Political Support and Candidate Choice∗

Political Support and Candidate Choice∗ Hannes Mueller† JEPS Working Paper No. 07-002 http://jeps.repec.org/papers/07-002.pdf May 2007 Abstract This paper proposes a simple model of political supporters in an environment of spatial political competition. We assume that supporters are driven by sympathy for a candidate with similar preferences on their side of the policy space and by fear of a candidate with different preferences on the other side. If parties maximize support in their candidate selection, political platforms can diverge significantly. We show that radical candidates have a positive effect on support for the other party. If candidate choice internalizes this externality, platforms converge and overall support decreases to a minimum. Keywords: Party competition, activism, conflict. JEL: D72, D74. ∗ I am grateful to Timothy Besley, Maitreesh Ghatak and Rafael Hortala-Vallve for comments and encouragement. I further thank the participants of the LSE work in progress seminar and two anonymous referees for useful comments. † STICERD, London School of Economics. E-mail: [email protected] "The political parties participate in the forming of the political will of the people." - Article 21, 1 of the German constitution 1 Introduction It is conventional wisdom that party platforms in many European countries have converged. Prominent examples are Spain, the UK and Germany. In the UK and Germany left parties under charismatic leaders have managed to break long periods of conservative rule by shifting away from old "left" ideals and risking a con‡ict within their own parties. In Spain it were the conservatives under Aznar who were able to beat the Spanish Socialist Worker’s Party by moving to the “centre”1 . The analysis of possible reasons behind converging or diverging political platforms has a long history in economic modelling. Most commonly it is analyzed in regard to voter preferences. However, the focal point in this paper is not the voter but the party activists, supporters or "multiplicators". They are individuals that are able to change voter turnout and possibly voter convictions through information or manipulation and are driven by a desire to help their preferred candidate or party. The reason these people are in the focus of this paper is simple - they seem to disappear. A crude measure for this is the decline of party membership. In their study on party membership in Europe, Mair and van Bietzen (2001) summarize "However, the most striking feature to be noted is the sheer extent and consistency of membership decline through to the end of the 1990s. [...] As these data clearly reveal, parties in contemporary Europe are rapidly losing their capacity to engage citizens."2 And membership decline is not the only sign of decreasing activism. Whitely and Seyd (2002) conduct several surveys with members of the two major parties in the United Kingdom. They …nd that the hours worked by activists decreased signi…cantly for Labour and the Conservatives throughout the 1990s. Looking at di¤erent activities, they …nd that generally the "high intensity"3 activities have decreased the most. 1 In Germany, the social democrats explicitly campaigned with stressing their role in the “neuen Mitte”. and van Bietzen (2001), Abstract 3 These activities are: attending meetings, canvassing voters and standing for o¢ce 2 Mair 1 While the decline of activism might not be such a dramatic phenomenon by itself, it could have drastic consequences for the political system. Due to their unique position within parties, activists are an important link between politicians and voters. Without their support, parties become more dependant on other sources for resources. The theoretical model brought forward in this paper describes the main motivations of party supporters and how these interact with considerations at the party level. It can be shown that a growing weight for strategic considerations in the candidate selection at the party level is linked to converging political platforms and decreasing political support. We start with the observation that voluntary support for parties is of considerable importance for their survival and success. Members alone contribute signi…cantly by providing income and labor resources4 . If a party wants to survive political competition it has to mobilize voluntary contributions. It is assumed here that this mobilization can only be achieved if the candidate of a party is supported by a group of engaged party members and volunteers. The amount of e¤ort these supporters put into mobilization depends on the strength of two psychological drives: sympathy and fear. A candidate with similar policy visions as the activist will motivate the activist to campaign for him (sympathy) while a candidate that has radically di¤erent views will motivate the activist to campaign for the opposition (fear). Given these main driving forces in activist motivation, we model two di¤erent factors for candidate selection at the party level: internal support and support raised in the other party. It seems evident that candidates with more support from their supporters have some edge in winning the internal struggle for leadership. It will be shown, however, that this factor is not consistent with a strategic approach of mobilization for winning an election against another candidate. The results of the model are simple. If parties select their candidates to maximize support from the multiplicators, candidates will diverge in their political platforms. If parties fully internalize the e¤ect a radical candidate on their side has on supporters in the other camp, platforms converge and support for the candidates drops to a minimum. This implies that if institutional, social and technological changes lead to a higher weight of strategic con4 See Scarrow (1996) for some evidence. 