[EM] some July 4 comments

Adam Tarr atarr at purdue.edu
Sat Jul 5 12:05:21 PDT 2003

Chris Benham wrote:

>"...Australia's  upper house..is the only one that uses IRV"
>This is the wrong way round, only the "lower"  house (the one on which the 
>government is based) uses IRV. It is called the House of 
>Representatives  and the other house is called the Senate and it is made 
>up of an equal numbers of members from each State, elected using the 
>Hare-Clarke PR method.

Thanks for the correction.  It seems a bit strange to me to use PR for the 
house that gives equal representation to each state regardless of 
population, and not the house that bases districts on population size, but 
I can see how things would end up that way.

>Assuming for a moment that a political landscape dominated by 2 major 
>parties  is a normal and happy situation,

That's a massive assumption, given that in proportionally represented 
parliamentary democracies that basically never happens.  But anyway,

>isn't it VERY important that the party which is prefered by an absolute 
>majority is the winner ?

Absolutely.  But preferred by a majority to what?  All IRV guarantees is 
that the winner is preferred by a majority (of those whose ballots are 
still active) to the candidate(s) remaining at the last stage.  Ranked 
pairs, or beatpath, on the other hand, give the election to the candidate 
that is preferred by a majority over ALL other alternatives, if one exists, 
and the one with the weakest defeats against him, barring that.  That's a 
much stronger protection of majority rule than IRV provides.

Both IRV and Condorcet will always elect any candidate who enjoys a true 
first-place majority.

>Also with IRV  the major parties are under more pressure to avoid being 
>challenged and maybe eventually supplanted  by  rivals from the same "side".

This seems to be a growing argument for IRV among those that understand its 
differences from Condorcet-compliant methods.  Looking at places where IRV 
is used, however, does not show this happening, and moreover the theory 
shows that the process of "supplanting" one party with a rival party would 
create the potential for monotonicity violations and all sorts of 
undemocratic results.  IRV elections become extremely unpredictable when 
there are three parties with comparable support.

>And of course not all  one-party systems are equivalent either. If the 
>method was Condorcet  in single seats each with approx. the same number of 
>electors, an insitutional anti-gerrymander in place, no corruption and a 
>regular high turn-out  and the (unlikely) result was that a sinlgle party 
>dominates by staking out the centre then arguably that is a much better 
>scenario than the present  Republicrat "two"-party  system in the US  today.

I suppose I'd agree, as long as the legislatures demonstrated good 
proportionality.  That said, I really doubt Condorcet will result in one 
party.  What seems more likely to me is Condorcet producing a few new 
centrist parties what hold different permutations of opinions on the main 
issues.  Having a one-party state is probably more about the political mind 
set of the country than the election method, and Americans enjoy a good fight.

>And yet on balance I prefer IRV to  Approval (which to my mind is 
>fundamentally silly).

I think it's hard to argue that approval is silly when compared to 
plurality.  At first glance, it does have a credibility problem when 
compared to ranked methods.  But the way approval counts the votes is so 
elegant and simple that it yields some nice and perhaps unexpected benefits.

>Recently while contemplating IRV and Coombs, I  thought of a new (to 
>me)  method as a joke which may not really be so bad.
>At each round, have an elimination runoff between the candidate with the 
>fewest first preferences and the candidate with the most last preferences. 
>("Elimination Runoff" ?)

An interesting idea, for sure.  Do you realize that this method would 
usually pass the Condorcet criterion?  After all, the only way a Condorcet 
winner can lose one of those elimination runoffs is if he runs against 
himself - i.e. the Condorcet winner has both the fewest first place votes 
AND the most last place votes at a given stage.  This is unlikely but possible.

That said, I'd reject this method because it fails clone-independence and 
because I don't see any benefits it holds over ranked pairs or beatpath.

>Adam wrote "..the Democrats and the Republicans are quite distinct on a 
>range of issues". mmm...Haven't you seen that  funny  "Rage against the 
>Machine" video ?  I think maybe you have a good microscope !

I actually haven't seen it, but I've heard of it.

One could make the argument that, at least in recent years, the dramatic 
swings of power that come with winning the presidency have served to 
radicalize the parties rather than make them more similar.  ("We better get 
everything we want NOW before we get kicked out!")  Just off the top of my 
head, here are some things that the Republicans are trying to push through 
or have pushed through, which Democrats unambiguously disagree with:

- Increased oil exploration
- eliminating estate taxes
- eliminating/reducing capital gains taxes
- lowered income taxes in general
- anti-abortion
- not raising minimum wage
- legal concealed weapon laws
- death penalty (some Dems do support)
- some of the post-9/11 civil liberty restrictions (although it's hard to 
say if Dems would oppose these, given a reversal of roles)
- allowing religious schools to receive federal funding

I could go on.  The parties do agree on lots of silent implicit issues, but 
those who argue that the parties are indistinguishable are experiencing 
some strong cognitive dissonance.

Sorry for the semi-off-topic-ness of this, but it does go to the point that 
parties remain distinct even when political scientists predict that they 
will become the same.


More information about the Election-Methods mailing list