my $Election = Vote::Count::Method::CondorcetVsIRV->new( ... );
my $result = $Election->CondorcetVsIRV();
my $Election = Vote::Count->new( TieBreakMethod => 'approval' );
my $result = $Election->CondorcetVsIRV( relaxed => 1 );
equivalent to default:
my $result = $Election->CondorcetVsIRV( relaxed => 0, smithsetirv => 0 );
Returns a HashRef with a key for winner.
Condorcet vs IRV Methods determine if the Condorcet Winner needed votes from the IRV winner; electing the Condorcet Winner if there was not a later harm violation, electing the IRV winner if there was. If there is no Condorcet Winner the IRV Winner is chosen.
To determine if there was a violation the ballots of one or more choices are redacted, with later choice on those ballots removed.
With these methods it is also possible to allow a tolerance for Later Harm.
The Double Redaction method (default) measures the later harm effect between a Condorcet Winner and the IRV Winner.
Considering the Margin of the Condorcet Winner over the IRV Winner and the number of votes needed by the Condorcet Winner from the IRV winner as measures of Preference for the Condorcet Winner and of Later Harm, it is also possible to establish a Later Harm Tolerance Threshold.
The Relaxed Later Harm option will select the Condorcet Winner when their margin of victory over the IRV Winner is greater than the number of later votes they need from the IRV Winner to be a Condorcet Winner. Although not presently implemented a different ratio or percentage could be used.
Because in most cases where the IRV and Condorcet winners are different there are Later Harm effects, without relaxed this method will almost always confirm the IRV winner.
This variation only redacts the ballots that choose the IRV Winner as their first choice. This gives the voters confidence that if their first choice wins by the later harm safe method, that their vote will not be used against that choice.
The simplest form is:
1. Determine the IRV Winner
2. Treating the ballots cast with the IRV Winner as their first choice as ballots cast only for the IRV Winner, determine the Condorcet Winner.
3. Elect the Condorcet Winner, if there is none, elect the IRV Winner.
Unfortunately, this simplest form, in cases where more than one choice defeats the IRV Winner in pairing and later choices of the IRV Winner's ballots determine which becomes the Condorcet Winner, removes the supporters of the IRV Winner from the final decision.
The form implemented by Vote::Count is:
1. Determine both the IRV and Condorcet Winner. If they are the same, elect that choice. If there is no Condorcet Winner, elect the IRV Winner.
2. Treating the ballots cast with the IRV Winner as their first choice as ballots cast for only the IRV Winner determine the Condorcet Winner.
3. If there is a Condorcet Winner, elect the first Condorcet Winner, if there is none, elect the IRV Winner. (The redaction cannot make the IRV Winner a Condorcet Winner if it isn't already one).
The double redaction version is later harm safe if the relaxed option is not used. The simple version later harm protects first choice votes only, it also does not protect the first Condorcet Winner's votes at all.
The simple version does not require Condorcet Loop resolution, and thus can be considered to be on par with Benham for complexity, and like Benham is Hand Countable. The double redaction version is more complex, but is perhaps more valuable as an approach for measuring later harm.
This method meets Later Harm with the default strict option.
The relaxed option allows a finite Later Harm effect.
Using the TCA Floor Rule and or Smith Set IRV add small Later Harm effects.
This method only meets Condorcet Loser, when the IRV winner is chosen instead of the Condorcet Winner, the winner may be outside the Smith Set.
Because this method chooses between the outcomes of two different methods, it is subject to the consistency failings of both. Given that Cloning is an important consistency issue in real elections, the clone handling should be an improvement over IRV.
Details specific to this implementation.
The Tie Breaker is defaulted to (modified) Grand Junction for resolvability. Any Tie Breaker supported by Vote::Count::TieBreaker may be used, 'all' and 'none' are not recommended.
Runs the election, returns a hashref containing the winner, similar to how other Vote::Count Methods such as RunIRV behave.
The first three behave as normal Vote::Count::Log methods, except that the default is /tmp/condorcetvsirv.
LogRedactedTo defaults to appending _redacted into the log names for the redacted election, it can be overridden by setting a value (which should be /path/basename) like LogTo.
WriteLog behaves normally, there is a log set for the CondorcetVSIRV object as well as child logs for the Election and RedactedElection, each of which has a set of logs for PairMatrix as well. WriteAllLogs will write all of these logs.
John Karr (BRAINBUZ) firstname.lastname@example.org
Copyright 2019-2021 by John Karr (BRAINBUZ) email@example.com.
This module is released under the GNU Public License Version 3. See license file for details. For more information on this license visit http://fsf.org.
This software is provided as is, per the terms of the GNU Public License. Professional support and customisation services are available from the author.
To install Vote::Count, copy and paste the appropriate command in to your terminal.
perl -MCPAN -e shell
For more information on module installation, please visit the detailed CPAN module installation guide.