2 siderations in the candidate selection, less political support for parties on both sides of the spectrum will be the result. This paper is structured as follows. To motivate our approach and main assumptions we discuss related literature in section 2 and review some evidence in section 3. Section 4 introduces the model and derives the main results. Section 5 concludes. 2 Related Literature The approach and topic chosen in this paper is linked to several streams in the literature. First of all it is closely related to the classical contribution of Downs (1957). While Downs is best known for postulating convergence in political competition he recognized that parties face two political margins: an extensive margin at which parties compete against each other for voters and an intensive margin where each party attempts to bring his own voters into the voting booth or elicit support from members. It is important to note that if one assumes spatial competition, these two forces determine convergence and divergence of policy platform. If competition at the extensive margin is more important, parties will converge because they seek to please the median voter. If competition at the intensive margin determines election fortunes parties will diverge in order to increase turnout. A conclusion which is recon…rmed in our model. The problem with an explanation of divergence through competition at the intensive margin is that turnout to elections remains very hard to model5 and often require ad hoc working assumptions. Glaeser at al (2005) for example state “Our key assumption is that awareness of a politician’s message is higher among the politician’s supporters than among his opponent’s supporters. This asymmetry means that when a politician’s policies deviate from those preferred by the median voter, he energizes his own supporters (who are more likely to be aware of this deviation) more than he energizes his opponent’s supporters (who are less likely to be aware of this deviation).” While our model follows a very similar logics, we will not assume di¤erences in awareness. 5 See Feddersen (2004) and Merlo (2005) for an overview. 3 Instead, we refer back to another branch in the literature reaching back to Coleman (1971). This branch has recognized that parties and politicians often do not follow the narrow goal of winning a federal election. Firstly, candidates have to be chosen within parties before they can compete against candidates of other parties. And secondly, individuals in politics - like voters - are at least partly motivated by a …ght for the “right” policies not power6 . We follow the former line of thought by modelling the struggle for power within parties as a struggle for support from activists. Starting with the observation that politicians need internal support in order to become candidates (see next section) we show that this dependence on internal support might lead to diverging platforms if political activists can only be motivated through political con‡ict. Our way of modelling activist motivation closely resemble the concept of intrinsic motivation described in Besley and Ghatak (2005). In their contribution on public organizations, the authors stress the importance of intrinsic motivation for resolving agency problems within organizations. Activists in parties are often solely motivated by through a close ideological match between them and their candidate and the idea of a “mission” as described by Besley and Ghatak certainly applies. But we stress here that it is not only the mission match within the party that motivates activists but also the mission mismatch to the opposite party. In order to be able to focus on activists we assume a very simplistic voter. In our model voters exist only at the intensive margin - the only decision they make is whether to turn up on the voting booth or not. Our view on voters links this paper to Shachar and Nalebu¤ (1999). They assume that votes are a function of the e¤ort put into campaigning by the political leaders. The e¤ort exerted by politicians in their model depends on the probability of being pivotal and, thus, turnout is higher when elections are close. 3 Political Support in Politics This section discusses the evidence on the role of support in candidate selection and voter turnout. The analysis presented here does by now means prove the mechanisms proposed 6 See Roemer (2001, 2004) for a simple formal formulation of this arguement in a non-strategic setting. The citizencandidate model proposed by Besley and Coase (1997) has a similar assumption at its core. 4 and should only serve as a motivation for our main modelling assumptions. Already a brief look at Germany and the United Kingdom reveals that candidate selection in parties strikes a balance between electability and internal power struggles. Networking within the party is a crucial prerequisite for power in the European political landscape and if a politician wants to become a candidate he/she has to gather as much support within the party as possible. Getting to the top and staying there is unthinkable