Friday, January 8, 2016

Open-list mixed member proportional models: The Bavarian example

Canadians can consider several models of proportional representation, including Bavaria's.

Scotland’s regional mixed-member model

Scotland’s regional mixed-member elections are simple and practical: you elect a local MP plus regional MPs.

Every vote counts. If your vote didn’t count to elect a local MP, it will help elect a regional MP from your local region, to top-up the local results.

You have two votes, a local vote for a local MP, and a regional vote. Your local region has nine local MPs and seven regional MPs.

To elect an MP in a 16-MP region, your party will need about 5% of the regional votes. This is unlike the “pure proportional representation” system of the Netherlands, where they have no local MPs, and no threshold to win seats. In their last election they got 11 parties in Parliament. Voters for an Animal Rights party cast 1.9% of the votes and elected two MPs, while four other parties got fewer than 5% of the votes.

In Scotland you can vote for a local candidate of a different party than the party you want to see in government. New Zealand has a similar two-vote system, where about 32% of voters split their ballots that way. This makes it easier for local MPs to get the support of people of all political stripes. They can earn support for their constituency-representation credentials, not just for their party. This boosts the kind of support MPs bring with them into the House of Commons, thus strengthening their independence.

Improved Scottish model

But Canadians would want one big improvement on the Scottish model: no closed lists. You can vote for the regional candidate of your party that you like best.

The Law Commission of Canada recommended a model that, like Sweden’s, gives voters a choice: you can vote for a regional candidate personally, or for the party’s regional list. Watch this six-minute video.

No one knows how many voters would use the list option: with a similar choice in Brazil only about 10% do.
Yet many PR-sceptics will scream “backroom-dominated party list.” And they will ask a valid question: how many personal votes does it take to move a candidate up the list? The Law Commission left the “personal threshold” detail to be decided.

Bavaria’s model

German federal elections use a mixed member system like Scotland’s. But since 1949 the German province of Bavaria has used an open-list variation, with no party-list option at all. You vote for a local candidate in your riding, and for a regional candidate in your region. (They have seven regions.)

Even simpler than the Law Commission model: no personal threshold. The ranking is done entirely by the voters. As usual, you have two votes: one for a local MP, one for a regional MP. The list order has no legal weight.

Does this process result in the list order prevailing anyway? In 2003, preferences led to the defeat of 25 advantageously positioned regional candidates, to the benefit of 25 others who were less so, out of a total of 88 elected regional members. Just look at the 1998 charts for Munich on pages 55 and 56 of Prof. Massicotte’s working document.

Note the spectacular fall of Minister Ursula Männle from the 3rd place assigned by her party to the 42nd decided by the voters’ individual preferences. This ensured her defeat.

Prof. Massicotte concludes “As a result of preferences, electoral rivalry between parties has a new level of competition between candidates of the same party within each region. . . . The Bavarian experience shows that voters take the possibility offered them very seriously . . . and that the preference system has a major impact on which list candidates get elected.”

Although every local candidate is also a regional candidate, some regional candidates are “regional-only.” This is necessary in case a party sweeps a region so massively that it elects some regional MPs on top of all the local MPs.

This also let the party members nominate a slate better balanced by gender and minorities, creating more voter choice. The Law Commission of Canada wanted to “support voters who decide to trust a political party’s choice of list candidates by allowing them to vote for the party slate . . . to promote successful election of women.” But Bavaria’s model also does this. And the position of women in Canada’s Parliament has improved since the 2004 Law Commission Report.

How does the Bavarian model work?

Is the regional ranking done only by a small group of voters who didn’t vote for the star candidate? No risk of that in Bavaria. The ranking is by the total number of votes cast for a candidate in the local race, added to the total number of regional votes cast for that candidate in all the other seats in the region (but not in the seat where he or she ran locally -- that would give double weight to those voters, so the regional list in a riding excludes the local candidates). Note that every local candidate is also a regional candidate.

The party vote is also a total of the two votes. Unlike most MMP models, Bavaria counts your local vote as a vote for a party, just as it counts your regional vote. This reduces the scope for the “Berlusconi trick” that ruined Italy’s system (running local candidates under a different party name than the regional candidates, twin parties, one a clone of the other).

Some people like the “best-runners-up” model used in the German province of Baden-Wurttemberg: no regional candidates at all, no lists. The regional MPs elected to top up the local results are the local candidates who, in that region, got the most votes without winning the local seat.

