range voting + runoff
- A commenter on the FairVote blog has asked me about my views on runoffs.
Here is my response:
I don't have a position that I advocate with respect to runoffs. My
main concern is to replace the plurality voting system with something
better. I'm not a fan of IRV. I am a fan of Approval voting and
could be supportive of Range and Condorcet voting. I feel that
runoffs do add value when the people's votes have not determined a
winner which clearly has majority support. For example it may be
appropriate to have a runoff election after an approval or range
election where the winner received less a score below 50%.
This by the way is consistent with material at rangevoting.org:
<li><a href="http://rangevoting.org/HonestRunoff.html">Delayed vs
<li><a href="http://rangevoting.org/TTRvIRVreasons.html">Analysis of
why top-2 runoff does not lead to 2 party domination but IRV does</a>
(and <a href="http://rangevoting.org/TTRvIRVstats.html">here</a>)</li>
<li>One of several pages demonstrating that <a
href="http://rangevoting.org/StratHonMix.html">Approval or Range
followed by a runoff by the top two scoring candidates does better</a>
(*) For this last study I would like to see if these results hold
under a more robust set of assumptions, including assumptions that try
to model actual behavior that tends to happen in real world runoffs.
Specifically, lower turnout and, I suspect, turnout rates varying
significantly between different factions.
What work has been done so far on evaluating runoffs in the context of
Approval and/or Range voting with respect to Plurality? Critics of
runoffs point out that they are expensive, poorly attended, and
unevenly attended with respect to other elections. Much of the
momentum behind IRV is due to it being sold as a way to avoid having a
runoff (or in some cases as a way to avoid having a primary).
I would like particularly like to see:
* What the rangevoting.org "official position" on runoffs is (if there
is one; I assume that this means whatever Warren and Jan agree to as
the organization has no way of taking official positions that I am
aware of.) There is a reasonable argument that a candidate who gets
the most votes but does not achieve a majority ought to be sent to a
* any evidence suggesting that runoffs are less likely in approval or
range elections then they are in plurality elections
* Simulations that touch on runoff elections after approval and range
voting elections with various rules and parameters:
** vary what triggers the runoff an who participates:
*** always have a top-2 runoff (I believe this is simulated)
*** only have a runoff if the top rated candidate did not get a >=50%
score. Then have a top-2 runoff
** vary how well attended the runoff is
*** have optional randomized reweighting of voters utilities for
candidates (representing reassessment in the context of a new election)
*** have various models for whether voters bother to show up for a
runoff (and be sure to simulate parameters that are realistic)
*** have voters for some factions be more likely to show up for a
runoff then others
- On Mon, 17 Sep 2007 03:02:29 -0000 AllAboutVoting wrote:
> A commenter on the FairVote blog has asked me about my views on runoffs.Top two runoffs can be useful when the election method has identified
> Here is my response:
> I don't have a position that I advocate with respect to runoffs. My
> main concern is to replace the plurality voting system with something
> better. I'm not a fan of IRV. I am a fan of Approval voting and
> could be supportive of Range and Condorcet voting. I feel that
> runoffs do add value when the people's votes have not determined a
> winner which clearly has majority support. For example it may be
> appropriate to have a runoff election after an approval or range
> election where the winner received less a score below 50%.
which two candidates include the one with greatest support, but:
Cannot provide true recovery for a method that does not dependably
identify such a candidate:
Plurality has been demonstrated to fail, such as in a recent
IRV fails in believable simple demonstrations.
Approval is often better than the above, yet does not allow a
voter to indicate multiple candidates as among the best, while indicating
one of them as the best.
Have little or no value when the method has identified the single
candidate with greatest support. Condorcet and Range do well at selecting
a single winner, leaving little benefit to gain from runoffs, considering
their expense - even when the identified support is less than a majority.
Runoffs among more candidates are possible, but have their own problems.
Dave Ketchum 108 Halstead Ave, Owego, NY 13827-1708 607-687-5026
Do to no one what you would not want done to you.
If you want peace, work for justice.