without internal support. But even enemies within parties are all united by the aim to win the election - an interesting con‡ict of interest. This con‡ict is visible in several pairs of politicians as they struggled and struggle for power. In Germany these pairs are for example Gerhard Schroeder vs. Oskar Lafontaine for the Social Democratic Party (SPD) or Angela Merkel vs. Edmund Stoiber in the conservative camp (CDU/CSU) while in the British context the pair Tony Blair vs. Gordon Brown in the Labour party tells a similar story. All these couples contain a signi…cantly more moderate and a more radical candidate and while electability usually speaks for the more moderate candidate the outcome of internal power struggle is not clear.7 We argue that one of the main driving forces deciding these con‡icts is the attention parties put to the external con‡ict - the upcoming election - when choosing their candidate. If we take this viewpoint it is no surprise that both Tony Blair before elections in 1997 and Gerhard Schroeder before the 1998 campaign stressed electability as the determining factor of candidate selection.8 In line with this shift of focus, both politicians stressed the importance of “modernization” and “professionalization” before and during election campaigns. The terms “spin doctor” and “think tank” gained importance and Gerhard Schroeder’s election campaign team even moved geographically from its traditional spot in the party centre into a new building - the “Kampa”. For both candidates it was a matter of political survival to break traditional forces that had governed party politics. Interestingly, commentators in the following years saw the political landscape in a crisis. At the outset of the German election campaign 2002 supporters were described in the following way “Former activists appear tired, speechless, without impetus and orientation, often passive, 7 Merkel lost her …rst struggle against Stoiber in 2002 while Blair might even loose his Premiership to Brown in 2006. did this quite dramatically by putting himself to a test in Niedersachsen. 8 Schroeder 5 almost standing aside."9 and the turnout in British elections decreased drastically. While turnout had been constantly above 70 percent for decades it dropped to 59 percent in 2001 and only rose slightly to 61 percent in 2005. The low turnout of 2001 triggered discussion about the reasons. According to a Brian Wheeler, BBC News Political Reporter, “The problem could lie with the parties themselves - and the perceived lack of di¤erence between their policies.” 10 Wheeler’s statement points to a central mechanism in determining support - ceteris paribus individuals will be more interested in politics if politics is dominated by a con‡ict. One example of such a con‡ict could be observed in the 2004 presidential campaign of the United States. In the campaign the Republican candidate George W. Bush and his proposed policies polarized the nation along partisan lines11 . And while causality is not necessarily clear, turnout sky-rocketed in that election year. Data from US statistics shows that voter turnout increased drastically in 2004. While it had already marked a signi…cant increase from 1996 to 54 percent of eligible voters in 2000, turnout soared to 61 percent of eligible voters in 200412 . But could the motivation of rank and …le within the party be responsible for changes in turnout? Gerber and Green (2000) stress the importance of direct canvassing for turnout and provide strong evidence for it. If the provision of cheap labour by activists facilitates for this kind of contact within the community, activism and turnout could be closely linked. However, there is some contradicting evidence on the overall e¤ect of voter mobilization and turnout13 . In any case, decreasing membership and support for parties threatens their power and source of funding. A staggering 30 to 40 percent of the total income in 1996 of the two biggest German parties for example were coming from members14 . If membership numbers drop, parties will have to look for alternative sources of funding. This opens the doors to interest group in‡uence and further diminishes the role of parties. 9 Walter (2002), http://www.bpb.de/publikationen/CLAJO6,0,0,Die_Mitte_im_Programmloch.html Wheeler (2001) UK politics: Dead or dormant? 1 1 See Abramowitz and Stone (2006) for evidence of this polarization and correlation with activism and turnout. 1 2 See McDonald (2004). 1 3 See Goldstein and Ridout (2002) and Green and Smith (2003) for two examples. 1 4 Scarrow (1996), p. 178 1 0 Brian 6 4 The Model We assume a one dimensional policy space X = [0; 1]: Two political parties compete for votes by choosing a candidate with policy preferences xL ; xR 2 X: Candidates are assumed to be policy motivated and follow their preferred policy once elected. Candidates are chosen by the parties. The time structure is as follows. In a …rst period, parties on each side of the spectrum the two parties (L and R) choose candidates with xL ; xR . After each party announced their candidate, activists choose a level of support they give to one of the two parties. The level of activism mobilizes the partisan voters. Voters give their vote to one of the candidates. The candidate with most votes wins the elections and implements her preferred policy. 