Talk:Duverger's law

Obsoletion of the law
I'm proposing this law be classified as obsolete. I know this is a big step and probably not going to pass, but the data does not suggest this "law" exists anywhere expect for the United States, and I don't believe case studies have ever become laws. Just like the miasmatic theory of disease's Wikipedia page references it as obsolete, I think this theory is pretty much equally debunked and the article should reflect such. -- Adamopoulos 0:17:14, 7 Feb 2021 (EST).

Two-party system
Is it really necessary to link to two-party system twice in this article? This has been added and reverted quite a few times. -- RobLa 03:41, 1 Mar 2004 (UTC)

Other methods
One thing I've always wondered is how Duverger's law applies in the case of other single-winner methods. Did Duverger just make a dichotomy between first-past-the-post and proportional representation? -- Dissident 19:04, 1 Mar 2004 (UTC)

Counterexamples
"There are numerous counterexamples: Malta has a single transferable vote system and what seems to be stable two-party politics;"

This is a logical fallacy. "...first-past-the-post election system naturally leads to a two-party system..." is not the same as "two party system requires first-past-the-post". The fact that Malta has two parties doesn't contradict the 'law', at least as it is explained in the current article.

Unfortunately, I don't know whether the mistake is in the statement of the principal of the description of the Maltese political system, so I'll have to leave this for someone else to fix. --Rory &#9786; 14:20, Mar 22, 2004 (UTC)


 * I agree, and further assert that it isn't merely a logical fallacy, but in fact, an inappropriate example. There are many (including myself) that have come to believe that single transferable vote also leads to a two-party system. -- RobLa 08:13, 23 Mar 2004 (UTC)


 * I won't press the point in the article - moving Malta down works for me - but I read Duverger's law as saying that first-past-the-post is more likely to produce two-party results than other voting systems are; it doesn't say FPTP inevitably leads to two parties. So there are two types of exceptions (if you can have exceptions to a principle): FPTP with several parties, and two party systems with a more proportional voting system. Incidentally, Australia only uses STV for the Senate, but AV for the more important House of Representatives. --Henrygb 23:06, 23 Mar 2004 (UTC)

Scotland
Scotland is not an example, in any respect, to a refutation of Duverger's Law. The law states that FPTP leads to a two party system. Scotland's Parliament is eight years old. The next example is the United Kingdom, which does not use FPTP, but mixed representation. Canada has a very regulated, non-partisan system.

It looks like every counter example is bogus. India, for example, has regional parties, but is there any constituency with more than two main parties? If not, Duverger's Law would appear to apply.

There also ought to be an explanation why Duverger's Law works. Namely, strategic voting in a unidimensional sytem under FPTP results in a simple, rational aggregate "hedging" strategy. -- JoshNarins 20:17, 1 November 2005 (UTC)


 * Duverger's Law is explained partially by strategic voting but also by the incentive for parties to improve their changes of winning a seat through merging or coalitions (and conversely, by providing a disincentive for an internal faction to split off). See, e.g., "Structure and Behaviour: Extending Duverger's Law to the Japanese Case" by Steven R. Reed [British Journal of Political Science, Vol. 20, No. 3. (Jul., 1990), pp. 335-356.] This article also generalizes Duverger's law to multimember districts: for n seats there will be n+1 candidates at equilibrium. - ChrisKennedy(talk) 03:30, 3 December 2006 (UTC)


 * I don't know about the other examples, but the UK does use FPTP for their lower house (and the upper house is not elected at all). Though this may change in 2011... --Roentgenium111 (talk) 17:12, 20 July 2010 (UTC)

Jenkins Committee
Any anglo-american attempt on analyzing, describing FPTP, single-seat-district, winner-takes-all, two-party-systems seems to have to wait until UK finally does their promised Jenkins_Committee reform of their domestic election system.

just check out citizens assembly, the struggles in Canada.

Note, if an election-system has two elected institutions, for example "house" and "senate", "lower" and "upper" parliament, the result becomes more complex. Adding a system for electing a president makes it more complicated. However, the basic mechanism is "at the local level", single-seat, FPTP, winner-takes-all mechanisms.

