[EM] IRV vs. Plurality

[Bart Ingles]

I had intended to reply to this thoughtful but long essay, but haven't
had time to go through it in detail before now.

On the question of IRV vs Plurality, I would like to first point out
that pure first-past-the-post is not really the norm for U.S.
elections.  For nonpartisan local elections, the question should really
be IRV vs. Runoff.  And for partisan state and federal elections, we
generally have primary and general elections, which are similar in
effect to runoffs. 


James Green-Armytage wrote on 8/3/03:
>         I am interested in knowing how everyone here feels about this question:
>         Which is better, IRV or Plurality?
>         The reason I think that this question is important is because it has
> immediate relevance for the election methods movement, if there is to be
> one. That is, do the IRV people, Approval people, and Condorcet people
> necessarily have to be blood enemies, or can they find a common ground in
> their critique of plurality?
>         Personally, I am hoping that we can find a common ground, because I fail
> to see how the movement could make much progress if it is so sharply
> divided against itself this early in the game.

I don't share this pessimism.  I think that so long as the various
factions debate honestly, this can only improve public awareness of
voting systems.  Even contentious debate will have the effect of
educating the public, which in the long run is more important than
implementing a method which is acknowledged below to be a mere stepping
stone to better methods.

Sentiments sometimes expressed on this list notwithstanding, I haven't
really seen Approval or Condorcet advocates following IRV people around
in order to campaign against them.  What I *have* seen are efforts to
correct bad information, or fill in omissions.  Failure to provide
correct information whenever feasible could be considered unethical,
especially as part of a conspiracy of silence among advocates of the
various methods.

In some cases, when responding to a pro-IRV statement, the intent is to
educate the IRV advocate as well as the public (after all, I think most
of us started out as IRV advocates before learning about the
alternatives).

I have a question of my own:  What kinds of pro-IRV activities do you
think should be protected from interference by advocates of other
systems?  Does this apply only to efforts to pass a pro-IRV law or
ballot initiative, or should we also refrain from responding to pro-IRV
editorials, for example?  Or how about the recent League of Women Voters
Election Systems Studies, which were in part promoted and financed by
pro-IRV groups-- should these be considered 'IRV territory' as well?


> 
>         I will try to start off the discussion with some of my own thoughts.
>         First of all, there are two basic categories of discussion: One is the
> technical evaluation of IRV versus Plurality, that is simply which one is
> a more desirable voting system, leaving all other possible systems aside.
> The other category deals more with the practical context. What are the
> possible benefits or pitfalls of IRV advocacy? What is the relationship
> between advocacy of IRV and advocacy for other, more desirable, systems?
> (I assume that most people reading this know that I prefer Condorcet's
> method to IRV.)
> 
> PART ONE: TECHNICAL EVALUATION OF IRV VS. PLURALITY, LEAVING OTHER METHODS
> ASIDE
> 
>         First, the technical side. To summarize, it seems to me that, all else
> aside, IRV is a substantially better voting system then plurality.
> 
>         To begin, let's assume an American-style two-party system as a starting
> point. Unlike plurality, IRV allows people to vote for third party
> candidates for their early choices, and still have the full power of their
> vote to help decide between the two major party candidates, assuming that
> the third party candidates are eliminated.
>         This is something in and of itself, because in many circumstances it
> allows people to vote their conscience without casting a vote that is
> irrelevant to the actual winner. This will allow people to get a somewhat
> more realistic idea of what kind of support third parties have, and it
> should add depth to political debates. It will also allow third parties to
> run genuine campaigns without being demonized as spoilers, which would be
> a pretty significant step towards eroding the two-party duopoly.

But the question is, 'how significant?'  On one hand, Merrill's graphs
show Condorcet and Social Utility efficiencies approximately midway
between those of Condorcet & Approval, and those of Plurality.  But the
relationship between any mathematical properties and the method's actual
effectiveness in promoting multiple parties is obviously non-linear. 
Starting with a fatal dose of rat poison, what are the odds of a
half-dose being fatal as well?  We can't know the answer without knowing
how much overkill was present in the original dose.

The only large-scale demonstration of IRV we have is Australia's lower
house, where district elections are virtually all bipartisan (there are
apparently three parties represented in the legislature, but only two of
the three are prominent in any given district).  This in spite of the
fact that Australia has a strong multi-party system fed by proportional
representation in its upper house.


