1. Features

The egregious flaw in standard Mixed-Member Proportional Representation (MMPR) is that the constituency elections are done by FPTP.

This means, as before, that there is a high probability that candidates who are NOT preferred by the majority are elected to these seats.  It’s the selfsame largest-minority victory problem — feeding into voter cynicism, disillusionment and disengagement — that drives so many to ask for proportional representation in the first place!

A later top-up of additional seats based on a party-preference vote does not correct this fundamental problem.  It can easily be addressed, however, simply by replacing FPTP with Condorcet/Ranked-Pairs for the constituency elections;  problem solved.

Keeping in mind the caveats in respect of PR in general, MMPR melded to Condorcet/Ranked-Pairs offers the best of both worlds:

  • Electing the bulk of the members in an accountable directly-representative way (Condorcet/Ranked-Pairs), while
  • Avoiding the distorted outcomes of first past the post, and, at the same time
  • Keeping PR fans happy at the price of selecting some members from a party-list.

2. Particulars

Such a Condorcet-MMPR system might look as follows:

Establish how many party-list seats are to be allocated.

  1. If the total number of seats is not to be increased, the current constituency seats must be reduced, requiring redistribution and adjustment of boundaries of the remaining constituencies.
  2. If these are to be simply new seats added-on, then no redistribution nor boundary adjustments are required.  This is the easiest change path, at the price of increasing the size of the legislature (and, consequently, on-going operating costs — as well as additional office space and staff for the additional members).

For each general election each party proposes a set of list-candidates (leaving aside, for now, whether the proportionality, and the party-lists, would be regionalized).

Party leaders might, perhaps, both run in an constituency, and be placed first on the party-list — then, even if the leader loses his or her constituency, he or she would be elected from the party-list (providing that the party is proportionally-entitled to additional seats).

For each constituency, the ballot would have:

  1. A candidate-specific section, listing the individual candidates (with their party affiliations, if any, else “Independent”) who are running in the constituency, as well as
  2. For general elections — A party-specific, section, where the various eligible parties would be listed.  A given voter’s most-preferred party choice does not need to match his or her most-preferred candidate from the constituency-vote section.
  3. Note that Independents cannot participate in the list-vote, but can still be elected to constituencies.

The voter would indicate:

  1. In the constituency section, his or her order of preference among the individual candidates for that constituency, and
  2. In the party-preference section (general elections only), his or her single most-preferred party, for the over-all proportional vote.

For a Condorcet-MMPR (general) election, such a ballot might look like this:

← Preferred →

[party 1]
[party 1]
[party 3]
[party 2]
[party 3]
[party 2]
  • The “Preferred Party” section would appear, of course, only for general elections.  Here we show a closed-list situation.  (For open lists the ballot would also show, by party, the list candidates, so that the voter could vote for one or more list-candidates of that party, which could alter their position in the list.  (This is in addition to voting for constituency candidates — this can make the ballot quite large and complex, as well as possibly confusing.))
  • We can see in this example that for the constituency election [candidate 1] is this voter’s most-preferred candidate, and, while indifferent between [candidate 3] and [candidate 4], he or she prefers them each more than [candidate 2] and less than [candidate 1]. [candidate 2] is this voter’s least-preferred constituency candidate.
  • We also see here that irrespective of the given voter’s constituency preferences he or she prefers [Party 2] for the party-list election.  This party-preference has no bearing on the outcome of constituency elections, but contributes to the over-all party-proportionality outcome.

For each individual constituency, an individual winner is determined by tallying the constituency votes by Condorcet/Ranked-Pairs, choosing the candidate most-preferred by the majority instead of by the largest minority.

And for General elections, for the election as a whole:

  1. Constituency candidates who are already elected in the given general election are removed from any party-lists in which they occur.
  2. Party-proportionality is determined by aggregating the party-preference votes from all ballots across all constituencies.
  3. Eligibility of parties to participate in the proportional allocation is determined.  This might, perhaps, be based on their percentages of the party-preference vote, or having elected or at least run some number or percentage of constituency candidates, or some combination of these.
  4. The determination of how many seats to be allocated proportionally is the (nominal) prescribed number of legislature seats less the number of constituencies that elected Independents, or that elected candidates of parties not eligible to participate in the proportional allocation — otherwise the proportional allotment can become skewed and allocate to party-vote parties more seats than they are proportionally entitled.
  5. Given this proportional-seat total, the number of seats to which each party is proportionally entitled is determined, according to the given party’s proportion (with respect to other eligible parties) of the party-preference vote.
  6. This party-total is then reduced by the number of constituencies that elected candidates of that party to determine the additional seats, if any, to allocate to it.  If this results in a negative value, then zero is used;  no seats are taken-away, and no additional seats are allocated.  (This is an “overhang” situation — the party has more seats than their proportional vote.)
  7. For each party entitled to such additional list seats, candidates from the given party’s list are then elected, in list-order, up to the given number of additional seats, or until the given party-list is exhausted (an under-hang situation — the party has fewer seats than their proportional vote), whichever comes first.

