APPROVAL VOTING WITH CLUSTER SEATS (AV/CS) (2007)
A Non-Competitive Voting System for Ontario
(Presentation to the Ontario Citizens Assembly on Electoral Reform)
Chris Bradshaw, Ottawa, Ontario, January 15, 2007
I have been interested in voting systems since I attended university – where we elected our student council using something called “fractional redistribution” – and since coming to Ottawa in 1969 and experiencing a voting system which allowed citizens to vote for two ward-council candidates and for four board of controllers. More recently, I have been active in the Green Party, which has championed alternative voting systems that would reduce strategic voting that Greens feel deprive it of votes, let alone any chance of electing MLAs. I have a degree in political science and worked most of my life doing public consultation in municipal planning. Also, I have run provincially twice, in 1999 and 2003, in the Ottawa Centre seat, finishing with the highest % of votes of any GPO candidate both times. I am not, however, speaking for the party tonight.
It was the old City of Ottawa voting experience that is at the basis of the system I propose today. On the one hand, it preventing voters from marking more names than there were seats to be filled, but, on the other, candidates and their supporters urged voters to mark fewer names, a practice called “plumping.” Why should a voter’s ballot be declared ‘spoiled’ if they marked more choices than permitted, but not if they voted for fewer? How many voters marked more names than seats and effectively lost their voting franchise? In fact, why would we devise a voting system that would declare any ballots spoiled? Does marking more than one co-equal choice give that voter more (read undemocratic) power than another voter marking fewer than allowed? No, since marking extra choices simply dilutes one’s vote, while still adhering to the democratic principle of voter fairness: no voter should be able to affect the relative standing between the candidates by more than a unit of one.
When we force voters to either mark only one candidate or even to rank-order them, we are making the candidates more competitive than we need to, and that behaviour taints both election campaigns and the behaviour of parties and their parliamentarians, worried that if they do something good with another party, they could be helping their ‘enemy.’ Minority governments would work better – and last longer – if parties could more easily find common ground on a limited legislative agenda for as long as it took to get it into law, after which the largest party could find new coalition partners for additional legislative work. When candidates see a vote for another candidate as being a lost vote for himself, it induces attempts to get voters to dislike the alternatives; and failing that, to at least discourage the voter from casting a ballot at all! There is too much winner-take-all attitude in politics; only the most partisan voters subscribe to that principle; the rest are just turned off by it and by the behaviour of politicians and their supporters to gain – or stay in – power at all costs. Voter participation is hurt by this.
Ironically, this forced choice on the ballot is not reflected in our other laws and practices: voters can financially support more than one party or candidate, and still receive full tax credits; voters can work for more than one candidate; voters can post signs for multiple candidates on their lawn, and even hold a membership in more than one party (although the parties specifically forbid it, it is impossible to enforce); and they can mix and match. The supposedly ‘normal’ human propensity to search for the one choice that is “best” was successfully challenged by Herbert Simon of Carnegie-Mellon University won the Nobel Prize in Economics for discovering that people do more ‘satisficing’ (finding the most available option that is ‘good enough’) than optimizing. I urge the Assembly to opt for my proposal that includes approval voting; allowing voters to mark each choice that they are satisfied with seems more practical than forcing them to arrive at “the best,” while still accommodating those that do want to make the extra effort.
THE PROPOSED SYSTEM (Approval Voting with Cluster Seats, or AV/CS1):
I have for some time concluded that all system used by countries or their provinces/states had serious limitations. I believe I have one that avoid their drawbacks, a unique system that could become Ontario’s gift to the democratic world. It consists of the following:
* The total number of seats will be divided up: my preference is for 2/3 of seats being local constituency seats and 1/3 being “cluster seats” determined by recycling ‘unused’ votes.
* Ballots allow voters to mark as many candidates as they want (including marking all or marking none). Each mark is called an “approval” since voters are only indicating which choices they approve of. [cf. http://www.approvalvoting.org/%5D
* The local seat is awarded to the candidates with the most ‘approvals,’ but only if that candidate is approved by a majority of those casting ballots. Seats in constituencies that don’t meet the majority criteria become – until the next general election – a “cluster” seat.