The effect of the Bavarian model is a hybrid with the best-runner-up model. You have regional candidates, but those who did well locally without winning the local seat have an advantage for the regional seat.

In the 2013 Bavarian election, of the 90 regional deputies, 13 were regional-only candidates who won election from regional votes alone, despite not receiving any local votes. Four regional candidates standing at the head of the regional list were not ranked first in the final count. (The regional candidates included all 977 local candidates -- an average of 11 in each riding -- plus 785 regional-only candidates, an average of  10 candidates from each party in each region.) 

Bavaria’s 5% threshold rule applies to parties and “organized groups” such as the “Free Voters” movement which denies it is a party, but has met the 5% threshold in the last two elections. In 2008 it became the third party, with more seats than the Greens, repeated in 2013.

The trend to having two votes, now universal in Germany, began in Bavaria in 1949. They wanted to let voters not only elect a local MP, but also express a preference for one of the regional candidates of the party they were voting for.

One reason Bavaria likes this system is that it tends to be as much of a one-party province as Alberta used to be, so it needs competition within its conservative party. The CSU governed in a liberal-conservative coalition in 2008-2013, but otherwise has governed alone since 1962.

How big are the regions?

Bavaria’s stable model has used seven regions since 1950, with an average of 23 MPs each. Some are as small as 17 MPPs (9 local, 8 regional), while the largest region (Munich) has 58 MPPs.

“I support reforms to add elements of proportional representation that also ensure that Members of Parliament remain directly accountable to their constituents” said Dominic LeBlanc. “Do we want to keep that bond between a specific MP and a specific group of Canadians?" asked Justin Trudeau.  “Canadians voted for real change, for a government that puts their needs ahead of the party's. A government that celebrates our diversity, and not takes a divisive approach” said Minister Maryam Monsef.

The regions have to be small and local enough to make that true, while still large enough to be at least moderately proportional.
Polls show over 70% of Canadians support proportional representation.

The mixed member proportional system used in Scotland and Wales, and recommended for Canada by the Law Commission of Canada, has all MPs tied to ridings or local regions; MPs would be elected in each riding but they'd be augmented by regional MPs elected from candidates in a local region. Justin Trudeau has never ruled that out at all, as long as the regions aren't too large.

Is Bavaria’s regional ballot too large and complex? In 2003 the number of spoiled or rejected votes was 1.2% for the local vote, 1.9% for the regional vote. Munich, with 58 MPPs, had a ballot with 469 names. No wonder some were rejected.

I suggest no regions larger than 18 MPs, preferably about 12 MPs on average.

More details of Bavaria’s model

Bavaria used to use the Scottish "highest average" (D'Hondt quota) but in 1990 this was declared unconstitutional, so they switched to "largest remainder" (Hare). Wouldn't it be lovely to live in a country where lack of full proportionality is unconstitutional?

Although they use a 5% threshold for representation, they give state funding to any party that reaches 1% of the regional votes.

Bavaria has almost half their MPs elected from the regions. Canada would use a higher ratio of local MPs.

Bavaria originally imposed a 10% threshold at the regional level, but in 1973 opted for a 5% threshold applicable to the whole province.

Bavaria’s 5% threshold rule has an unusual feature: a party needs to pass the 5% threshold province-wide to win even a local seat; no independents. Canada would never copy that feature.

Unlike German federal elections, the ballot had no circle for the voter to mark after the party name. If voters nevertheless marked the party name alone on the regional ballot, it was counted as a valid vote for the party, but had no effect on the ranking of the regional candidates. Only 1% of voters did this in 2013.

Update: the 2018 election:

How well did Bavaria’s open-list MMP system work in 2018? They elected 205 MLAs: 91 directly elected, 114 regional top-ups elected from open regional lists. The governing CSU lost their majority, and formed a coalition government with the centre-right “Free Voters.” This was not simple, since the “Free Voters” had never whipped their members votes, but now had to.

2018 results:
Christian Social Union: 85 MLAs, all local (was 101 in 2013)
Greens: 38, 6 local, 32 regional (was 17)
Free Voters: 27, all regional (was 17)
AfD: 22, all regional (was 0 – new party)
Social Democrats: 22, all regional (was 42)
Liberals: 11, all regional (had been 0, below 5% threshold)

Of the 114 list seats, 31 were elected thanks to voters moving them up the list, while 83 would have been elected with closed lists. 

