How many losers do you need?

For a project I am currently working on, regarding district-level vote fragmentation, I am toting up the number of losing parties (with any vote share) in electoral districts in various countries.

Here are a few cases that I have completed. This first list of countries gives the number of losing parties, averaging across districts and (usually) more than one election per country.

Albania, 36.2
Australia, 2.5
Barbados, 1.4
Canada, 4.2
Czech Republic, 15.5
Israel, 11.6
New Zealand, 2.3
Spain, 14.7
UK, 4.6

The figure for New Zealand refers to several pre-MMP elections.

This is a small set of countries, and I am currently expanding it. But what really jumps out here is how many partisan choices Albanian voters have. This refers to the 2009 and 2013 pure-PR elections only. The mean district in these Albanian elections has a magnitude of 12, a mean number of winning parties of only 3… and thirty-six losing parties!

Some of the other PR cases also have a lot of losers, but none like Albania.

It is interesting to see how many losing parties are typical of the two big FPTP countries, Canada and UK, with both averaging more than four per district.

Another interesting summary statistic is what vote share the losing parties average. In Albania it is 1.05% per party. That sums to a LOT of wasted votes! Spain, too, has a high average vote share for losing parties for a PR system, at 2.8%. In the Czech Republic it is 1.2% and in Israel (with nationwide PR, rather than districted) it is only 0.4%.

Thus we should also consider the average percentage of the first losing party in a district:

Albania, 4.6
Australia, 35.8
Barbados, 39.1
Canada, 28.5
Czech Republic, 4.7
Israel, 0.94
New Zealand, 32.9
Spain, 6.4
UK, 28.6

Of course, in FPTP systems the average shares for first losers and thus for all losers are especially high, but the larger losing parties in any given district tend to win some seats elsewhere.* In the PR cases, a lot of these losers are not winning anywhere (or might win in just one to two high-magnitude districts); they are just small parties that have no chance at all. I wonder what it is about Albanian party law, or other features of the country’s politics, that contribute to so many micro-parties running.

_______
* The third party (second loser) in the FPTP cases of Canada, UK and (formerly) New Zealand averages 14-15%. The average winner in Canada and the UK, by the way, is at almost precisely 50% (in NZ it was 52.8%).

Most of the raw data from which I calculated the above numbers come from the Constituency Level Electoral Archive, although I am augmenting it from various other sources. The list of countries shown here is the subset on which I am currently working: parliamentary democracies with “simple” electoral systems by Taagepera (2007) criteria, meaning no second round or upper tier. Thanks to Cory Belden for her research assistance; some day she will get a much better acknowledgement than one on a blog post…

Large countries do not use closed lists

I mentioned earlier that Indonesia was the largest democracy to use proportional representation (PR), and that it uses open lists, as does the fourth largest democracy, Brazil. India, the largest, uses First Past the Post (FPTP, or plurality), as does the USA, the second largest democracy. By my count, all the world’s largest democratic countries use FPTP, OLPR, or mixed-member majoritarian (MMM) until we get to Germany (16th largest country, democratic or otherwise), which is mixed-member proportional (MMP). One does not encounter the first purely closed-list system till Italy (bonus-adjusted PR), the 23rd largest country. And, of course, Italy was OLPR till 1994 and then MMM till 2006. So the first country down the list to have been entirely closed lists for at least twenty years (upon its vote, due later this year) would be South Africa. Along with Spain (#28), these are the only closed-list systems among the world’s top 30. After that, Argentina (#32) and Mozambique (#50) are the only other closed-list systems in the top fifty.* It seems we have an almost-iron law that big countries do not use closed lists.

_______
* Yes, I skipped right over one very big country, Russia, because I don’t consider it a democracy, and in any case it is soon to abandon closed lists for a return to its former MMM system.

April 2014: lots of elections

April is a good month for election-wathchers.

Today Hungary votes for the first time since the constitutional and electoral reforms imposed by Fidesz following the two-thirds majority granted it by the country’s mixed-member very-unproportional electoral system.

