Disciplined? Whether Proportional Representation Increases or Decreases Party Discipline Depends on the System

How proportional representation will affect parties’ control over their MLAs has been a significant issue in the BC referendum on electoral reform.  Opponents of proportional representation have suggested that that proportional systems increase the extent to which MLAs are beholden to parties.  Proponents argue that proportional systems may have the opposite effect, increasing the ability of voters to get rid of MLAs that they may not like but whom a party might.  The extent to which proportional systems strengthen or weaken party discipline depends on the system’s design.  Closed list mixed member proportional systems will lead to higher levels of party discipline, dual member proportional systems will likely not change the level of party discipline, and open list mixed member proportional and rural urban proportional systems will weaken party discipline.

There are two mechanisms through which electoral systems can affect party discipline.  First, electoral systems can affect how much control parties have over which candidates are likely to get elected.  The more parties determine who is most competitive from their party, the more disciplined parties will be.  In such cases MLAs that toe the party line are likely to be rewarded with a higher likelihood of re-election.  MLAs that break with the party leadership can be punished with a lower likelihood of re-election.  Second, electoral systems affect the ability of voters to choose between candidates in the same party.  When voters have a choice between multiple candidates from the same party they can punish candidates they do not think are responsive enough to their interests without having to defect to a party they dislike.  This gives MLAs more of an incentive to represent their constituents even if doing so leads them to break with the party leadership from time to time.

First past the post offers some, but limited, ability for MLAs to resist party discipline on the first mechanism, and no incentive to resist party discipline on the second.  Under first past the post, parties’ candidates in particular constituencies are chosen by local party members in the riding.  At this stage, parties are limited in their ability to decide who runs for office and, in particular, who runs in competitive ridings.  This being said, there are ways in which parties can exert significant control over who wins nominations.  Parties vet candidates and can decline to allow candidates with problematic pasts or viewpoints that party opposes to run for the nomination in a particular riding.  Party leaders can also refuse to sign the nomination papers of candidates that break with the party on too many issues.  Regardless of the electoral system, parties have relatively broad leeway to remove MLAs who break with the party too much from the party caucus (these MLAs remain in the legislature, but as independents).  Under first past the post parties relinquish some control over candidate choice to local party membership, but retain an ability to veto candidates leadership does not like.

On the second mechanism, first past the post offers very little to voters looking to punish MLAs for being too responsive to party discipline.  Because only one candidate runs for any given party in any riding, voters cannot vote against an MLA from the party they prefer without also having to change the party they support.  Under first past the post if you like a particular party but do not like their candidate you are out of luck.  You either have to vote against the party you like to get a local candidate you prefer or for a local candidate you dislike to get the party you prefer.

Dual member proportional does little to change the incentives that MLAs face.  Under this system individual candidates would still run in ridings, though most ridings would have two candidates from each party.  Parties would choose a primary and secondary candidate for each riding, likely using processes that are very similar to the ones used under first past the post.  They would likely retain the ability to vet candidates and to refuse candidates that do not fit with the party leadership’s beliefs.  Voters would not be able to rank different candidates from the same party against each other.  They would have to accept whichever candidate is listed as the primary candidate.  As a result, voters would still face a difficult decision if their preferred party nominated a candidate they do not like as their primary candidate.

Mixed member proportional (MMP) could see some change in the level of party discipline in the legislature, though whether party discipline would increase or decrease depends on the design of the system.  MMP compensates parties that receive fewer seats than their share of the vote by giving them extra MLAs off of a party list.  The higher up a candidate is on the party list, the greater the chance they have of being elected.  If parties decide on the order of the list, as is the case in closed list systems, list MLAs have a strong incentive to toe the party line.  This is because parties are likely to reward well-disciplined MLAs with high positions on the party list and punish candidates that break with the party with low positions on the party list.

