A. Downs and the Spatial Theory of Party Competition 23 Chapter 2 Anthony Downs and the Spatial Theory of Party Competition Overview: In this chapter I examine the behaviour of political parties in two-party representative democracies such as the United States and the United Kingdom. The central question to be addressed is the following: In what circumstances will competition force parties to converge upon the electoral centre-ground? The initial answer to this question is provided by Anthony Downs's (1957) An Economic Theory of Democracy. Within political science the argument that parties will move to the electoral centre in an effort to maximize their vote has acquired the formal title of the median voter theorem. It is a theorem routinely [inked to Downs; when textbooks offer an account of the median voter theorem, it is Downs's name that appears in the first paragraph; and when theorists provide criticisms of and alternatives to the theorem it is Downs's name that appears in the first footnote. I shall set out the basic terms of Downs's argument and identify its intellectual precursors. Downs's argument rests upon a particular set of assumptions. Alternatives to these assumptions and explanations as to why parties might sometimes retain distinctive policy positions are explored. This exercise takes up the single largest part of the chapter and offers an introduction to more recent rational choice scholarship. In terms of the broader political science context, I start by looking at the way in which our understanding of democracy has changed over the last few centuries, and shall conclude the chapter by contrasting accounts of democratic legitimacy developed by political theorists with the model of democracy analysed by Downs. Setting the stage: the demands of democracy We are used to describing as democratic regimes in which decisions are taken by elected representatives (Pitkin, 1967). Yet at the rime of its inception following the English, French and American revolutions, the founders of what we now call representative democracy presented their 22 preferred method of decision-making as an alternative to, rather than as a particular form of democracy (Manin, 1997; Dupuis-Deri, 2004). At that time democracy was equated with direct democracy and direct democracy with anarchy. Ordinary citizens would, it was argued, be too easily swayed by populist.rhetoric. For proponents like James Madison, one of the authors of the Federalist (1751-1836), representation would refine and enlarge the public views by passing them through the medium of a chosen body of citizens, whose wisdom may best discern the true interest of their country and whose patriotism and love of justice will be least likely to sacrifice it to temporary or partial interests. (Federalist, 10) Although this distinction between republicanism, as it was known, and democracy lives on in the names of America's two largest political parties, it has, in other respects, been largely forgotten. We tend nowadays to judge the credentials of a democratic system, representative or not, in terms of the tightness of the fit it provides between public opinion and policy outputs. The idea that representation is to be favoured because it loosens this fit seems increasingly alien. It was in the later part of the nineteenth and early part of the twentieth century that theorists first began to argue that the practice of representative democracy was flawed because elected agents did not pursue the interests of their voting principals. Marxists argued that representative or bourgeois democracy was a sham and that policy outputs reflected the interests of business. Elitists like Pareto suggested that 'we need not linger on the fiction of "popular representation" - poppycock grinds no flour' (quoted Duneavy and O'Leary, 1987: 140). In the interwar years such arguments contributed to a more general loss of faith in democracy and to the rise of parties openly extolling alternatives to it. In the immediate postwar years, the academic task of defending democracy's credentials initially fell to pluralists like David Truman (1951) and John Kenneth Galbraith (1953). Against Marxists and elitists alike they argued that ordinary citizens could and did influence government policy but that they did so primarily through their membership of pressure-groups. Policy outputs, they argued, broadly reflected the inputs of competing pressure-groups and because pressure-group membership was broadly representative of public opinion, this meant that policy-making could be described as being democratic. Pluralists 24 Rational Choice A. Downs and the Spatial Tbeoiy of Party Competition 25 accepted that some groups, most notably business groups, would dominate policy-making in particular sectors. But they denied either that the state was structurally predisposed to favour business interests or that business held a dominant position in every policy sector. But notice what is missing from this account: elections. Pluralists did not altogether ignore elections. The single most influential pluralist, Robert Dahl (1956:131), argued that the prospect of having to fight elections forced political leaders to anticipate and so respond to public opinion. But pluralists tended not to dwell very long on elections. With the publication of An Economic Theory, this was to change. The precursors of party competition Downs's argument can be approached and best understood in terms of two earlier pieces of work. The first is Joseph Schumpeter's (1942) Capitalism, Socialism and Democracy!. Schumpeter [1883-1950] was an economist who, shortly before presiding over a period of disastrous hyper-inflation as Austria's Finance Minister, proclaimed himself to be the greatest horseman in Austria, the greatest lover in Europe and the greatest economist in the world. Most of Schumpeter's work revolved around the subjects of economic development, entrepreneurship and the business cycle. But his writing on democracy, although amounting to no more than a couple of thousand words, has proven to be his most lasting contribution to the social sciences. Schumpeter is usually regarded as an elitist and certainly an unambiguous elitism underpins one of his most famous arguments that public opinion is ill-informed, fickle and easily manipulated. The notion of there being a settled and reasonable public will which it is the duty of politicians to discern and respect is, Schumpeter argues, nonsense. The popular will is 'the product and not the motive power for the political process' (1942: 263). Elitists like Pareto, Michels and Mosca, who made a number of similar observations, theteby concluded that representative democracy was a sham. Schumpeter did not. Democracy was that 'arrangement for arriving at political decisions in which individuals acquire the power to decide by means of a competitive struggle for the people's vote' (Schumpeter, 1942: 269). Parties have to compete with each other to get elected, and competition forces them to select policies they believe voters will find attractive. Downs (1995: 197) credits Schumpeter with providing 'the inspiration and foundation for my entire thesis' and is, in particular: much taken with Schumpeter's argument that policy emerges as the by-product of the competitive struggle for votes. But whilst