Scorecard

How do the various proportional voting systems that have been proposed for use in Canada compare with one another?  While each achieves the goal of significantly reducing the disparity between vote share and seat share by party, they do so in different ways, emphasizing somewhat different electoral values.  The table below summarizes some of the key aspects of voting systems that various groups or parties have articulated as being important to them.  Brief descriptions of the voting systems and qualities assessed as well as justifications for the scores assigned follow.

MMMP (M3P) STV+/ RUPR STV MMP FPTP
Voter Equality
Direct Representation Score (% of voters with specific MP they voted for) ~95-98% ~95+% ~90-95% ~65% ~50%
Proportionality (Vote vs Seat Share by Party) mismatch <2-3% mismatch <2-3% mismatch ~5-6% mismatch ~3-4% mismatch ~12-20%
Voter Choice
Type of Choice Rank or pick 1-3+ local region candidates Rank or pick 1-3+ local region candidates Rank 1-3+ local region candidates Pick one local, one regional candidate Pick one local candidate
Choice Within Party Yes Yes Yes Only on regional list No
Voter Experience
Ballot Size
(# of names)
~5-20 ~5-15 ~10-15 Local: ~4-5
Regional: ~20-40
~4-5
Incentive for Strategic Voting Low Low Low Moderate Strong
Geographic Representation
Prevents Regional Sweeps Yes Yes Yes Yes, using top-up MPs No
Local Connection (first tier) Most voters support one of the MPs in a multi-member riding* Most voters support one of the MPs in a multi-member riding* Most voters support one of the MPs in a multi-member riding* Half the voters support the only MP in a riding ~1.67X current Half the voters support the only MP in a current riding
Regional Connection (top-up seats, if applicable) Top-up MPs are in multi-member ridings (~4-8 seats) Top-up region covers ~8-9 ridings N/A Top-up region covers ~8-14 ridings N/A
Smallest Riding Size** 1.0X ~1.15X 2X* ~1.67X* 1X*
Indirect Effects
Women Elected STV examples in range of 40-67% STV examples in range of 40-67% examples in range of 40-67% examples in range of 31-50% 26%
Overall Score
Rating: Excellent Excellent Very Good Good Poor

* With voting systems that use multimember ridings, all MPs derive support from across the riding and are responsive to their supporters there.  Voting rules are available that can ensure, if desired, that there is one MP elected from each current riding (M3P explicitly has this feature), so in principle these STV-like systems can provide the same local connection that our current voting system does while giving almost all voters in these multi-member ridings access to an MP they support and markedly more diverse and representative policy perspectives.

** The smallest riding size for STV would have two MPs; those of the other systems would have only one.  We have flagged the riding size for FPTP because, despite being as small as possible, half the voters there have no access to an MP they support.  With STV, typically about 75% of voters in a two-seat riding would support one of the two MPs, and with M3P, all voters would have access to MPs from the larger region.

Scores in the table above are colour-coded to indicate how the values of the metrics are usually perceived (we acknowledge that there are some disputes as to the relative merits of different systems, so readers should feel free to interpret the scores according to their own set of electoral values).

Voting Systems Assessed

The systems presented in the table above are:

  • M3P (Mixed Multi-Member Proportional, or MMMP) – this is a hybrid system co-designed by Profs. Antony Hodgson (president of Fair Voting BC) and Byron Weber Becker at the request of the federal Electoral Reform Committee (originally called Riding-Centric PR).  It is a minor variant of the STV+/RUPR model (see below) designed to work with the existing riding boundaries;  it can also ensure that there is one MP explicitly identified with each current riding.
  • STV+/RUPR (Single Transferable Vote+/Rural Urban Proportional) – this closely-related pair of voting systems were co-designed by Fair Voting BC and Fair Vote Canada to address concerns expressed by some MPs that the most rural ridings might be too large for them to feel comfortable with under the more well-known STV and MMP voting systems.  With STV+/RUPR, we would have larger multi-member ridings in the more urban areas and smaller multi-member ridings in the more rural areas, with the possibility of some single-member ridings in the most remote parts of the provinces, and a small number of top-up seats (~15% of all seats).
  • STV (Single Transferable Vote) – as recommended by the BC Citizens’ Assembly on Electoral Reform, with STV we would have multimember ridings throughout each province (more members per riding in more urban areas).
  • MMP (Mixed Member Proportional) – as recommended by the Ontario Citizens’ Assembly on Electoral Reform, with MMP we would have about 60% of the seats laid out as enlarged single member districts and about 40% of the seats as compensatory (‘top-up’) seats assigned at the regional level based on the party preferences expressed by voters.