4.1 Voters Voters are modelled as passive partisans15 . These individuals always vote for “their” party in case they cast their vote at all. Once contacted by an activist, the probability that a partisan voter goes voting is signi…cantly increased. It is this ability to mobilize voters that gives activists their role in the political competition. We assume for now that votes by partisan voters are proportional to the time and e¤ort an activist puts into his activities. This has the important implication that votes from partisan voters are proportional to the sum of support from party activists. Adding well informed, independent voters does not modify our results qualitatively. Assuming the standard setting, these voters would create a pressure on the parties to chose a moderate candidate with a platform at x = 0:5 and the more important the group of independent voters, the less will results deviate from that policy platform. For simplicity, we focus on the partisan voters and contrast our results to the standard median voter result of converging platforms. 1 5 See Green et al (2002) for the importance of these voters in determining the outcome of political competition. 7 4.2 Activists In the categorization given by Grossman and Helpman (1995), activists are not driven by an in‡uence motive but by an electoral motive. They are assumed to be “policy takers”, i.e. they do not intend to change the policy of the candidate they support by casting support, they just want to make her win. The intensity of their support is a genuine reaction to a certain constellation of candidates. More speci…cally, the e¤ort activists exert depends on their sympathy towards the candidate in their own party and the fear of the candidate in the other party. We assume a continuum of activists with policy preferences xi 2 [0; 1] ; who are able to mobilize partisan voters. Activists always support the candidate that o¤ers the policy platform closest to them. The activist with the policy preference xi = xL +xR 2 supports none. Assume a linear utility function in policies so that the expected utility of an activist of party L can be written UiL (ei ; xi ; xL ; xR ) = p(ei ) [ jxi xL j] + (1 p(ei )) [ jxi xR j] c 2 e 2 i Where ei is the e¤ort activists i exerts when campaigning for party L, 2c e2i denotes the cost of e¤ort and p(ei ) is the resulting probability that party L wins the election. Note that the activist believes that she will have an impact on the election outcome. This assumption is analogous to the assumption that voters believe they have an impact on the election outcome. While there is considerable controversy over this assumption, it seems reasonable in the context of activists who spend considerable time and money campaigning for their preferred party. Left activists maximize their expected utility by picking a level of e¤ort according to the …rst order condition cei = p(ei ) [ jxi ei For simplicity, we assume that p(ei ) ei xL j] p(ei ) [ jxi ei xR j] = c: The optimal level of e¤ort is then 8 eiL (xi ; xL ; xR ) = for every xi < xL +xR 2 xL j + jxi xR j (1) (i.e. all activists that support the L party) and analogously eiR (xi ; xL ; xR ) = for every xi > jxi jxi xR j + jxi xL j xL +xR : 2 Thus, given the two candidates with policy preferences xL ; xR the e¤ort of multiplicators is dependant on two factors - sympathy and fear. Support derived from sympathy is proportionally increasing in the similarity between the multiplicator’s personal bliss point xi and their preferred candidate. The maximum sympathy can be gained by a candidate whose position coincides with the one of the multiplicator. The further the candidate is away in terms of ideology the less support the multiplicator provides. The amount of support driven by fear is determined by the distance from the worse option to the multiplicator. The further away this option, the more the multiplicator fears that candidate wins the election and therefore exerts a higher e¤ort in his support activities. An important implication of this support function is that support for one party is determined by the platforms chosen by both candidates. It also means that the share of the multiplicators supporting one or the other party is driven by the candidate choices of the two parties. Candidate choice has therefore an impact on who in the population hands out lea‡ets or publicly calls for support of a party and how motivated these supporters are. We follow this thought now. 9 0 xL‘ xL b‘ b xR 1 Figure 1: The e¤ect of a radical candidate L on aggregate support in party L 4.3 Aggregate Support Given the individual support decision described by equation 1 we now turn towards the aggregate level of support. Figure 1 gives a graphic illustration of the di¤erent intensities of support created by two candidates. Given candidate R prefers the platform xR ; a candidate with platform xL would rally support from activists between 0 and b. Since the multiplicator at point b prefers a policy exactly between xR and xL ; she is not willing to support any of them. This is because sympathy and fear are equally weak, i.e. the two terms in equation 1 cancel. However, any supporter to her left will support the candidate that proposes xL : The di¤erence in support is caused by both an increase in sympathy for candidate L and increasing fear due to the growing distance to candidate R’s preferred policy. To the left of candidate L; sympathy for her decreases again and that e¤ect cancels with the increasing fear of the policy xR : In the …gure 1 this results in a constant level of support to the left of point xL : If another candidate with preferences x0L < xL is evaluated, both the border of the supporters shifts towards the left from b to b’ and the maximum support increases. The program x0L leads to less but more motivated supporters. This is an important trade-o¤. The next section shows how it determines candidate choice. 