>         So, IRV works well when only two candidates are serious contenders.(In
> those cases, it should always select the Condorcet winner.) However, when
> there are three or more serious candidates, IRV can get somewhat chaotic.
> Therefore, it may tend towards fairly binary political divisions, as does
> plurality.
>         Also, IRV does not allow a smooth transition from a given set of two
> major contenders to an alternate set, because any party attempting to
> replace a current member of the dominant pair set may have to pass through
> a stage where they stand a good chance of throwing the race to the other
> member of the pair set, thus going against the will of their would-be
> supporters.

An excellent way of putting it.


>         This danger can be avoided if the supporters of the original pair set
> member rank the replacement party second over the other pair set member in
> sufficient number. Also, it can be avoided if the other pair set member is
> severely weakened.
> 
>         As far as far-wing candidates go, far left Naderish candidates for
> example, they would have a lot of trouble winning under IRV, because even
> if they did manage to eliminate the Democrat, they would throw the race to
> the Republican unless nearly all of the Democratic voters ranked the
> Naderish candidate above the Republican. Of course, any far-wing candidate
> would have trouble winning with Condorcet as well, but the difference is
> that IRV offers a very strong incentive for voters to betray a favorite
> far wing candidate if there is a danger that s/he might eliminate the
> compromise candidate and throw the race to the other side.
>         As a result, the benefit of IRV in providing a clearer impression of
> support for third parties should begin to falter at this point.
> Nevertheless, if the support does reach this point, then that is a pretty
> strong message in itself.
>         Also, this barrier of one party trying to take the place of another as a
> member of the duopoly, while it is still quite steep under IRV, is not
> nearly as steep as using plurality. That is, using plurality, third party
> candidates have to build up their support from zero under the constant
> adversity of being labeled a spoiler, and under the constant ambivalence
> of voters as to whether to stick with the lesser of two evils. With IRV,
> at least third parties can build up a reasonable-sized support base before
> the second-order spoiler effect comes into play, and therefore they can
> get themselves within 'striking distance' of one of the major parties
> while still under the umbrella of non-spoilerhood.
>         A centrist candidate may be less likely to cause the same kinds of
> problems using IRV as a far-wing candidate might cause. That is, a
> candidate who is ranked second by nearly all of one wing party's voters,
> or ranked second by a good portion of both wing party's voters. If such a
> candidate manages to eliminate one of the major party candidates, then
> they have a much better shot at winning the whole race than a Naderish
> candidate would. Still, of course, a centrist candidate has a much steeper
> hill to climb using IRV than using Condorcet.

>         To its credit, IRV never gives strategy incentives for truncation, unlike
> Condorcet, although of course it does more frequently offer incentives for
> favorite betrayal and order-reversal in general. Still, that is an
> attractive anti-strategy measure, that later choices in IRV can never
> either help or harm earlier choices. (Approval also fails on this measure,
> if you translate it as the fact that adding a sincerely lower-preference
> candidate to your list of Approved candidates can cause a sincerely higher
> preference candidate to lose.)

Why is this important?  In general, a slight incentive for truncation
would help weed out low utility candidates.  When there are no widely
accepted compromise candidates, IRV can "scrape the bottom of the
barrel" in order to fill its 50% quota.


>         In general, IRV has significantly better Condorcet efficiency than
> plurality does.

...About midway between Plurality and Condorcet, according to Merrill's
simulations.  But this assumes sincere voting.  Strategy would tend to
improve Plurality's CE more than IRV, so that in practice the two would
be closer.


>         If accurate polling data is available, it is possible that strategic
> voting will lead to Condorcet-like results, although at the expense of
> favorite betrayal, and possibly some nasty strategizing within the polls.

No different from Plurality here, except that it's easier to do with
Plurality.

 
>         In contrast to plurality, IRV allows for ranked ballots. This allows
> voters to communicate in much more detail than on a plurality (or
> Approval) ballot, which is in general a good thing, although admittedly
> voters may either vote strategically, thus distorting their communication,
> or regret their choices later. [...]

I don't generally favor ranked ballots, so this would not be a point in
favor of IRV for me.


