The Introduction of Bill 39 (September 25, 2019) specifies a mixed compensatory electoral system to appear in a referendum as an option for future elections to the National Assembly of Quebec. Unfortunately, the proposal includes several ill-advised features. Some of these features will make the system appear rigged against supporters of small parties. Some of these features will create an incentive for parties to engage in undesirable campaign tactics, which will further undermine the legitimacy of the system. The proposed system can and should be fixed. This article identifies the smallest set of changes necessary to offer Quebecers a sound and respectable alternative to the status quo.
The Proposal in Brief
The proposed system has the following characteristics:
- Quebec will be divided into 17 regions.
- Each region will be further partitioned into divisions.1 There will be 80 divisions in total, which implies an average of 4.7 divisions per region.
- In each division, the candidate with the most votes will be elected as the division Member.
- In addition to division seats, each region will have a number of regional seats. There will be 45 regional seats in total, implying an average of 2.6 regional seats per region. Regional seats will be allocated to parties using a compensatory formula that favours parties underrepresented at the division level.
- A party’s regional seats will be assigned to the candidates at the top of its regional list. These candidates will become regional Members. Parties will submit their regional lists to the Chief Electoral Officer prior to each election.
The proposal raises two concerns. The first concern is that the system can be considered only moderately proportional at best. There are too few seats per region to ensure highly proportional election outcomes, and too few of those seats are regional seats. The second concern is that the proposed system includes several ill-advised features: a 10% Quebec-wide threshold, a non-standard seat allocation formula, a ban on double candidacy, and a failure to guard against undesirable campaign tactics. Because these features act against the interests of voters, they are likely to lead to a perception that the system is rigged in favour of certain incumbents or parties.
How to Fix the Proposed System
With only a small number of straightforward changes, Quebecers could be offered a voting system that is a clear improvement over the status quo. What the Quebec Government must do is eliminate the questionable features that have been included in the current proposal. Here are the most important changes that should be made.
- Drop the 10% threshold.
- Use the standard D’Hondt formula.
- Permit double candidacy.
- Make division votes count toward regional seats.
Even with these four changes, the system can be considered only moderately proportional. It is possible to go further and offer Quebecers a fully proportional voting system. The number of regions could be reduced from 17 to 5, for example, though this would require substantial revisions to the current proposal. The four changes listed above are intended to be the smallest set of changes necessary to produce a sound and respectable voting system. The rationale for each change is explained below.
1. Drop the 10% threshold.
In the current proposal, a party that receives less than 10% of the votes Quebec-wide cannot be awarded any regional seats, not even in a region where it receives considerably more than 10% support.
An election threshold as high as 10% is virtually unheard of anywhere in the developed world, and would be viewed as anti-democratic on the international stage. A practical consequence of this feature is that it may introduce new forms of tactical voting. Some voters may feel compelled to switch their vote away from a regionally popular party if the party could fall below 10% across all of Quebec. Other voters may switch their support to a party that is close to the 10% threshold in an effort to push the party over the top and swing a half-dozen or more seats. Another consequence is that it will frustrate legitimate citizen-backed efforts to bring new parties with new ideas into the National Assembly.
There is no need to create controversy with a Quebec-wide threshold. The small regions of the proposed model already establish natural regional thresholds of about 10% on average.
2. Use the standard D’Hondt formula.
As specified in 379.1 of the Introduction to Bill 39, the proposed system is to use a modified version of the D’Hondt formula where every second division seat is ignored when calculating the number of regional seats to allocate to a party. If a party wins two division seats in a region, it is treated as if it had won only one. If a party wins three or four division seats, it is treated as if it had won two. In other words, the formula benefits parties that win two or more division seats in a region, especially if they win an even number of division seats.
The modification is peculiar, and may lead to strange campaign behavior at election time. Consider this scenario:
- Party A receives 100,000 votes in a region, and wins 1 division seat.
- Party B receives 100,001 votes in the region, and wins 2 division seats.
Since the two parties have received essentially the same number of votes, the compensatory formula should try to leave them with the same total number of seats. So if the two parties end up next in line to receive a regional seat, the seat should go to Party A. However, because Party B’s second division seat is ignored by the modified D’Hondt formula, Party B gets the regional seat and ends up with three times as many seats as Party A. The regional seat assignment makes the outcome less proportional, which is the opposite of what regional seats are supposed to do. Parties may adjust their campaign strategies to capitalize on these types of situations.