10 Given the individual support it is reasonable to assume that the support for a party can be modelled as the sum over individual support described in equation 1. Under the assumptions xL ; xR 2 [0; 1] and xL xR one can therefore write SL (xL ; xR ) = xL (xR xL ) + 1 (xR 4 xL )2 (2) Where the …rst term describes the rectangular area to the left of xL in …gure 1 and the second term the triangular area between xL and b. As long as xL is not equal to xR both of these areas exist. For candidate R support can be written as SR (xR ; xL ) = (1 xR )(xR xL ) + 1 (xR 4 xL )2 (3) With these support functions, the trade-o¤ between motivation of activists and the number of activists described above can be made clear. The con‡ict in politics is given by the distance xR xL , the larger the distance the more motivated are activists in both parties. On the other hand, a radical candidate in the L party, i.e. a low xL in equation 2 leads to a lower number of activists and therefore harms support. These two e¤ects can also be understood graphically in …gure 1. If the internal battles within the parties lead to a radical candidate the rectangular area will be very high but also very thin. If a more centrist politician succeeds, the rectangular area will be very low but therefore much broader. The result of this trade-o¤ in terms of aggregate support is shown in …gure 2. It displays the aggregate support function in equation 2 for a given xR . Overall support is relatively low at xL = 0: This is due to the fact that there are relatively radical supporters but only a minimal number of them. At xL = 0 the candidate is more radical than anyone else in the population and sympathy for the candidate is relatively low at that point. Support increases …rst as xL grows because multiplicators are still fairly motivated but also grow in numbers. But as xL gets closer and closer to xR = 0:6 multiplicators get so de11 0.14 0.12 0.1 Support 0.08 0.06 0.04 0.02 0 0 0.1 0.2 0.3 0.4 0.5 0.6 xL Figure 2: Aggregate support for party L at xR =0.6 motivated that their increasing number cannot make up for the decrease in e¤ort per supporter. When xL = xR multiplicators are indi¤erent and do not support the candidate. If the candidates follow identical platforms, no multiplicator wants to put e¤ort into convincing her surrounding of any of them. Note that as long as xR > 0 the support function, SL (xL ; xR ) always has a maximum. This fact proves helpful when thinking about the mechanism for candidate choice within the parties. 4.4 Parties and Political Equilibrium We assume that candidates entering the race have their own preferred policy and activists know that politicians cannot commit to any other policy than their favorite policy. This assumption gives very little role for the party in terms of deriving a policy. However, the choice of candidate becomes a crucial factor. We assume that inner party politics can me modelled as some maximization problem over aggregate support which picks one candidate from a spectrum of candidates with preferences on X 16 . While the mechanism for choosing 1 6 Following the citizen candidate model one could think of the process as having more than one stage. Endogenous candidate entry, entry costs close to zero and non-strategic behavior by supporters would probably lead to the described 12 the candidate is not modelled in detail it can be thought of as a mixture of public discussions and political intrigues. A major problem in analyzing the rules of candidate selection within parties is that signi…cant parts of it happen behind the scenes. Even in countries that have primary elections in place, the run up to these elections is still governed by intransparent mechanisms. This section models two major criteria in candidate selection. As discussed in the European cases, one factor in candidate selection seems to be the support from within the own camp for a candidate. It should be clear that politicians are support maximizers to a certain extent. Partly due to reasons of political survival within the party and partly for the pure bene…t of enjoying support17 . On the other hand, if the goal is to win the elections against another party, support in the other camp has to be taken into account, too. A party is not likely to succeed in electoral competition if its candidate mobilizes large shares of the population against herself. The second criterion in candidate selection is therefore the e¤ect on support for the opponent. 4.4.1 Maximizing Support A …rst approach to candidate selection is the assumption that parties are relatively inward looking and the success of a candidate is only dependant on support for her. The politician with most support from multiplicators is then chosen to become the candidate. In that case the candidate selection process can be modelled by the maximization of equations 2 and 3. In short, party L solves the maximization problem maxSL (xL ; xR ) xL (4) on the other side of the spectrum party R solves maxSR (xR ; xL ) xR (5) result. 1 7 When an interviewer asked Gerhard Schröder about his main concerns before giving a speech to a larger audience he replied that he always peeks to see whether the venue is packed. 