Mixed Transferable Voting

From electowiki

This is a form of one-vote Mixed-member proportional (MMP), using delegated transfers. It's designed to be simple for voters; familiar to those used to FPTP; highly proportional; and favorable to relatively large parties, with internal diversity. The acronym for this method is MTV, but the hashtag would be #MTVoting to distinguish it from the music television channel.

Here's how it works:

  1. Seats are divided into two groups: district-based (50%-67%) and at-large (33%-50%). One equal-population district is drawn for each district-based seat.
  2. There is an open primary in each district, which guarantees and allows 2-4 winners per district. Winning candidates then affiliate with parties (which may or may not have pre-set rules limiting who can affiliate with them).
    1. For instance, a simple primary election method would use choose-one ballots, and the winners (those who progressed to the next steps) would be those who ((got over 20% of the primary vote) and (got over 10% of the amount of votes in that district from the most-recent general election)) and/or (were in the top two candidates in this district's primary).
  3. Each candidate pre-rates other candidates on a 0-4 star scale (based on the assumption that they'd save the 5-star rating for themselves). Party-affiliated candidates must give other same-party candidates 3 or 4 stars, and different-party candidates 0-2 stars.
  4. Ballots list the candidates in the local district, with a write-in line for each party. Write-ins may identify candidates by party and name, or by party and district. Lists of all candidates, and their pre-ratings of each other, are available in polling places.
  5. Voters choose a single candidate, either by choosing one of the local candidates or by writing in a candidate from another district. Voters may also vote for a party without choosing a specific candidate, by putting a check next to that party's write-in line without writing anything in.
  6. Each ballot is "filled in" with the pre-declared ratings of its chosen favorite candidate. Ballots that choose a party but not an individual candidate are filled in with ratings of 5 stars for every candidate in that party and 0 stars for all others.
  7. Any candidates who get over 50% of the votes from their local district are seated. Each seated candidate uses up up to one Droop quota of their direct votes. If they have more votes than that, their remaining voting weight will be used up in step 8 or transferred in step 9.
  8. For each seated partisan candidate who did not use up a full Droop quota, use up the remainder of their quota if possible, spread proportionally across votes from their party.
    • For instance, say candidates W, X, Y, and Z were from one party; and that W won with 120% of a quota, X won with 70% of a quota, and Y and Z survived with 60% and 40% of a quota. That means that after step 7, the party would have would be 120% of a quota of votes remaining — 20% leftover from W, and 60+40% from Y and Z. From this, you'd need to use up 30% that X was lacking. That would leave 15% of W votes to transfer, and 45% and 30% with Y and Z respectively. The party's remaining total is now 90% of a quota; if their last-surviving candidate gets just 10% votes transferred from independents or from another party's leftovers, they will win one more seat.
  9. All remaining votes are used to fill the remaining seats by some proportional method, with the restriction that any district that does not yet have a winner must get at least one winner.
    • For instance, the underlying method might be ER-STV with fractional transfers (as explained here: Single transferable vote#Ways of dealing with equal rankings). Note that ER-STV here is chosen to be a familiar method, not an optimal one, as the difference between different methods at this stage is relatively minor.
    • Note that this requires the underlying method to be modified in two minor ways. First, it must guarantee one winner per district (as in Local PR: http://localpr.ca/basics/overview/); and second, there must be a clear way to decide who gets seated even if step 9 did not use up enough votes, so that there are more quotas than seats. In most cases, there is an obvious way to make such modifications. For instance, for ER-STV, you would add the following three rules:
      • A candidate is seated as soon as all the same-district rivals are eliminated, and vice versa.
      • As soon as the number of un-filled seats is no greater than the number of districts without a winner, all candidates from districts with a winner are eliminated.
      • If more candidates have a quota than there are empty seats, the seats are allocated sequentially, in descending order of number of votes.

The key differences from a vanilla MMP method are:

  • Unlike most MMP methods, voters need vote for only one candidate, who may be local or from another district.
  • This method uses delegated transfers (see steps 3, 6, and 9). This allows proportional outcomes (intra-party as well as inter-party) from the simple choose-one ballot.
  • There is no party threshold for winning seats; a party can win with just one quota of votes. However, because of the open primary step, parties which don't have at least 20% popularity in at least one district would not have any candidates survive to the general election. In practice, this would probably tend to mean that most (but not quite all) seats would go to the two or three largest parties, but that independent and/or minor-party candidates would still have a fair chance to win seats occasionally. Smaller minority groups (ethnic, ideological, or other) would still be represented insofar as they voted together, but they'd be more likely to be factions within a party than separate parties.

Some comments are in order:

  • The district-based open primary in step 2 would keep fringe parties — those which can't break 20% support in any district — from winning. However, those voters' votes would not thereby be wasted; they'd be able to choose, and very probably successfully elect, the independent or major-party candidate who comes closest to representing their views.
  • If 2/3 of seats are district-based, then a quota would be just under 66% of the votes from the average district. If 1/2 of seats are district-based, then a quota would be under 50%, so there would be no overhang issues.
  • Independents would have a fair chance of winning a seat. They could do so by getting over 50% of the local votes, or by getting under 50% locally but the remainder of a full quota of direct write-in votes from other districts. However, merely getting a plurality locally, without votes from other districts, would not suffice.
  • Step 8 will tend to prevent overhang seats in most cases, except for parties that only get votes in places where they can narrowly win a district seat. This does leave a possible "dummy list"-type strategy, where the party strategically splits into two; one for candidates who will win district seats with more than a quota and those who will get under 50% of district votes, and another for those who will get over 50% but less than a quota. Note, however, that this strategy is dangerous for candidates on the second list; they will not win seats if they get under 50% in their district, even if they are the plurality winner there. I believe that this danger is serious enough that this strategy will be rare; certainly, it should be less of a problem than under two-ballot MMP.