We note that other voting systems have been proposed (e.g., the Dual Member Proportional system in Prince Edward Island, the Single Member District PR system, and various forms of weighted voting).  We have not assessed these alternatives here because they have been less widely discussed, but they are certainly worthy of consideration.

For more details on the above systems, please check out our more detailed descriptions at  Fair Ways to Vote.

Qualities Considered

Voter Equality
Direct Representation Score This metric measures how many voters end up casting votes directly for an elected candidate.  Our current voting system has a value of about 50%.  A pure List PR system would have low party disproportionality, but a 0% Direct Representation score.  The more voter-centric systems can have scores close to 100%.
Proportionality (Vote vs Seat Share by Party) Proportionality is the principal issue most people think of when considering proportional voting systems.  There are metrics that express the difference between vote share and seat share by party (e.g., the Gallagher or Loosemore Hanby Indices).  A completely proportional system would have a disproportionality index of zero.  Our current system tends to have values in the high teens, while highly proportional systems tend to have values below about 5%.
Voter Choice
Type of Choice Different voting systems offer voters different kinds of choices.  Some allow voters to pick only one candidate, while others allow voters to rank one or more (in case their first choice does not have enough support to be elected).  We are not considering here systems that only allow voters to pick a party rather than a candidate (e.g., Closed List PR).
Choice Within Party Some systems offer voters the ability to choose between candidates from the same party.
Voter Experience
Ballot Size (# of names) Ballots should offer voters meaningful but not overwhelming options.  The number of names on a ballot should be reasonable, such that a voter can be reasonably expected to identify their top one or several choices.
Incentive for Strategic Voting Our current voting system encourages strategic (dishonest) voting, in which a voter perceives an incentive to vote for someone other than their true preference in order to avoid a worse outcome.  Voting systems should encourage honest voting and make it difficult to attempt strategic voting.
Geographic Representation
Prevents Regional Sweeps One significant failing of our current voting system is that it allows single parties to win all or virtually all seats in particular regions of the country.  A voting system should enable accurate representation of diverse perspectives within each region.
Local Connection (first tier) MPs should have a meaningful connection to the local region from which they have received voter support.  Our current system is said to offer local representation, but half the voters have no-one representing them.  Multi-member systems offer local representation over moderate-sized regions.
Regional Connection (top-up seats, if applicable) If top-up seats are used, they should ideally be associated with reasonably local regions.
Smallest Riding Size** Some people, most notably more rural MPs, express concern about how the size of their ridings might change under proportional voting.  We measure the size of the smallest possible riding under various systems relative to current single member ridings.
Indirect Effects
Women Elected Many people would like our voting system to encourage improved gender balance (amongst other diversity goals).  Research suggests that the best ways to improve representation of women are (1) to encourage or force parties to put forward multi-candidate slates that are visible to voters and (2) to avoid requiring women to challenge male incumbents to win a nomination.

Score Justifications

The disproportionality and direct representation numbers in the table above are largely derived from the Elections Modelling project by University of Waterloo computer scientist Byron Weber Becker.

Disproportionality:  We use the standard (country-wide) Gallagher Index as a measure of disproportionality. Becker reports results based on recent election data (2006-2015) to derive likely values we might expect over multiple election cycles.