Schumpeter waxes lyrical about the nature and meaning of democracy he is remarkably vague about the tactics parties will adopt in order to get elected. Using a military metaphor he at one point suggests that parties will fight to gain control of 'hills' that afford them 'strategic advantage'. But it was another and very different metaphor which Downs eventually employed to analyse party competition. The use of left and right as general analytical contrasts is long-standing. In the Pythagorean table of opposites, left is associated with darkness and evil and right with light and goodness (Lloyd, 1962). Yet as a term of political description the origins of the spatial metaphor are more recent. At the very start of the French Revolution delegates to the Estates-General broke away to form a National Assembly. Because voting within this Assembly was conducted by physically standing at required moments, representatives started to sit themselves next to like-minded colleagues on, literally, the left and right-hand sides of the Assembly floor. Because Assembly rules prevented representatives from describing each other as belonging to named political factions such as the Girondists, left and right were soon being used as terms of description and abuse. Out of this simple and exclusive contrast between left and right there soon developed a conception of political space as a continuum with a centre, centre-left, centre-right and so on (Hindmoor, 2004: 3-4). Within a few months, Mounier, for example, emerged as the leader of a faction sitting at the physical centre of the Assembly and advocating a kind of English constitutionalism as an alternative to both the absolute monarchism of those on the right and the republicanism of those on the left. Propelled first by the Revolutionary and then Napoleonic wars, the use of the spatial metaphor spread first to Scandinavia and the low countries, then to Southern Europe, and, eventually, to Britain. By the start of the twentieth century the spatial metaphor offered voters, politicians and commentators a kind of political Esperanto, a universal language of politics which could be used to describe political processes and outcomes in seemingly very different countries. The first attempt to understand political competition in spatial terms is generally credited to another economist, Harold Hotelling. The question Hotelling addressed himself to, might, initially, seem 26 Rational Choice A. Downs and the Spatial Theory ofPaity Competition 27 incredibly obscure. Imagine there is a one-dimensional space, perhaps the 'main street' of a town, across which customers are equally arranged (Hotelling, 1929: 46). Where will two profit-maximizing shops locate? Most people's intuition is that one will locate to the left-of-centre and the other to the right so minimizing the average distance customers have to travel to the nearest shop. But Hotelling shows that this is not so. Each shop will actually move to the centre of the street. For it is only then that each will be able to prevent its rival from gaining a larger share of the market. Consider, for example, the position of a shop which moves to the left of the centre. Its rival could locate immediately to its right and in doing so acquire the business of the majority of those customers who live to the right. The crucial point in Hotelling's argument then comes at the very end of the article. Extending his discussion from economics to politics he suggests that so general is this tendency [to converge upon the centre] that it appears in the most diverse fields of competitive activity, even quite apart from what is called economic life. In politics, it is strikingly exemplified. The competition for voters between the Republican and Democratic parties does not lead to a clear drawing of issues, an adoption of two strikingly contrasted positions between which the voter may choose. Instead each party strives to make its platform as much like the other's as possible. (Hotelling, 1929: 54) Hotelling did not actually use the terms left and right to describe the position of parties in political space, did not seek to represent the position of parties using a linear scale and did not use the nomenclature of the median voter. Nevertheless, it is his spatial analysis which Downs (1957:115) sees himself as 'borrowing and elaborating' upon. Political parties will, as Schumpeter suggests, compete to attract the support of voters. This competition, as Hotelling argues, pushes the parties toward the centre-ground. The median voter theorem Stated more formally, the median voter theorem rests upon the set of assumptions listed below. Whilst these all obviously call for elaboration and perhaps qualification, I want, for the moment, to concentrate on their implications. 1 There are only two parties. 2 Political space is one-dimensional. 3 Parties can move to and occupy any point in this one-dimensional • space. 4 Parties are vote-maximizers. 5 Voters, if they vote (see Box 2.1), vote for the party closest to them in political space. 6 There is perfect information. 7 Voters' preferences are fixed. In Figure 2.1 the horizontal axis shows a series of positions in political space running from left to right. Voters' preferences can be mapped on to this one-dimensional scale (assumption 2} and are fixed (assumption 7). The vertical axis shows support for these alternative positions. In Figure 2.1(a) the aggregate distribution of preferences shows a situation in which most voters are clustered at or near the centre of the horizontal axis and in which there are relatively few voters on the far-left or far-right. As it happens, surveys have shown this is pretty much the kind of distribution routinely found in Britain and America. Figure 2.1(b), on the other hand, shows a very different distribution in which there are a large number of left-wing voters, very few voters at the centre and a cluster of voters at the far-right. The important point to note here is that whatever the distribution of voters, there will always be a median voter, a person whose preferences are such that there are exactly as many voters to their right as to their left. In Figure 2.1(a) the median voter is located at the very centre of the horizontal axis, whilst in Figure 2.1(b) the median voter is located further to the left. For the same reason Hotelling argued that competition forces businesses to locate at the centre of a street, the median voter theorem holds that competition forces parties to move to the position of the median voter. We know that there are only two parties (assumption 1) and that these parties will seek to maximize their vote (assumption 4). Consider now what would happen if party A were, in Figure 2.1(a), to locate to the left of the median voter at A'. We know that the second party, B, can move anywhere in political space (assumption 3) and that it knows the distribution of voters' preferences (assumption 6). If it were to move immediately to the right of A', to B', it would attract the votes of the majority of voters to its right (by virtue of assumption 5). So B would win the election. This would give A an incentive to move immediately to the right of B so regaining majority support. But B could then move 28 Rational Choice Box 2.1 The paradox of not voting Voting is a costly activity. 