[EM] IRV - disincentives

[James Gilmour]

Eric Gorr> Sent: Wednesday, March 31, 2004 12:40 AM
> Consider a case with four candidates and a voter who casts a ballot:
>     a > b > c > d
> but does not want to see either C or D win the election. It does not 
> seem to me that the voter would be any better off if they had cast a 
> ballot:
>     a > b
> preventing the possibility of their ballot being used to increase the 
> vote total for C.

Your chosen voter had TWO objectives, as defined by the complete (assumed sincere) list of
preferences:  a > b > c > d.  This sequence of preferences tells us this voter does not want D
elected (anyone in preference to D), AND that were it come to choice between C and D, this voter
would prefer C over D.

The truncated list "a > b" tells us nothing about the voter's preferences as between C and D.  By
truncating, the voter may have ensured that his/her vote never counts for C or D, but it fails to
register part of the voter's sincere preferences, ie c > d.  In IRV (and STV-PR) it is impossible to
mark your preferences to help prevent the election of two candidate: you can put only one candidate
in the last position (and I have done that in real elections).

I would suggest that a voter with the sincere preferences you suggest should always mark a
preference for all candidates, or at least, a > b > c, because that has the same effect in most
implementations of IRV and STV-PR.

James Gilmour