* All approvals not used to determine the winner go into the provincial ‘pot.’ Included are approvals for those not winning, as well as approvals cast for winning candidates that exceed either the majority criterion or exceed those garnered by the second-place candidate, whichever number is smaller.
* Elections Ontario divides the ‘pot’ among the ‘unused’ approvals, after adding to the cluster seats those local seats that could not be awarded, due to the lack of a majority for any of the candidates. Each party then divide the constituencies they did not win into the number of geographically contiguous (regional) clusters equal to the number of cluster seats they earned. Elections Ontario then names that party’s constituency candidate in that cluster of constituencies that had the highest percentage of approvals in their own race (using percentage rather than absolute number of approvals ensures that different populations or different voter turnouts do not become a factor).
The result is that each voter has a local MLA for their constituency/riding — unless no candidate got an approval from a majority of voters — and a cluster MLA for the parties whose candidate they approved of. There are no province-wide MLAs or double ballots (e.g., MMP).
Approval voting is an old concept that has only recently become used in professional societies, together representing 450,000 professionals. The secretary-general of the U.N. is also elected via AV. The use of ‘unused votes’ is borrowed from the single-transferrable vote (STV).
Using the guide’s list of criteria (plus two additional criteria), here is my assessment of my system:
1. Legitimacy – This is probably the most difficult to predict, since the confidence of the electorate comes only with experience. It is the simplest and most transparent.
2. Fairness of Representation – AV/CS is better than any at reflecting every voter preference with equal weight. It also provides every MLA with a clear constituency, either all the voters in a constituency, or all the voters in several adjacent constituencies who gave his party’s candidate an approval. And, because all ‘unused’ approvals get a second life, no vote is wasted and no ‘safe’ (i.e., lopsided and unvarying support for a particular party) vote or electoral district is taken for granted, and therefore ignored. FPTP is also unfair because it discourages small parties with wide geographical appeal and a positive message, while it encourages smaller parties with regional appeal and negative messages. It also creates many “safe” constituency which results in such predictable results that no party makes efforts to win votes, like adding planks to their platform to meet a local priority or having the leader or senior official visit the constituency.
3. Voter Choice – I find both FPTP’s one-choice limitation and STV’s and alternative-voting’s rank-order voting to be confusing, confining, and intimidating. Approval voting overcomes these problems. First, there is no spoiled ballots, since multiple choices are allowed, and any kind of unambiguous mark is acceptable. Second, there is no chance that, on long rank-order ballots (e.g., STV), a voter will use a number twice, or leave out a number in sequence. Third, the voter is not placed into a moral dilemma, where he has to act ‘strategically,’ either by eliminating in his mind less popular (in other voters’ minds) choices, or to rank the choices in a way to have the most impact on the results. With my system, every mark has an equal impact, with no ‘pecking order.’ With AV, voters are able to give their support to candidates that are effective legislators and to those with their policy priorities. To be able to voice the ‘what’ (legislative priorities) as well as the ‘who’ of the next government is liberating to not just the voter, but the people elected.
4. Effective Parties – I would ask, if FPTP is so satisfactory, why is it not used by any Canadian party to elect its leaders or its candidates? And, compared to list systems, AV/CS denies parties the prerogative to guarantee the election of unpopular candidates by putting them at the top of their fixed-order lists.
5. Stable and Effective Government – FPTP does best of any to create majority governments, but it does so only by skewing election results in a way that growing numbers of voters find to not be legitimate. On the other hand, since all the alternatives reduce the chance for majorities, you want a voting system that engenders parties and candidates to better prepare for working across party lines, to find a majority of parliamentarians who can agree on a particular initiative. AV/CS is the most successful, as the voter does not have to treat the choices as mutually exclusive. No other system does that.
6. Effective Parliament – My system does this best, thanks to the cooperative nature of AV/CS ballot structure. Candidates and parties that are both clear about their priorities and committed to work with other parties will be the most successful at getting ‘approvals’; those who campaign negatively will be ‘punished’ as not having the attitude to allow a coalition government to work. At the same time, parties that try to please all voters’ priorities with overly long platforms will do more poorly than those with clearer agendas that suggest what their priorities are. Although FPTP is given credit for parliamentary stability, its ‘reverse Robin Hood effect’ (taking from poor parties and giving to rich ones), gives inordinate power to ‘swing’ voters. Just 5% of voters shifting from the government party to another can not only cause a change in government, but large changes in the MLA population, resulting in an inordinate loss of experienced parliamentarians.
