Consensus?: IRV vs. Primary w/Runoff

[Forest Simmons]

On Thu, 31 Jan 2002, Steve Barney wrote in part:

> remains true that some methods are more manipulatable than others. I am only
> claiming that IRV is LESS manipulatable than the primary w/runoff procedure
> when there are 4 or more candidates. 

Why do you believe this?

Here's a rule of thumb test that you can apply:

The more manipulable a method is, the easier it eliminates the sincere
Condorcet Winner.

A method that places a premium on first place preferences and still
eliminates the CW, even when the CW has the greatest first place
preference, is a very manipulable method. That's IRV.

As Mike noted, primary with runoff will not eliminate the CW if it has the
greatest number of first place preferences.  Even if the CW comes in
second place in number of first preferences, it will still win.

That's a pretty good indicator of the relative manipulability of the two
methods.

The problem with IRV is that strategic ranking of candidates extends down
below the first level, so that potentially several levels are sensitive to
information.

In primary with runoff only the first choice is sensitive to information,
and it is no more sensitive than the IRV first choice.

Suppose that your sincere preference order is

ABCDEFG

and the polls say that G and E are running neck and neck with D not far
behind, and C not too far behind D in fourth place, A not too far behind
D, while only B and F have poor showings. 

In primary with runoff you would would have to decide on your primary vote
by considering the poll and whatever information you could glean from
other sources (some of them inevitably misleading) while taking into
account your preference strengths, utilities, etc. 

At the runoff stage your best strategy is to vote your sincere preference
without regard to popularity polls. 

Under IRV you not only have to make the agonizing decision of whom to rank
first, but make equally agonizing decisions on whom to rank as second,
third, and fourth choices.  Since these choices have to be decided up
front, they are sensitive to the available information, and hence, subject
to manipulation. 

Does it still seem to you that IRV is less manipulable than primary with
runoff?  If so, why does it seem that way?

Forest