6 Part II focuses on the political consequences of candidate selection methods through an analysis of four democratic dimensions: participation, representation, competition, and responsiveness. All of these dimensions are part of our basic understanding of modern representative democracy - a regime in which the citizens participate in choosing among parties or candidates who compete with each other in an attempt to be the representatives of the people, and who are expected to demonstrate responsiveness to their demands after they are elected. However, modern representative democracy is implemented at the national level. What happens within parties? The answers are not the same as they are for the national level, because parties are - as their name suggests - part of a democratic whole, not the democratic system but a subsystem within it. Our analysis mainly involves the relationship between the level of inclusiveness of the selectorate and these four dimensions. We would have liked to address other relationships, and to analyze such intriguing issues as the influence of exclusive candidacy requirements on representation (do strict candidacy requirements breed unrepresentative candidates?); the impact of decentralization on turnout (does decentralization increase- turnout in-eandidate-selectioa, - especially in the cases ■ of the more inclusive selectorates?); and the effect of the intraparty voting systems (is it similar to the national level, where proportionality encourages participation?). Such questions, and many others, can be asked only in the future, after candidate selection is sufficiently researched to the point that scholars will have enough comparative data to begin answering them. We hope that this part of the book will ignite enough interest to lead down such a path. Participation Participation is a central dimension of democracy. In a modern representative democracy, the entire adult citizen population has the right to elect the representatives who will govern them. Democracy at the national level requires universal participation, that is, maximum inclusiveness. But what about participation at the inrraparty level? In this chapter we address this issue, looking at participation as inclusiveness and as turnout, as well as the question of the quantity versus the quality of participation within parties. In many democracies, candidate selection methods are becoming more inclusive (Bille 2001; Hazan 2002; Kittilson and Scarrow 2003; Scarrow, Webb, and Farrell 2000). If in the past most party selectorates were composed of party delegates - standing agencies such as national or local committees and congresses, or special nomination or selection conventions - today more and more parties give rank-and-file members the right to influence candidate and leadership selection. This general trend makes analyzing the expansion of the selectorate - particularly the adoption of the highly inclusive method of party primaries - relevant for virtually all democracies, whether they are moving in a similar direction or if their political actors are contemplating such.a shift. Although there is much debate over the extent of the decline in party membership and its interpretation, its occurrence - indicated by both absolute and relative measures - is a clear empirical finding (Mair and van Biezen 2001; Scarrow 2000). In light of this phenomenon, one of the ways that citizens are brought back in by political elites is through increasing their role inside parties (Scarrow 19996). This is often expressed by giving rank-and-file party members influence over important intraparty decisions, among them candidate selection. While this phenomenon is recognized in the research literature, its political consequences are still in need of systematic evaluation. In this chapter, we are concerned with the actual impact of democratizing candidate selection methods on patterns of political participation - specifically the political consequences of expanding the selectorate from an exclusive to a more inclusive body of participation. For example, what are the ramifications of a party's decision to shift its selection of candidates from a small nominating committee to a larger group of party delegates? Or alternatively, what is the effect of shifting from the latter to party primaries? In party primaries, the focus of this chapter, party members' votes decide who the party's candidate will be in 90 Democracy within Parties a single-member district or the candidacies (and often the ranking of candidates) on the party list for general elections. However, before analyzing the political consequences of intraparty democracy for participation, we must distinguish between two very different concepts related to participation: inclusiveness and turnout. rNCLUSIVENESS VERSUS TURNOUT Candidate selection methods can be distinguished according to the four dimensions outlined in the earlier chapters. This chapter focuses on the party selectorate - the body that selects the party candidates. Party selectorates are classified according to their level of inclusiveness. At one extreme, the selectorate consists of the entire electorate of the nation; that is, all citizens who are eligible to participate in general elections. At the other extreme, the selectorate - or rather the selector - is a single party leader. Between these poles, we can find various alternatives, from a relatively inclusive body of party members, through party delegates, and up to a small, exclusive nominating body that is composed of just a few leaders. A completely different concept is that of turnout in the candidate selection process, which ranges from very high to very low, regardless of the size of the selectorate. A very inclusive selectorate, such as that of all parry members, can produce either high or low turnout. By the same token, a relatively exclusive selectorate may also shift along4he4urnotrt^^ a clear example of how inclusiveness and turnout can vary independently of each other (Norrander 1986). Turnout increases in those primaries where there is competitiveness and decreases in those where the race is not even close, regardless of the level of inclusiveness - even across states or time where this level is constant. In those states where one party clearly dominated, and the primaries determined the eventual winner of the general election, turnout for selecting the candidate was sometimes even higher than in the general election (Key 1949). In addition, variables such as education, age, income, and profession influenced the level of turnout across similar candidate selection methods, even more than they affected turnout in general elections (Crittenden 1982). One can envision the inclusiveness—exclusiveness dimension as the horizontal axis and the level of turnout as the vertical axis. These two can vary independently, since expanding the amount of people who are allowed to take part in the candidate selection process does not require them to actually participate. While it is quite likely that the absolute number of people participating will be higher in the more inclusive candidate selection methods, it is likely to be relatively lower than in the more exclusive arenas. We can predict this negative correlation f: Participation 91 I between inclusiveness and participation by following Olsonian logic: the more t potential participants there are, the weaker the incentives are to participate actively j (Olson 1965). Indeed, Tan (1998) and Weldon (2006) found that the larger a party is (in terms of membership), the less active its members are. Evidence that is presented later in this chapter points in the same direction - turnout tends to decline as we move from an exclusive selectorate to a more inclusive body. rNCLUSIVENESS Israel's experience with the democratization of several of the parties' candidate selection methods is a case in point. In the pre-state years, and in the first decades after independence (1948-73), the parties used highly exclusive selectorates, such as nomination committees composed of a few leaders, to select their candidates. In the 1970s and 1980s, many parties transferred candidate selection to their wider and more representative institutions, composed of party delegates. This was the first step toward expanding participation, which was followed in the 1990s by several parties further opening up with the adoption of inclusive party primaries. The absolute selectorate figures clearly show that Israel's main parties underwent a dramatic increase in political participation when selectorates ranging from 1,269 (Labor's Central Committee delegates in 1988) to 3,153 (Likud's Central Committee in 1992) were replaced by those ranging from 178,852 (Likud's membership in 1996) to 261,169 (Labor's membership in 1996). This was the essence of the adoption of party primaries, hi this sense,, absolute political participation increased dramatically. Moreover, since the Israeli electoral system is a national closed-list system, which gives no say to the voters as to the composition of party lists, the party primaries provided an important new venue for increased participation. In the 1992-2009 period, the average ratio of dues-paying members (those who were eligible to participate in party primaries in the main parties in Israel) to party voters stood at 1:5.3 (Table 6.1). In the nine cases of party primaries in Israel to date, between one-third and one-ninth of the main parties' voters - that is, the potential population of party members, if one assumes that a party member is likely to be a party voter - chose to register as members. Of these, an average 57.6 percent turned out to vote in the primaries, which means that on average approximately 1 in 9.8 party voters (10 percent) took an active part in the candidate selection process. These figures indicate the creation of a new, significantly more inclusive arena of political participation. Did the more inclusive candidate selection methods, such as party primaries, bring citizens back into party politics? Rahat and Kazan (2007) analyzed the Israeli case and concluded that primaries brought about only a provisional 92 Democracy within Parties Table 6.1. Number of party members, participants in party primaries, party voters, and their ratios in Israel's main patties, 1992-2009 Party Year Number Number of Number of Ratio Ratio of party participants in voters in general members: participants: members candidate selection elections voters voters Labor 1992 164,163 118,197(72.0%) 906,810 1:5.5 1 7.7 Labor 1996 261,169 194,788 (74.6%) 818,741 1:3.1 1 4.2 Labor 1999 163,044 101,087 (62.0%) 670,484 1:4.1 1 6.6 Labor 2003 110,988 58,783 (53.0%) 455,183 1:4.1 1 7.7 Labor 2006 116,948 68,331 (58.4%) 472,366 1:4.0 1 6.9 Labor 2009 59,025 31,789 (53.9%) 334,900 1:5.7 1 10.5 Likud 1996 178,852 91,907 (51.4%) 767,401 1:4.3 1 8.3 Likud 2009 98,492 48,458 (49.2%) 729,054 1:7.4 1 15.0 Kadima 2009 79,649 35,125 (44.1%) 758,032 1:9.5 1 21.6 Year indicates the year of the general election prior to which the primaries took place. Source: Data from the political parties and newspapers. resurgence in party membership.1 Israel's experience with trying to meet the challenge of declining party membership by empowering party members is quite common. Research on participation and activism in German parties in the 1960s found that party members did not take advantage of the participation mechanisms that were available to them (Gunlicks 1970). Nevertheless, in the 1980s and 1990s, the veteran German parties adopted additional participatory devices, including member participation in candidate and leadership selection, as a reaction to declining membership and electoral Kisses, and in response to demands for more direct political participation. Scarrow (1999Z>, 2002) concluded that reforms meant to bring about more inclusive parties did not succeed in bringing more citizens into party politics, but did empower those who were already members. The German Green Party institutionalized a series of organizing principles inspired by grassroots participatory ideals, but still failed to attract large numbers of party activists. Poguntke (1992) argued that those people who could be expected to be active in party participatory democracy are loyal to specific policies and not to a party organization. They can be mobilized for the promotion of specific goals, but not for continuous partisan activity. In addition to the cases of Israel and Germany, Britain (Webb 2002), France (Rnapp 2002), the Scandinavian countries (Sundberg 