Monday is the general provincial (or “national”?) election in Quebec. The final ThreeHundredEight.com projection shows the Liberals most likely will win a majority, although the estimated range includes the possibility of a minority government. The Parti Quebecois (PQ) can be said confidently to be in second place, according to the projection, while the Coalition Avenir Quebec (CAQ) has closed in on the PQ in votes. The PQ’s lead over the the third place CAQ is now only around five percentage points, but it is projected to get only one fifth as many seats as the PQ. With projected votes breakdown of 39.0-27.6-22.7, but seats of 69-45-9, it is quite a “non-Duvergerian” result in the votes, but fairly Duvergerian in seats. The mechanical effect of FPTP will make the legislative party system much more two-party than the votes, by significantly over-representing both of the larger parties. The Liberals and PQ have been targeting ridings (districts) held by the CAQ, but it may not be working. There are indications that the CAQ is the party with the momentum, though getting into second place in votes is both unlikely and would not get them out of third place in seats. The fourth party, Quebec Solidaire, appears to have two safe seats despite only 8.4% of the projected votes; it is great to be small party under FPTP if your supporters conveniently concentrate their places of residence! (There were two earlier F&V posts about this election campaign: on by JD on the pro-PQ bias, and one by me on the surprising mid-campaign swing against the PQ.)

Also coming right up are general elections in two of the world’s largest democracies. Indonesia’s legislative elections are this Wednesday, 9 April. Indonesia uses open-list PR,* and is the world’s largest democracy to use any kind of PR (just ahead of Brazil, which also uses open lists, and votes later this year). Indonesia uses a counterhoneymmon electoral cycle, with the presidential election coming on 9 July.

Also this week marks the beginning of the biggest voting exercise of them all, India. The general election will take place in nine phases,** starting 7 April and ending 12 May (see map with schedule). Results will be declared in mid-May. Indications from polls are that the opposition Bharatiya Janata Party and its National Democratic Alliance (NDA) of various state-based parties will win, and that it will be the worst outcome ever for the Indian National Congress and its ruling United Progressive Alliance. The NDA likely will be short of a majority, however, and will need outside support from other state-based parties that have joined neither alliance for the election. BJP leader Narendra Modi, current Gujarat state Chief Minister, is the likely Prime Minister. However, if the NDA is short of a majority, it can’t be ruled out that the support parties could demand a different PM, given what a lightning rod Modi is for communal tension.

________
* And SNTV in the largely powerless second chamber, also being elected Wednesday.

** Or “stages”. Please don’t call them rounds.

Western Australia Senate re-run and the Joint Committee on Electoral Matters

The following post was originally a comment by Chris Curtis.*

The rerun of the WA Senate election is to occur on Saturday, 5 April. General opinion from the “experts” is that people are horrifed that micro-party candidates (voted for by 23.5 per cent of the national population in 2013) actually won seats and that people will realise their “mistake” and flock back to the “proper” parties. They predict that the result will be two or three Liberals, two Labor, one Green and, as an outside chance, maybe one other. I say the 23.5 per cent of the population are delighted that their vote actually elected someone and that they will do the same thing again, though I do not predict how the preference flows will work in WA.

The Joint Committee on Electoral matters has started taking submissions and conducting hearings into the 2013 election, the one that led to so many complaints because the “wrong” sort of people managed to get into the Senate (with more initial votes than most of the “right” sort of people had won from without complaint in every Senate election since 1949).

Submissions can be found at Joint Committee on Electoral Matters Submissions. Mine has yet to receive approval and thus parliamentary privilege (not that it is the sort that really needs it). Submissions particularly worth reading are those from Malcolm Mackerras (no. 7) and Michael Maley (No. 19). One worth reading get the flavour of those who want to turn the single transferable vote into a party list system to shut the candidates of micro-parties out – the effect of which is to favour the Greens – is that from George Williams (No. 23). You can follow the links at the top of the page to get transcripts of public hearings. Malcolm Mackerras appeared on 7/2/2014 and George Williams on 13/3/2014. (BTW, they are Australian date orders, not US ones: as with “tomato”, you say 12/31; we say 31/12.)

__________

* which somehow got intercepted by the spam filter. I was so close to clicking “empty spam” when I noticed this was most certainly not spam!

Colombian president’s reform proposals

The Colombian president, Juan Manuel Santos, currently seeking reelection, and his new running mate, Germán Vargas Lleras, have proposed a package of institutional reforms.

The proposals are:

    Abolish reelection eligibility of the president;
    Extend the term of the president from 4 to 6 years;
    Unify the terms of the president with those of regional and municipal elected executives (governors and mayors);
    Abolish the national district for the Senate;
    Abolish the preference vote.