MMP systems can also be designed in a way that allows voters to determine the order of the party list.  These types of systems are referred to as open list systems.  In these systems voters cast an additional vote for a candidate on their preferred party’s list.  The more votes a candidate gets, the higher they end up on the list and the more likely they are to get one of the extra seats the party needs to be proportionally represented.  Where closed list MMP systems give MLAs a strong incentive to fall in line with their parties, open list systems give them an incentive to break with their party if doing so will lead voters to push them higher up the party list.  For example, an MLA who breaks with their party to support an issue important to women may win over large numbers of feminist voters in the party, and as a result end up higher up on the party’s list.

Additionally, open list MMP allows voters an opportunity to vote against candidates they dislike without defecting from their preferred party.  If one does not like a candidate on the party list one can simply vote for someone else on the list.  If a disliked MLA does not get enough votes from their parties’ supporters they fall to the bottom of the party’s list and have a lower likelihood of being elected.  This gives MLAs another incentive to be responsive to their party’s voters even if doing so leads them to break with party leadership at times.  This is only the case, however, for MLAs elected off party lists.  MLAs elected in ridings would face the same incentives and party discipline that MLAs face under first past the post.

John Horgan has indicated a strong preference for open list systems making it likely that this type of system will be used if voters vote for MMP.  The referendum question and Elections BC information on the options on the ballot do not commit the government to either an open or closed list system.

Rural urban proportional systems offer the greatest opportunity for voters to reward MLAs that push back against party discipline.  Under this system rural parts of the province would use MMP, and as a result, rural MLAs would face the same incentives discussed above.  Urban areas, however, would use a single transferable vote (STV) system that allows voters to rank order candidates from different parties against each other.  This weakens party discipline through two mechanisms.  First, by rank ordering candidates from the same party, voters would exert control over which candidate from a party stands the best chance of getting elected.  The more first place votes a candidate receives, the more likely they are to win election regardless of what the party leadership thinks of them.  As a result, MLAs would have an incentive to break with their party if doing so means getting more first place votes.

STV systems also allow voters to rank candidates from different parties.  This gives voters a great deal of control over how they weigh the pros and cons of voting for candidates they like from parties they dislike and candidates they dislike from parties they like.  A voter who dislikes a candidate from their preferred party may choose to rank all of the candidates from their preferred party above the disliked candidate.  The voter, if they choose, could go further and start ranking candidates from less preferred parties above their disliked candidate.  This is something that the voter would not be able to do under an open list MMP system.  This gives MLAs strong incentives to be responsive to voters, even if it means breaking with party discipline.  Voters have plenty of other options to choose from should an MLA disregard their interests in order to toe the party line.

It is important not to overstate the influence that electoral systems have over party discipline.  Parties control a significant number of MLAs’ opportunities for career advancement.  Control over cabinet and committee appointments, speaking time, and the ability to vet who runs for a party offer party leadership a number of levers through which to punish MLAs that break with party discipline regardless of the electoral system.  None of this would change under a different electoral system.  At the same time, by influencing which MLAs are more likely to be elected, electoral systems will affect party discipline.  Closed-list MMP systems will likely have the strongest levels of party discipline, followed by both first past the post and dual member proportional systems.  Open list MMP systems will likely have lower levels of party discipline than first past the post systems and rural urban proportional systems will likely have the lowest level of party discipline of all of the systems on offer in the BC referendum.


Where Voters Are: Geography and Electoral Reform in British Columbia

Throughout British Columbia’s electoral reform referendum campaign a lot has been said about the way proportional representation will affect the emergence of new parties and the representation of under-represented groups.  Debates over these issues often ignore the important role that geography plays in the way that electoral systems affect voters’ and parties’ representation.  As a result, campaigners often claim that different systems will have the same effects on voters with different geographic distributions and miss the extent to which systems will affect parties with the same geographic distributions of voters the same way.  It is a mistake to assume that electoral systems will affect women and minorities in the same way and that electoral systems will affect the Green party and far-right parties differently.