'It takes rime m register, to discover whni pňrtiei are running, to deliberate, 1:0 go to the polls, and to mark the .-' ballot1 (Downs, 1957:166): The existence of these costs is:significant '■ .because whilst.ir may matter a great deal tb people whether or not their preferred party is elected, rational actors will realize that. the. chances of their vote making any difference to the eventual outcome are incredibly low: lower indeed than their chances of being knocked down by a car whilst crossing the road to get'to the polling station (Meehl, 1977). Now, as Downs (1957: 266) and, more recently, Bernard Grófman (1993) have observed, the existence of;such costs ■ gives rational choice theorists the opportunity.to explain why turnout - increases-when the costs of voting are. lower either because the weather is sunny, the polls „arerop_en longer or poscal.voting is possible. Hut there is an* obvious problem here. In any reasonably-sized -electorate the chances of any one person's vote making a,difference to the result are so small that the costs of voting are almost certainly going to exceed the possible discounted benefits of doing so. This means that for. most people it is simply irrational to vote. But.of course millions of people do vote. This is the 'paradox of not voting', or, as. it is sometimes called, the 'paradox that ate rational choice' . (Fiorina, 1989). ? ' ' '..''■■' \' '■ ■ • So whjrdo people vote? Possible answers include the following. (1) People vote because they val uc democracy and recognize thatdemocracy .: will collapse if nobody votes (Downs, 1957:267-B). But this is hardly an to the right of A and so on. This process of competitive 'leapfrogging' would, however, come to an end when both parties straddled the position of the median voter, for neither would then be able to increase its share of the vote by moving to either the left or right. Convergence upon the position of the median voter is in this way a stable equilibrium. Political commentators in America and Britain have frequently remarked upon the electoral significance of the centre-ground. In 1992 Bill Clinton's unexpected victory over George Bush Senior was credited to his success in shifting the Democrats towards the centre-ground. Four years later Clinton secured an equally unexpected reelection by pursuing a strategy of 'triangulation' (Morris, 1996). By portraying a Republican Party then led by Newt Gingrich as hopelessly right-wing whilst tacitly accepting the excesses of his own party, Clinton positioned himself between whilst also rising above A. Downs and the Spatial Theory of Party Competition 29 . answer- because, as Downs (1957: 270) himself recognizes, voters will 'actually get this reward [of living in a democracy] even if |they] do not vote as long as a sufficient number of other citizens do'. (2) Voters sire - rlsk:nversc and are driven to the polls by the fear of not voting and then . seeing theit preferred candidate lose by one vote (Ferejohn and Fiorina, 1974). Yet if people are so risk-averse why would they ever leave their house (Beck, 1975)? (3) Voters are driven to the polls out of a sense of duty, out of a sense that they ought to vote (Rikcr and Ordeshook, 1968). But if people arc prepared to eschew self-interested calculations here, why assume that in other areas of their economic and political, life they act in a purely instrxtraental and self-interested mannefi (4) People vote 'expressively' to demonstrate (to themselves and others) their . support for a particular candidate (Brennan and Lomasky, 1993). (5) Peoplevote because they think that other people, will calculate that it is not nationalI _ro vote jind that rfjerV-vqte.iwY/ therefore bedecisivc (see Mueller, 1989: 351-2). But would rational people'"really think it likely that nobody would voce when millions of people have consistently done so in previous elections? [6] People vote because* the costs of voting are extremely low, and below some threshold people do not bother to calculate whether the costs of low-cost activities are greater than their benefits (Barry, 197H: 23), To put the same point in n slightly different way, people vote because they 'satisfice' and the costs of voting are so small, that people do not bother to adjust their behaviour to maximize 'their1 ■ ■ returns (Bendor et aL, 2003). (7) People vote because they want .to demonstrate to others that they can be trusted to behave in a non-selfish ■'. manner (Overbye, 1995). the available alternatives. A few months later New Labour in the UK won an election landslide by erecting, in the words of a Sunday Times (1997) editorial, a 'razor wire' fence around the centre. According to one of Tony Blair's biographers, John Rentoul (2001: 197), it was as a visitor to the Democrat's 1992 campaign that he, Blair, had learnt the 'important lesson that all politics is a battle for the centre-ground'. Precisely because the electoral significance of the centre-ground is and always has been so widely recognized, it is open to critics to argue that the median voter theorem is 'banal' (Green and Shapiro, 1994: 6). But Downs is not simply arguing, as political journalists routinely argue, that parties can gain an advantage by moving to the centre. Downs can be understood as attempting to specify the reasons why and the necessary and sufficient conditions in which parties find it in their interests to converge upon the centre-ground. 30 Rational Choice Figure 2.1 Two-party competition with standard and bimadal distributions (a) standard distribution Number of people preferring policy position Left-wing | | Median voter Party A' Party B' (b) bimodal distribution Number of people preferring policy position Right-wing Left-wing Median voter Right-wing Qualifying the argument: accounting for divergence Critics routinely argue that rational choice models have little or no explanatory value because they are constructed from a series of A. Downs and the Spatial Theory ofPaiiy Competition 31 implausibly heroic assumptions, most notably that individuals are rational.' One possible defence against this argument is the instrumentalist one that theories ought to be judged in terms of the accuracy of their predictions rather than the realism of their assumptions (see Box 8.3). To this end, we might re-cast the median voter theorem as a prediction about the way in which parties will behave. As Colin Hay (1999) observes, there is plenty of evidence that the members and leaders of political parties in Britain and America believe that something like the median voter theorem holds. Members often choose parry leaders on the basis of judgements about which candidate will prove most appealing to the wider electorate; newly elected leaders routinely talk about the need to appeal to voters at the centre-ground and berate activists who want to return the party to its ideological roots; and party leaders use opinion polls and focus groups to road-test policies. Politicians' faith in the veracity of the median voter theorem is not without foundation. There is plenty of evidence that parties can usually increase their share of the vote by moving closer to the position of the median voter (see, for example, Sanders, 1999). The problem here is, however, that the theorem does not simply predict that parties will converge towards the position