Direct Representation:  For STV, every elected MP is elected solely with votes cast explicitly for them.  For STV+/RUPR and M3P, this metric is a bit harder to calculate or estimate, but a significant fraction of the votes that help elect the top-up candidates are also cast explicitly for them (likely between about one third and two thirds).  For MMP, about 50% of the votes are cast explicitly for the local riding winners, but only about a third of the remaining votes would be explicitly cast for the winners from the open list, so we estimate a direct representation score of about 65%;  the remaining 30-35% of voters who contribute to electing MPs do so through intra-party transfers.  That is, their vote is used to help elect another candidate of the same party as the candidate they named.

Smallest Riding Size:  Our current voting system nominally divides the country into equally-sized electoral districts averaging about 100,000 people (note that there can be considerable variability both between and within provinces).  For MMP, we have assumed a 60-40 split between local and regional seats.  For STV+/RUPR, we have assumed about 12.5% top-up seats.  For STV, we note that the smallest multimember riding would have to have two MPs.  We note that the main arguments in favour of smaller rural ridings are related to the difficulty of providing ombudsperson service and ensuring that there are people in parliament who are familiar with the issues of the area, but the principle motivation for adopting proportional voting is to ensure that as many voters as possible are represented by an MP they support on legislative matters;  since not all voters in a given riding share the same political leanings, it is necessary to have multiple MPs representing each geographic area to provide legislative representation, which is the primary constitutional purpose of MPs.

Ballot Design:  Our current voting system uses a very simple ballot – voters are asked to mark an X beside one of typically about four or five names.  While this is simple, it does not give the voters any choice between candidates from each voter’s preferred party, nor does it encourage parties to put forward more diverse slates of candidates.

Three of the voting systems presented here could all use the same kind of ballots – STV, STV+/RUPR and M3P would all typically use a ballot that would list one or more candidates from each party, arranged in columns.  Typical ballots would list about 10-15 names in total, though there could be more names listed in the largest urban areas (perhaps up to 20 or so).  On average, we might expect the number of names to equal about 2.5-3 x N, where N is the number of seats available in the district (e.g., about 10-12 candidates for a 4-seat district).  With M3P, there is an incentive for each party to run a candidate in each current riding, so the number of names might be a bit higher – closer to 4N for the party candidates, plus any independents.  There can be more smaller districts with STV+/RUPR and M3P, so the low end of the range can be smaller.

The usual recommendation for MMP would be a two-part ballot:  the first part of the ballot would be essentially identical to our current ballot (i.e., about 4-5 candidates, one from each party), while the second would typically list all candidates running in the region.  For an 8 seat region with 5 local seats, there would likely be about 4-5 candidates per riding, or 20-25 names.  For a 14 seat region with 9 local seats, there would typically be closer to 40 names listed.

With open list MMP, the voter would typically be asked to choose two candidates – one in the local riding and one from the regional list.

With STV, STV+/RUPR or M3P, the voter would typically be asked to rank one or more candidates (it is also possible to use a ‘Mark an X’ rule with these ballots) – no-one is advocating any system in which the voters would be required to know or rank all the candidates shown on the ballot.

STV has sometimes been criticized for having a counting process that is somewhat complex to describe, but a number of ballot marking and counting variants are possible that are very straightforward to use – for example, one could simply use the standard runoff process of sequentially eliminating the lowest-ranked candidate and transferring those ballots to the next-named candidate on each ballot, stopping when there are only as many candidates left as seats.  We therefore don’t mark STV ballots down on this issue.