First past the post awards seats to parties based on their ability to win votes in particular ridings, and not across a province or country.  This means that in first past the post the geographic distribution of voters will affect their influence over elections to a greater degree than in proportional systems.  In first past the post groups that are geographically concentrated tend to have a greater influence on election outcomes than those that are not.  Such voters will make up large portions of the electorate in particular districts.  This will give them a large sway over outcomes in those districts and parties a strong incentive to respond to their interests.  The opposite is true for voters that are geographically dispersed.  These voters make up small portions of the population in most ridings and as a result have a more limited influence over who wins the ridings.  This is the case even if the geographically dispersed voters make up a reasonably large share of the provincial population.  The same principle applies to small parties.  Small parties with geographically concentrated voters will have the ability to win ridings even if they have a relatively small share of the province’s voters.  Small parties with geographically dispersed voters will struggle to win seats.

Because ethnic minorities and immigrants tend to be geographically concentrated, they tend to benefit from first past the post systems.  Under such systems, minorities and immigrants make up majorities or near majorities in a reasonable number of ridings.  In order to win those ridings, parties have to demonstrate that they are responsive to immigrants and minorities.  Additionally, immigrants and minority candidates can use this geographic concentration to help win nomination contests.  This in turn increases the likelihood that minority candidates will run and be elected to the legislature.  Just as candidates from labour unions will be more likely to win party nominations in districts with large numbers of unionized workers and farmers will have an easier time winning nomination races in ridings with large numbers of farmers, immigrants should have an easier time winning nomination contests in ridings with large numbers of immigrants.

The importance of geographic concentration to immigrants’ and minorities’ influence over federal Canadian and British Columbian parties is further increased by the extent to which districts with large numbers of minorities are swing ridings.  At the federal level, the need for the Conservatives to win swing ridings in suburban Toronto and Vancouver led the Conservatives to commit significant resources towards campaigning towards immigrants and minorities.  Similarly, the extent to which there are swing ridings with large numbers of immigrants and minorities in places like Surrey and South-Eastern Vancouver forces BC’s political parties to demonstrate that they are responsive to immigrants and minorities.

While geographic concentration helps immigrants and ethnic minorities in first past the post systems, geographic dispersion hurts women’s representation.  Because women are evenly distributed across the province, there are not districts with more women than others.  While women make up a majority or near majority of the population of every riding, women who vote primarily on women’s issues usually do not.  Unlike with immigrants and minorities, there are few ridings in which such voters hold the balance of power.  This is the case despite the fact that there are a large number of women in British Columbia that care about and vote on women’s issues.  Their lack of geographic concentration leads them to be under-represented by first past the post systems.

The corollary of the two paragraphs above is that immigrants and ethnic minorities will not be represented as well under a proportional system while women will be better represented.  Under proportional representation, immigrants and ethnic minorities will lose many if not all of the ridings in which they are a majority, and with it, the incentives that such ridings create for parties to pay particular attention to their interests.  This is even the case in proportional systems that use ridings, such as mixed member proportional or dual member proportional.  Under both systems ridings would have to be combined in order to add the extra members (either off of party lists or as second members in ridings) that ensure each party is proportionally represented.  This will mean that urban areas will have half as many ridings and, as a result, there will be at least half as many ridings with a majority or near majority of immigrants and ethnic minorities.  On top of this, the fact that parties’ overall seat shares will be determined by the province-wide vote and not how many ridings they win will reduce the importance of winning swing ridings with large numbers of immigrants. On the flip side of this, in proportional representation a party that increases it vote share amongst women will increase its seat share even if those women are not concentrated in particular ridings.  This will increase the incentive for parties to speak seriously to women’s issues and to run women as candidates.*

The same principles regarding geography that affect voters’ representation will also affect parties.  Small parties with geographically dispersed voters tend to win a small number of votes in a large number of ridings, but only have a few ridings in which they have enough votes to be competitive to win seats.  Small parties that have geographically concentrated supporters, like the Bloc Quebecois in federal politics, have enough votes in the ridings where they are strong to win seats.  This is the case even if they do not have a great deal of national or province-wide support.  Large parties, in contrast, tend to benefit from geographic dispersion and are hurt by geographic concentration.  Such parties have a large enough province-wide vote share that they can remain competitive in large numbers of ridings even if their voters are geographically dispersed.  If their vote becomes too concentrated, they run the risk of winning more votes than they need in areas where they are strong, and not enough votes to be competitive in ridings where they are weak.