7. Stronger Voter Participation – The trend is toward continued dropping rates of voter participation. This is due to: a) the growth of negative campaigning (see point 11); b) failure to hold parties accountable (next point); c) decline in civic engagement (see Putnam), and d) a limited choice of candidates and parties that are thought of as ‘elect-able.’ One of the features of AV/CS is the ability of voters who don’t find any choice acceptable to take a ballot and then deposit it without marking any choice, thereby decreasing the chance that any candidate will reach the majority threshold. It is interesting to note that the Green Party provides for this in its internal elections by always include the choice, “NOTA” –None Of The Above. The GPO also is the successful party after the NDP to have a clear political philosophy that voters know it is deeply committed to.
8. Accountability – AV/CS gives voters the means to support more than one choice, allowing the voter to use one or more of his choices to support a choice more for its priorities than for its ‘govern-ability.’ Without this, voters usually ignore smaller parties with clear priorities, to avoid ‘wasting’ their ballot. Under one-choice ballots or rank-order systems, those choices are discriminated against, reducing the chance that the winning party will learn about voter preferences by looking at support received by smaller parties (e.g., A winning Conservative government would become more environmental if the Green Party attracted a lot more votes that usual – but under FPTP, the Green support is always eroded at the last minute in the voting booth by citizens worried that a vote for the GPO would be wasted in deciding who the government is. Under AV/CS, the voter can influence both the who of the next government and what).
9. Simplicity, Practicality, Etc. – [Kudos to the Assembly for adding these criteria]. There is nothing simpler than voters being able to look at each name on their ballot, one at a time, and decide whether or not to give it a ‘thumb’s up’ or not, without considering the effect of their other choices. And when ballots are counted, all marks on the ballot are counted at the same time and given equal weight without considering their rank-order. And voters will feel less nervous when they know their ballot cannot be spoiled.
10. Role of MLA and Constituents (Not an Assembly criterion) – One of the strengths of FPTP is the strong unique relationship between an MLA and 40,000-60,000 voters that share a common geography. PR systems with list seats create two kinds of MLAs, one with this kind of relationship and the other without it. And STV creates large, overlapping multi-member districts that represent a single type of MLA, but a bad compromise). AV/CS, in contrast, retains both the local seats (only slightly larger than now – unless the total number of seats is increased to closer to what Ontario once had before federal ‘harmonization’) and creates cluster seats that will give each party an MLA for its supporters in that region.
11. Style of Campaigning (Not an Assembly criterion) – There has been much talk about the growth in negative campaigning. Candidates focus more on their major competitors’ faults than their own strengths and defending their own record. Under all systems except AV/CS, supporters use negative info about opponents, rather than selling their own. This creates a ‘politics of resentment’, and a cycle in which the next government only has the mandate to not be like the previous one.
I want to leave the Assembly with the clearest of messages: As much as politicians might like a winner-take-all political system, only the most partisan voters share their preference. Ironically, under FPTP, such voters’s support is taken for granted, leaving the majority of voters – who I consider non-competitive voters – to being the ‘swing’ voters that politicians during a campaign focus their platforms on. If these jaded voters were to appear on the Dr. Phil Show, he would ask them, “How’s it working for you?”
Annex A: SPECIAL CIRCUMSTANCES:
1. By-Elections: When a local MLA resigns, dies, or is barred from serving for the rest of the legislative session, Elections Ontario will hold an election in that constituency, using the same rules as for general elections, with the exception that unused approvals will go into the next general election’s provincial approvals-pool, rather than be used during the by-election, since there will be no cluster ‘pot’ activated between elections (see next). The rule requiring the winner to have majority support will not be applied in by-elections.
In the case of a cluster MLA seat being declared vacant, the seat goes – without the need for a by-election – to the candidate of the party of the departing MLA who finished second in that cluster in the previous general election.
2. A Local Race Ending in a Tie: If the tied candidates both have a majority that exceeds any other candidate’s majority, those candidates tied would simply flip a coin. The approvals of the “loser” (along with his name) would then go to the provincial pool, while those of the winner would not. No tie is possible for cluster seats.