Canada
In Canada I think that the regional politics, which are mentioned in passing in this article, have a much larger influence. At the moment each political party has a regional bias, most especially the Bloc Québécois which is only in Quebec, but even the NDP, Liberals, and Conservative parties have regional biases. I suspect this is true in many other countries that have strongly differentiated regional cultures. Sbwoodside 19:50, 12 March 2006 (UTC)

With the election in 2010, and the demisce of the BQ, regional politics no longer plays a role. — Preceding unsigned comment added by 76.28.101.152 (talk) 02:25, 31 August 2011 (UTC)

Finland
Finland is not a counterexample, they have a proportional representation system. http://en.wikipedia.org/wiki/Parliament_of_Finland#Parliamentary_elections

Graph
I had an idea for a visualization of this, showing the proportion of votes that went to each party for each US presidential election, and showing how two parties take the majority of the vote, and those two parties only change occasionally, as described in the article about the Whig party. (Inspired by these graphs, though an implementation of this would be percentage-based, and so constant-width.) But the actual data is a little more complex, with multiple people running for the same party, the original elections with no parties, etc. Any ideas? — Omegatron 01:43, 2 September 2007 (UTC)
 * In my opinion stacked bar graphs make comparisons across years difficult. The linked graphic makes it hard to tell how Republican votes compare to Democratic votes because the independent candidates are always listed on the top. I would prefer a non-stacked bar graph with parties listed next to one another and grouped by election. - ChrisKennedy(talk) 02:40, 2 September 2007 (UTC)
 * Additional chart suggestions (based on Edward Tufte's writings): remove the bar border, remove the two decimal places from the y-axis labels, change the y-axis interval from 10% to 20%, remove the ticks from the x-and y-axis labels, make the axis lines a light gray (50% or so) to reduce their prominence (dashed might be good too), and remove the border from the legend. The color difference between Libertarian and Green party is also negligible - I'd make Libertarian orange so that it stands out more. - ChrisKennedy(talk) 02:47, 2 September 2007 (UTC)

You're reading way too much into that image. That's just a very crude example of some actual data using default bar graph settings, but isn't even complete (only shows the top 3 or 4 parties) and nothing like I was actually imagining. Originally I was imagining something like this, that emphasizes popularity trends over time:



But for the elections, it would be a constant width because it would be shown as percentages, and the amount of votes for each party would be shown shifting with each year, showing how two parties share the majority of the vote for long periods of time, and then upheavals shake things up and then it settles out so that a different two parties share the majority of the vote, but still two parties.

So that's what I'd like to show with the visualization, but the real life election data doesn't exactly fit into a graph like that. it only exists on election years, not continuously; the parties change names and split up and such over time, multiple people run under the same party, etc. So this is just an idea. I am sure there is a way to show a nice clear visualization of actual data, but I'm not sure exactly how to show it. — Omegatron 03:05, 7 September 2007 (UTC)


 * Well, you could do absolute population, which would show (mostly) the trend in population increasing, but also give a sense which elections got a lot of people out to vote, and which party that might have helped. —Preceding unsigned comment added by JoshNarins (talk • contribs) 23:30, 10 February 2008 (UTC)

The term "law"
Why is it called a law? Shouldn't it be called Duverger's Theory? —Preceding unsigned comment added by JoshNarins (talk • contribs) 00:54, 9 April 2008 (UTC)

Dubious counterexamples removed
I removed the following long unreferenced and disputed (see above threads) "counterexamples" from the article:


 * India, the world's largest democracy, has multiple regional parties, especially the Communist Party of India (Marxist) that has been strongly entrenched in three states - West Bengal, Kerala, and Tripura for nearly three decades. It may be argued that the NDA (National Democratic Alliance) and the UPA (United Progressive Alliance) multiparty coalitions serve as cognates of the two parties of Duverger's law.
 * Scotland has had until recently SMDP and similar systems, but has seen the development of several significant competing political parties.
 * In Canada, the New Democratic Party and its predecessor, the Co-operative Commonwealth Federation, have had a constant presence in Parliament since the CCF's first election in 1935. At least four - and sometimes five - political parties have been represented in the Canadian parliament at any given time since the 1993 election. In addition, the now-defunct Social Credit Party of Canada also maintained itself in Parliament nearly consistently from 1935 to 1979, often resulting in Parliaments with four national parties represented. Most successful third and fourth parties have been regionally based, however, such as the Canadian Alliance/Reform Party and the Bloc Québécois.  The Bloc only runs candidates in Quebec, where competition is primarily between the Bloc (and its provinical counterpart, the Parti Québécois) and the Liberal Party, and the Conservative Party holds third-party status.

