With two representatives per district this is a pretty good method, if we want
a two-party system and if we accept the idea of having representatives with
different weights. Spoiler and gerrymandering related problems are greatly
reduced, and the method allows also third parties to grow.
With more than two representatives (3, 4, 10,...) per district the systems
becomes more multi-party oriented and more proportional. The proposed approach
of using a Condorcet method to pick one of all possible candidate sets is
however quite heavy computationally. For that reason the number of candidates
(and representatives) in each district should be kept quite small.
Juho
On 5.11.2011, at 2.06, capologist wrote:
> In this post I discuss a proportional representation system called
> Interactive Representation (IR). A brief description of the system is
> followed by a discussion of some characteristics compared to traditional
> systems such as single-representative districts (the dominant paradigm in the
> United States) and proportional representation systems like STV.
>
>
> OVERVIEW
> --------
> In traditional systems, a state is divided into e.g. 100 districts, and each
> district elects one representative to the legislature, where each
> representative gets one vote.
>
> In IR, each district has multiple representatives. I believe that three or
> four would be a good number, but for simplicity I will assume two
> representatives per district for most of this post.
>
> Voters are presented with ranked ballots. By a mechanism I will describe in
> an appendix, the winning candidates are selected. The highest-ranked winner
> on each ballot is the preferred winner on that ballot. Each representative
> casts a number of votes equal to the number of ballots on which he was the
> preferred winner.
>
> A voter can leave candidates unranked; if no winning candidate appears on a
> ballot, then that ballot has no preferred winner. This is effectively the
> voter saying that he would rather be unrepresented in the legislature than be
> "represented" by someone who would vote contrary to his wishes.
>
>
> CHARACTERISTICS
> ---------------
> PRECISION OF REPRESENTATION: EVERY VOTE COUNTS
>
> In a single-winner district, your candidate may win or lose by, say, 5000
> votes. If you had not bothered to vote, the margin would have been 4999 or
> 5001, and the end result would have been the same. Your one additional vote
> is effective discarded at the end of Election Day. This leads to a "My vote
> doesn't make a difference" mentality that discourages participation in the
> election. Even in STV, the proportion of elected representatives with a given
> political philosophy is a rounding of the proportion of votes cast for them.
>
> In IR, by contrast, that additional vote sticks, and your chosen
> representative is one vote stronger in the legislature than he would have
> been without your vote. I believe this would encourage voter participation.
> Similarly, increased voter participation in your district means more
> representation for your district in the legislature, which I believe would
> also serve to encourage participation.
>
>
> LOCALITY OF REPRESENTATION
>
> In STV, it takes a large number of representatives to capture small changes
> in voter support. For example, it takes at least 12 representatives to
> reflect any differences between 45%, 50%, and 55% support. (Even at 12, those
> proportions get rounded to 5/12=42%, 6/12=50%, and 7/12=58%. A more precise
> reflection of voter support requires more representatives.)
>
> In IR, those differences can be captured with as few as two representatives
> per district. This makes it feasible to have a large number of small
> districts. This has several advantages:
>
> * It empowers the voter to select a candidate who reflects both his political
> philosophy and the interests of his particular district.
>
> * It reduces the number of candidates and representatives with which the
> voter must be familiar in order to vote effectively.
>
> * It makes each representative answerable to a relatively small constituency,
> which in turn gives each voter better access to his representative. By
> contrast, in STV each representative is answerable to a much larger
> constituency and each individual voter is therefore less significant. In
> single-winner districts, each representative answers to a small district, but
> those voters who are politically opposed to his positions have effectively no
> representative to go to.
>
>
> IMMUNITY TO GERRYMANDERING
>
> With single-winner districts, if Big-endians constitute 51% of the electorate
> in each of two districts and 0% of the electorate in a third, they elect two
> out of the three representatives from those three districts, and get two out
> of three votes in the legislature, despite constituting barely one third of
> the electorate. With IR, they would get 34% of the legislative votes from
> those three districts, the same as their proportion in the electorate.
>
> A similar feature is that representation automatically adjusts to reflect
> changes in the size of the electorate. Traditionally, districts are redrawn
> every so often in such a way that all districts have about the same
> population size. In between such redistrictings, population shifts can lead
> to some districts being overrepresented and others underrepresented. In IR,
> the redistricting process would be similar, but in between redistrictings, a
> population increase in a district would automatically mean a representation
> increase, assuming the additional population votes.
>
>
> STABILITY OF LEGISLATURE MEMBERSHIP
>
> I suspect opinions will be mixed on whether this is a good thing or a bad
> thing.
>
> With single-winner districts, as the political views of the electorate
> vacillate between 51% Big-endian and 51% Little-endian, the Big-endian
> representative loses his seat to a Little-endian, who is then voted out of
> office in favor of a Big-endian, etc.
>
> With IR, the Big-endian and Little-endian representatives would both be
> secure in their seats. As the politics of the district vacillates between 51%
> B and 51% L, the voting strength of the two representatives changes, as 49%
> support is more than enough to maintain a representative's seat. The
> Big-endian representative will by voted out of office if support for
> Big-endianness drops to a very low level, or if Big-endian voters choose
> another Big-endian representative they like better.
>
>
> NO PLACE FOR STRATEGIC VOTING (?)
>
> STV is claimed to have little room for strategy, but there is certainly some.
> If a candidate needs 10,000 votes for election and your first choice receives
> 15,000, then 2/3 of your vote counts for your first choice and 1/3 of your
> vote is transferred to your second choice. If you are confident that your
> first choice will be elected without your vote, it makes sense to demote him
> and transfer your entire vote to your second choice.
>
> In IR, I can think of no realistic scenario in which insincere voting makes
> sense.
>
>
> APPENDIX
> ----------
> SELECTION OF WINNING CANDIDATES
>
> Each voter ranks the candidates. From the ranking of candidates, we construct
> an implicit ranking of candidate sets, where the voter's preference between
> two candidate sets is defined as his preference between the most-preferred
> candidate in each set.
>
> For example, let's suppose the district gets two representatives, and Joe
> Voter casts a ballot listing four candidates as such:
>
> A > B > C > D
>
> Since there are two seats available, we look at candidate sets of order two.
> Joe's implicit ranking of the candidate sets is:
>
> {A,B}={A,C}={A,D} > {B,C}={B,D} > {C,D}
>
> The Schulze Method is then applied to determine the winning candidate set.
> With modern technology, this is not difficult.
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info
----
Election-Methods mailing list - see http://electorama.com/em for list info