    (Ensuring that the party Leader is at the top of the list would largely eliminate situations where the Leader, having campaigned mightily for the party, in the end loses his or her own constituency seat — whereupon some party stalwart falls on his or her own sword and resigns his or her freshly-elected office, thus triggering a by-election in which the party Leader then contests the seat.)

For vacancies between general elections:

  1. Once seats are allocated in the general election, whether by constituency election or election from party-lists, they are not reallocated as members’ affiliations or party allegiances might change.
  2. In the event of a constituency vacancy, a Condorcet/Ranked-Pairs by-election for that district is held (no party-list vote for by-elections), the outcome of which does not affect the already-established party-list-assigned seats.
  3. In the event that a party-list-assigned seat is vacated, the next person on the party-list for that party for the prior general election, if such person is still a member of the party and willing to serve, is selected, and otherwise the next person after him or her, and so on.  If there is no such person remaining on the list, the seat remains vacant (“under-hang”).

3. Results

This gives us a legislature with:

  1. A substantial number of members with (geographic) constituencies elected by Condorcet/Ranked-Pairs as the (true) majority preference in each case, and, in each case, directly accountable to the citizens of their respective constituencies.

  2. A (probably lesser) number of party-list members — not associated with any particular (geographic) constituency but representing “ideological” or, perhaps, “philosophical” constituencies in terms of the parties from whose lists they were elected.  These representatives are arguably more accountable to the party than to the voters themselves. (If proportionality is regionalized the list-members would nevertheless also represent a particular region.)

  3. A Proportional-Representation legislature, where the number of representatives from eligible partiess closely reflects the voters’ “Preferred Party” votes.

4. Implementations for Canada

Implementation of such a system for Canadian federal or provincial elections might unfold as follows:

Ranked-Pairs Implementation Flowchart
  1. Convert existing electoral-district (constituency) elections to Condorcet/Ranked-Pairs — No change to electoral-districts — just a change in how we mark the ballots, and how we count them.

    Significant immediate reform irrespective of the MMPR decision or delays in its implementation.

  2. Correct technical constitutional issues arising re: Senate seats vs. House of Commons seats for various regions due to possible MMPR regional-proportionality complications.

    “Theoretically, it would be possible to create five or six regions in Canada corresponding to ‘natural’ geographic divisions:   the Atlantic provinces, Quebec, Ontario, the prairie provinces and British Columbia.  However, there is at least a possibility that the creation of such supra-provincial districts might require a constitutional amendment.  The so-called Senate clause stipulates that every province has a right to ‘a number of members in the House of Commons not less than the number of Senators by which the province is entitled to be represented …'”2004 Law Commission report

  3. Undertake a national referendum for Condorcet-MMPR.

  4. If passed, and the total number of seats is NOT increased — this means a reduction in the number of geographic constituencies to accommodate the requisite number of list-seats — electoral-district adjustments would be required.

    “Two-thirds of the members of the House of Commons should be elected in constituency races [using the first-past-the-post method], and the remaining one-third should be elected from provincial or territorial party lists.  In addition, one list seat each should be allotted to Nunavut, Northwest Territories, and Yukon.” 
    2004 Law Commission report, Recommendation 4 (FPTP proposed to be amended to Ranked-Pairs (an option apparently not considered in the report)).

    Under the current distribution order (2013), Canada, federally, has 338 electoral districts (of which BC has 42).

    Under the Law Commission proposal with this same total number of seats, we would have 226 geographic constituencies, and 112 party-list seats.  BC, as a region, would have 28 federal geographic constituencies, and 14 federal party-list Seats.

    The 2021 national census would allow this redistribution to be done properly in time for parliaments elected following, say, 2023.

  1. Convert existing electoral-district (constituency) elections to Ranked-Pairs — No change to electoral-districts — just a change in how we mark the ballots, and how we count them.

    Significant immediate reform irrespective of the MMPR decision or delays in its implementation.

  2. No Constitutional Issues for Provincial Elections.

  3. Undertake a Provincial Referendum for Ranked-Pairs MMPR.

  4. If passed, and the total number of seats is NOT increased — this means a reduction in the number of geographic constituencies to accommodate the requisite number of list-seats — electoral-district adjustments would be required.

    For BC, for example, assuming (roughly) one-third of the existing 85 provincial geographic constituencies are eliminated, becoming instead list-seats, the BC legislature would then have 57 geographic seats, and 28 list seats.

    The 2021 national census would allow this redistribution to be done properly in time for legislatures elected following, say, 2023.

Next: New Zealand MMPR

Share Button