> PART TWO: CONCERNS RELATING TO IMPLIMENTATION OF OTHER METHODS
> 
>         So, I don't think that there is much ambiguity about the technical
> superiority of IRV to plurality, however marginal some might argue it is.
>         What is more uncertain for me is the relationship between IRV advocacy
> and the advent of other systems that are more effective, especially
> Condorcet, STV, and CPO-STV.
> 
>         Of course, IRV may also help lead toward STV, which I think would be a
> good thing in itself. I imagine that this is a part of the reason that the
> Center for Voting and Democracy decided to advocate IRV in the first
> place, seeing as they started off as PR advocates with STV as their
> favorite method. (The original name of the organization was "Citizens for
> Proportional Representation.")
>         That is, of course, being familiar with IRV makes STV much, much easier
> to understand, as they are essentially the same system.
> 
>         Also, there is a chance that IRV may lead to Condorcet's method in the
> long run.

I don't consider the "stepping stone" argument to be a valid one.  It's
impossible to know whether IRV would help promote other systems more
than it would serve to delay them.  It could go either way.


>         That is, that advocacy for IRV and use of IRV can introduce many of the
> relevant concepts to voters that they will need in order to better
> understand Condorcet.
>         IRV introduces people to the idea of ranked ballots, of course. It

This carries no weight for Approval advocates, of course.  The actual
implementation of ranked ballots might be a plus for Condorcet
advocates, but only if the implementation is capable of making pairwise
comparisons and building the matrix.

For those who like both Condorcet and Approval equally, the
implementation of ranked ballots is at best neutral.  For those who are
primarily Approval advocates, ranked ballots would be a step backward.


> introduces people to the idea that people's second and third choice votes
> can have the same weight as their first choice votes, yet their voting
> power is still essentially only one vote strong.
>         Also, IRV advocacy introduces a critique of the plurality system that
> also provides the justification for Condorcet's method. It works to show
> people that the composition of government is contingent on the voting
> system, and it will help wake people up to the fact that they could have a
> change in government through a change in the voting system.

Not just IRV advocacy-- even the most bitter debate between advocates of
different voting systems would have the same effect (and may even
attract more press coverage).


>         Basically, IRV advocacy makes promises which in fact are fulfilled better
> by Condorcet's method than by IRV, such as ending the spoiler effect. By
> convincing people that it is important to end the spoiler effect, IRV
> proponents are perhaps advocating Condorcet more than they are advocating
> IRV, at least in the long run.
>         I know that I heard of IRV before I heard of Condorcet's method.
> Understanding both the benefits of IRV and the flaws that remain in IRV
> was what enabled me to appreciate Condorcet's method.

So if you and I were able to make the leap, it should be possible for
other IRV advocates to see the light as well.  But would you have
learned about IRV's flaws if not for the IRV detractors on this list and
elsewhere?


>         If IRV does become widely used, then third parties will at least have a
> marginally greater chance to participate in politics. They will be more
> likely to be eligible for matching grants, access to debates, etc.
> Hopefully this will smash the myth that the votes that third parties have
> been getting under the plurality system are indicative of their true
> support; that is if they start gaining more votes, it will be clear that
> the voting system was holding them back. This is obvious to most election
> methods fans, but most Americans probably do not think of it this way.
>         Perhaps if, after making big gains, their support reaches a second
> stagnant equilibrium at the point just before they cause a major party
> candidate to get eliminated, it will be fairly obvious that IRV is at
> fault, and a move to Condorcet will seem natural (at least to the third
> parties!).

I would rather point this out ahead of time.


>         Alternately, if people do vote honestly under IRV, and a big second-order
> spoiler effect occurs, then the flaw in IRV will be completely transparent
> to any commentator, and this could also provide a very clear opportunity
> for Condorcet to come to the fore.
> 
>         Ironically, the fact that IRV isn't as radical as Condorcet in
> undermining two-party dominance could conceivably be a benefit in
> practical terms (although it is an obvious drawback in normative terms).
> IRV gives third parties less power to upset an otherwise winning democrat
> or republican, which I think that many democrats and republicans actually
> would appreciate. If the mainstream politicians do accept IRV, though,
> alternative election methods have a pretty big foot in the door, in terms
> of public awareness, etc.
> 
>         Okay, those are some of the more hopeful appraisals of the relationship
> between IRV advocacy and Condorcet advocacy. Now I will list a few of my
> worries on the subject, about how IRV could possibly do more harm than
> good.
>         One worry is that IRV could help to further legitimize the two-party
> system and absorb the momentum needed for deeper reform.
>         That is, it is possible that we may all get behind IRV, and it has a
> great success, but then it gets 'locked in place,' and the power to make
> any changes to is taken above our reach for an extremely long period of
> time.

Agreed.  This does seem possible.