This geographic dynamic means that both the Green party and far-right parties will be under-represented in first past the post systems and better represented in proportional systems.  Neither environmentalist nor anti-immigrant voters tend to be geographically concentrated.  While there are certainly ridings with more environmentalist or anti-immigrant voters than others, the differences are usually not large enough to prevent first past the post from hurting them.  This is demonstrated by green parties’ under-representation in pretty much every jurisdiction that uses a first past the post system and by the UK Independence Party’s performance in the 2015 British election (where the party won 1 seat out of 650 on 12.6% of the vote).  In proportional systems the geographic dispersion of these parties’ supporters does not matter, and so both will win more seats on lower shares of the vote than they would have under first past the post.

Two things lead electoral systems to treat groups of voters differently, the size of the group of voters and their degree of geographic concentration.  Electoral systems do not care if a group has traditionally been under-represented in politics or what ideologies parties have.  Groups that are geographically concentrated, be they ethnic minorities and immigrants or farmers, will tend to benefit under first past the post systems.  Groups that are geographically dispersed, such a women or young people, will tend to benefit under proportional systems.  The same goes for parties.  Small parties with geographically concentrated voters, such as the Bloc Quebecois, will benefit from first past the post systems while small parties with geographically dispersed voters, such as the Green party and the UK Independence party, will benefit from proportional systems.  Ignoring the central importance of geography to the way that electoral systems affect voters and parties can lead one to make mistakes about the way both will be treated by different systems.

*It should also be noted that the advantages that first past the post creates for immigrants and ethnic minorities only hold for immigrants and ethnic minorities that are geographically concentrated.  Those that are geographically dispersed would benefit from proportional representation.

**For more on how first past the post affects parties’ responsiveness to immigrants and ethnic minorities see the following sources:

Marwah, Inder, Triadafilos Triadafilopoulos, and Stephen White. (2013). “Immigration, Citizenship, and Canada’s New Conservative Party.” In James Farney and David Rayside (Eds.) Conservatism in Canada. Toronto: University of Toronto Press. Pg. 95-119.

Triadafilopoulos, Triadafilos. (2012). Becoming Multicultural Immigration and the Politics of Membership in Canada and Germany. Vancouver: UBC Press.

Westlake, Daniel. (2018) “Multiculturalism, Political Parties, and the Conflicting Pressures of Ethnic Minorities and Far-Right Parties.” Party Politics. 24(4): 421-433.


Two Ships Yelling At Each Other in the Night: The BC Electoral Reform Debate Seemed to Be More About Entrenching Sides Than Convincing Voters

Throughout my time as a university student I was a member of the debate club.  More than a few times I heard judges (or was a judge) that described debates that lacked engagement as rounds that felt like two ships passing in the night.  Watching John Horgan and Andrew Wilkinson debate electoral reform this past week brought back memories of some of the more unpleasant or frustrating debate rounds I saw.  Though there was some attempt throughout the debate to touch on some of the very broad issues surrounding electoral reform, most of the debate was about two leaders trying to play to their bases.  There was not much in the debate to help voters who are trying to decided between proportional representation and first past the post.

Much of the half-hour debate between Horgan and Wilkinson was dominated by exchanges over whether the choices put to voters were clear.  Wilkinson was right to point to concerns with the level of detail provided with respect to some of the systems on offer but demonstrated a severely limited understanding of the options.  Wilkinson repeatedly asked about how many MLAs would be in each riding when that information is made fairly clear in the Elections BC guide mailed to voters.  Mixed member proportional (MMP) would have larger ridings with one MLA per riding while dual member proportional would have larger ridings with two members per ridings.  The only system where the number of members per riding is unclear is the single transferable vote (STV) portion of rural urban proportional, and it is hard to imagine ridings with any more than 6 or 7 MLAs in that kind of system.  Wilkinson could have pointed to other issues where lack of clarity in the systems presented is a serious concern, for example the lack of clarity with respect to whether lists in MMP would be open or closed.  Instead, he spent much of his time obfuscating about an aspect of the question answered in a pamphlet that has been mailed out to everyone in the province.