2002), and Ireland (Murphy and Farrell 2002) all provide examples of parties that expanded and empowered their selectorates. Yet in these cases too, their efforts railed to enlarge the number of members significantly. ' More recent data on membership in 2006 and 2009, included in Rabat (Forthcoming), substantiates these findings. '< Participation 93 • The parties' adoption of more inclusive participatory arrangements in their ! candidate selection methods seems to have had only partial success in combating 1 declining membership and in engaging citizens to participate in party politics. ; Attempting to be modest in our expectations, we can indeed argue that adopting "J more inclusive candidate selection methods, such as party primaries, resulted in '': two positive developments for participation, and therefore for democracy: First, i although it did not bring about a serious and sustained increase in party member- S ship that brought citizens back into party politics, it did stem the decline in the j number of party members, or at least slowed it down. Second, a new arena for I political participation was created wherein citizens could influence aspects of politics that they were unable to before, which was especially important in those institutional settings where voters previously lacked such influence (Cross 2008). TURNOUT The average turnout of party members in Israel's party primaries is 57.6 percent, ranging from 74.6 percent in Labor in 1996, to 44.1 percent in Kadima prior to the 2009 election (Table 6.2). Turnout in Labor has declined over time, while in both Likud and Kadima it never reached high levels - with more recent figures at around 50 percent or below. Table 6.2 shows that in Israel there were differences in the rate of participation between the parties, and also for the same party at different times. One is tempted to argue that such political variables as competitiveness, the party's being in government, or the party's public support affect the rate ..of member participation. Yet, it appears that the best predictor for turnout rates is the relative distribution of voting stations. This explains the higher levels in Israel's Table 6.2. Turnout in party primaries for selecting the candidate lists compared to general elections in Israel, 1992-2009 (in percentages) Year Labor Likud Kadima General elections 1992 72.0 — — 77.4 1996 74.6 51.4 — 79.3 1999 62.0 — — 78.7 2003 53.0 — — 68.9 2006 58.4 — 63.5 2009 53.9 49.2 44.1 64.7 Average 62.3 50.3 44.1 72. i Year indicates the year of the general election prior to which the primaries took place. Voter turnout in general elections is from the Central Elections Committee (Tittp://www. knesset.gov.il/eleetionsl 8/heb/history/PercentVotes.aspx), last accessed 24 Fehraary 2010. Source: Data from the political parties and newspapers 94 Democracy within Parties Labor Party compared to Likud; and the decrease in participation in Labor primaries over the years is correlated to die adoption of a more frugal policy concerning the spread of polling stations. The fact that this "technicality" is the best predictor for turnout stands in contrast to the turnout rate in party congresses, which take place in one location yet manage to draw participation of more than 90 percent. When one remembers that members pay for the right to participate in these party primaries, turnout seems quite low. Similar, and even lower, rates of membership turnout in candidate selection were recorded in other countries. Obler (1970) reports - on the basis of data from fifty cases - that turnout in the Belgian party primaries in 1958-65 was 58.9 percent on average; in only 16 percent of the cases was it more than 60 percent. Membership turnout in candidate selection in Canada was between one-third and one-half (Cross 2002,2004). In Finland, it stood at 39-45 percent prior to the 1979 elections, while prior to the 1995 elections the averages for the four largest parties in the various districts were between 20 percent and 63 percent (Helander 1997; Kuitunen 2002). In 1999, two-thirds membership turnout was recorded in Fine Gail (Gallagher 2003). In Taiwan, the turnout in the Kuomintang primaries stood at 45.1 percent in 1989,29.1 percent in 1991, and 29.6 percent in 1992 (Wu and Fell 2001), while in the Democratic Progressive Party in 1989 it stood at 70 percent (Fell 2005). In Denmark, membership participation through postal voting - in those parties in which voting is not just consultative - was between 53 percent and 76 percent (Pedersen2001).2Inthe more inclusive cases, turnout was lower. In Iceland's party primaries, turnout in 1983 was 26 percent, and in 1987 it was only 16 percent (Hardarson 1995). In Argentina, between 1989 and 2003, the Peronist Party had an average of 14 percent of registered voters participating in its "open" primaries, and only 7 percent when it used closed .primaries,, while the Radical Civic Union experienced even lower figures - 5 percent and 2 percent respectively (Jones 2008). In the United States, turnout for the primaries for the House of Representatives in twenty-four states with closed primary systems stood at 27-37 percent of those voters who registered as Democrats and Republicans (Ezra 2001). The turnout statistics given above reflect various countries and parties in diverse settings (from voting in a meeting to voting through the mail), which allows us cautiously to propose two conclusions: First, overall turnout of party members is somewhere around 50 percent, or somewhat lower, with a wide range of 20-75 percent. Second, turnout is lower in the more inclusive cases where participation is open to all voters. In order to give these turnout statistics more meaning, we now compare them to turnout in general elections and to turnout in the more exclusive selectorates. Participation rates for events other than candidate selection are similar. For example, rates of participation in mayoral candidate and leadership selection in German parties were between 34% and 57% (Scarrow 19996). Weldon's (2006) data on party members' activities in general, based on surveys, is similar. 1 Participation 95 I If we evaluate turnout in party primaries in Israel vis-a-vis turnout in the general j elections, we see that in all cases the latter is higher than the former (Table 6.2). j While party members invested more than the voters in securing their right to vote i (they even paid dues in order to become members), they still demonstrated lower \ turnout levels. As far as we can tell, this is true for practically all political parries in the democratic world, even those where selection was the true election.'' If we I follow Olsonian expectations based on the size of the selectorate (assuming that rewards are equal), then we would expect a higher turnout among party members than among regular voters. We can thus conclude that the incentives for participating in party primaries are perceived as lower than those in general elections, 1 somewhat similar to second-order elections in which such levels of turnout are I quite common. A comparison of turnout in party primaries to turnout in the more exclusive selectorates composed of party delegates tells us that Olsonian logic works well when the return is similar - that is, turnout in candidate selection among party delegates is much higher. Data collected in Israel are revealing because although they do not systematically cover all cases - due to the poor levels of documentation in intraparty selection - the picture is clear. For example, while turnout in Likud primaries in 1996 and 2009 was around 50 percent, turnout in its central committee selection was 92 percent in 1992 (first selection round); 88 percent in 1999; and 91 percent in 2006. In 1988, the only time that Labor selected most of its candidates using its central committee, turnout was 98 percent (first round), compared to 53-75 percent recorded in the primaries. Turnout in the cases of selection by party agencies in other Israeli parries was also similar: 79 percent, 87 percent, and 89 percent turnout in Meretz in 2003,2006, and 2009, respectively; 92 percent in the National Religious Party's centra! committee in 2003: and 92 percent in Hadash. in 2009...... A comparison with data from the United Kingdom reveals a similar picture. The British Conservative Party used a general meeting of party delegates in the last stage of choosing its candidates for the European Parliament in 1979. Turnout was higher than 70 percent in two-thirds of the cases, and over 90 percent in one-quarter of cases. At the same time, the Liberal Party used party primaries in some constituencies, and turnout was quite low - the highest turnout was only 34 percent (Holland 1981).4 The Kuomintang in Taiwan used two selectorates at the same time: party activists (cadres) and party members. These two selectorates, each with a 3 The exception is the southern United States, where turnout in the primaries was higher than the general elections during the hegemony of the Democratic Party (Key 1949). 4 The Labour Party also used party delegates in the final stage, with somewhat lower turnout. Compared to the Conservatives, where only 5% of cases had a turnout of lower than 50%, in Labour it was 10% of cases. Regardless, the two parties combined show that the lowest level of turnout for party delegates - even after the vast majority of candidates had already been filtered out - was almost the same as the highest turnout exhibited by the more inclusive party primaries of the Liberal Party. 96 Democracy within Parties different level of inclusiveness, showed a large gap in turnout prior to the 1991 and 1992 elections - just under 30 percent among the party members, but almost 70 percent among the party activists (Wu and Fell 2001: 32). Data on turnout in leadership selection in several countries paints a similar picture. Turnout in the exclusive selectorate of the parliamentary parry is higher than turnout among party delegates; turnout among the latter is, in turn, higher than turnout in the more inclusive selectorate of party members (Kenig 2007). Turnout among party members is lower in comparison with turnout in general elections and in candidate selection among party delegates. This might not be a surprise since candidate selection is seen as less significant than general elections and party members are expected to be less committed, or less motivated, than parry activists. If inclusiveness was expected to produce a new and more involved arena for participation, then this expectation was not met. If the expectation was more modest, such as the creation of just one more arena for participation, then it could have been fulfilled if we would have been able to avoid the problems presented in the next section. QUANTITY VERSUS QUALITY: THE PATHOLOGIES OF INCREASED TNCLUSIVENESS Beyond the positive desire to expand participation, for whatever reason, there are political consequences that need to be taken into account by parties before they become more internally democratic. The inability to retain an increase in membership (if there was any such increase), and to convince new members to indeed turn out in higher numbers, are two indicators of problems resulting from expanded participation, but there are more phenomena that must be considered. The overarching question, which will be assessed here on several levels, is: Does adopting more inclusive participatory methods of candidate selection influence the quality of party participation? We now look beyond the overall numbers to analyze the question of the quality of membership and its meaning. Duvergefs (1954: 90-116) taxonomy of degrees of participation in political parties (Figure 6.1) sets reasonable expectations of party members. It is built from concentric circles of increasing affiliation and participation. The widest circle is that of voters, citizens who merely vote for a given party. The next is that of supporters, voters who also acknowledge that they favor a particular party and may occasionally speak on its behalf. The third circle is the party members, who are at minimum supporters who are formally registered with the parry, and a minority of whom actually takes an active part in party activities (Selle and Svasand 1991; Heidar 1994). Participation 97 Figure 6.1 Degrees of participation in political parties Source: Elaborated from Duverger (1954: 90-1) Finally, we find the innermost circle of mihtants or activists, members of the party who see to its organization, operation, propaganda, etc. Following Duverger's taxonomy, and that of other scholars (Seyd and Whiteley 1995), party members are expected at minimum - even when parties intentionally lower the barriers for entrance in order to recruit "supporters" (Scarrow 1994) - to be loyal voters for the party and to be affiliated and engaged with it for more than a short period. The following discussion points out several disturbing pathologies concerning inclusiveness within parties: the majority of party members might not forge any long-term affiliation with the party, but rather register with the sole purpose of taking part in the more inclusive candidate selection process; many party members might not fulfill even the minimal requirements of being party voters, and the incentives for mass registration may even encourage corruption. In short, primaries may result in instant, opportunistic, and corrupt membership. 