Political science has very few ‘laws’, perhaps explaining why the European discipline has so stubbornly clung onto its most famous product – Maurice Duverger’s claim that countries using ‘majoritarian’ voting systems will always have two party politics. This ‘Law’ has underpinned numerous completely ineffective efforts by European politicians in PR systems to create party consolidation by changing their voting laws. With European political scientists meeting this week in Berlin, Patrick Dunleavy explains why such ‘reforms’ have had zero success. Modern theory and better evidence now show that the alleged ‘Law’ has lost all credibility in FPTP countries – and works only in the USA.
All elections (indeed all competitions) have a horse race element in which most attention from the media, elites and voters themselves tends to focus on the top two contenders, as much in proportional representations contests as in British-style voting systems (often called ‘first-past-the-post or FPTP systems). Yet the most famous ‘law’ in political science, coined by the Frenchman Maurice Duverger in the early 1950s, holds that the operations of first-past-the-post voting system with single-member districts directly and strongly tends to cause two party politics.
So the USA, UK, India, Canada and other Westminster systems should all be two-party polities in this perspective, if not at the aggregate national level, at least at the level of contests in each local constituency. Modernized and made more operational by Gary Cox in an influential 1997 book, Making Votes Count, Duverger’s Law still has so many thousands of cites in the political science literature that no practicable citations count is really feasible. (Google Scholar yields over 37,000 references, and Google Books nearly 9,000 more, but both cover mostly since the mid-1990s).
In Europe, the Law was energetically criticized by a few political scientists (notably Giovanni Sartori), but also endlessly taught and repeated to generations of politicians, especially on the right wing of politics. Many European elites have longed to consolidate party politics into simpler patterns than usually occur with proportional representation systems. So inspired by Duverger’s Law, many different reforms have been tried across Europe, designed to introduce ‘majoritarian’ elements and so capitalize on the anticipated consolidation of politics. These include de Gaulle’s re-founding of the Fifth Republic in 1958; the 1990s shift away from PR in Italy (pushed by Stefano Bartolini); various ‘reinforced majority’ provisions in Italy, used first for electing city councils and then scaled up by Berlusconi into a national fix in the noughties. Most recent has been a vote in the Romanian parliament to adopt full-blown FPTP. All these ‘constitutional engineering’ efforts have failed completely, with European voters stubbornly remaining multi-party in their afffiliations – for a very good reason. Even in its supposed heartlands, Duverger’s Law doesn’t work.
Duverger claimed that his effect operated in two ways. First, politicians know that with UK- or US- style voting systems a party can only ever win a seat by coming top in a local election district (or constituency) – that is, by winning the largest pile of votes (a plurality), whether or not the party has a local majority. So if you are not going to be in the top two parties locally, why stand? Nationally, what’s the point of coming second, third, or fourth in lots of seats without winning any – the traditional fate of liberals in the UK since the 1930s? As a result, third and fourth parties don’t stand, and if they do, they quickly die out.
Second, voters know that only the top two parties are contenders, so why ‘waste’ their one and only vote on supporting an also-ran party that is going to come in at a third, fourth or lower place? The pressure to make a difference pushes voters instead to back the top two and leave the rest to wither on the vine.
This logic still works in the USA, and my first chart shows how the resulting American party system looks. (This chart and both subsequent ones are drawn from joint work with Rekha Diwakar). Going down to the level of 450 sears in the House of Representatives, the bottom axis shows seats where Republicans are in the lead on the right hand side of the chart, and seats where Democrats lead on the left-hand side. The vertical axis shows the combined proportion of votes in each locality backing third, fourth or fifth parties.
Chart 1: District outcomes in the US House of Representatives, 2006
Duverger’s prediction is completely borne out here. All the seats are piled along the bottom axis, often with zero support for third parties in two-party only contests. When there is any degree of support for third and subsequent parties, it is almost always tiny or very minimal. The two coloured triangles show zones where one of the top two parties wins a clear majority (50% + 1) of local votes in the election district – the red zone being Republican seats and the blue zone being Democrat ones. Virtually all the 450 available seats in the House of Representatives lie in one or other of these coloured zones. Indeed in quite a large number of seats the leading party is piling up super-majority vote shares in the 70 to 90 per cent range.
The only odd seats with higher level of third party support actually occur not in the middle white areas (as one might expect) but in these high dominance areas. These outlier results do not genuinely mean high levels of third party voting however – instead they just reflect a local Republic dominance that is so complete that no Democrat candidate will stand and lose their money doing so, and vice versa in the Democrat majority areas. Here voters denied a major party choice plump for the closest they can get to a viable opponent – for instance, rightwing voters denied a Republican candidate to support will back the Libertarian. So even these cases do not undermine Duverger, but instead underline one-party dominance areas.
But the fact that one country appears to offer strong support to Duverger’s Law is not in itself very helpful, since the USA has many other features that might conduce to the same effect – including a Presidential system, the absence of socialism, a political plutocracy, an absence of much limits on campaign spending and political advertising, etc.
Two other large FPTP countries
So the question that needs answering is whether this same pattern occurs at the local district level in all plurality rule voting systems, as Duverger deemed it would? My second chart shows the situation in Great Britain and it should be immediately obvious that the pattern here is completely different from America.