The proposed system should just use the standard D’Hondt formula and avoid unnecessary peculiarities.
3. Permit double candidacy.
The following quote from the explanatory notes of the Introduction to Bill 39 suggests that a division candidate cannot be included on a regional list:
Lastly, a person is not allowed to seek nomination for two seats simultaneously.
In other words, double candidacy is prohibited. This interpretation is supported by the following statement from 247.1, a restriction that would not make sense if double candidacy was permitted:
The list may contain a number of names that is equal to or less than the total number of regional seats to be filled for the region.
It is standard practice to permit double candidacy, meaning that a division candidate can also appear on their party’s list for the encompassing region. If such a candidate does not win their division seat, it is possible for them to be elected as a regional Member. If double candidacy is permitted, then regional lists should be allowed to have as many names as the total number of seats in the region, including both division seats and regional seats. It is not sufficient for a list to be only as long as the number of regional seats, as in 247.1 above, since some of the list candidates might be elected at the division level.
Double candidacy should be permitted. Otherwise, small parties may be forced to withdraw their most qualified candidates from division races so that the candidates are eligible to appear on regional lists. This makes division races less competitive, which may be helpful to certain parties but is not in the best interests of voters. A thorough argument for double candidacy can be found in Louis Massicotte’s 2004 report (page 71).
If double candidacy is permitted, and if the maximum length of a list is increased to the total number of seats in the region, then it becomes unlikely for any party’s list to become exhausted. Nevertheless, it is worth examining the following text from the explanatory notes:
Rules are prescribed for a regional seat that becomes vacant. If the seat was held by a party, it is filled by a candidate from the same regional list as the outgoing Member or, in an exceptional case where the list is exhausted, by an elector designated by the party.
In the event a regional Member leaves their seat vacant and the party’s regional list is exhausted, the seat could simply be awarded to the next eligible candidate based on the election results. The next candidate would likely be from the party with the next highest quotient. It is worth making this very small additional change to reassure Quebecers that only individuals nominated prior to an election can become Members.
4. Make division votes count toward regional seats.
The proposed system requires voters to mark an ‘X’ on both a division ballot and a regional ballot. In the current version, division votes do not count toward regional seats. The proposal should be modified so that all votes cast for party lists or for party-affiliated division candidates contribute to the quotients that determine which parties receive regional seats.
The simplest way to make division votes count toward regional seats is as follows: instead of using only the regional votes as the upper part of the quotient, use the average of the total number of division votes and the number of regional votes that the party receives in the region. That is, add up all of the party’s division and regional votes and divide by two. It is necessary to use the average of the two vote counts, rather than the sum, to avoid disadvantaging independent regional candidates who cannot receive division votes.
The reason for this change may not be as obvious as the reasons for the three previous changes. The 10% threshold, the modified D’Hondt formula, and the ban on double candidacy are all non-standard features, so getting rid of them would bring the proposal more in line with other jurisdictions. The opposite is true for this change. It is common for division votes to be excluded from the calculations that allocate regional seats to parties. Yet a deeper look at the potential consequences of this convention reveals why Quebec should deviate from common practice in this aspect of the system.
With two votes and small regions, the proposal most closely resembles the mixed compensatory system used for elections in Wales. In the 2016 election to the Welsh Assembly, Labour received 48.3% of the seats with 31.5% of the vote, a difference of 16.8%. The large discrepancy is partly due to the small regions and low percentage of regional seats, and partly due to tactical campaigning and tactical voting that emerge as a result of excluding division votes from regional seat calculations. The problem is that large parties benefit mostly from division votes, and small parties benefit only from regional votes. Both academics and parties can be found advocating that electors split their tickets based on this tactical consideration, and there is evidence these appeals influence voters to a degree. This form of tactical voting exacerbates discrepancies between vote shares and seats. It also reduces competition at the local level, since division votes are ignored by small parties.
Quebecers should be spared the campaign tactics, the lack of competition, and the diminished level of proportionality that result from excluding division votes from regional seat calculations. Quebecers should also be protected from the manipulation strategy that undermined the legitimacy of elections in Albania (2005), Venezuela (2005), and Lesotho (2007). The strategy relied on the fact that large and small parties had little incentive to compete for the same votes. If the formula is changed so that division votes contribute to a party’s regional quotients, all parties have an incentive to compete for all votes and the manipulation strategy becomes infeasible.