98 Democracy within Parties Participation 99 Pathologies of Mass Registration Over eighty years ago, Merriam and Overacker (1928: 5) pointed out the abuses that arose with the adoption of primaries. "In the first place it soon became evident that there was no guaranty that participation in a party caucus or primary could be confined to members of the party immediately concerned____Parry primaries were invaded and controlled by men of a different or of no political persuasion, and from other districts—" Duverger (1954: 361) described the final stage of candidate selection by the Belgian Christian Social Party in 1949, which included a general poll of the party members on the rolls during the current year. "In certain arrondisse-ments, candidates were known to have organized a member hunt and to have enrolled hundreds of members a few days before the closing date for participation in the poll. Some even tried to buy books of blank members' cards."5 The abuses encountered when parties become more inclusive have already received significant scholarly attention across a variety of cases. The most blatant of these is the resulting massive registration drives of party members by the candidates, which produce not only inflated membership figures but also "instant" members who come and go quite rapidly. Instability is another sign of the problematic quality of membership, as it indicates that people join the party, or are recruited, yet do not stay affiliated after the primaries are over. A measure of this is the difference between the number of party members during primaries and the number of members when parry primaries are not taking place. Several scholars conducting research on the Canadian parties have provided ample evidence of this. Malloy (2003), for example, states that Canadian party membership is extremely volatile, with cyclical patterns that peak as candidate selection nears due to massive enlistment campaigns, and most do not renew their membership. His description is worth citing in fidi: It is established practice in Canadian politics for potential nominees... to recruit and transport large numbers of new constituency association members to local meetings solely to support their candidacy, to the point that in 2002 one of the chief concerns ... was disputes over the rules for handling members forms and how many could be copied or distributed at a time. These "instant Liberals," "instant Conservatives," and so on, often have no previous connection to the party and are commonly recruited in mass numbers from ethnic groups, senior citizens, and youth. Most sitting MPs are able to recruit sufficient numbers of local members to overcome any challengers recruitment efforts, sometimes to ludicrous extents as local associations increase by an average of 300 per cent for nomination meetings. Once they have obtained membership and supported their candidate in the nomination meeting, few renew their membership or otherwise participate in the party. (Malloy 2003: 126) Duverger cites an administrative report of the General Secretary to the party congress. In the elections of 1993, in 84 percent of the Canadian constituencies with competition, the winning candidate reported enlisting new members into the party; and in one-third of the constituencies with competition, the number of new members enlisted by the candidates more than doubled the existing membership (Erickson 1997). Carty, Cross, and Young (2000) reported an increase in party membership of 60 percent and 70 percent, and up to 300 percent in some instances, prior to candidate selection. These new members enlist only in order to take part in candidate selection and do not renew their membership (Carty and Cross 2006). While instant membership doubles the number of party members in Canadian parties, within two years it returns to the earlier figure (Carty and Eagles 2003). Moreover, not only did the candidates conduct a massive enlistment of members into the party, some candidates also paid the membership dues (Cross 2006). The problem of instant members has brought about the creation of internal and external committees in the Canadian Liberal Party to look into the matter, but many of their proposals have not been implemented (O'Brien 1993). Registration campaigns in Israel prior to the primaries led to an increase of between 59 percent and 332 percent in the number of party members (Table 6.3). When there were no primaries on the horizon, the number of members dramatically decreased back to former levels. It would appear, therefore, that most of the members joined the parties - or were recruited - with the sole intention of participating in the more inclusive candidate selection process, and not in order to create a significant link between themselves and the party.6 Table 6.3. Membership at the beginning and at the end of registration campaigns in Israel's main parties, 1991-2008 Party Year Number of members at beginning Number of members at end Growth rate from beginning (%) Decline rate til! next beginning (%) Labor 1991-2 80,000 164,163 105 -51 1995-6 80,000 261,169 226 -73 2001-2 70,000 110,998 59 -57 2005 48,000 119,717 149 -50 2007 60,000 103,568 73 — Likud 1992-3 50,000 216,000 332 -58 1995-6 90,000 178,852 99 -44 2001-2 100,000 305,000 205 — Source: Data from the political parties and newspapers * Another indication of the problematic nature and quality of party membership is the phenomenon of "double registration," the simultaneous enrolment of citizens in more than one party. While this is usually not against the law, it might he prohibited by particular parties (such as the New Democratic Party in Canada). In Israel, it stands not only against the rules set in the parties' constitutions, but is also against the Parties Law. Regardless, the cross-referencing of party members' names conducted in 1996 revealed that 8% of Labor members and 12% of Likud members were also members of another party (Israel, Party Registrar, 13 March 1996). 100 Democracy within Parties Participation 101 Scherlis (2008: 587) describes the Argentine experience with primaries as follows: Soon, party primaries became blatant clashes between electoral machines. Victory in primaries depends on the capacity to mobilize clienteles. This entails significant spending for a variety of goods and services, from the salaries of the local brokers, who establish the link with the clients, to the funds for buses, taxis and drivers needed to transport voters to the polling stations. In these primaries ideological issues are not at stake, nor do the contestants' backgrounds matter. Hence, who wins is solely determined by the resources at their disposal and their efficient use. In Taiwan's Kuomintang Party, whoever succeeded in enlisting more new members to take part in candidate selection won {Fell 2005), while in the Progressive Party members were recruited and their fees were paid, and there was even bribery (Baum and Robinson 1995). The adoption of one-member-one-vote party primaries in British Labour resulted in pressure for the massive registration of members - for example, the party's National Executive Committee (NEC) reported that 800 out of 4,000 members in four constituencies in Birmingham were not even registered as voters. Another example is that of 217 membership forms received with a single personal check; and after a two-year investigation, the NEC decided to accept only nine as members - the other 208 either did not want to be members or did not exist (Criddle 1997:192-3). Party primaries also produced massive ethnic registration -for example, Asians in Labour by Asian candidates - a phenomenon that also took place in Canada (Carty and Cross 2006). These could be perceived as channels for marginalized groups to influence politics, but in reality they are better captured as classic patron-client relations. An examination of the 1996 Israeli election results uncovered thirteen towns in which the number of Labor Party members was larger than the actual number of voters (Rahat and Sher-Hadar 1999a). A shared characteristic of most of these towns - low socioeconomic standing - indicates that media allegations concerning patron-client methods in the registration campaigns were evidently grounded in reality. Back and Solomos (1994) argued that in the British Labour Party, candidates paid membership dues, members were enlisted without their knowledge, and often registration was based on patron-client relations. Intraparty democratization in Mexico also encouraged the creation of patron-client relations between members enlisted from weaker socioeconomic strata and the candidates and their affiliates (Combes 2003). In 2005, the Israeli Labor Party decided to nominate a committee, headed by a former judge, to ensure that all new registered members were recruited according to the law - that is, that they voluntarily joined the party, paid their membership dues, and were not members of other parties. As a result, almost one-half of the 90,000 new memberships were canceled on legal grounds. We argue that a similar Pandora's box is likely to be opened elsewhere, and that parties around the world choose to ignore these problems, just as they do in the case of the financing of intraparty selection (Hofhung 2006). It appears that many citizens register with a party in order to select a particular candidate, with no intention of voting for this candidate's party in the general election. In some Icelandic districts, the number of participants in the primaries outnumbered the total number of party voters - in one district the former was 140 percent of the latter (Kristjansson 1998, 2002). Hardarson (1995: 163^1) examined which percentage of those who took part in the Icelandic primaries also voted for that party in the general elections. The results were between 75 percent arid 90 percent, depending on the party, for the 1983 and 1987 elections, with the overall average being 81.1 percent in 1983 and 82.5 percent in 1987. This means that almost one out of five who took part in a particular party's candidate selection process did not vote for that party at the subsequent election. These phenomena do not result from the adoption of primaries per se, since such occurrences are also evident in parties that do not hold primaries - for example, in the massive registration of members prior to the selection of delegates for positions in a party agency. However, they provide an indication that the quality of empowered membership is problematic and may even decline following the strong incentives that such a system creates to register a large quantity of new members, regardless of their quality. Uninformed and passive members In Israel, many who were recruited in massive registration campaigns were uninformed to the point that they did not know that by registering to take part in the party's internal contest they became party members (Rahat and Sher-Hadar 1999c).7 This is but one indication of the problematic nature of the relationship between the parties and their newly empowered members. Kristjansson (2004) showed a similar gap between the number of party members listed by the parties in Iceland (just over 50,000 - which is one-quarter of the electorate) and the result of a survey from which one can conclude, on the basis of the respondents' answers, that only about 34,000 considered themselves to be party members. He concluded that 40 percent of the members listed by the parties "entered the party, without any feeling of attachment, in order to vote in the party's primary" (Kristjansson 2004: 65). Taking into account Duverger's (1954) taxonomy (Figure 6.1) - together with the fact that many registered members joined, or were recruited, only for the sake of the primaries - we would expect the members' participation rate in party 7 A survey prior to the 1996 elections in Israel (Arian and Amir 1997) found a relative gap of almost 50% between those who declared that they were party members (9%) and those who declared that they participated in the party primaries (13%). 