3. Party Status: It is proposed that a party must have two or more seats to qualify as a party after an election. Those seats can be local or cluster. For a party to be awarded a cluster seat, it must have at least 6% of the cluster vote (equal to two cluster seats), or have won one local seat, in which case, it would need 3% of the cluster vote to be awarded one (province-wide) cluster seat.
4. “Orphaned” Approvals cast for Independents and for candidates whose party did not get party status: Those approvals in the provincial pool that did not win a party seat will be clustered together to create a non-party cluster seat, and the CEO will award it to the independent or non-party-status candidate with the largest voter approval. If more than one such seat is awarded, Elections Ontario will do the clustering.
5. MLAs Changing Party Affiliation – Sitting MLAs who change party affiliation would be allowed to do so (but have to face the electorate in the next election), but cluster-seat MLAs would not, since they were awarded their seat using approvals from the ‘pot’ which mostly is made up of those cast for other candidates in their party.
NOTE: Could approval voting and alternate (rank-order) voting be combined in the same ballot? Apparently they can. See Annex B for a URL reference.
Annex B: SOURCES:
* http://www.nyu.edu/gsas/dept/politics/faculty/brams/approval_preference.pdf (discussion of two ways to combine preferential and approval voting)
* Policy Options, August 2001 [my speech to the Institute on Research in Public Policy on electoral reform, as leader, Green Party of Canada]
* Putnam, Robert (2000) Bowling Alone: The Collapse and Revival of American Community
* Milner, Henry (1999) Making Every Vote Count
Annex C: “Electing” the Electoral System
Why should FPTP get a ‘bye’ into the ‘finals’ by a process in which only the ‘best’ of the alternatives is put onto a ballot with the existing system? I strongly suspect that this process will eliminate AV/CS, which is a new system with no “field experience.” Consider how many candidates enter the leadership races in major parties, thanks to a voting system on which they all appear on the first ballot. But if the government used a different ballot structure, such as AV/CS, it could include all distinct alternatives and still not ‘split’ the anti-FPTP sentiment.
To this end, I propose a process that will meld two challenges: 1) to test how people’s voting behaviour is affected by the ballot structure, and 2) to arrive at the best Ontario electoral system by using the best electoral system.
A major disappointment in the literature on electoral systems is the fact that the different outcomes asserted by those proposing new systems cannot be proven, since only one voting system can be used for each election. But since the possible electoral-reform referendum is not a regular election, and its results somewhat advisory (since there is an arbitrary 60% threshold requirement, below which the government says it only might order a change), it might afford a chance to some a relevant experiment.
This could be done by listing the election alternatives on three ballots on the same piece of paper given to each voter. I suggest three, because between all the alternatives, there are really only three ballot structures used (the MMP ballot is not relevant, since there is no party lists at play). Each ballot sections would have its own instructions: 1) the first would direct the voter to mark only one name on the ballot (FPTP), 2) the second would direct the voter to mark a priority by each name, with “1″ being the first, and “6″ being the lowest (alternative voting), and 3) the third directing the voter to mark as many choices as were acceptable to him as the constituent’s representative (approval voting).
Such a system would not allow for a test of list seats, list choices, or multi-member districts that are part of the single-transferrable vote (STV), but it would be good enough to use in the 2007 election to sort out voters’ preferences for the electoral system. They would be given the list of alternatives and allowed to mark their preferences under the three ballot-rule regimes that are possible. Since the results are only advisory to the Cabinet, anyway, the Cabinet would have the three outcomes to compare.
Another advantage of this process is that it emphasizes the learning aspects of this process, allowing the government to insert another process before reform is initiated (or eliminated), as a result. Surely, it is not appropriate to use a process that is less democratic than even the existing unfair system!