of the median voter. It predicts that parties will move to the position of the median voter and so adopt identical policy positions. Now parties do sometimes signal a change in position by adopting some of their opponents' policies. In the early 1990s, for example, New Labour in Britain moved itself to the right by embracing Conservative spending plans and promising to leave income tax unchanged and the basic terms of trade union legislation in place. In a similar way, the Australian Labor Party undercut its Liberal opposition in the 1980s by embracing free trade, privatization and financial deregulation (Jaensch, 1989). But no matter how intent upon outflanking their opponents they are, parties do not simply adopt other parties' policies wholesale. Detailed surveys of manifesto commitments (Budge, 1999) and policy outputs (Chappell and Keech, 1986) show that political parties retain distinctive characteristics. Left-wing parties tend, for example, to pursue more expansionary monetary policies, run-up larger budget deficits, spend more on welfare and preside over falling unemployment. Conversely, right-wing parties tend to pursue more restrictive monetary policies, cut budget deficits, spend more on defence and preside over falling inflation (see Mueller, 2003: 447-50 for a review of the evidence). 32 Rational Choice A. Downs and the Spatial Theory of Party Competition 33 Instrumentalists maintain that theories ought to be judged in terms of the accuracy of their predictions rather than the realism of their assumptions. Yet the obvious rejoinder to this is that bad assumptions lead to poor predictions. Given that political parties do not converge upon the position of the median voter and adopt identical policy positions, do we therefore need to abandon Downs' assumptions in order to understand the process of party competition? In some ways I think the simple answer to this question is that we do. But this does not mean that Downs' theory is a poor starting-point for a discussion of party competition. The assumptions employed by Downs are valuable not necessarily because they are accurate, they are often not, but because they serve as explanatory prompts which we can use to account for the behaviour of actual parties. Down's model of party competition specifies the necessary and sufficient conditions in which parties will find it in their interests to converge upon the position of the median voter. Having identified these conditions we can therefore start to explain why parties in the 'real world' do not always converge upon the position of the median voter in terms of variations in these conditions. Over the following pages I will look at and consider the plausibility of each of the assumptions Downs makes and show how alternative assumptions can be used to account for policy divergence. Assumption 1: there are only two parties In a now classic study of the impact of voting systems upon the number of parties, Maurice Duverger (1951) first formulated what has since become known as 'Duverger's Law' and 'Duverger's Hypothesis' (Riker, 1982b), Duverger's Law holds that plurality (or first-past-the-post) voting systems tend to lead to two-party competition. Duverger's hypothesis holds that proportional voting systems tend to be associated with multi-party competition. This argument rests upon the identification of 'mechanical' and 'psychological' effects associated with the use of plurality systems. The mechanical effect refers to the way in which plurality systems discriminate against third parties or candidates whose vote is evenly divided across a number of constituencies. In the 1992 and 1996 US Presidential elections, for example, Ross Perot acquired 19 and 9 per cent of the national vote respectively. Yet because he came in third in most states, Perot did not acquire a single vote in the Electoral College. The psychological effect refers to the tendency of voters, knowing that third parties are discriminated against in this way, to avoid 'wasting their vote'. Does Duverger's Law hold? The answer to this question depends upon the way in which the number of parties in any one country is counted, and there are a number of possibilities here. At one extreme we could count any party which put forward any candidate in any election. In this case there would be no two-party systems. Indeed judging by the number of parties officially registered with, for example, the British Electoral Commission, Britain would be a 27S-party system with the 'No Candidate Deserves My Vote Party' led by Amanda Ringwood of St Albans counting equally with the Conservatives and Labour. At the other extreme, and following the suggestion of Giovanni Sartoria (1976), we might only count a party if it has a realistic chance of governing alone. But realistic is a term obviously open to interpretation and this method would have the unfortunate consequence of rendering some stable democracies, including Britain's in the 1980s, as one-party states. The compromise suggested by Rein Taagepera and Matthew Shugart (1989) is therefore to count the 'effective' number of parties (Box 2.2). Using this method, parties are counted in proportion to their size in such a way that small parties, although counted, do not count to the same extent as larger ones. By the usual standards found within the social sciences, the evidence for the existence of Duverger's law and hypothesis is quite strong (Riker, 1982b). As Table 2.1 shows for the case of both votes cast and seats taken, the effective number of parties in plurality voting systems is significantly lower than it is in proportional voting systems. This is not necessarily to say that Duverger is beyond reproach. For it may be that countries in which there is only one salient political cleavage, and which therefore naturally lend themselves to two-party competition, choose plurality voting systems for this reason. Conversely, it may be that countries with multiple cleavages - socio-economic, religious, linguistic and territorial - which naturally lend themselves to multiparty competition choose proportional voting systems. The number of parties in a country may, in other words, cause the voting system rather than the voting system causing the number of parties (Bogdanor, 1984; Colomer, 2005). But we do not need to resolve this issue here. What matters is that Downs's assumption of two-party competition is, in those countries using the plurality voting system, a reasonably plausible one. 34 Rational Choice A. Downs and the Spatial Theory of Party Competition 35 The.following, account is drawn from Taagcpcra and Shiigarr (1989: 77-80)» litiagine four possible party systems jn each of which'there are •five parties attracting the following vote shares: ^P^I":^,i' Tarty A H ■ C D E ' System 1 51% 42% 5% 1% 1% System 2 51% 2fi% 11% 11% 1% System 3 40% 37% 11% 1% System 4 4(1% 37% 9% 5%. :•. Intuitively, we might describe system 1 as a. two-party system. But it is far - lassclcur hpw:we might describe system's 2,3 and 4. In each of these, the tota| votes cast. Yet in each "case the smaller parties attract some support, til order to resolve this problem we cnuk! establish n.'cut-off* point below which a party would hot be 'counted'. But any such number would dearly be arbitrary. If, for example, we set this cut-off at 10% a then.system 3 would,he a four-parry.system and system 4 a two-party system eyeu though there.is very little actual difference between rhrm. The best way uf calculating the effective numher of parries :s to let the vote shares determine their own weights in the foliowaig manner: 1; Multiply the fractional shnre.of cia'ch party against itself. So in the case ;: .:. of system 1, the fractional share of part)' A is 0751 and 6.5 1. x 0.51 = 2 Add Together the resulting figures for each party in this system: 0.2601 i- 0.176 +■ 0.0325 + 0.0001 + (1.0001 _ 0.4388. 