Chart 2: Constituency outcomes in the 2005 general election, Great Britain
Source: Dunleavy and Diwakar, 2011
Here no local contests at all produce results either on or even near the bottom axis of the chart. And very few lie within the shaded triangle areas for Conservative and Labour local majorities. There are almost no seats where any party gets more than two thirds of the vote. All these effects occur because the whole cluster of seats is lifted off the bottom axis, with more than 15 per cent of the available votes going to third, fourth and smaller parties in virtually every constituency across the country.
There is clearly still some two-party focusing going on in British elections, as there is any election. But in addition, of course, the upper part of the overall triangle here is thickly populated with seats where the Liberal Democrats, Scottish Nationalists and Welsh Nationalists have come first and won the constituency (shown as brown dots). So in every respect, this is a very different chart from that of the USA.
Let’s look now at the world’s largest plurality rule voting country, India, shown in my third chart. There the pattern is radically different again. Here there is quite a big bunching of seats close to the bottom axis, and some even on the axis itself. But even in this minority of seats, almost nowhere does any candidate get more than two thirds of votes. And seats in the areas where the two leading party blocs have local majorities are clearly a minority – the BJP being the orange shaded area and the Congress being the green. Even at the bottom of the chart, most of the seats lie in the middle white area where neither of the big party blocs has a local majority.
Chart 3: District outcomes in the Indian Lok Sabha elections in 2004
Source: Dunleavy and Diwakar, 2011
There is a second prominent feature of the India chart, namely the large ‘mushroom cloud’ of seats that surges up from the white areas at the bottom to the top of the overall triangle. The top zone here encompasses seats (shown as blue dots) where parties other than Congress or BJP win or dominate local voting, including the Communists and many different regional and local parties.
Facing facts, revising theories
Any physical scientist looking at these three charts could tell straight away that we are looking at three radically different systems. The idea that parties or voters are behaving in the same ways across them is deeply unlikely. The factors leading to perfect two party politics in the USA cannot be general to all plurality rule systems – they must instead be specific to the American political context. Incidentally perfect two-party systems like this are now found almost nowhere outside the USA, except for a few small Caribbean nations. In particular, all the major Westminster system countries have shown strong trends towards multi-partism. For a time in 2010 indeed the UK, Australia, Canada, India and New Zealand (which moved to PR) all had coalition or minority governments – subsequently Canada moved back to a majority Conservative administration.
There are probably many reasons why political scientists have clung on to the bogus ‘Law’ for so long. American scholars are notoriously prone to ethno-centric thinking, and for them two party politics seems the ‘natural order’ of things, and coalition government dangerously fuzzy and anarchic. This slant is reinforced for rational choice theorists like Cox, who want neat mathematical models with clear equilibrium predictions. And the intuitive tug of Duverger’s two mechanisms (discouraging small parties from standing and voters from backing them) is still strong.
Here, though, new theory has hugely compromised the scope of the Law’s operations. For example, in a fine 2010 paper the US political scientists Eric Dickson and Ken Sheve use rational choice proximity models to predict that no contest should end up with more than two thirds of votes for one party. The logic here is that a local majority of 67 per cent or more can afford to split its vote across two parties, knowing that its biggest faction will still always win over any opposition party. By doing so, the majority of the large local majority can always advance their welfare. This logic works perfectly in the UK and India, as the charts above show (with no seats above the 67% majority level), but not at all in the USA.
There is a huge debate to be had about why Duverger’s law now is junk. Did the Duvergerian mechanisms once work, but have now ceased to apply? Have voters and parties ceased to worry about ‘wasting’ their efforts? Were Duverger’s ‘theory’ elements always wrong intellectually? And so on…
But for now, as our continent’s leading political scientists gather at the European Political Science Association conference in Berlin, could we not just accept (at last!) that the ‘law’ is bust and cannot be revived or defended any further? Arguing this case with colleagues, I still feel like John Cleese in the famous Monty Python sketch, insisting on the completely, blindingly obvious. Duverger’s law is extinct. It has ceased to be. It is no more. This is a dead parrot.
This blog draws on recently published research by Patrick Dunleavy and Rekha Diwakar, ‘Analysing Multi-party Competition in Plurality Rule Elections’, in the journal Party Politics. Sadly this is behind a paywall, but if you can’t get access please email Patrick.
An earlier version of this post appeared on LSE’s British Politics and Policy blog.
The programme for the European Political Science Association conference is online here.
Note: This article gives the views of the author, and not the position of EUROPP – European Politics and Policy, nor of the London School of Economics.
Shortened URL for this post: http://bit.ly/LBlcI1
Professor Patrick Dunleavy – LSE Public Policy Group and LSE Government Department
Patrick Dunleavy is Professor of Political Science and Public Policy at the London School of Economics and Political Science, where he has worked since 1979. He was educated at Corpus Christi College and Nuffield College, Oxford, where he gained his D.Phil. He has authored and edited numerous books on political science theory, British politics and urban politics, as well as more than 50 articles in professional journals. His current research includes a seven country study of how central governments relate to the IT industry for the UK’s Economic and Social Research Council.