Apparently congressional terms, currently four years, would not be changed. While Colombia does not use concurrent elections, the terms of office for both congress and president are both four years. (At this moment we are in the period in between the congressional elections, held in March, and the presidential elections, the first round of which will be 25 May.)

In general, I do not like different term lengths for president and assembly in pure presidential systems. When combined with “permissive” rules such as relatively high-magnitude PR and the election of the presidency by majority runoff, different electoral cycles for president and congress promote too much fragmentation.* The last thing Colombian politics needs is more encouragement to fragmentation.

Abolition of the national Senate district would be a movement in a less permissive direction, which might by itself be desirable, but at the cost of removing the current beneficial effect of allowing for minority political views to aggregate support across regions. (Aside from major urban centers, most house district magnitudes are in the 2-7 range.) Abolition of the preference vote would probably encourage more splitting of some existing parties that manage to cooperate only because various candidates, and the factions they belong to, can cultivate votes independently within lists, while still pooling for their common seat-maximization. Again, Colombia hardly needs devices to encourage fragmentation.

These proposals would be, in my assessment, retrogressions. (In case that was not clear by now.)

___________
* Under current rules, governors and mayors (and those levels’ respective legislative bodies) are elected to four-year terms, but in odd-numbered years (2007, 2011, etc.). The linked article mentions the possibility of extending the terms of those regional and local executives elected in 2011 until 2018, when the next presidential election is due.

Quebec pro-PQ bias and electoral reform

By JD Mussel

The Montreal Gazette has reported about the Mouvement Démocratie Nouvelle (MDN), which recently launched its campaign for proportional representation in Quebec, specifically MMP.

The article mentions a number of factors that may be at involved in creating the apparent pro-Parti Quebecois bias in First-Past-the-Post in Quebec. Firstly, ridings in Montreal traditionally favour the Liberals, where their votes are very concentrated, to the party’s disadvantage. Secondly, the article highlights malapportionment in the Province, which may be one of Canada’s worst. The overrepresented ridings are generally more likely to be rural and francophone, while the underrepresented ones tend to be more urban and are likely to have a larger anglophone population, as illustrated here.

There have been three plurality reversals in Quebec since the rise of the Union Nationale (1936), all of which occurred in elections where the Liberals had received a plurality of the vote.

The article also included projections, produced by Wilfred Day of what the 2012 result would have been under pure PR and MMP. I am very curious as to the exact model used for the latter projection!

Meanwhile, the MDN’s website (in French) is worth a browse, in particular its historic overview of all Quebec elections since Confederation, showing vote shares vs seat shares as well as some historic background.

Quebec election, 2014

Quebec’s general election will be 7 April. From my cursory reading of the news prior to the election call, I had the impression that a majority for the Parti Quebecois (PQ) was all but in the bag. (The PQ currently heads a minority government.) However, sometimes funny things happen during campaigns.

Looking at the projections at ThreeHundredEight.com, which are based on aggregating and weighting polls, one can see that the Liberals are now ahead of the PQ in the vote projection. As of today, they are even ahead outside the projection tool’s estimated confidence intervals*, which range 41%-47% for Liberals and 32%-36% for the PQ.

Seat projections, however, remain closer: 61-76 for Liberals, 46-59 for PQ. If the race tightens even a little bit, the prospect of a plurality reversal becomes real, given the near overlap in project ranges for seats despite the existing minimum projected gap of five points in votes. If we convert the projections into advantage ratios (%s/%v)**, we get a range for the Liberals of 1.18 to 1.28, but for the PQ of 1.16 to 1.31. We can see that either party would benefit from the over-representation expected from a plurality electoral system, but the PQ benefits slightly more–from the model’s projections–when it is at the higher end of its vote range, even though its current maximum projection would not give it a plurality of the vote. Extrapolating from these figures, the PQ might be able to win a majority of the seats on only around 38% of the vote and with the Liberals still slightly ahead.

A reversal–and not the first in the province–happened as recently as 1998, when the Liberal party had 43.55% of the vote and 48 seats, while the PQ had 42.87% but 76 seats. Yes, 76, for a really large majority despite losing the province-wide vote.

I think I will start paying more attention to the Quebec campaign now.

_______
* Taking the confidence interval to be the range from “low” to “high” rater than minimum-maximum.

** Using low-end or high-end projections for both seats and votes in all cases.