3 Divide 1 by mis number = 2.278. So restricting ourselves to 1. cecimal place system 1 has 2.3 effective parties, system 2 has 2.S effective parties, system 3 lias 3.1 and D has 34/Note that die effective number of parties can be calculated in terms of the yores cast (as it is here) or in terms of the .-.umber of ssats each Assumption 2: political space is one-dimensional Downs (1957: 115) assumes that 'political preferences can be ordered from left to right'. Some social scientists (Giddens, 1994) and occasional politicians (Blair, 1997) have argued that, in a post-industrial, Table 2.1 The effective number of parties Country Year Effective number of parties by vote Effective number of parties by seats Countries using plurality voting Bahamas 1982 . 2.1 1.6 Canada 1984 2.8 1.7 India 1984 3.8 1.6 Malaysia 1982 2.3 1.4 New Zealand 1984 3.0 2.0 South Africa 1981 2.6 1.5 Sri Lanka 1977 2.8 1.4 United Kingdom 1983 3.1 2.1 United States 1984 2.0 1.9 Average 2.0 1.8 Countries using proportional voting Belgium 1985 8.1 7.0 Denmark 1984 5.8 5.5 West Germany 1983 2.6 2.5 Switzerland 1983 6.0 5.3 Australia 1984 2.8 2.4 Finland 1983 5.4 5.1 Ireland 1982 2.7 2.6 Netherlands 1982 4.2 4.0 France 1981 4.0 2.8 Sweden 1985 3.5 3.4 Average 4.5 4.0 Note: An index of voting systems used in each country is available at http://www. worldpolicy.org/globalriEhts/prindex.html. Source; Data from Taagepera and Shugart (1989: 82-3). post-communist, post-modern society, the 'terms left and right no longer have any relevance'. These arguments are, I think, overdone (see Bobbio, 1996). Left and right continue to form a staple part of political discourse, but it is clear that political competition in Britain and America is no longer, if it ever were, exclusively one-dimensional. Before proceeding any further we need to be clear about the terminology being 36 Rational Choice A. Downs and the Spatial Theory of Party Competition 37 used here. Dimensions refer to policy issues over which voters or politicians have connected beliefs. Recent studies of British electoral behaviour (Heath etal., 2001: 78-81) have revealed the existence of at least two and possibly three dimensions. The first and still most salient of these is the left-right dimension; the second is a libertarian-authoritarian dimension taking in such issues as racial equality, the death penalty and stiffer sentences; the third, most recent, and still weakest dimension is a British nationalist one composed of attitudes towards devolution, Northern Ireland and Britain's nuclear deterrent. What happens to the dynamics of party competition when there is more than one dimension? Figure 2.2 shows a situation in which there are three voters (A-C) whose preferred positions (marked as A, B and C) are mapped against a socio-economic and libertarian-authoritarian dimension. Voter A is extremely left-wing and authoritarian, voter B is moderately right-wing and libertarian, and voter C is extremely right-wing and moderately authoritarian. Assume that there are two parties, X and Y, and that X has initially located itself at point X,. Where could Y position itself in order to attract more votes? To answer this we need to construct a set of indifference curves (a.^, b,b,, CjC,) intersecting Figure 2.2 tarty competition in two dimensions (i) ■ at X, and showing points in political space between which each voter is jhdifferenI:- Look at, for example, the semi-circle axar This shows a set of points between which A is indifferent. To put the same point in a different way, all the points along this line, including Xp are an equal distance away from A's ideal or 'bliss' point. What about all those other points in this two-dimensional space? If voters vote for the party closest to them in political space (assumption 5), A will vote for a party positioned anywhere along this line in preference to any beyond it and to the right because any such party will be closer to its bliss point. We can also say that A will prefer any party positioned at any point inside the line to any party on the line because any such party will be closer to its bliss point. The next thing to look at here is the two shaded areas or 'winsets'. The first and larger of these, on the lower right-hand side of the figure, shows those points voters C and B prefer to Xr We know that B prefers any point in this area to X.j because these points are closer to its bliss point. We know that C prefers any point in this area to Xj for the same reason. The second and smaller winset at the top shows those points voters A and C prefer to Xr We can now return to the question with which we started. Where could Y position itself in order to attract more votes? By moving to anywhere within either winset it could attract the support of two of the three voters. Assume that Y actually positions itself at Yj in Figure 2.2. Where could X now position itself to attract more votes? Figure 2.3 shows a set of indifference curves (a^,, bjb,, c^) this time intersecting at Yr There are, once again, two shaded winsets. The first, at the bottom of the diagram, shows those points B and C prefer to Yr The second, running towards the top of the diagram, shows those points that A and C prefer to Yr So X could move anywhere within these areas and attract the support of two of the three voters and so win the election. Now we could repeat this exercise indefinitely. With this particular distribution of preferences no matter where one party positions itself the other party could always attract more votes by moving to a different location. The discovery, in the late 1960s and early 1970s, that there is usually no stable equilibrium when there are two or more dimensions came as something of a shock to rational choice theorists. Far from being inherently stable, party competition, it now appeared, was 'chaotic' (McKelvey, 1976). In recent years die argument has, however, swung back. Theorists accept that there will not usually be a stable equilibrium when there is more than one dimension, but they maintain that for most plausible distributions of preferences, parties will find it in 38 Rational Choice A. Downs and the Spatial Theoiy of Party Competition 39 Figure 2.3 Patty competition in two dimensions (ii) Authoritarian Libertarian Left-wing Right-wing their interests to adopt positions within a relatively small region known as the 'uncovered set'. What is the uncovered set? A position Px is said to cover position P, if P, is majority-preferred to P2 and if all the alternatives which are majority-preferred to P: are also majority-preferred to P2. There will usually be a set of positions that Pj does not cover and to which it is therefore particularly vulnerable. These positions are the uncovered set of Pr Assume now that the uncovered set