Annex D: List of AV/CS’s Major Advantages,
in Comparison to FTPT, STV, and Alternate Vote with Lists
FACTORS FPTP STV Alternative Vote w/Lists AV/CS
1. Competitiveness of Candidates, Parties Very Less competitive (rank-ordering in multi-member districts allows voters to support more than one party) Moderately, but with coalition governments more likely, voters look for ‘reasonability’ Much lower, as getting vote doesn’t require taking it away from another choice
2. Comfort Level of Voters Ballot simple, but strategic voting required. Ballots with no preference or more than one are considered “spoiled” and not counted Most intimidating. (ballot is very long and voter required to rank-order); this type of ballot is very easy to “spoil” without computer assistance Rank-ordering subject to errors, risk of spoiling ballot; also, voters don’t like including (‘approving’ of) low-ranked choices Simplest. Each choice is given thumb-up or not, without regard to other choices’ acceptability. Impossible to “spoil” ballot. Negative voting possible without ref. to polls (mark all choices but one)
3. All Citizens’ Votes of Equal Value Voters have to ‘strategically’ vote to avoid wasting, but all voters in “safe” seats get ignored Yes. This system uses the idea that excess support gets moved to other candidates with lower rank-order (but this is very complex) Many safe seats, but “list votes” are all equally valuable. All votes either help win local seat or get used to determine awarding of cluster seats (similar to STV system, but far simpler)
4. Opportunity for Voters to Indicate Mandate Priorities Single mark limits ability to send more than one message Doesn’t support one-issue parties. Extra marks on ballot not equal in status Issue parties can be marked along with candidates with a “chance”
5. Effect on Smaller Parties Discourages them significantly Not as proportional as lists, but better than FPTP. Smaller parties must strategically be careful not to have too many candidates Smaller parties still rarely win local seats, but campaign for “list” votes; list seats used to overcome unfair local-seats results. Smaller parties can get “approvals” for many reasons, and if reasonable, might win local seats; some smaller parties will win local seats when major candidates are too competitive
6. MLA Relationship with Voters A MLAs get a local constituency, but often not a majority-mandate Medium, as all MLAs will represent, jointly with other MLAs, large multi-member seats. Same as with FPTP, plus list seat members have no constituency relationship 2/3 of MLAs have local constituency and support of a majority of voters. All cluster winners have relationship with regional constituencies
7. Levels of Voter Participation “Wasted votes” and strategic voting, and unfair outcomes discourage it. Also fewer candidates result in reduced voter choice. Breeds cynicism. Fairness is attractive, but ballots have very large numbers of candidates, making it hard for voters to have enough info to vote intelligently Outcomes of local MLAs still w/o majority; list MLAs distant and cynicism about party control of list candidates (although it guarantees more minority and women MLAs). Highest, because no votes “wasted”, ballot is simple, and strategic voting not required.
8. Stability Within and Between Governments Because such a small number of voters have such a big effect on party rep., when there is a shift, huge numbers of experienced MLAs are “retired” and huge number of “rookies” come in. Not efficient. Stability lacks integrity. More stability than with FPTP. More stability in MLA personnel, but still more competitive than AV/CS, making coalition governments harder to form and maintain. High stability in personnel; and lowest level of competitiveness means that more parliamentary collaboration will occur.
9. Vote Counts Clearly Documented Vote counts are simple to express, but strategic voting makes them suspect. Also lower voter participation undercuts validity of results. This system’s ballot and tally process are the most complex, making it hard for parties to gauge support. Hardest to document (unless using fractional redistribution process), since dependent on a long, iterative process or computers (for which voter trust is an issue) Each mark on ballot (an “approval”) is counted regardless. No need for iterative counting.
10. Majority Guaranteed for Single-member Seats No By using priority voting for MLAs (like mini-lists), getting a majority is not an issue Not for the majority of seats (German system, though, still uses FPTP for local constituencies) Yes, guaranteed, for local seats. Not an issue for cluster seats.
[table that accompanied the above did not carry through to WordPress]
[post-script: with the defeat of Ontario selected alternative to first-past-the-post (FPTP), as well as similar defeats since in other Canadian provinces, I realize that my exotic system will never get enough support to garner a chance in the sweepstakes for an alternative. Instead, I designed an even simpler system for introducing alternatives: let a couple of popular ones, say STV and Party Lists, co-exists with FPTP. This system would allow each voter to select which system’s ballot to use to vote, this choice would, in the succeeding election be used in the aggregate to allocate seats for the three systems. For instance, if each of the alternatives were started off with 10% of the seats for the first election, and they attracted a larger share of ballots (say, 15% and 20%), then in the next election, the number of seats each would get would increase accordingly. I presented this to a national workshop of Fair Vote Canada but have seen no interest in pursuing it since.]