102 Democracy within Parties Participation 103 primaries to be somewhere between the level of participation in candidate selection by party agencies (activists/delegates) and participation in general elections (voters). That is, party members should be less motivated than party activists, but more so than the average citizen. However, as already shown, the average turnout rate for primaries appears to be lower than for the general elections. Parties, as voluntary associations, must have the ability to use selective incentives in order to reward their activists and encourage them to continue to work for the party organization. From this perspective, the adoption of party primaries — which can bring about mass registration on the eve of intraparty elections - is damaging for the party. Registration that suffers from the pathologies outlined above does not serve the party, but rather the immediate personal needs of the candidates. Enhanced and equivalent inclusive political participation in candidate selection damages the differential structure of rewards in parties - the privileges of long-time loyal activists are equal to those of new, temporary, and unfaithful registrants. The strategic few and the passive many With the adoption of a more inclusive selectorate, all interested players - the candidates, the members, and the parties - place an emphasis on the quantitative side of the registration campaign rather than the qualitative. They interpret numbers as political power, and seek to reap the immediate rewards vis-á-vis each other, concentrating on both the intraparty struggle and the forthcoming general election. The phenomenon of instant, opportunistic, corrupt, and unobligated (and sometimes even uninformed) membership, together with the fact that approximately one-half of party members do not bother to participate in party primaries, indicates that most members are not strategic actors who decide to take advantage of the opportunity that party primaries grant them. On the contrary, the new party members play the game as relatively passive participants, and this is despite holding positive attitudes about the new participatory devices and calling for further reforms in this direction (Young and Cross 2002). The gap between idealistic stands concerning democracy within parties and real-life abuse of these participatory opportunities is a result of the interactions between a few interested strategic actors — such as interest group leaders, vote contractors, and the competing candidates - and a largely passive and uninterested public. Many party members do not join a party at their own initiative, but rather are mobilized by these few strategic actors.8 E An interesting counterexample, showing what party membership could be like in the off-years when there are no primaries and the "hard core" of party members is what remains, can be seen in a survey conducted in Israel (Citizens' Empowerment Center in Israel 2007). In 2007, when the Likud Party did not have primaries, its party members responded that 50% of them had been members for over a decade; 77% were sure they would remain members; their reasons for joining were mainly due to identification with the parry platform and to advance certain public policies, while support for a particular candidate was far behind. Von Beyme (1996: 147) called parties in the postmodern era "omnibus parties," where "people enter the vehicle, are carried for a while and drop out when they do not see any reason to go further.'' This concept is carried even further in the case of party primaries. Kitschelt (1988: 130) encapsulated the major points of our argument when he claimed, "The very emphasis on individualist, participatory norms and ideologies is likely to create unexpected perverse effects in the parties' behavior, such as a lack of activists' commitment to party work, high turnover, and the rise of informal party elites." Candidate selection in an inclusive selectorate may thus become a personal enterprise rather than a partisan matter, an enterprise of instant members recruited by individual politicians, and not one of attracting new active members. When parties do not make a serious effort to control registration, and prefer to focus on recruiting more members than any other party - what Scarrow (1994: 46) calls to "improve membership statistics" - the demonstration of their public "credibility" could prove to have perverse long term political consequences. WHAT CAN BE DONE? Various solutions can be suggested for the problems of membership quality that stem from the adoption of party primaries. The first is to adopt the logic of "if you can't beat them, join them," implying a further opening of the parties, which would enable even nonpartisans to participate in party events, such as policy decisions ■ifirl rortHi'^t'^ ^If^tinrt (T>f\(r\ir\$h-t*. 1 GQ9.Y T« tfiic pacp inrrf*3»:f*H mprhr'iTSphriTi 1« seen as an end in itself, with no aspirations for enhancing the power of parties as collective associations. As Dalton (2008) argues, we may have less participation per event, but more participatory events. This direction could lead to the Americanization of politics — parties becoming arenas, rather than associations with a substance of their own. Another solution (Teorell 1999) might be to reject the participatory democracy model as too demanding (and maybe too naive), leaving no room for the choice not to participate. Instead, parties should adopt the model of deliberative democracy - a model that does not reject representative government, yet suggests adding aspects that the "competitive model" lacks and that go beyond the electoral process. Leaders, members, and supporters would be linked by a deliberative poll - for example, a statistically representative sample of supporters would deliberate with party leaders on policy issues, and a statistically representative sample of members would deliberate on candidate selection. This recommendation seems to solve the problems of quantity versus quality and free-riding, yet is prone to many other problems. The bias of group thinking and the problem of legitimizing the poll in 104 Democracy within Parties Participation 105 the eyes of most citizens, who never learned the basic rules of statistical probability, are just two such problems. Other, less sweeping solutions suggested in the literature, or through the experience of parties, include the freezing of membership when the selection process begins (Lovenduski and Norris 1994; Norris and Lovenduski 1995, a step taken by British Labour the Welsh Plaid Cymru).9 The Canadian New Democratic Party closed registration ninety days before selection, not twenty-one days before (and sometimes even less than that) like the Liberals and Conservatives (Cross 2006). In the Belgian Social Christian Party, candidate selection was open to those who were members for at least one year (Obler 1970). The Irish Fine Gael also limited the right to participate in candidate selection to members who registered at least a year before the selection took place (Galligan 2003). The Progressive Democrats in Taiwan reduced the weight of the members' votes to offset problems of massive membership enlistment and the paying of membership dues by the candidates (Fell 2005). Mair (1987) and Farrell (1994) described a process of centralization of candidate selection in Ireland, in order to offset manipulation and inflation of membership numbers. Over eighty years ago, Merriam and Overacker (1928: 6) poignantly stated the following about the pathologies of a more inclusive selectorate. "In short, the primary election, having become one of the most important steps in the process of government, was open to every abuse that unscrupulous men, dazzled by prospects of almost incredible wealth and dictatorial power, could devise and execute." Following this, they proposed a list of specific recommendations aimed at solving the problems of the American primaries through state legislation. Our suggestion, which we elaborate in Chapter 10, is to enable the various circles of participation inthe party 4o4ake-part-in intraparty politics — that-is5 meaningful participation should be granted to rank-and-file members - but to maintain a structure of intraparty selective incentives at the same time. In the case of candidate selection, this could be achieved as long as parties involve several party agencies in the process, in a multistaged method, granting the more exclusive circles the ability to screen candidates but giving members the right to decide between a few viable options. This is the current tendency in many European parties, for example in the British parties. It may not save them from a decline in the number of members, nor from some of the pathologies of primaries that were mentioned earlier in this chapter, but it seems to provide an optimal balance between wider participation and the needs of the party as a voluntary association, and to somewhat weaken some of the incentives for misbehavior that are encouraged by pure primaries (Rahat 2009). hi short, before going too far with the opening up of candidate selection, one should ask whether positive intentions concerning participatory democracy can lead to the demise of meaningful, qualitative participation. In addition, one should inquire whether a more inclusive selectorate has an impact on other aspects of party politics, such as representation, competition, and responsiveness. These issues are the focus of the following chapters. * The rule that determines the freezing of membership in the selectorate when a contest is announced can give an advantage to a retired incumbent in deciding his or her inheritor. That is, incumbents can recruit selectors before they officially announce their resignation (Back and Solomos 1994). Representation 107 Representation If there is an agreement in the political philosophy and political theory literature concerning the interpretation of the concept of representation, it is about the multiplicity of its meanings. These meanings are not only different, but at times can even be contradictory. For example, there is the classical and well-known distinction between the notion of the representative as a delegate versus a trustee (Pennock 1968), known also as the mandate/independence controversy (Birch 1993). Beyond these classical notions, scholars have added almost endless distinctions between types of representation. This multiplicity is described as stemming from different philosophical approaches (Birch 1971) and from the historical evolution of the concept (Martin 1997). In the framework of these theoretical distinctions, the notion of representation as a reflection of society, that is, as representing a microcosm of it (Birch 1993), is only one of several. However, when dealing with representation empirically, rather than theoretically - in the context of recruitment and electoral studies - the notion of representation that is used almost universally and uniformly is that of reflecting the demographic composition of society (or, in the case of parties, reflecting the demographic composition of^ their, voter...groups)..In.the. recruitment literature, an. institution is considered more representative if it reflects society in terms of gender, class, education, ethnicity, religion, etc. (Best and Cotta 2000£; Norris and Lovenduski 1995; Norris 2006; Patzelt 1999; Putnam 1976). In the electoral studies literature, an electoral system is seen as more representative if it more accurately translates votes into seats - the closer the proportion of votes is to the proportion of seats, the more "representative" the electoral system is considered to be (Gallagher 1991; Lijphart 1985, 1994; Loosemore and Hanby 1971; Rae 1967; Riedwyl and Steiner 1995; Taagepera and Shugart 1989). Studies of recruitment also identify a connection between demographic representation and the type of electoral system that is used. Women's representation was found to be higher in parliaments that were elected by proportional electoral systems than in those parliaments elected by majoritarian systems (Kittilson 2006; Kunovich and Paxton 2005; Matland 2005). However, even parliaments that are