of each and every position has been identified. By looking at the intersection of all these points (the set of points they have in common) it will be possible to identify the uncovered set for the whole policy space. Parties will find it in their interests to adopt positions within this space for two reasons. (1) The number of positions that will be majority-preferred to those within the uncovered set will usually be quite small. (2) Positions within the uncovered set will be majority-preferred to most of those positions outside of it. Whilst parties may, in the short-term, adopt positions outside of the uncovered set, in the longer term competition will usually lead parties to return to it (see Tullock, 1967; Cox, 1987; Miller, Grofman and Feld, 1989; Mueller, 2003: 236-41 for a review). The uncovered set will often be located in and around the position of the dimension-by-dimension median. In Figures 2.2 and 2.3, the Figure 2.4 Dimension-by-dimension median Authoritarian Libertarian Left-wing Right-wing median voter on the socio-economic dimension is B (A is to their left and C is to their right). The median voter on the authoritarian-libertarian dimension is C (A is more authoritarian and B is less authoritarian). Figure 2.4 shows two lines, the first intersecting point B and the second intersecting point C. The dimension-by-dimension median is the point at which these lines intersect. So the existence of multiple dimensions may not always make a significant difference to the dynamics of party competition. Assumption 3: parties can move to and occupy any point in this one-dimensional space The median voter theorem requires that parties can move anywhere in political space. What might prevent them from doing so? One possible answer is party activists. Downs (1957: 24-5) defines a political party as a 'team of men seeking to control the governing apparatus by gaining office in a duly constituted election'. He then goes on to assume that 'members agree on all their goals' and that each party can therefore be treated as if 'it was a single person'. But this is simply implausible. Intra-party conflict can frequently be as intense as inter-party conflict. Such conflicts arise because ordinary party members tend to have different political views from party leaders and ordinary voters 40 Rational Choice A, Downs and the Spatial Theory* of Party Competition 41 (May, 1973). That the beliefs of party members should differ from, those of ordinary voters is hardly surprising; people will presumably only join the Labour or Democratic Parties if they regard themselves as being to the left of the Conservatives or Republicans. So given this self-selection filter, the median Labour or Democratic Party activist is bound to be to the left of the median voter in the country as a whole. Such differences will not matter if party activists have little or no power and party leaders are able to move their party to the position of the median voter. Neither will it matter if activists are prepared to set aside their policy preferences in order to secure their party's election. But if neither of these conditions holds, party leaders may be unable to move to any point in political space. If the power activists have is the power to elect the party leader, vote-maximizing candidates for the party leadership may well have to adopt positions appealing to the median voter within the party rather than the median voter within the wider electorate. Candidates in left-wing parties will have to present themselves as belonging to the left whilst candidates in right-wing parties will have to present themselves as belonging to the right. This will not matter if, once elected, party leaders can steer their party back to the centre. But if voters value responsibility, such a strategy might be costly. Party leaders may find promises they made during a leadership election thrown back at them by a rival during a subsequent national election campaign. If activists have some direct control over policymaking, the opportunities for a party leader to move towards the position of the median voter may be even more limited. Assumption 4: parties are vote-maximizers Downs (1957: 27) maintains, rather gruesomely, that politicians, all politicians, 'act solely in order to attain the income, prestige and power which come from being in office". Because they have no preferences over policy they 'formulate policies in order to win elections rather than win elections in order to formulate policies' (ibid.). Now it is certainly the case that voters have an overwhelmingly poor view of their elected representatives and to this extent Downs's assumption may ring true, but judging by the lengths diey go to champion their favoured policies in internal party debates, the assumption of vote-maximizing nevertheless looks like a caricature. As I have previously suggested, it is, with sufficient ingenuity, always possible to account for any pattern of behaviour in terms of the assumption of self-interest. So it might be argued that politicians argue about policies because doing so allows them to, for example, publicly express their allegiance to factions within a party they believe can further their career. But it is incumbent upon social scientists not simply to find explanations which are consistent with the facts but which best explain those facts. As rational choice theorists have subsequently come to accept (Wittman, 1977; Laver and Hunt, 1992), the most plausible explanation of most politicians' behaviour is they care about both policy and votes; that they have both policy-seeking and office-seeking motives. Politicians have principles. They will not say or do anything in order to be elected, but their principles are not cast in stone. Politicians will formulate their policies with at least one eye upon the prevailing public mood. From this point it is of course easy to see how we might account for the fact that parties adopt differing policy positions in terms of their distinctive policy preferences. But such an argument obviously risks looking entirely ad hoc. In recent years the argument that preferences over policy can be used to explain policy divergence has, however, been given greater empirical and analytical bite through a challenge to another of Downs's assumptions. Assumption 5: voters vote for the party closest to them in political space Downs maintains that the position a party occupies depends solely upon the policies it has adopted and that voters only care about policies. Consequently voters will always select the party closest to them in political space. Now surveys have shown that voters are, statistically, far more likely to vote for the party closest to them in political space (for a review of the evidence see Merrill and Grofman, 1999); but they have also shown that the relationship between position and vote is an imperfect one. Voters, it would appear, care about policy but they do not only care about policy. At this point, rational choice theory rubs up against an older, behaviourist, political science tradition which emphasizes the extent to which many voters identify with and vote for parties that, in policy terms, they may not actually be closest to (see Campbell etal.,1954, I960; Butler and Stokes, 1969). Voters, it is argued, do not step back and look at the