Did the list-topper (first on the list) always get elected? Almost. In the region of Lower Bavaria, the liberal FDP elected only 1 MLA, and he had been second on their regional list.

One of the more unusual new Green MLAs is Paul Knoblach, a 12th-generation Bavarian farmer. Knoblach was among the CSU’s longtime supporters and had volunteered for its campaigns. But in 2018, at age 64, he ran for the legislature for the first time. And he did it as a Green. He was number 12 on the Green regional list for Lower Franconia, but he got enough attention that the voters moved him up to 3rd place and into the legislature.

Did the open lists hurt women? I did not check most results, but the SPD zippers their lists, and I noticed in Upper Palatinate the SPD elected 2 MLAs, list numbers 1 and 3 (two women). Conversely, in Middle Franconia the SPD elected 4 MLAs: 1, 2, 3, and 5 (three men). 

All CSU members were elected directly, no list seats.

Here are the details of three of the seven regions:

In Upper Palatinate:
3 Free Voter list seats: 1, 2, 4
2 AfD list seats: 1, 3
2 Green list seats: 1, 2
2 SPD list seats: 1, 3 (two women)
1 FDP list seats: 1 

In Lower Franconia:
3 Green seats: 1 direct (was #2 on list), 2 list seats (1 and 12)
2 AfD seats: 1, 2
2 SPD list seats: 1, 2
2 Free Voter list seats: 1, 5 (5: Anna Stolz, lawyer, Mayor of the City of Arnstein; had been elected Mayor in 2014 as the joint candidate of the Greens, SPD, and Free Voters; the local Greens said they were very proud of her as Mayor.)
1 FDP seat: 1

In Upper Bavaria (which includes Munich):
17 Green seats: 5 elected directly (they had been 1, 2, 5, 10, and 38 on the list) and 12 elected from the list: (3, 4, 6, 7, 8, 9, 12, 13, 16, 22, 25, 36)
8 Free Voter list seats: 1, 2, 3, 4, 5, 6, 9, 12 (12: Hans Friedl, “a socially ecologically liberal voice, an immigration law based on the Canadian model, no privatization of the drinking water supply, a clear rejection of the privatization of motorways” so the voters moved him up to #8)
7 SPD list seats: 1, 2, 3, 4, 11, 14, 23
6 AfD list seats: 1, 2, 3, 5, 9, 22
5 FDP list seats: 1, 2, 7, 13, 16

Canadian adaptations

As well as allowing independents, Canada could need other changes to the Bavarian model.

We would need smaller regions than they use.
They use a 5% threshold province-wide for a party’s voters to elect a Regional MP. The effective threshold for smaller regions would already be at least 5%. But we might want some guarantee that a regional micro-party might not run in some region of Ontario or Quebec and elect a single regional MP with only 1% of the vote in the province, if the public expects this would not happen. A threshold of 4% or 5% applied province-by-province would prevent this. In Europe, the threshold is generally 4% or 5%, and New Zealand has been debating which of these numbers to use.
Take the Green Party in Quebec. With PR, assume their vote doubles, to 4.4%. If Montérégie is a 12-MP region (excluding a couple of ridings in Montérégie-est), and the Green vote doubled, Green voters there would have cast enough votes to elect their star Quebec candidate JiCi Lauzon. The same in 12-MP Montréal-est (deputy leader Daniel Green), in 11-MP Estrie—Mauricie—Centre-du-Q—Montérégie-est (Corina Bastiani?) and someone in 13-MP Laval—Laurentides—Lanaudière. Four MPs, just as they would deserve. A threshold of 4% lets them be elected, 5% does not.

(Note: updated after 2018 election)


Gregg Hill said...

Very informative. I'm surprised Wilf that in your reference to the LCC report you didn't mention Belgium which is referred to in an end note rather than the main text itself. It seems to me if we (more likely the parties) want flexible lists rather than open lists like Bavaria has the Belgian (as well as other) model of a non-preferential single transferable vote mechanism based on a quota formula (in Belgium candidate & party votes divided by one more than the number of seats allocated to the party). That would definitely be more flexible than Sweden's system. The LCC report incidentally had the interesting notion of an open list that unlike in Bavaria candidate votes and party votes would be strictly separated with the former used strictly to rank candidates and the latter used strictly to allocate seats to the party. Interesting that the world's most widely used PR formula would be declared unconstitutional. But they would have been better off to follow New Zealand's lead and adopted pure Ste Lague. You know why...

Boyd M L Reimer said...

Thank you very much for this.