For his part, Horgan could have responded to Wilkinson’s questions about the number of MLAs relatively easily, basically by saying what was in the Elections BC handout.  Instead he just asserted that the system was simple and easy to understand and by emphasizing that the most important thing was the systems’ proportionality.  Horgan had a chance to address a serious concern that many British Columbians have about regional representation by providing a bit of detail about how the systems on offer worked.  Instead he played to his base by emphasizing proportionality as a core value.  That may have went over well with someone who already supports proportional representation, but likely did little to convince people who are on the fence or are concerned that they do not have a complete understanding of the systems being voted on.

Wilkinson’s scatter-shot critique of Horgan also verged into some odd territory.  His concerns about majority governments were on point (there are good reasons to prefer majority governments and good reasons to prefer minorities/coalitions- and proportional representation will create minorities/coalitions).  At the same time, it was odd to have Wilkinson criticize the 5% threshold by arguing that it leads votes to get “thrown in the trash.”  His concern about votes below the threshold not affecting election outcomes is a valid one, but is a larger problem in first past the post than in the systems Horgan is defending.  The remedy to the concern Wilkinson is raising with respect the threshold is to adopt a Dutch-style list proportional system.  This system is not only not on the referendum ballot, but is the polar opposite of the first past the post system Wilkinson is defending.

Horgan’s assertion that the referendum is a simple choice between proportionality and first past the post is also overly simplistic.  It ignores valid concerns that voters have about regional representation and the control parties have over who is nominated in ridings and MLAs once they are elected.  Like with the number of MLAs from each riding, these concerns could have been addressed in the debate.  Horgan could have pointed to the fact that members from the same party would have to run against each other in the STV portion of rural urban proportional as something that weakens party discipline and increases regional representation.  He could made similar arguments with respect to open list MMP systems (though here the fact that government has not committed to an open list becomes a problem).  Debates over electoral reform are about more than proportionality, and voters on the fence between different systems likely care about a range of issues.  Horgan had a chance to address those voters by talking about the way different proportional options affect things like regional representation and party discipline.  Instead he played to his base by focusing the majority of his time on proportionality when most people who place a high value on proportionality will have already made up their mind to support proportional representation.

It is hard to see this debate as anything more than a missed opportunity to improve voters’ understanding of the arguments for and against proportional representation.  Wilkinson could have raised legitimate concerns about how party lists will be decided. He also could have done a lot more to make the case for first past the post as a system them emphasizes fine-grained local representation and creates majority governments.  Instead the questions he raised have largely been answered by the material produced by Elections BC and some of his arguments were so scatter-shot that they would be best used to defend a proportional representation that is more radical than anything on offer in the referendum.  Horgan had an opportunity to reach beyond his base and address concerns about regional representation and increased partisan control over democracy that are often associated with proportional systems.  Instead he played to his base by repeatedly emphasizing proportionality.  The result was a debate that was two ships passing in the night as opposed to a serious engagement with the arguments for and against proportional representation.  The discourse over electoral reform in BC will be worse off as a result.


Choosing Between Proportional Systems Part 5: A Final Thought on Proportionality

Over my last four posts* (on province-wide MMP, regional MMP, dual member proportional, and rural urban proportional) I looked at the different proportional system being proposed in the BC referendum on electoral reform.  In this post I will do a quick comparison of the extent to which each system produces proportional results.  I find that dual member proportional systems produce the most proportional outcomes.  The degree to which mixed member proportional (MMP) will produce proportional results depends on the extent to which list seats are allocated based on region-wide vote share or province-wide vote share.  Rural urban proportional contains some disproportionality but is not necessarily less proportional than a regional mixed member proportional system.  Having said all of this, each of the systems being proposed produces far more proportional results than first past the post systems.

To compare the proportionality of the different systems I looked at the difference in vote share and seat share in each of my simulations from my previous posts.  I also looked at the differences in the actual elections held under first past the post to ensure I had a point of comparison for the proportional systems.  To get the total disproportionality for each election I added the differences between vote and seat shares for each party.  Crucially, I added together the absolute value of these differences.  I treated under-representation by two points the same way as over-representation by two points.  If one party was under-represented by two points and another over-represented by two points, the total disproportionality of that election would be four points.  I calculated disproportionality for each election as well as an average over the three elections I ran simulations for.