policy position of each party and then rationally choose between them. Because they are rationally ignorant they usually vote for the party they have always voted for. It would appear that the strength of such partisan identifications has waned in recent 42 Rational Choice A. Downs and the Spatial Theory of Part}' Competition 43 decades. In Britain, for example, the proportion of voters who 'strongly identified' with a political party fell from around 44 per cent in the early 1960s to around 12 per cent by the late 1990s. In other words, it may be that voters are becoming more Downsian in their behaviour. But the relationship between policy position and vote nevertheless remains a probabilistic rather than deterministic one. What difference does this make to party competition? If party leaders only care about getting elected it makes no difference. If voting is probabilistic, parties can still increase their chances of getting extra votes by moving towards the position of the median voter. But, as James Enelow and Melvin Hinich (1982, 1989) have shown, if candidates have both office-seeking and policy-seeking motives, parties will be more reluctant to sacrifice preferred policy positions for the uncertain prospect of acquiring more votes. Probabilistic voting sustains equilibria in which parties converge towards the median voter but do not adopt identical policy positions. Assumption 6: there is perfect information Parties can only move to the position of the median voter if they know where that voter is located. Equally, voters can only vote for the party closest to them if they know where the parties are located. Perfect information is the lubricant that keeps the median voter theorem running smoothly and predictably. It is therefore not difficult to see what difference a little bit of uncertainty - which Downs (1957: 77) defines as 'lack of sure knowledge about the course of past, present, future or hypothetical events' - might make. If, for example, parties disagree about where the median voter is located, they will obviously end up adopting different positions. This much seems obvious. But uncertainty also matters in a more interesting way. In committing themselves to particular policy positions, parties are making promises. They are promising, if elected, to implement one set of policies rather than another. But voters cannot know with absolute certainty whether a party intends to or will be able to fulfil its policy promises (Hinich and Munger, 1996), and for this reason in an uncertain world it matters a great deal whether voters regard a party as being trustworthy. Downs recognizes this and suggests that competition leads parties to act reliably and responsibly. A party is reliable if 'its policy statements at the beginning of an election period - including those of its preelection campaign - can be used to make accurate predictions of its behaviour during the [subsequent] period' (Downs, 1957: 104-5). The easiest way for parties to acquire a reputation for acting reliably is to keep their policy promises. Because 'rational men will vote for an unreliable opposition party only if the [alternative] parties have' such abysmal proposals that random policy selection is preferable to them' (Downs, 1957: 107), the desire to be re-elected gives parties an incentive to act reliably. What then of responsibility? A party is responsible if 'its policies in one period are consistent with its actions (or statements) in the preceding period, i.e. if it does not repudiate its former views in formulating its new programme' (Downs, 1957: 105). A responsible party retains faith with its policy position over a long period of time; an irresponsible party constantly changes its position. But if voters value responsibility parties may not always be able to maximize their vote by moving to the position of the median voter. Consider the position of a party which has, wittingly or unwittingly, strayed to the left of the median and so lost an election. On the one hand it can gain votes by moving to the right and toward the median voter. This much we already know. But if voters value responsibility it will risk losing votes by changing its position. What its vote-maximizing strategy will be will depend upon the precise number of votes it risks losing. But it is not hard to see why the need to appear responsible might lead parties to retain distinctive policy positions. It is easy to see why voters might care about whether a party has a reputation for acting reliably. But what is unclear here is why they should care about whether a party has acted responsibly. Why does it matter what a party has said or done in the distant past? The best answer to this question is, I think, one which draws us back to the earlier discussion of politicians' motives. If a party has retained its policy position over a long period, voters may infer that it has a genuine, policy-seeking, commitment to its position and that the party can, if elected, be trusted to do what it says it is going to do. If, on the other hand, a party has only recently adopted a policy position and only done so when that position became popular, voters might worry that its commitment is purely instrumental and vote-seeking and that the party might, at the first sign of trouble, renege upon its promises. This suggests a striking and seemingly paradoxical conclusion. Parties which attempt to maximize their vote by constantly changing their policies to suit the message of the latest opinion poll or focus group, might actually risk losing support. Parties which adopt policy positions because they genuinely believe in them may end up attracting more 44 Rational Choice A. Doivns and the Spatial Theory of Party Competition 45 votes. The dichotomy previously posited between vote-maximizing and policy-seeking may therefore be a false one. The attempt to vote-maximize might be self-defeating. Assumption 7: voters' preferences are fixed Downs (1957: 55) assumes that the preferences voters have over policy positions derive from their underlying 'fixed conceptions of the good society'. But Downs simply does not say where these conceptions come from or why they should be considered as fixed. Some voters, it would appear, are simply right-wing and others are simply left-wing. The blame here does not simply lie either with Downs or, more generally, rational choice theory. For, by and large, preference-formation remains as much of a mystery to psychologists, sociologists and historians as it does to rational choice theorists. Social scientists now know a great deal about what people want. They know far less about why they want it. At the crudest of statistical levels one thing we do, however, know is that people from similar socio-economic backgrounds are more likely to have similar political stances. From mass surveys of voting behaviour we know, for example, that people who went to state schools, have manual jobs, belong to a trade union, work in the public sector and live in council houses are more likely to regard themselves as being on the left than people who went to public schools, have professional jobs and so on. As Patrick Dunleavy (1991) argues, knowledge of such relationships, whether simply intuitive or confirmed by polling data, gives incumbent parties the opportunity to pursue policies