Essential Questions

This is a dynamic list of possible statements about what single-winner election methods should be like. It is meant to give a survey of the EM list members' basic opinions.

Each list member can add their own column and express their degree of agreement below each of the statements. For reasons of space, please just put your initials in the column's head.

When you add a new statement about some essential property of election methods, please try to formulate it as clear as possible, using as few ambiguous terms as possible, and keep the list sorted by groups of related statements.

Please do not change the wording of statements as soon as someone expressed a degree of agreement. If you do, please announce on EM list, add a new line of degrees of agreement and put the old line of degrees of agreement in brackets.

Degrees of agreement

++  I agree strongly
 +  I rather agree
 0  I am indifferent about this
 -  I rather disagree
--  I disagree strongly

 ?  I am undecided about this

Participants

SR  Stephane Rouillon
JH  Jobst Heitzig
JG  James Green-Armytage
KV  Kevin Venzke
MO  Mike Ossipoff
JL  Juho Laatu
CB  Chris Benham
RL  Rob Lanphier
JF  Jeff Fisher

Statements and agreement by category

STATEMENT
  DEGREE OF AGREEMENT BY... (INITIALS)

What are the goals of single-winner election methods?

  SR  JH  JG  KV  MO  JL  CB  RL  JF
to elect a winner
  ++  ++  ++  ++  ++  ++  ++  ++  ++
to provide a social order (=ranking)
  ++  --  0   0    0   +   0   0  - 
to make it probable that voters vote honestly
       +  +   ++      ++  ++  ++  +
( ++   +  ?   ++   0  ++ )
to get rid of the lesser-of-2-evils problem
       +  +   ++  ++   +   +  ++  ++
to gain detailed information about voters' preferences
   +   +  +   +   ++  ++   0   +  0 
to give voters with no information about others' preferences equal power
  ++   +  +   +    +  ++  ++   +  + 
to give both majorities and minorities a fair amount of power
  ++  ++  -   ?    -  ++   -   +  ++
to provide majority rule when broader consensus cannot be reached
       -  +   +        0       +  + 
to avoid discouraging candidates (even unlikely winners) from running
       +  +   ++          ++  ++  ++
to accommodate write-in votes
                                  +

What information should be asked for and used?

  SR  JH  JG  KV  MO  JL   CB  RL   JF
Pairwise preference information (e.g. rankings) should be used
  ++   +  ++   +  ++   +   ++   +   ++
Approval information (e.g. cutoffs) should be used
   +  ++   ?   0   +   0    +   0   ++
Cardinal ratings information should be used
   -   -   +  --   +   0    +    +    - 
Strategic information (e.g. AERLO) should be used
  --  --   ?  --  ++   -   --   0   --
It should be possible to rank X and Y equal independently of whether they are approved
   +  ++   +   0   +   0    ?  ++    - 
It should be possible to rank X over Y without the need to either rank Z over Y or X over Z
   +  ++   ?  --   ?   0    +   -   --
It should be possible to rank X over Y and Y over Z without the need to rank X over Z
  --   +  --  --  --   0   --   -   --
Candidate withdrawal options should be used with some methods
   ?  --   +  --   ?   -   --   -    ?
Two or more rounds of voting should be used in some cases
       ?   ?   -       0    -   +   --
Each ranked ballot must be complete (no ties and no truncation)
              --           -         ++

How should this information be interpreted?

  SR  JH  JG  KV  MO  JL  RL  JF
Approval information should be interpreted as cardinal rates of, say, 0 or 1
   +   -   +   +   +   0   +  ++
Ranking X and Y equal means X and Y should get the same probability of winning
  ++   +   ?   ?   -   +   +  ? 
Ranking X and Y equal means the decision about X and Y should be delegated to the other voters
   -   -   -   -   +   0   +  + 
Expressing undecidedness between X and Y means this decision should be delegated to the others
  ++  ++   -   -   +   +   +  + 
It is preferable to measure defeat strength in pairwise methods by winning votes rather than margins
          ++  ++       -  ++  + 
Ranking X and Y equal in first-place means neither should lose to the other pairwise
           ?  ++           ?  -

What about certain types of "winners" and "losers"?

  SR  JH  JG  KV  MO  JL  RL  JF
Beats-All Winners (=Condorcet Winners) should always win with certainty
  ++   -  ++   0  --   +  ++  + 
Beats-All Winners should never lose with certainty
  ++  ++  ++   0   ?   +  ++  + 
Approval Winners should never lose with certainty
   -   +  --  --   ?   0   +  ? 
Beaten-By-All Losers (=Condorcet Losers) should never win
  ++  --  ++   +  --   -  ++  ++
A Beaten-By-All Loser should never win unless s/he is an Approval Winner
  --  --   -   --      -   -  - 
Beaten-By-All Losers should always have winning probability less than 1/2
   +   +   ++  +  --   -  ++  +
Approval Losers should not win
   -   -  --  --  --   0   0  +
An Approval Loser should not win unless s/he is a Condorcet Winner
   +   -  --  --  --   ?  ++  +
When >50% of voters rank X and don't vote for Y, Y should never win
      --   +  ++          ++  0

What other special properties should the winner have?

  SR  JH  JG  KV  MO  JL  RL  JF
The winner should always belong to the Smith/GeTChA/Top Set
  ++   -  ++   -  --   -  ++   +
The winner should always be top on at least one ballot
  --   ?   -   0  --   -   0  --

What effects should certain manipulations have?

  SR  JH  JG  KV  MO  JL   CB  RL
Raising X on one ballot without changing anything else should never decrease X's winning probability
  ++  ++   +  ++   +   +   +   ++
Adding a ballot which only ranks X should never decrease X's winning probability
  ++  ++   ?   +   +   0   ++  ++
Adding a ballot saying "X>(whatever)" should never decrease X's winning probability
  ++   ?   ?   +  --   +   +   +
Changing a ballot which only ranks X to "X>(whatever)" should never decrease X's winning probability
  --   -   ?   +   -   +       +
Changing a detail "X>Y" to "Y>X" on one ballot should be unlikely to change the winner from W to Z
  ++   +   ?   0   +   0   +   0
Cloning should never affect the other candidates' winning probabilities
  ++  ++   +   +   -   +   +   +
Nominating "noise" candidates which are not liked much should be unlikely to change the outcome
  ++   +  ++  ++   +   +   ++  +
A voter with several "favorites" shouldn't be able to get one elected by not voting for one
           +  ++               ++

Questions of trade-off

   SR  JH  JG  KV  MO  JL  RL
 Freedom of preference expression is more important than anti-strategic properties
    -   +   ?   ?  --   ?  --
 Reduced need for strategy is more important than methods' "vulnerability to strategy".
        ?   -      ++   ?   ?
 Efficiency is more important than simplicity
    ++  ?   +   ?  --   ?  ++

See also

• Method evaluation poll