Women Elected:  Many people support proportional voting systems because they would like to see women elected in closer proportion to their share in the overall population.  More women are typically elected under proportional voting systems than under our current system (26% in 2015 Parliament).  Although nomination practices and male incumbency appear to play the dominant role in how many women are elected [see perspectives from the UK and Canada], voting systems that encourage parties to put forward gender-balanced multi-candidate slates visible to voters can help.  With MMP, the open list ballot is the mechanism by which parties can show voters that they are offering more women, while with STV, the single ballot will show multiple candidates from most parties.  Below, we show the percentage of women elected under some STV and MMP systems:

  • STV:
  • MMP:
    • Wales (closed list):  42% (2016), down from 50% in 2003;  interestingly, in 2016, women won 48% of the constituency seats, but only 30% of the list seats;  in 2003, women won 55% of the constituency seats, but only 40% of the list seats.  Both these results are contrary to the expectation that women will win more frequently on the lists than in the constituency seats and show the importance of other factors.
    • Germany (closed list):  36.5% (2013)
    • Scotland (closed list):  35% (2016), down from 40% in 2003;  as with Wales, in the 2003 election, women won a higher proportion of constituency seats than list seats (44% vs 34%), demonstrating that it is likely not the list element of MMP that is the principal reason behind the higher election rates for women in Scotland.
      • Note:  an interesting article from Scotland argues women are elected there in close proportion to how frequently they are nominated;  SNP nominated over 40% women, 43% of SNP MSPs are women;  Labour nominated over 50% women, 46% of LP MSPs are women
    • New Zealand (closed list):  31% (2014), down from 34% in 2008;  New Zealand’s experience of adopting MMP offers a warning for Canada – in the last parliament prior to adopting MMP, women held 21% of the seats, but in the first election under MMP, women won only 14% of the constituency seats [ref] – likely because they lost nomination battles to male incumbents when the number of single member ridings was cut almost in half.  In the years since then, women have won an average of only 24% of the constituency seats (very similar to Canada’s experience with single member ridings at 26%), but 43% of the list seats [report].  Since the government tends to be comprised mainly of MPs who have won in constituencies, women in New Zealand have tended to be under-represented in the governing party.
    • Bavarian State Parliament (Landtag) – open list:  31% (2008)

In summary, we see that women have won at relatively high rates in a variety of STV elections in Australia (40-52%) and at even higher rates in European Parliamentary elections (up to 67% in Malta).  Women have also won a higher percentage of seats under MMP than under our current system (in the range of 31% to as high as 50%, though their election rate off the list has typically not been quite as high).  We also have only a single data point for open list MMP (the Bavarian Landtag).  Critics of STV sometimes point out that Ireland and Malta do not elect many women in their parliaments using STV, but there are other cultural reasons that likely explain those results.  In addition, there are countries that use or have used MMP or related parallel systems that elect few women (e.g., Albania, Romania, Hungary), so individual counter-examples do not negate the claim that these systems would be likely to enhance the proportion of women elected here in Canada.

In short, both STV and MMP have a mechanism that encourages parties to present slates of candidates that are more gender-balanced.  This dynamic likely increases representation of women in larger parties under STV (i.e., those that will typically form government) and for the second and smaller parties under MMP.  We expect the variants of STV (STV+/RUPR and M3P) to have similar levels of performance.

Resistance to Strategic Voting:  Our current system is well-known to be susceptible to strategic voting – i.e., where a voter feels that the way to generate the best outcome from their perspective is not to vote honestly, but instead to vote for another candidate who they prefer less to prevent a candidate they like even less from winning.  STV is widely regarded as being highly resistant to strategic voting [see reference] – the voter’s best strategy is almost always to vote sincerely as it is normally far too difficult to acquire the information necessary to employ a strategic voting strategy.  MMP involves the same strategic considerations as our current system for the local contest, but does encourage voters to vote honestly at the regional level, though there are at least three additional strategic considerations at that level:

  1. if an open list is being used, voters might consider not voting for their preferred candidate if they are particularly popular as that vote is likely to be unnecessary – they might instead be better off to vote for a less popular candidate in hopes that that person will be elected along with the most popular one
  2. if a voter’s preferred party is likely to be under the threshold to win list seats, a voter might consider voting for their second preference party
  3. if a party is likely to win so many local seats that they would not be eligible for any top-up seats, they might consider running a ‘clone’ party to pick up additional seats at the regional level [see reference for more information].

In summary, STV is considered most resistant to strategic voting, MMP reduces the need for strategic voting due to the second vote, and our current system is notorious for encouraging it.