that will increase their vote. By giving larger tax breaks to public schools, reducing the powers and so attractiveness of trade unions, privatizing firms and selling council houses, right-wing parties can increase the number of people going to private schools, owning their own homes and so forth and, by doing so, lead some people to change their conceptions of the good society (see Stubager, 2003 for a detailed empirical analysis). The preferences voters have and, by extension, the location of the median voter is not, it must be concluded, fixed. Parties have a choice. They can either accommodate themselves to the preferences voters have or they can try to 'shape' those preferences to suit their policy preferences. Incumbent parties also have an advantage over opposition ones in that they can maximize their chances of re-election by manipulating the economy to their advantage (see Box 2.3). Assessment In some subsequent chapters, principally but not exclusively those on social choice theory (Chapter 4) and rent-seeking (Chapter 7), I express a number of reservations about the trajectory of rational choice theory. An Economic Theory of Democracy is, however, a book I believe students of politics can all learn from. In the first place, the issue the book addresses is an important one. It obviously matters a great deal whether competition leads parties to move towards or away from the median voter. In the second place, Downs's demonstration that parties will, when a particular set of circumstances hold, converge upon the position of the median voter remains a compelling one. Thirdly, and finally, Downs's argument is an attractive one because it is presented so simply. The literature on party competition An Economic Theory has inspired is also interesting and valuable. It is certainly the case that an increasing level of technical sophistication has made some of this material less accessible to outsiders. But technical sophistication has not become an end in itself. This remains a 'problem' rather than a 'method-driven' area of research (Shapiro, 2005). The issues theorists are continuing to address, about the dynamics of multi-dimensional competition, about the extent and impact of uncertainty and so on, retain an obvious relevance to politics in the 'real world'. In the final part of this chapter I do, however, want to briefly identify one way in which developments within the political science discipline have robbed An Economic Theory of at least a part of its significance. The distinction between positive and normative, between is and ought, is a long-standing and important cornerstone of the way in which we think about the world. Within political science departments, the distinction is usually manifested in the work of political theorists, who think about the way the world ought to be, and others, public policy analysts, area specialists and comparative politics experts, who look at and try to understand the world as it is. Clearly this is a crude division. To the extent that 'ought implies can', political theorists have an obvious responsibility to consider whether their proposals are feasible (see Goodin, 1995). But in terms of the way mostpolitical scientists approach their task, the division is nevertheless a recognizable one. It is tempting to regard Downs's work and the literature on party competition it has inspired in exclusively positive terms as being about the way parties behave. In the introduction to his book Downs certainly encourages such a reading. Casting an envious eye towards the rigour 46 Rational Choice _-_-—--- ■ ; I Box 2.3 Political business cycles ' Political business cycle theory; is associated'with die claims that (i) the - vote an incumbent party receives in-any election depends, at least in part, /■' upon the economy's performance at that time, and (u) that incumbent governments can manipulate die economy in the run-up to an election in order to maximize their chances of re-elecrion. Manyifnotrnosttheorists- '.. alsoclaim that; (iii) manipulating the economy in this way causes long-term economic damage. Stated in this way, political business cycle theory is neither counter-intuitive nor startlingly original. The eighteenth-, century Scottish political philosopher David Hume (1711-76) repeatedly expressed his concern '.hat it 'would be scarcely more impudent to give n prodigal son a credit in every banker's shop in London* than to empower ■ a statesman to draw bills, in this manner on posterity' (quoted, Hanipsher-Monk, 1992:151). Rational choice theorists have, however, i been able to provide a more sophisticated analysis fif the circumstances and ways in which politicians will seek to manipulate the" economy. William Nordhaus (1975) argues that there is a stable and inverse :; relationship between inflation and unemployment which incumbent politicians can exploit by either increasing public'-expenditure or reducing taxesln the run-up to an election.This generates an economic boom [ .;-;leading.tdlowei«unempIoyment-andyeventuaHy,rbut-only.onccianrelec-:.=-'-tion is over, higher inflation. Douglas Hilibs (1977, 1992) argues that political business cycles take a 'partisan' form: that left-wing parties concentrate upon reducing unemployment -and,thatTigh&wing*parties • and status of general equilibrium theory within economics he writes that little progress has been made toward a generalised yet realistic behaviour rule for a rational government similar to the rules traditionally used for rational consumers and producers. As a result, government has not been successfully integrated with private decision-makers in a general equilibrium theory .. . this thesis is an attempt to provide such a behaviour ride for democratic government. (Downs, 1957: 3; emphasis added) Yet, as I argued in the introduction to this chapter, An Economic Tbeoiy can also be understood as offering a normative defence of representative democracy. For if what we require of democracy is a 'tightness of fit' between public opinion and policy outputs, the median voter theorem shows that competition gives parties an incentive to A. Downs and the Spatial Theory of Party Competition 47 pattempt to reduce mflatiomln a sustained onslaughton the assumptions"; .and practice of;Keynesian economics, James Buchanan and Richard Wagner■ .•(19.77-,-1978) argue that incumbent.governments cr.n, in the short term at least, appease voters by offering them both higher public expenditure and lower taxation through the accumulation of ever-increasing budget deficits paid by fmure generations of voters. : ; There are two problems with these models. The first is theirassump-tion that incumbent politicians will always seek to manipulate the ccon-:omyforelectoral:gain;Jf,showever an incumbentpartybelievesthatitis* going to win or lose a forthcoming election come what may,it will have no incentive to try and engineer a boom (S'chultz, 1995). The second problem is ■ ■char these models tend to portray voters as being basically stupid and easily and repeatedly fooled by politicians. If, however, ■people are rational surely they:should expect• politicians., to -engineer-a: boom and if they expect them to do so why would diey dien reward •themac the"polls?.*ln*recent'years'-theorists have sought-to recancilesuch 'rational expectations' With tlu\ existence-of