Please don't re-add them without giving references. (BTW, the article could also do with a list of examples!) --Roentgenium111 (talk) 17:10, 20 July 2010 (UTC)

please don't change the hilariously huge glaring error
hint: the plural nature of wikipedia keeps me in stitches — Preceding unsigned comment added by 75.66.100.92 (talk) 22:59, 25 May 2011 (UTC)

The M + 1 Equilibrium
The current lede refers to something called "The M + 1 Equilibrium", which sounds like an episode of Star Trek. It seems to have been added two years ago with no further explanation anywhere in the article or any obvious links to other articles explaining it.

Can somebody with knowledge of the subject please add an explanation, a link to another article, or just delete it if it's not relevant? - IMSoP (talk) 16:09, 15 April 2019 (UTC)


 * I've deleted this reference. Dan Bloch (talk) 23:33, 15 July 2019 (UTC)

Britain isn't really a counterexample
The article does not mention another anomaly of British national politics that accounts for its multiparty Parliament (although since either Labour or Tory has (2010–15 excepted) had an absolute majority for most of the last century, the UK has been a two-party democracy for all practical purposes): Northern Ireland.

Neither of Britain's two major political parties contests seats there (the Tories' recent use of the DUP as a semi-coalition partner notwithstanding), thus leaving space for that country's local parties, one of which includes Sinn Fein, whose members make a point of not taking their seats.

In fact, it occurs to me that if you factor out Scotland and Wales as well, since I don't foresee anyone from Plaid Cymru or the SNP standing for election outside those countries, England would be your better test case for this, and frankly it seems like it mostly proves the point: the vast majority of seats at the last election were won by either Labour or Conservative, with the Lib Dems and the one Green largely irrelevant politically (at least at the moment). Daniel Case (talk) 18:49, 20 January 2021 (UTC)

UK and India Examples
Not sure I'd agree with the dubious tag there: the last time someone other than the Conservatives or Labour has formed a government in the UK was 1910 (though it lasted until 1918), while literally no one other than the INC and the BJP has formed a government in India (unless you count the BJP's predecessor as a separate party). It should also probably be pointed out that most individual districts have only one or two competitive parties, and the majority of third-party representatives are from regional parties.98.250.177.207 (talk) 13:16, 18 October 2021 (UTC)

Australia has also been dominated by a 2 party system since federation
Add Australia to that line please 124.150.86.247 (talk) 13:52, 5 January 2022 (UTC)

Mathematical Proof
Dear User:HudecEmil, thank you for improving the section, esp. adding a link to a description of Palfrey's paper. As far as i understand, there are even more premises in the paper: At least, "rational strategic voting behavior" is one.

One more notice: using Google Scholar, i have found https://scholar.google.com/scholar?as_sdt=0%2C5&q=A+Mathematical+Proof+of+Duverger%27s+Law&btnG= and https://authors.library.caltech.edu/83112/1/sswp688%20-%20published.pdf which appears to be a photo-scan of the paper. I also found https://authors.library.caltech.edu/81155/1/sswp688.pdf which is another reproduction of the same paper(?) --Himbeerbläuling (talk) 12:03, 18 August 2022 (UTC)


 * Yes, maximizing individual utility/rational choice is another assumption. HudecEmil (talk) 14:42, 18 August 2022 (UTC)

September 2023 cleanup
Noting that I've been (and will continue) removing a number of unsourced (or innappropriately sourced) examples and significantly copyedited for clarity, grouping related topics, reducing redundancy and adding reliable sourcing.

Am posting a user-created (@quinnnnnnby) graph here that was removed from the article in case others think it merits reinclusion. Was not very clear what the graph illustrated based on its description or the sources used to generate it. Superb Owl (talk) 22:36, 25 September 2023 (UTC)