proportionally elected are unrepresentative in demographic terms, which leads to the question: Are the parties that select candidates from a large pool of aspirants responsible for this phenomenon? Because it is party affiliation that influences the voters and gives them a strong cue concerning whom they should vote for - even when the electoral system is highly personal - the decision of the party selectorate(s) concerning who will be the party representative in a realistic seat or list position is often the decisive one. Scholars acknowledge the role of parties in influencing representation, specifically their role in reducing the candidate pool from all (party eligible) aspirants to only the party candidates. Yet, the treatment of candidate selection methods in such studies is rather simplistic. Some who bypass the problem of the lack of accessible data tend to address candidate selection indirectly by liriking the nature of the electoral system to the kind of candidate selection method that is in use (Matland and Studlar 1996), thereby overlooking the possibility that the intraparty arenas of competition may independently vary across countries, parties, and time. Others treat candidate selection as an expression of the centralized/ decentralized party structure (Kittilson 2006), and thus ignore the possible consequences of the different dimensions that distinguish between candidate selection methods. The nature of candidate selection methods is, of course, not the only factor that influences representation, nor is it necessarily the central one. Other factors, as shown by studies of women's representation, may also influence it, such as the low supply of women candidates (Norris and Lovenduski 1993; Shepherd-Robinson and Lovenduski 2002), the financial costs of ronrring a campaign (Cross 2004), or increased primary competition (Lawless and Pearson 2008). Yet even these factors can be linked to the properties of candidate selection methods. The low supply of women may stem from their assessment that a specific method does not give them a fair chance to win, while financial costs are clearly influenced by the level of inclustveness of the method — the more selectors there are, the more, costly... an effective campaign will be. This chapter assesses how responsible the parties' candidate selection process is for representation by analyzing the links between the various aspects that delineate candidate selection methods and representation. It argues that the various dimensions of candidate selection can each have an impact on different aspects of representation. The first section examines representation from a theoretical perspective, links it to the study of candidate selection, raises several methodological issues, and presents two representation indices that are available for cross-national comparative research of party-level representation. Then, the chapter assesses the relationship between each of the four dimensions of candidate selection and representation: the obstacles raised by candidacy requirements, the inclusiveness of the selectorate and the representation of ideas and of presence, the social and territorial tradeoffs resulting from the decentralization of candidate selection, and the nature of the appointment/voting system vis-a-vis representation. Whenever sufficient data are unavailable, we present the little available empirical data. When even this is impossible, we suggest theoretical propositions. The final section 108 Democracy within Parties discusses the dilemma for those parties that aspire to be both inclusive and representative at the same time. REPRESENTATION^) - THEORETICAL AND METHODOLOGICAL ISSUES A perspective of representation that is relevant for our study is that of the representation of ideas versus representation as presence. The representation of ideas, as described by Pitkin (1976), sees representatives as reflecting the political beliefs of their voters. Representatives are responsive to their constituents when they support policies that are consistent with the platform on which they were elected. The drawback here is that the representatives do not have to be anything like their constituents. They could be all male, all white, all members of the elite -as long as they act in accordance with a specific set of ideas. This drawback may be overcome by viewing representation as presence, focusing not on ideas but on descriptive characteristics. In other words, what is important according to this perspective is not what is being represented, but the social identity of the representative. From Lijphart's (1969, 1977) works on consociational democracy, in which stability is said to be attained by guaranteeing a presence for all significant subgroups within society, to Phillips' (1995) comprehensive argument for the politics of presence, this view of representation sees the presence of a representative from a particular group as a crucial element that must be considered in any calculation ©^representation............................................ If the representation of ideas could lead to a group of representatives who are quite distinct from their voters, then representation as presence could result in a group of representatives who are unaccountable. If voters choose representatives because of who they are - according to their race, gender, religion, etc. - without an indication of how they are likely to act on the most important issues, then how can accountability be maintained? As Phillips (1995: 24-5) declared: While the politics of ideas is an inadequate vehicle for dealing with political exclusion, there is little to be gained by simply switching to the politics of presence. Taken in isolation, the weaknesses of one are as dramatic as the failings of the other. Most of the problems, indeed, arise when these two are set up as exclusionary opposites: when ideas are treated as totally separate from the people who carry them; or when the people dominate attention, with no thought given to their policies and ideas. Since neither view is truly sufficient, what is needed is a combination of both. Only in the interplay between the representation of ideas and representation as Representation 109 presence will the party become a more representative body, in a politically meaningful way. We can see expressions of these kinds of representation at the interparty level. Aggregative parties attempt to represent various ideological currents and identities within their ranks, while typical sectarian parties stand for a specific identity (and sometimes also a resulting unitary ideology), hi general terms, representation within parties may coincide with notions of representation among parties. That is, both kinds of representation are relevant to candidate selection because parties -in their attempts to address the electorate and to control, or at least regulate, intraparty conflicts - are likely to try to balance their lists of candidates in terms of both notions of representation. A party could strive to have candidates who hold a range of ideological stands that the party claims to represent (representation of ideas) and also a range of identities that the party thinks should be represented (representation as presence). In some parties it is important to give representation to ideological currents (usually defined as the "left" and "right" wings of the party), while in others identity will be seen as central. There are some identities that are important in almost all parties, like gender, and there are identities that have to do with the particular character of the party. A left-wing party may grant more significance to having blue-collar candidates, while right-wing parties may consider it important to have businessmen as their candidates. While the representation of minorities seems to interest many parties around the globe, the identity of these minorities reflects a specific societal context. Furthermore, particular parties may believe that certain nuances are more important than others — aspects that other parties tend to ignore. For example, Shas, an ultra-religious Israeli party, is unrepresentative of its voters in terms of gender and-religiosity — it has no women on its list of candidates and its candidates, are much more religious than its voters. On the other hand, it is highly representative of the different streams within the ethnic group it claims to represent, and appoints candidates with different backgrounds to realistic positions. The complexity of the notion of representation and the problems of its assessment are clearly exhibited in its operationalization. First, most studies tend to use representation as presence because it is the cheaper, easier, and more accessible way to operationalize the concept. It is easier to count men and women than to assess the ideological position of each representative. Limited by what the existing literature offers, the lion's share of this chapter is dedicated to representation as presence. Second, because representation as presence is sensitive to national and even intraparty political culture, it is hard to conduct a cross-national comparison, except for the almost universal case of women's representation - and that is what is usually done. Third, when one wants to analyze the influence of candidate selection methods on representation, it is important to look at realistic candidacies rather than at the composition of parliament - which is influenced by electoral results and not only by the selection results. A few studies go beyond looking at the composition of parliaments to analyze the composition of the pool of 110 Democracy within Parties Representation 111 candidates (Holland 1987; Norris and Lovenduski 1993, 1995). Fourth, when estimating representation, there is a need not only to distinguish realistic candidacies from token candidacies, but also to find a way to measure their relative strength. That is, a list of twenty candidates in which women occupy positions 11-20 is clearly less representative than a list in which women occupy all odd or even positions (Hazan and Rahat 2006; Rahat, Hazan, and Katz 2008). There are numerous studies which have shown that members of the legislature do not accurately reflect their society in terms of gender, race, religion, education, socioeconomic status, etc. (e.g. see: Putnam 1976; Best and Cotta 20006; Gallagher, Laver, and Mair 2006). There are very few studies, however, devoted to how candidate selection methods affect representation. This chapter aims at filling this gap. Operationalizing intraparty representation In order to operationalize representation in candidate selection, we developed two indices designed to measure the level of representation in parties that compete in list electoral systems (Hazan and Rahat 2006). They are sensitive to differences in party size, provide reasonable operational definitions for representation in list systems, and allow for adding data in a weighted manner so as to address parties of different sizes. They can also be used in the case of single-member districts, and can thus serve as useful tools for cross-national comparison. Our two indices of representativeness take the representation of women as an indicator of the representativeness of the candidate list. The first index relates to the proportion of women in realistic positions on the party list. This is done by measuring ihe share of women out of the total number of the party's realistic candidates, not those who appear on the list as a whole. The second measure also relates only to realistic positions, but unlike the first takes into account the relative position of women on the list, giving a higher value to higher positions on the party fist The index of representation (IR) simply calculates the percentage of women in realistic positions on the party list by counting the number of women in realistic positions divided by the number of realistic positions, multiplied by 100. The formula is: IR = 4^„ x 100 5> Wrp is the number of women in positions equal or higher in rank to the number of realistic positions. Rp is the number of realistic positions (which we defined earlier as the number of seats the party won in the previous election). An example of the index of representation - in a single selection event - is as follows: Party A won 5 seats in the previous election; women appear in the third and fifth positions in the candidate list for the elections for the next parliament; the index of representation is thus 2/5 x 100 = 40%. Like the index above, the weighted