On the surface, it may appear that making division votes count toward regional seats will favour large parties, which tend to receive slightly higher vote shares at the local level. The truth is more complicated, since the change will also help small parties compete for local votes. In any case, the impact of this change on parties is not nearly as important as the impact on voters, and voters simply benefit when potential tactics are eliminated and more parties are competing for every vote.
A final point is that there are two alternative ways of changing the proposal to make division votes count toward regional seats. Averaging division votes and regional votes, as explained above, entails the smallest change to what is already proposed. However, the following alternatives also solve the problem.
- The first alternative is to adopt an open-list mixed compensatory system based on the model used in the German state of Bavaria. Similar to the change described above, division votes and regional votes would be averaged together to determine which parties receive regional seats. The difference is that every regional vote cast for a party would also be for a specific candidate chosen by the voter. Most candidates would contest both a division and the encompassing region. Candidates’ positions on the party list would be determined after the election based on the total number of votes they receive, including both division and regional votes. More information on the Bavarian system can be found in a blog post by Wilf Day (January 8, 2016).
- The second alternative is to drop the regional ballots altogether, and simply use division votes when calculating each party’s quotients. A party’s regional seats would still be awarded to its candidates in the order they were placed on the regional list. A single-vote mixed compensatory system of this nature was recommended for Quebec by Louis Massicotte in his 2004 report (page 125). Although it requires a significant modification to the current proposal, it is a robust option that would greatly simplify the proposed system.
I would like to thank Réal Lavergne for reading several drafts of this article and helping me organize and clarify the recommendations. The responsibility for these recommendations and all information provided in the article is my own.
I still do not understand why only using regional votes helped Labour In Wales. Small regions and insufficient list votes seem to me what explains it.
The small regions and insufficient list seats are of course a major factor. But allocating list seats using only regional votes has caused the competitiveness of the Welsh system to degrade over time. One of the consequences is that Labour’s seat bonus has grown, but there are other consequences.
Here’s the theory:
A) Labour has little incentive to compete for regional votes. They have never received a list seat in 4 of the 5 regions, the regions where they dominate locally.
B) Other parties have little incentive to complete for district votes. Even if they manage to wrest an extra district seat from Labour, which is extremely difficult in most cases, it will likely be of no benefit because they will likely lose a list seat to a different party.
C) Voters slowly learn that their district vote will be wasted if it is not cast for Labour, and their regional vote will be wasted if it is cast for Labour. So more and more of them begin to split their tickets for tactical reasons.
D) The more voters engage in ticket-splitting tactics, the more likely Labour is to retain their district seats, the less likely they are to receive list seats, and the stronger the incentive becomes for other voters to engage in the same tactic. The fact that the tactical incentives reinforce themselves in a positive feedback cycle is what makes this a serious concern.
If the theory is valid, we would expect to see a number of effects:
1. Labour’s seat bonus will rise over time as their district vote performance increases relative to their regional vote performance. And that is what we see, with the seat bonus starting at 11% in the 1999 election, increasing to 13% for the next three elections, then rising to almost 17% in the most recent election.
2. Labour will receive more district votes than regional votes in the 4 regions where they dominate locally, but fewer district votes than regional votes in the remaining region. This is exactly what happened in the 2016 election.
3. Small parties will campaign only for regional votes and encourage tactical voting. We see an example of this behaviour in the video linked below.
4. Parties will exhibit little desire to campaign against Labour. They are not fully competing for district votes, which are of no value to them. They are not fully competing for regional votes, which are of no value to Labour. There is some anecdotal evidence for this in the video above, where the Green Party refers to themselves as “Labour voters” (1:55).
5. List representatives will hardly compete locally, and will therefore be seen as having little contact with constituents. There is some evidence for this in the Richard Commission’s report. It may also explain Louis Massicotte’s observation that list representatives are treated as second-class members of the Assembly.
If MMP is adopted somewhere in Canada, that jurisdiction may end up with small regions featuring insufficient numbers of regional seats. As long as the district votes count as party votes, such a system should reduce our existing problems without introducing serious new problems. The concern is not just a potential seat bonus for one party. Introducing a new form of tactical voting, reducing local competition, and creating an upper and lower class of representative are also potential consequences of excluding district votes from regional seat calculations.