
penalty, with margins.

Furthermore, why shouldn't their be, in certain cases, some penalty to 
truncation?  It basically means you're displaying less preferences -- 
you're voting less.  It seems very natural that this could hurt you on 
occasion.  The idea that more fully expressing your preferences can 
hurt you bothers me a lot more.

>So, if people are better off believing that ranked ballots are not 
>allowed, why not just ban partial rankings?  I mean, rather than allow 
>partial rankings, but covertly penalize them, why not just be up front 
>about it and say they aren't allowed? 

Well, if you force full ranking, then almost all the differences 
between Ranked Pairs and SSD dissolve.  ALL the differences between 
Plain Condorcet and Ranked Pairs dissolve.  So making this argument 
sort of defeats the purpose of arguing between Ranked Pairs and SSD in 
the first place.

That said, I think that as a matter of voter freedom, we should not 
force the voter to fully rank the candidates.  This makes voting far 
more of a chore in a place where a bunch of fringe candidates that the 
voter has not studied (nor does he/she care to study) make the ballot.  
This could make voting a very frustrating experience for many people.  
An election reform proposal that forces full ranking would also be hard 
to sell to the public.

Finally, this could create some ugly strategies in a real election.  In 
my example, for instance, the Republicans would want to ask their 
supporters to either vote random second places, or some other scheme 
designed to produce an even split.  This hardly seems democratic, and 
the public reaction would be very negative.

>Is it because you're hoping to 
>fool people who are trying to use strategy themselves?

I'm not sure I follow this suggestion.  There is strategic value to 
truncation in Ranked Pairs, and possibly SSD.  You're going to have to 
explain to me the scenario where some people try to strategically 
truncate in SSD, only to be burned for it.

-Adam



More information about the Election-Methods mailing list