index of representation (WTR) counts the number of women in realistic positions, but this time takes into consideration their relative positioning. Higher values are given to higher positions on the list. Thus, each position on the list, up to the number of realistic positions, is given a value in descending order - the last position is given one point and each higher one merits an additional point. The formula is: WIR s [(g)* fr" x 100 Wp stands for the value of the positions won by women in each selection event. Vpi stands for the total value of the positions in the specific selection event. Rp stands for the number of realistic positions available in each selection event. The example of the weighted index of representation - in a single selection event - is based on the data given in the previous example: First, the sum of values of the positions on the list is calculated. The total value of the list is 5 (for first position) + 4 (for second position) + 3 (for third position) + 2 (for fourth position) + 1 (for fifth position) =15. Women won position 3 on the list with a value of 3, and position 5 on tire list with a value of 1. The sum of these values is 3+1=4. The weighted index of representation is thus 4/15 x 100 = 26.7%. CANDIDACY REQUIREMENTS AND REPRESENTATION: OBSTACLES AND BARRIERS The political consequences of candidacy requirements for representation within parties is a topic that has yet to be addressed, and thus our arguments can only be tentative in nature, hi the chapter on candidacy, we outlined several common requirements that are imposed by numerous political parties, all of which serve as barriers for potential candidates. Each one of these requirements restricts the possible pool of candidates and thus offsets the possibility that a party will accurately represent the electorate in general, and its voters in particular. For example, a maximum age requirement will exclude the representation of older voters, a membership period of a year or two will hurt the representation of those members who have recently joined the party, a substantial monetary deposit could eliminate the representation of the lower classes, etc. While each requirement 112 Democracy within Parties Representation 113 above - and the level set by the party - could serve a specific purpose, they will likely impact negatively on representation as presence. More exclusive candidacy requirements reflect an attempt by the party to control the supply side of potential candidates. This may be due to a desire to maintain party cohesion, so that those who fulfill the enhanced eligibility criteria -and are subsequently both selected and elected - will behave according to party dictates once in office. A party with exclusive candidacy requirements can arrive in office as a cohesive unit, manifesting a patent party culture. But this attempt to ensure party-centered representation might bias representation as presence. Demands for a background of long-time party activity - which can be seen as a sign of loyalty and dedication - obviously work against younger aspirants, who will find it more difficult to prove such a lengthy record. Women and lower-class aspirants may also be hurt if their time resources are lower than those of middle-class men (Norris and Lovenduski 1995, 1997). If a party adopts more inclusive candidacy requirements - or erases them altogether - where every voter can stand as a party candidate, the party has little to no influence as a gatekeeper for potential candidates. In other words, aspirants for office practically impose themselves on the party, which must accept their candidacy. While this appears to remove the barriers of representation concerning candidacy, it does raise other obstacles that can affect the level of representation. For example, the United States displays some of the most highly inclusive candidate requirements, and as a result politics there have been described as candidate-centered (Wattenberg 1991). Open candidacy has, in turn, brought about an ever-increasing role for financial supporters of prospective candidates, which has negatively influenced the representation of several social groups in American politics. Thus, fonT)al,.explicitcandidacy. requirements-can be seen as an attempt to regulate the supply side of recruitment, which has its own biases when there is no such regulation. Norris (1997,3, 19976), Patzelt (1999), Best and Cotta (2000ft), Gallagher, Laver, and Mair (2006), and other scholars found that in most parties the incumbents are not representative of their parties' voters. This means that in order to create a more representative group of party candidates, some incumbents must be replaced by new candidates from those social groups that are underrepresented. Once the obstacle of an existing incumbent is removed, the barrier to representation can be more easily overcome. For example, Burrell (1992) found that in the primaries for the US House of Representatives between 1968 and 1990, women did just as well as men when there was no incumbent candidate. Thus, a party's treatment of incumbency may also affect representation. It is much harder to transform an unrepresentative group of candidates into a more representative one if incumbents are easily and constantly reselected; it is easier to do so when there is higher, party-controlled turnover. Candidacy requirements, such as automatic renomination, can ease the reselection of candidates, or they can be neutral vis-a-vis incumbents, in which case they will likely continue to limit the ability of the party to enhance representation. Candidacy requirements can also work against incumbents, the best example being term limits, which artificially open up positions or constituencies that can be filled by more representative candidates - but this is not a necessary outcome, and may at times work to the contrary. A party that offers its incumbents almost automatic candidacy for upcoming elections evidently wants to win those elections - incumbents have already proven their electoral potential. Also, since incumbents already have a base of support in their party or in their constituency, parties - wanting to avoid internal conflict on the eve of a general election - will refrain from exposing their incumbents to internal challenges. This is especially true for candidacies in single-member districts, but also for candidates in multimember districts where the electoral systems in use encourage personal voting (open ballots, single transferable vote, or single nontransferable vote). On the other hand, a party may place restrictions on incumbents in order to ensure some turnover - as is the case with term limits or demanding that an incumbent's candidacy be ratified by a special majority. Such an attempt might be made in order to prevent the creation of an autonomous power center by the party in government, and to keep power in the hands of the extra-parliamentary party. Restrictions on incumbents can also be adopted in order to create opportunities for improving representation. Incumbency can be a major obstacle for parties that are concerned with enhancing representation, and they can cope with this by encouraging turnover. Whatever the intention is, higher turnover creates opportunities to enhance representation. Parties react in different ways to the dilemma of choosing between incumbency and representation. In general, it seems that the clash between these two is more apparent in single-member districts than in list systems. That is, enhancing-women's representation by pushing a male incumbent one position down the list is easier than mandating a woman candidate in a single-member district with an incumbent man. One case in point is France, where the adoption of quotas through legislation failed to lead to a large increase in women's representation at the national level - where single-member districts are in use - but did significantly increase their representation at the local level - where list systems are employed (Murray 2007). Quotas were also effective at the national level in Belgium, where a list PR system is in use (Meier 2004). In Britain, the Labour Party chose to bypass this issue by sticking with its incumbents, but ran all-women shortlists in about one-half of the open seats for the 1997 and 2005 elections (Criddle 1997; Cutts, Childs, and Fieldhouse 2008). The use of the concepts "closed seat" (a district where an incumbent tries to get reelected) and "open seat" (a district where the incumbent does not compete) in countries that use single-member districts illustrates the preferred status of incumbents. There is no parallel notion of "open" or "closed" position(s) when it comes to party lists. Indeed, incumbents are less safe in list systems (Matland and Studlar 2004). In such systems, incumbents sometimes even face restrictions or rules that 114 Democracy within Parties Representation 115 place additional requirements on their reselection. In list systems there is more room for maneuver, and representation can be increased significantly without "directly" challenging incumbents. Yet, incumbent men can still lose their safe position if significant gender quotas or zipping - women in eveiy other position -are adopted. This, in turn, leads to struggles over the magnitude of the quotas. Thus, the issue in countries where there are list systems is not that of candidacy requirements (the treatment of incumbents), but rather what we call the level of (social) decentralization. In Israel, for example, incumbent men in both Labor and Likud successfully guarded their interests when most positions that were allotted in the name of representation (territorial or social) - those likely to be taken by newcomers - were not at the top of the list. It seems that in both single-member districts and list systems, incumbents do not pay the price for the underrepresentation of social groups; it is usually newcomers who have to compete within the confines of the need for representation. Thus, while representation is an ideal to which parties may aspire, its magnitude is often a result of a down-to-earth struggle between incumbents with vested interests in the. existing system and newcomers whose interest is to change the rules of the game - concerning candidacy but also other aspects of the selection method - so that they can make room for themselves. LNCLUSIVENESS AND REPRESENTATION: NEGATIVE FOR PRESENCE, NONLINEAR FOR IDEOLOGY Smaller, exclusive selectorates are more capable of balancing representation, in both senses (ideas and presence). As Matland (1993:740) puts it, "The power to actually balance the ticket is greater in systems where the nomination process is much more closed than the very open process found in the United States." When selection can be controlled by a party oligarchy that appoints candidates - and to a lesser extent, when voting takes place in a party agency and can be somewhat coordinated - there are more chances that representatives of different social groups (women, minorities, etc.) and ideological currents within the party will capture realistic positions on the party list, or realistic constituency seats. However, when parties allow their members or supporters to produce the list of candidates, the result could be unrepresentative lists, because such a vast selectorate cannot be coordinated or instructed to select a socially (or ideologically) representative group of candidates. When the parties in Taiwan moved to more inclusive primaries, one of the first problems they encountered was the lack of representation of the various intraparty groups among the selected candidates (Fell 2005). In Iceland, the People's Alliance increased the representation of women from the 1970s, but when it adopted primaries to select its candidates it could no longer do this - that is, the more inclusive primaries became a barrier for the representation of women (Kristjansson 1998). In Israel - which had one of the highest levels of women's representation in the world in the 1950s - the adoption of more inclusive candidate selection methods over the years decreased the ability of parties to ensure sufficient representation for women. It now has one the lowest levels of women's representation among the established democracies (Rahat, Hazan, and Katz 2008). The parties themselves, by their behavior, appear to validate the claim that inclusiveness can harm representation. The process of democratization of candidate selection methods in Western Europe took place parallel to increases in the use of representation