13 Given that the support functions in equations 4 and 5 are dependant on the policy choice of the other party, we need to specify what assumptions parties make about the competitor’s platform choice. Under the assumption that candidate selection takes into account that the other party chooses a candidate following the same mechanism we have to …nd a Nash Equilibrium in a simultaneous draw from candidates on X to solve equations 4 and 5. An an equilibrium is then de…ned by a situation in which no candidate on either side of the political spectrum can rally more support for himself given the rational expectations about the candidate on the other side of the spectrum. The solution to this problem is summarized in the following proposition. Proposition 1 Assume the candidate selection process solves the maximization problems described by equations 4 and 5. In addition, candidate selection is made under complete information about the selection process in the other party. Under these conditions there exists a unique Nash equilibrium with xL (xR ) = 1 4 and xR (xL ) = 43 : A proof of this result can be found in the appendix. Interestingly the maximization of support seems to lead to a certain degree of divergence. To understand this result it is useful to regard …gure 2 once more. The maximum in support for party L is de…ned by a mixture of maximizing the sum of aggregate sympathy and fear. Clearly, complete convergence at xL = 0:6 = xR causes indi¤erence with all multiplicators and therefore no support. If an equilibrium exists, it cannot be one with complete convergence. Since aggregate support functions feature one unique equilibrium it is straightforward to show that a simultaneous solution of both maximization problems implies a unique equilibrium. 4.4.2 Maximizing Support and Minimizing Resistance Up to now we assumed that the intra-party race for candidacy is depending only on the support that party members can gather from multiplicators. Parties therefore choose a candidate who maximizes support within the party. However, as shown in the German and British case this criterion of candidate evaluation seems to be joined by an evaluation of how the candidate will fare in the political competition against the other party. What is the di¤erence? If 14 0 xL‘ xL b‘ b xR 1 Figure 3: The e¤ect of a radical candidate L on aggregate support in party L and R. winning against the other party is an important factor in choosing the candidate the party members will also evaluate what e¤ect their candidate has on the support on the other side of the spectrum. The di¤erence between the two motives can be made clear with …gure 3. Like …gure 1 it shows the reaction of supporters for the left party in case the party chooses a more radical candidate x0L instead of xL . Only this time the reaction of supporters of candidate R to the change in candidate L is included in the picture. It can be seen that in the camp of supporters for candidate R grows and puts in more e¤ort. The reason is that while sympathy for candidate R has stayed constant the fear of the policy applied in case of a victory of L has grown. If the party has to decide on their candidate it is possible that it takes this e¤ect into account. Denoting this internalization of the external e¤ect with 2 [0; 1] the candidate platform of party L now solves maxSL (xL ; xR ) xL SR (xR ; xL ) and the platform by candidate R, with internalization 15 2 [0; 1] solves (6) maxSR (xR ; xL ) xR Note that = SL (xL ; xR ) = 0 implies the basic model set up. The case (7) = = 1 on the other hand implies a selection process of a candidate that is entirely based on the di¤erence in support between the candidates. It should be stressed that the optimal xR and xL are now functions of and The outcome of this model can be summarized as follows. Proposition 2 Assume the candidate selection process can be described by equations 6 and 7. In addition, candidate selection is made under complete information about the selection process in the other party and the parameters and . Under these conditions there exists a unique Nash equilibrium with xL ( ; ) = 1+2 + 4+2 +2 xR ( ; ) = 3+2 4+2 +2 and A proof of this result is given in the appendix. It can be shown that positive values of ; the partial derivative xL is also positive. That means that the more the candidate selection in party L internalized the e¤ect of their candidate choice on support for candidate R the more moderate will be their candidate. In fact, it can be easily seen that the equilibrium platform for candidate L converges to 1 2 for ! 1: That is, party platform L converge towards the median multiplicator if the party internalizes the e¤ect of their candidate choice on support for candidate R. This is true regardless of the value for . Equivalent results hold !1. It is obvious that the party internal selection processes with internalization leads to a more moderate candidate than the candidate selection based only on maximization of support. This implies that support drops with growing internalization. On the other hand the reaction of the candidate L, to growing internalization in party R is xL = 2 2 2 (4+2 +2 )2 which is negative for all values of 16 < 1. This is a rather surprising e¤ect. It means that with increasing internalization on the other side of the spectrum candidate L becomes ceteris paribus more radical. To understand this result it is helpful to remember that the derivative xL describes a change in equilibrium behavior. In