IRV mailing party adventures (fwd)

Forest Simmons fsimmons at
Thu Apr 5 15:50:40 PDT 2001

Oops, I wrote P3 in place of P1 in statement 1 below:

"1. At first the members of P2 are more sympathetic to P3 on average than
are the members of P3."

Should read

1. At first the members of P2 are more sympathetic to P3 on average than
are the members of P1.

---------- Forwarded message ----------
Date: Thu, 5 Apr 2001 09:02:01 -0700 (PDT)
From: Forest Simmons <fsimmons at>
To: election-methods-list at
Subject: Re: IRV mailing party adventures

Demorep has a good point here, namely that election dynamics tend to make
certain kinds of distributions of candidates and voter preferences more
likely than others, as they evolve with time.

For example, suppose that we start with two major parties, P1 and P2, and
that in the process of competing for corporate support, etc. they become
more and more alike.  Other parties start to arise to fill in the vacuum
of representing the needs of the common folk. One of these, P3, eventually
emerges as the new party with greatest support. 

P1 and P2 have maintained rhetorical differences and other superficial
differences in order to give the illusion of choice to the people. One of
these parties, say P2, is superficially closer to the new party P3, than
the other one is.

Therefore ...

1. At first the members of P2 are more sympathetic to P3 on average than
are the members of P3. <---- should read P1, not P3 

2. As P3 grows it tends to get more conversions from P2 than from

3. Members of P3 tend to rate P2 candidates above P1 candidates.

4. P3 grows by depleting P2 of its members that are sympathetic to P3.

5. As P3 becomes more of a threat to the powers that be, P1 and P2 tend to
close ranks in opposition to the upstart. They become a sort of de-facto

6. Within this coalition P2 is the weaker member because it is the one
most depleted by conversions to P3.

7. Yet P2 has nearly all of the second place support because on the one
hand it is in the same coalition as P1, and on the other hand most P3
members still prefer P2 to P1. 

8. As long as P2 has virtually all of the second place support (while
neither P1 nor P3 has a majority of first place support) P2 is the
Condorcet winner.

8. When P3 first surpasses P2 in size, it is still far from the strength
it would take to surpass the coalition {P1, P2}. 

9. At this point, if voters vote sincerely, IRV will give the win to the
stronger member of the majority coalition, P1, not the Condorcet winner

10. So we see that under IRV at this point of the evolution there are
two alternatives for the members of P3:
(i) vote sincerely and get their worst nightmare the P1 candidate.
(ii) vote insincerely and get their compromise candidate P2.

11. This dilemma continues until P3 has a clear majority of support over
the coalition, because if P3 has only a few percent majority, sincere
voting by P3 members still entails a high perceived risk of getting the
worst result. 

12. In other words this strong downward pressure continues as long as P3
is between about 30 and 55 percent of the voting population. 

13. How does a party grow from 30 to 55 percent of the population when
there is such a strong downward pressure?

14. It doesn't, unless some charismatic returning war hero gives it a
quantum boost.

In summary, under the rules of IRV, an initial two party system naturally
evolves to a state where the weaker member of a majority coalition is the
Condorcet winner. The evolution stagnates at this stage until some
catastrophe or miracle jolts the system into some other strange

Hence we see that under IRV, not all configurations are equally likely.

The random simulations referred to in recent postings are oblivious to
these over-riding dynamics, and therefore make IRV look much better than
it really is.

Demorep's intuition is correct (this time).


On Wed, 4 Apr 2001 DEMOREP1 at wrote:

> DEMOREP1 at wrote:
> > That is, IRV will wipe out a compromise choice somewhat often (how often 
> ???)
> Mr. Harper wrote-
> 3 candidates, 100 voters, IRV, ~20,000 tests
> Random electorate: in around 2-3% of elections.
> Two-dimensional electorate with a gaussian distribution: less than 1%
> That is, in that percentage of the elections, there will be a Condorcet 
> winner,
> and IRV will elect a different candidate.
> ---
> D- A slight problem.  The electorates will NEVER be *random* in the *real* 
> political world in *real* public elections.
> In rough and tough political -economic times (such as 1858-1860 or 1930-1932 
> in the U.S.A.), the electorate *may* become very polarized.  
> IRV will very likely produce a whole bunch of extremists taking office and 
> claiming majority *mandates* for their extremist agendas.

More information about the Election-Methods mailing list