[EM] Consensus?: IRV vs. Primary w/Runoff

[Forest Simmons]

On Tue, 29 Jan 2002, Forest Simmons wrote:

> On Tue, 29 Jan 2002, Steve Barney wrote:
> 
> > Do we have a consensus that the instant runoff vote (IRV) is MATHEMATICALLY
> > better than the common two step plurality vote (primary) with a follow-up
> > runoff between the 2 top plurality vote getters? It seems to me that it cannot
> > be worse, given that manipulation is bad. As far as I can see, the only
> > mathematically provable difference between them is that IRV is less
> > manipulatable. They share the same faults, including non-monotonicity. With 3
> > candidates, I believe they are mathematically identical.
> > 
> 
> I'm not sure what you are trying to say here. You conclude by saying that
> you "believe they are mathematically identical" immediately after opining
> that there is a "mathematically provable difference" between them. I
> suppose that's as good as any way of expressing confusion and bewilderment
> after being exposed to the stormy sea of opinion on the topic :-)
> 
> It seems to me that there is very little difference between the actual
> plurality runoff and the simulated one (IRV) in the case of a three way
> race, because in either case there is just an initial choice (with no
> ostensible information advantage in either case) and the other choice,
> which is hardly subject to manipulation in either case because it is a
> final head-to-head determination of the winner. 
> 
> So in a three way race virtually all of the manipulation occurs before the
> first vote, whether we use IRV or an actual runoff.
>  
> The main difference is this: in an actual runoff if one of the finalists
> behaves badly (or even heroically) between initial vote and runoff vote,
> you still have a chance to change your mind, which could be good,
> depending on the accuracy of the new information (always a judgment call,
> whether before or during the game). 
> 
> I'm not sure what you are comparing IRV with in the case of four or more
> candidates.  On the one hand you refer to a two step runoff, and on the
> other hand you talk about mathematical equivalence. Obviously a two step
> plurality runoff cannot be mathematically equivalent to a plurality runoff
> wherein several candidates are eliminated sequentially. 
> 
> Since this more elaborate, many step, plurality runoff is the type that
> IRV simulates, that is the type of "actual" runoff that I will compare IRV
> to in the following comments. 
> 
> If there are four or more candidates, there is a big difference in the
> actual and the simulated. 
> 
> In the actual runoff, the early results (suitably interpreted) serve as
> fairly reliable polls to inform strategy for later votes.  This advantage
> is not available under IRV.  IRV voters have to rely on the corporate
> pollsters, news media, and other rumor mongers which may have a vested
> interest in down playing the chances of a third party challenger, for
> example. 
> 
> Under IRV all of the manipulation happens ahead of the game. There's no
> chance to adjust and compensate for that later in the game, as more
> reliable information surfaces.
> 
> Of course, some (superstitious) IRV proponents actually believe that their
> best IRV strategy is to always rank the candidates in their sincere order
> of preference, regardless of their chances of winning.  For them, none of
> my preceding comments about strategy and manipulation would make any
> sense. 
> 
> IRV may be more convenient than an actual runoff, but I don't see any
> other advantage, mathematical or otherwise.  And it seems to me that IRV 
> comes out on the short end of the stick strategically. So count me out of
> any rumored consensus to the contrary :-)
> 
> Some of the other EM list members surely have insights that I've over
> looked in this regard, or at least, pithier ways of expressing them.
> 
> Forest
> 

Another angle just occurred to me: a simulation of the two step runoff
might be preferable to the IRV simulation of the many step runoff.

Among other reasons, it would be more obvious in the two step runoff that
the simulation doesn't eliminate the need to rank a fully viable candidate
as your first choice to keep your viable compromise from being eliminated
on the first step.

Some IRV proponents still believe that with IRV everyone can safely give
their fantasy candidate a go, because their compromise will automatically
still be in the running when it comes time for their vote to transfer. 
[When everyone thinks that way their compromise candidate gets squeezed
out by the fantasy candidates.]

So this two step runoff would make obvious the importance of getting down
to serious business on the first step, which is not so obvious under IRV,
but just as true under IRV.

Don't get me wrong. I'm not advocating this two step runoff simulation. 
I'm just saying it probably would not be as bad as IRV. 

Forest