case of a rise of internalization in party R from to ; the party will put more weight on eroding support in party L and therefore move towards the median multiplicator. As long as < 1, party L is still putting a higher weight on maximizing support than on eroding support in party R and will therefore move to the left. How much the optimal platform hinges on the degree of internalization is becoming even more clear if one assumes equal internalization across parties. For xL = = , xL simpli…es to 1+ 4 and analogously xR to xR = 3 4 These results show in a very simple way how the di¤erent weights on support in the own and opposing camp shape the selection process within parties. Parties choose candidates with identical platforms if they fully internalize the e¤ect of their candidate on support in the other camp. It is important to note that this complete convergence implies collapse in overall support. This last result is due to the fact that identical preferred platforms of the two candidates lead to an exact neutralization of fear and sympathy for all supporters. They are indi¤erent to which candidate wins the election and will therefore not put any e¤ort into making one of them win. As shown in this section the parameters and have a huge impact both on political platforms and the support from multiplicators. One way of distinguishing between this model and a more standard approach where parties try to satisfy a party median position at the same time as winning the election is that support in both parties goes down if one 17 parameter or goes up. Convergence by one party then leads to a support crisis in both. If the chance of winning an election is increasing in the distance between the aggregate support for the two candidates, we would expect the party with the higher level of internalization to win more elections. However, if we assume that the work of supporters is actually bene…cial to a democratic societies, i.e. based on information rather than manipulation, democracies could face a public good problem. In a way, con‡ict in politics is then bene…cial because it leads to participation and an incentive for multiplicators to get active. On the other hand the party that invokes con‡ict will probably loose the election18 . 5 Conclusion This paper o¤ers a simple model of political supporters and their motivation for political support. As opposed to the common view that e¤ort in a campaign or …nancial support for a party hinges on the ability to in‡uence policy, supporters in the model presented here are not interested in in‡uencing the policy choice of their own party. Their support is a genuine reaction to a certain constellation of candidates chosen by a party mechanism. We assume that activists are motivated by sympathy for their own candidate and fear of the policies proposed by the opposition. Given this motivation we derive two main results. Firstly policy platforms diverge if power struggle within parties lead to a candidate choice that maximizes aggregate support for the candidate. This divergence is not complete due to the fact that rising motivation of radical supporters is weighed against loosing more moderate supporters to the other party. Secondly, party platforms converge if candidate selection internalizes the e¤ect of candidate choice on support for the political enemy. If aggregate support for the own candidate and for the opponent enter with equal weights, convergence is complete and overall support drops to a minimum. In the light of our discussion in section 3 the model postulates a link between decline in activism and changes in candidate selection in parties in Germany and the UK. There are 1 8 As noted in the introduction, this arguement needs an objective dimension of policy like quality to work. 18 indicators that both Blair and Schroeder won internal party struggles by stressing electability concerns. Convergence therefore went hand in hand with more “spin doctoring” - the concentration on polls to …ne tune political statements and a high weight for media appearance. This change of the parties communication channels, away from the grass roots level towards the mass media re‡ects a less inward-looking approach by party functionaries. One ambition of this paper was to model a basic trade-o¤ governing candidate selection in parties. In order to further our understanding of this selection we have to understand the combined impact of two important factors governing individual behavior in politics: policy and power. We assumed here that supporters are driven by policy only - this prevents motivation through power, i.e. career concerns. An assumption that might be justi…ed in many cases but neglects an important part of party politics. If being a supporter is seen as a precondition to becoming a politician, we would expect di¤erent policy platforms to attract di¤erent types of supporters and therefore to create a di¤erent type of politician. Con‡ict could then indeed be a mechanism to attract policy motivated individuals to parties. A dynamic aspect that resembles the way Caselli and Morelli (2004) model the emergence of “bad politicians”. There is a certain notion of a public good in the model presented in this paper. The public good could be called "political con‡ict". A party that chooses a radical candidate creates con‡ict and therefore political support on both sides of the spectrum. It can be argued that this support is actually a "good". For example, support could mean that more voters are informed about the di¤erent policy choices and therefore cast their vote on a more informed basis. But as shown above, creating con‡ict always implies that the political enemy gains support, too. As a result, the provision of political con‡ict negatively a¤ects the probability of being elected. If this is true, parties could be caught in a prisoners dilemma. They converge and loose not only their base but also the ability to in‡uence the “will” of the people through voluntary support. 