correction mechanisms. Parties increasingly tend to restrict the choices of their more inclusive selectorates in order to ensure representation, particularly when it is defined as presence and concerns the representation of women. As Norris (2006:106) put it: Grassroots members in many European parties have gradually been given greater opportunities to nominate candidates. At the same time selectors are operating within a more constrained scope of decision-making, due to the simultaneous adoption of rules implementing positive action strategies. A wider number of members are therefore able to engage in selection decisions, but they face a more restricted range of choices. The Dutch Democrates 66, for example, allowed its members a direct role in candidate selection, and as a result produced unrepresentative lists. This led, in 1986, to the formation of a committee that produced a recommended list of candidates for the members. This committee took into consideration several of the candidates' attributes, from social background to media appeal, and produced a more representative list. These candidate lists were usually approved with few changes (Andeweg and Irwin 2002). The Belgian Social Christian Party became disenchanted with inclusive party primaries due to their inability to produce balanced lists of candidates, and its leadership subsequently tried to avoid using them (Obler 1970). In Finland, the same logic of balancing inclusiveness with representation was apparent in the law that both determined that party members would select the candidates and also allowed the party organization in the district to alter one-quarter of the list selected through primaries. Indeed, this type of intervention is often used in order to create a more representative list of candidates (Helander 1997; Kuitunen 2002). According to Narud and Valen (2008), the Norwegian parties' use of exclusive candidate selection methods - which allow the parties to control the process - enabled them to produce more representative candidacies than most other parties elsewhere. An interesting case is Mexico, where in 2002 a law was passed that 30 percent of candidates must be women, unless the party holds primaries to select its candidates (Baldez 2007). This law apparently saw a tradeoff between inclusive participation and representation, 116 Democracy within Parties Representation 117 and allowed the parties to choose for themselves which principle was more important to them. The implementation of more inclusive candidate selection methods in large voting entities requires a strengthening of the corrective mechanisms concerning representation of women, other social groups, and geographical peripheries (in the case of multimember districts). From the liberal perspective there is a problem when an advantage is given to the group over the individual (Htun 2004), and the advocates of a free market are also not fond of such mechanisms. Yet, the use of representation correction mechanisms seems to be a norm that has spread, being adopted also by center and right-wing parties, and even at the national level as a norm that obliges all parties. If we adopt the stand that intraparty democracy is not only about enhanced participation, then we should see representation correction mechanisms as a must in those cases where enhanced, inclusive participation is attempted. There is not much research - beyond the US case of highly inclusive primary systems - when it comes to the relationship between the representation of ideas and the inclusiveness of the candidate selection method. The only research we encountered regarding this relationship outside the United States (Mikulska and Scarrow 2008) found that more open selection methods in the United Kingdom created more congruence between the candidates and their voters on the most salient electoral issue concerning the economy. However, studies of the US primary systems found that this aspect of the relationship between inclusiveness and representation is not linear. When candidate selection is at its most inclusive pole, such as in nonparty and blanket primaries, candidates will pursue positions closer to the median voter (Persily 2001). In somewhat less inclusive methods, such as open primaries, the candidates will be moreideologiGaVdueto the-likely participation of more principled and strategic voters from outside the party. In more restrictive closed primaries, candidates will be less extreme than in open primaries, but not as moderate as in the semi-open primaries (Kanthak and Morton 2001). It is very likely that when we move toward the middle of the inclusiveness-exclusiveness dimension, the nonlinear relationship with the representation of ideas will continue. Candidates chosen by party members will be less ideological than those chosen by selected party delegates, for whom loyalty to the party and cohesiveness in office are important concerns (Scarrow 2005). Candidates chosen by the party elite will, in turn, be more ideological than those selected by the parry leader, for whom personal loyalty is paramount These propositions, while theoretically sound, must still be checked empirically. Scholars tend to analyze, and politicians tend to address, the visible and more easily recognizable distortions of representation as presence, while putting aside the harder to analyze and more complicated representation of ideas. When it comes to representation as presence, the negative relationship is not "accepted" but rather tackled by using correction mechanisms that are designed to deal with the imbalance created by increased participation. Those few who bother to address the relationship between inclusiveness and the representation of ideas usually see it as a byproduct of the system - not as something that can or should be fixed, but rather as something that can vary with changes in the levels of inclusiveness of the selectorate. DECENTRALIZATION AND REPRESENTATION: SOCIAL AND TERRITORIAL TRADEOFFS The decentralization of candidate selection, as denned in this study, can be social or territorial. We analyze each, in turn, vis-a-vis represention. We end this section by addressing the problematic relationship between these two elements - the tradeoffs between social and territorial representation. As already mentioned in the discussion of inclusiveness, there is a negative relationship between enlarging the selectorate and tire resulting representation of different social groups. This tradeoff produces an imbalance that has been rectified, at least partially, by the parties' instituting representation correction mechanisms such as quotas and reserved seats (Dahlerup 2006; Kittilson 2006; Krook 2009; Mateo-Diaz 2005; Tremblay 2008). Factors like party ideology and the power of intraparty groups affect the existence and the scope of these corrective representation mechanisms for the various groups. Moreover, a major factor in deciding to adopt, such mechanisms is the electoral appeal that is attributed to the presence of these representatives. Parties tend to ensure representation when it is electorally beneficial, or when they are convinced that ignoring such demands could cost them voter support. Intraparty group demands that are not perceived to be electorally advantageous are often rejected, or answered by symbolic gestures such as allotting them an unrealistic position on a list or a seat in a district with no chance of victory for the party. In any case, a price (arguably worthwhile) is paid because there is a tradeoff between more inclusive participation and representation. Yet, the very cause of representation might be damaged as a result of unwise use of corrective mechanisms, specifically, mechanisms that may damage the image of those whose representation is being corrected (Bacchi 2005). This is especially likely to occur when small quotas are used for a long period of time. In such a situation, the quota may become a mechanism to ensure the reselection of a specific person, rather than the representation of a social group. Moreover, candidates from the group whose representation was guaranteed will focus their competition on the few positions allotted. If the group competes amongst its own members, each candidate will ask selectors not only for their support but also to refrain from supporting other representatives of the group - in order to increase his or her chances of winning. 118 Democracy within Parties Representation 119 In that case - a likely development when quotas are small - the group will gain less votes and there will be an increased need to use corrective measures. The correction mechanism can thus be damaging because it leads to the segregation rather than to the integration of the social group that it is meant to promote. This can perpetuate the group's "competitive inferiority" - which was the rationale for adopting these mechanisms in the first place. A wise use of generous and/or gradually increasing quotas for a predetermined period of time might help overcome this problem. Territorial decentralization has a positive relationship with territorial representation. If more power in the candidate selection process is given to the regional and/or the local selectorates, at the expense of the national party organization, the likely result will be more candidates chosen who represent the regional and the local levels. They can be seen as territorially representative because they live and run their life in the specific location, or simply because they were selected by a regional or local selectorate. One should keep in mind that the mere fact that candidate selection is decentralized does not mean that interests at a lower level - which also have to be balanced - will necessarily be fewer than those at a higher level. Party officials within the constituency also have to take into account the different groups within the party, and they have many fewer candidacies to play with - and only one in the case of single-member districts. The most territorially decentralized systems, especially those that produce a single candidate, may achieve optimal representation for the locality, but it will be harder to promote social representation. As Matland and Studlar (1996: 709) argued: Centralized control over nominations means that party elites can increase the number of viable women candidates in response to pressure for greater representation. Most single-member district systems tend to have decentralized nomination structures; because of this central party organs wanting to increase representation have considerable difficulties in getting their wishes carried out at the local level. Indeed, in her cross-national comparative study, Kittilson (2006) found that higher women's representation positively correlates with centralized parties. In Germany, for example, since 1953, the ratio of women elected in the constituencies (selected by the more decentralized selectorates) to those elected from party lists (selected by the more centralized selectorates) has varied from 1:3.5 in 1965 to 1:6.5 in 1972 (Roberts 1988: 109). In Ireland's decentralized system, territorial local representation takes precedence over other types of representation, even within the small multimember districts that are used in the general elections. Even where the main parties can choose more than a single candidate to run in the multimember constituencies, the only attempt made to balance the selected candidates is in terms of territorial district interests, rather than according to gender, age, or socioeconomic status (Gallagher 2003; Marsh 2005). The result is that decisions made at these lower levels are independent and do not accumulate to produce a socially representative group of candidates overall, as evident from looking at the low level of women's representation in Ireland. There is a problem of coordination between the demands and needs of territorial and social decentralization, which can develop into intraparty conflict. This problem is illustrated by the failure of British Labour to coordinate in the 1970s and 1980s for the purpose of increasing women's representation (Denver 1988), and the conflicts between the national center and the local organizations in the 1990s over the implementation of women's representation mechanisms (Mitchell and Bradbury 2004; Quinn 2004). The decentralized selection process used in the Canadian parties was an obstacle to both female and minority candidates in the parties' attempts to increase their representation (O'Brien 1993) and they were able to advance