>         Also, I'm not sure whether anyone is doing it on purpose, but we have a
> real problem if IRV advocates are strawmanning Condorcet or Approval. Even
> if the intentions are good, this cannot be justified.
>         The argument that voting systems people should wait under IRV is fully
> entrenched before even bringing up alternatives does not hold water. First
> of all, that could well be a very long time, quite possibly forever, and
> in the meantime, there is a lot of benefit that can be had in using
> Condorcet's method for purposes other than elections for representative
> government. There is no sense in holding back that benefit to keep people
> from questioning the idea that IRV is the best.
>         Along the same lines, if IRV people actually devote effort to making
> Condorcet invisible, this is also unjustifiable. IRV people have no
> responsibility to devote effort to promoting Condorcet, if they don't feel
> like it, but there is no sense in their trying to 'keep Condorcet down.'
> Again, I have no idea if anyone is actually doing this, but I think that
> it is a problem if they are.
>         Of course, I think that this cuts both ways. I don't think that anyone
> should present straw man arguments against IRV, and I don't think that
> anyone should try to keep IRV out of the discussion.

Fair enough, I agree with the above.


>         In general, I think that it is very counter-productive for advocates of
> Condorcet and Approval to spend their efforts trying to block attempts to
> implement IRV. It seems obvious that their effort would be better spent
> trying to implement their own favorite system, rather than defending
> plurality against IRV.

I repeat my question from above:  What kinds of pro-IRV activities
should be protected from interference by advocates of other systems? 
Specific attempts at implementation, or any activity by IRV advocates? 
And what constitutes "blocking"-- actual lobbying against an IRV
initiative, or any public criticism of IRV?  It seems to me that there
is a fine line between "blocking" and educating the public.  Where would
you draw the line?  And if the attempt to implement IRV happens to be in
my own hometown, can I lobby my own council members against it, or
should I move to another city before contacting my representatives?

In general, I see nothing wrong with competition.  That's democracy.  In
some cases, the controversy could be beneficial.  While I don't favor
following IRV advocates around and attempting to counter all of their
efforts, I don't feel any particular obligation to engage in a
conspiracy of silence either.  In the extreme-- and I'm not saying
that's what's being advocated here, but I have seen it elsewhere-- the
suggestion that we should hide our differences from the public strikes
me as elitist and unethical.  



>         This is a very big country, and there are lots of people who use voting
> to decide things, and most of those people are still using plurality. Use
> of IRV hardly means blocking Condorcet. 99% of the time, it means
> overturning plurality or two round runoff.
>         Perhaps when we are deciding on which method to use to elect the
> president, this will no longer be the case, but we are quite a long way
> off that point, and for now the field is wide open.
>         It seems to me that there is tons of room for people to implement IRV in
> some places, Condorcet in others, Approval in others, and still more
> systems in others. Not only is this the most cooperative approach for the
> movement to take, it is also the one that provides for the most feedback.
> That is, if many groups are using each of the different methods, then
> voting methods organizers will be able to track the results and see how
> the different methods are working under different circumstances, and this
> will enrich the theoretical debate immensely.
> 
> CONCLUSION
> 
>         I have presented some of my opinions here. Although I am uncertain about
> some things, I feel fairly sure about others. I am fairly sure that IRV is
> technically better than plurality. I am fairly sure that IRV, Condorcet,
> and Approval people should try to treat each other as allies rather than
> competitors, and that they should not try to strawman or exclude each
> others' systems.
>         I think that the voting methods movement will be healthier given a
> pluralistic, multi-system approach, that is one in which different
> alternative systems are being advocated for and implemented
> simultaneously. I don't think that we should try to close off debate and
> resolve on a single system before we begin advocacy. I think that we
> should give people a chance to make intelligent decisions about voting
> systems, rather than only letting them know about the system that we like
> best. Of course, if they only ask for one recommendation, then we should
> give them the one we prefer. But if they are interested in the
> alternatives, then we should not hide them or give false arguments against
> them.

My opinion of IRV is lower than James's, so my approach to advocacy will
naturally differ accordingly.  In my view, while IRV might be a
technical improvement in some ways, it could also be considered a step
in the wrong direction, much as walking northward into a blind alley
would be an impediment to actually traveling north.

I pretty much agree with the rest of this conclusion, although I would
put it that "IRV, Condorcet, and Approval people should treat each other
as competitors rather than enemies."  But again, I think that
competition can be healthy & beneficial if conducted honorably.

Regards,
Bart