19 References [1] Abramowitz, Alan and Walter Stone (2006) The Bush E¤ect: Polarization, Turnout, and Activism in the 2004 Presidential Election. Presidential Studies Quarterly, 36(2), pp. 141-154. [2] Besley, Tim and Stephen Coate (1997) An Economic Model of Representative Democracy. Quarterly Journal of Economics, 108(1), pp. 85-114. [3] Besley, Tim and Maitreesh Ghatak (2005) Competition and Incentives with Motivated Agents. American Economic Review, 95 (3), pp. 616-636. [4] Caselli, Francesco and Massimo Morelli (2004) Bad politicians. Journal of Public Economics, 88(3-4), pp. 759-782. [5] Coleman, S. James (1971) Internal Processes Governing Party Positions in Elections. Public Choice 11, pp. 35-60. [6] Feddersen, Timothy (2004) Rational choice theory and the Paradox of not voting. Journal of Economic Perspectives 18, pp. 99-112. [7] Glaeser, Edward L; Giacomo A. M. Ponzetto and Jesse M. Shapiro (2006) Strategic Extremism: Why Republicans and Democrats Divide on Religious Values. Quarterly Journal of Economics, 120 (4), pp. 1283-1330. [8] Gerber, Alan and Donald Green (2000) The E¤ects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field Experiment. The American Political Science Review, 94(3), pp. 653-663. [9] Goldstein, Kenneth M. and Travis N. Ridout (2002) The Politics of Participation: Mobilization and Turnout over Time. Political Behavior, 24 (1), pp. 3-29. [10] Green, Donald; Bradley Palmquist and Eric Schickler (2002) Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. New Haven & London: Yale University Press. [11] Green, Donald and Jennifer Smith (2003) Professionalization of Campaigns and the Secret History of Collective Action Problems. Journal of Theoretical Politics, 15(3), pp. 321-339. [12] Grossman, Gene M. and Elhanan Helpman (1996) Electoral Competition and Special Interest Politics. Review of Economic Studies, 63, pp. 265-286. [13] Mair, Peter and Ingrid van Biezen (2001) Party Membership in Twenty Democracies, 1980–2000. Party Politics, 7 (1), pp. 5–21. [14] McDonald, Michael P. (2004) Up, up and away! Voter participation in the 2004 presidential election. The Forum: A Journal of Applied Research in Contemporary Politics 2(4): Article 4. [15] Merlo, Antonio (2006) Whither Political Economy? Theories, Facts and Issues. forthcoming in R. Blundell, W. Newey and T. Persson (eds.), Advances in Economics and Econometrics, Theory and Applications: Ninth World Congress of the Econometric Society. Cambridge: Cambridge University Press. [16] Niedermayer, Oskar (2003) Parteimitglieder seit 1990: Version II/2003, Berlin 2003. [17] Osborne, M.J. and A. Slivinski (1996) A Model of Political Competition with Citizen Candidates. Quarterly Journal of Economics, 111, pp 65-96. [18] Roemer, John E. (2001) Political Competition: Theory and Applications. Cambridge, Mass. & London: Harvard University Press. 20 [19] Roemer, John E. (2004) Modeling Party Competition in General Elections. Cowles Foundation Discussion Paper No. 1488. [20] Scarrow, Susan E. (1996) Parties and Their Members : Organizing for Victory in Britain and Germany. Oxford & New York : Oxford University Press. [21] Shachar, Ron and Barry Nalebu¤ (1999) Follow the Leader: Theory and Evidence on Political Participation. The American Economic Review, Vol. 89 (3), pp. 525-547. [22] Walter, Franz (2002) Die Mitte Im Programmloch. Aus Politik und Zeitgeschichte, B21 / 2002. [23] Whiteley, Paul and Patrick Seyd (2002) High Intensity Participation: The Dynamics of Party Activism in Britain. Ann Arbour: The University of Michigan Press. [24] Wittman, Donald (1973) Parties as utility maximizers. American Political Science Review, 67, pp. 490-498. A Divergence due to Support Maximization Given the two conditions for the political platforms of the party leaders xL (xR ) = arg max SL (xL ; xR ) and xR (xL ) = arg max SR (xR ; xL ) these platforms have to solve to …rst order conditions xR 1 (xR 2 2xL and 1 2xR + xL + xL ) = 0 1 (x 2 R xL ) = 0 Both second order conditions show that the extreme values xL and xR are maxima. An equilibrium requires that both platforms are best responses to the platform of the other party. Only then there will be no other candidate within the party that can threaten the position of the party leader. Solving for xL yields xL = 13 xR . Plugging the optimal decision rule of party L into the …rst order condition for xR yields the equilibrium values of 41 and 43 respectively. B Equilibrium Values with Internalization The …rst order condition for the optimal choice of xL is xR 2xL 1 (xR 2 xL ) xR 1 1 (xR 2 xL ) = 0 this simpli…es to + (1 ) 2 xR = (3 + ) xL 2 The …rst order condition for the optimal choice of xR is 21 (8) 1 1 2xR + xL + (xR 2 xL ) 1 xL + (xR 2 xL ) = 0 and simpli…es to 1+ (1 ) 2 xL = (3 + ) xR 2 (9) Solving for xR in equation 9, plugging into equation 8 and some algebra yields the equilibrium value for xL : If that result is then plugged back into equation 8 some manipulation reveals the equilibrium value for xR : 22