The graph below shows the proportionality for each system.  The most noticeable thing in the graph is that each proposed system produces a more proportional outcome than first past the post.  This is reassuring and not all that surprising.  The proportional systems live up to their name and design.

All Party Disproportionality

What is also important in the graph is that the dual member proportional and province-wide mixed member proportional systems provide the most proportional results of the systems on the ballot.  On average they get the distribution of seats within 5 percentage points of the distribution of votes across the province.  In 2017 they do particularly well, getting parties’ seat shares within 3 percentage points of their vote share.  They can do this because they take into account province-wide vote share when distributing seats.  With 87 seats to distribute either in ridings or on party lists, these systems can take into account small differences in parties’ vote share in a way that regionalized systems cannot.  A larger number of seats mean less need to round parties’ vote shares up or down so that their vote share matches a whole number of seats (as opposed to a fraction).

The two more regionalized proportional systems are less proportional.  Regional mixed member proportional systems average a difference in party seat shares and vote shares just under 10 percentage points.  Rural urban proportional systems average a difference a difference just under 8 points.  This reflects the fewer seats in each region that these systems have to allocate, it is hard to get a result that matches seat shares to vote shares with a high level of accuracy when one only has 10 or 15 seats to assign.  A party that wins 45 percent of the vote in a region with 10 seats cannot be perfectly proportionally represented.  They need to receive either 4 seats and be under-represented by 5 percentage points or 5 seats and be over-represented by 5 percentage points.  If seats were being allocated province-wide, the party with 45 percent of the vote could be given 39 seats, 44.8% of the 87 seats in the legislature.

So long as the number of seats in the legislature remains the same (which will be the case after the BC referendum), there is a trade-off between regional representation and proportionality.  The smaller regions are (and the more fine-grained the regional representation), the fewer seats are assigned to each region.  This means there has to more rounding done to turn vote shares into whole numbers of seats, and as a result, means a more disproportionate result.  Dual member proportional gets out of this to some extent by assigning the extra members if gives to parties to ensure they are proportionally represented as second MLAs in ridings.  The need to represent small parties that may be strong in regions with small numbers of seats, however, means that sometimes these extra members are assigned to regions where parties are particularly weak.  This is how the Green party gets the odd seat in Abbotsford under dual member proportional in some of my simulations.

It is finally worth noting, that this trade-off makes the ambiguity in the way the BC government has described the MMP system they are proposing particularly problematic.  Between the Attorney General’s report of electoral reform and the Elections BC guide it is unclear if list seats will be assigned based on party’s regional vote shares or province-wide vote share.  If it is the former, the MMP system proposed will be less proportional than DMP and may even be as disproportional as rural urban proportional.  If that is the case, voters that place a high value on proportionality may want to rank the system 3rd amongst the different proportional systems.  If, on the other hand, list seats will be assigned based on province-wide vote share, voters who value exact proportionality may want to rank MMP first or second amongst the proportional systems.

Two out of the three options on the referendum ballot offer a clear trade-off.  Dual member proportional offers greater proportionality.  It offers reasonable regional representation through the second seats that it assigns to ridings, though the regional distribution of small parties’ seats will be somewhat random, as noted in my simulation of the Green party’s seat distribution under the system.  Rural urban proportional offers better regional representation, but less proportionality.  In the single transferable vote ridings in urban areas, the need for candidates from the same party to compete against each other will force them to demonstrate that they are particularly good at serving their constituents.  The assignment of list members to particular rural areas of the province ensures that each rural area has a substantial number of MLAs speaking for it.  Where to fit MMP into all of this is difficult because of the ambiguity around how exactly the system will be implemented.  Depending on whether list seats are allocated based on regional or province-wide vote, it has the potential to favour either regional representation or proportionality.

*These posts also contain more detailed explanations of each of the electoral systems on the ballot in the referendum.