only a limited number of female candidates in vacant constituencies, mainly due to pressure from the national party organization (Cross 2006). This conflict can occur not only in single-member districts, but also in larger districts when the local party has other preferences, or when it feels that its autonomy is affected by the national headquarters' demands for social representation (Valen, Narud, and Skare 2002). In short, the almost unavoidable inherent tension between the interests and values of the local and the national levels is sometimes expressed by conflicts over social representation. The attempt to impose representation correction mechanisms on the lower levels implies some recentralization of the national party's power in the candidate selection process. Social representation can be ensured by using centralized and exclusive selectorates For example the greatest representation of women m Chilean parties m the 2004 local elections was in the most conservative party, the Independent Democratic Union, which used a centralized and exclusive candidate selection method; somewhat lower female representation was achieved in another right-wing party, the National Renewal, which used a decentralized but exclusive selection method. A relatively low number of women were selected in the center and left-wing parties that used decentralized and inclusive methods (Hinojosa 2009). However, while centralization (like exclusiveness) means that that there is more potential to ensure social representation, it does not necessarily lead in every case to such as a result (Ware 1996). Centralization (like exclusiveness) provides opportunities to enhance representation, yet there is still a need to motivate forces within the party to press for the use of these opportunities. In those parties where there are high levels of exclusiveness and centralization, cohesion is also likely to be high, and pressure for representation is less likely to originate from within the party. What might motivate the party elite to address representation are electoral considerations. In those parties where there are high levels of decentralization and inclusiveness, it is more likely that internal pressure for representation will be stronger, but the ability 120 Democracy within Parties Representation 121 to respond to it will be lower. Correction mechanisms are likely to be the answer in such a scenario. Matland's (1993: 753) account of the high levels of women's representation in Norway identifies an optimal balance between these two options: The system of county nomination meetings is sufficiently open to allow organized interests to enter. At the same time the system is sufficiently closed and small enough that a cohesive, well-organized interest can have substantial influence on the choice of candidates, even if they only represent an intense minority within the population as a whole. APPOINTMENT OR VOTING AND REPRESENTATION: MECHANISMS AND PREFERENCES Appointments give the (necessarily exclusive) selectorate the greatest ability to coordinate candidacies and thus to balance representation among different social groups and intraparty (personal and ideological) factions. The output, however, is dependent upon the will of the selectorate (and the pressures on it), as is the case with more exclusive and centralized methods. In those parties where only some candidates are appointed, if the bulk of the candidates selected by a voting system end up being unrepresentative, the remaining appointed candidates can help to balance the overall outcome. Recently, the Canadian New Democratic Party appointed several women, specifically in constituencies where an incumbent party representative had retired, in its attempt to balance representation (Cross 2006). The leader of the Liberal Party also u.->ed his powers k> appoint s'evera! women candidates (Carty and Eagles 2003; Erickson 1997). Some parties use appointments to increase their appeal to groups outside the party; that is, to project a wider ideological image. In the late 1970s and early 1980s, the Spanish Communist Party appointed candidates from other left-wing parties, and from trade unions, in order to bolster its image (Esteban and Guerra 1985). The Italian Communist Party did the same when it added leftist independents to its candidate lists (Wertman 1988). In Belgium, the leader of the Liberal Party appointed religious candidates in order to change the image of the party and gain support beyond liberal, anticlerical voters (Obler 1973). As we elaborated in previous chapters, candidate selection is often conducted by several selectorates. fn a multistage system, one selectorate can use an appointment system while another may use a voting system, thereby producing a mixed ' According to Cross (2006), the local organization of the New Democratic Party had to convmce the central party that it tried to find a candidate from an underrepresented group if it wanted to appoint someone who was either not a woman or not from a defined minority group. appointment-voting system. The typical mix is that of starting with appointments by an exclusive selectorate and ending with voting by the more inclusive one. Such an order makes sense, since it is easier to gain democratic legitimacy for early screening than for a veto on the decisions of a more inclusive selectorate. Thus, processes that allow a small selectorate to appoint a list give the party more control over representation, yet give the more inclusive selectorate a chance to democratically legitimize the earlier coordinated process. In Norway, a nominating committee took into consideration the need to balance representation in terms of geography, gender, age, occupation, etc. in designing a recommended list. This recommendation was men voted on (Matthews and Valen 1999). German parties used appointments to the party list - which are ratified by a delegate convention - in order to compensate for the underrepresentation of specific groups among the constituency candidates (Kitzinger 1960; Loewenberg 1966). In Belgium, the use of the model list system meant the appointment of a list that was then ratified (or, rarely, rejected) by party members. The central office agencies of the British Labour and Conservative parties screen candidates; that is, exclusive centralized selectorates actually appoint candidates to a list, and then more inclusive and decentralized selectorates choose the candidates from this list using a voting system. Such systems might be conducive to representation because small selectorates, in order to legitimize their appointments, are likely to respond to the demands of groups within the party for representation (Matland 1993). When it comes to the voting systems used to select the party candidates, the relationship with representation depends on the specific voting system. While appointment is clearly the easiest way to ensure representation, voting systems differ in terms of their results. The first distinction is between single-round and multi-round systems. The mulri-round system ~ the gradual selection of realistic candidates - allows the selectorate some control over the composition of candidacies. If there is a problem of representation - there are not enough women/ minorities/laborers - the multi-round system provides an opportunity to fix it. hi a single-round system, all realistic candidacies are filled at once, so there is no chance to fix distortions. The second distinction has to do with the proportionality of the voting system. A majoritarian voting system, where the number of votes is equal to the number of realistic candidacies, can allow the largest intraparty group to win all of the list positions, or all of the constituencies, and the result will not be representative of different groups within the party. A complete victory for the largest group can be mitigated if the voting rules are not plurality but majority (two-round, alternative, or elimination vote), unless the largest group actually holds a majority within the selectorate. While we do not have evidence of the influence of intraparty voting systems on representation, it would seem logical to argue that as in the case of electoral systems (Matland 2005), plurality and other majoritarian candidate selection methods are more likely to produce less representative party candidacies. That is, we expect social and factional representation to grow 122 Democracy within Parties with proportionality - from the semi-majoritarian voting system, through the semi-proportional system, to the proportional one. The choice of voting system can raise obstacles to the largest group's attaining all of the realistic positions. It does not, however, guarantee other aspects of representation. For example, a minority group could win realistic positions under more proportional selection methods, but could select a group of unrepresentative candidates in terms of presence, that is, white-collar, middle-aged, white men from the capital city. Overall representation - beyond possibly that of ideas, if it is ideology that divides the party groups - will be absent. Moreover, if the largest group is interested in advancing representation, adopting more proportional voting systems could actually offset this goal. An additional factor can be the number of votes allotted to each selector. Matland (1993) found that at the interparty level, the larger the party district delegation is, the greater the representation of women in the district. In a "translation" of this finding to the intraparty arena, we expect that the number of votes for each selector will affect representation. The more votes each selector has, the more he or she is inclined to include representational considerations in voting calculations. Thus, multi-vote systems - whether they are majoritarian or proportional -could produce more representative candidacies than a single-vote system. More representation can be achieved either through appointments or via a proportional selection method, but this is not necessarily the case. Appointing more representative candidates, or adopting a voting system that allows for a more representative outcome, are only mechanisms that allow for representation - they do not guarantee it. For more representative candidates to be appointed, representation has to be a significant preference for the party, and especially for its selectorate. The same is true for multi-Found and-multi-vote selection methods, in which representation can be balanced if this is indeed a priority for the selectorate. Representation 123 correction mechanisms together with the democratization of the selection process. There are conflicting relationships between territorial representation and social representation, which could require some centralization in order to be balanced. Appointment systems, mixed appointment-voting systems, multi-round and multi-vote voting systems, and more proportional voting systems could help produce more representative candidacies than a mere majoritarian, one-round, single-vote selection. Yet, these are all mere mechanisms that can only work if representation is an important principle for the party, or is seen as an electoral asset, and if there are strong groups that promote the cause. Those advancing the principle of representation will have to struggle against the vested interests of most of the incumbents, along with those groups whose representation was already recognized and who have won their share of the representation pie. The principle of representation presents an uneasy dilemma for democrats. If they opt for what seems to be intuitively a more democratic candidate selection method - an inclusive, decentralized method in which candidates compete for the secret votes of thousands of party members or even supporters - they might end up with the same old candidates, mostly white, upper-middle-class, educated men. Intraparty democracy has to be somewhat limited in order to achieve what they usually perceive as more democratic representation. Similar to the element of participation, the principle of representation is also perceived as both necessary and positive in a truly democratic society, hi order for these goals to be attained, it might be necessary for parties to involve several party agencies in the process, in a multistage method, granting the more exclusive circles the ability to balance candidacies but still giving members the right to decide between a few viable options (Rahat 2009). Opting for a simple, one-stage candidate selection method, removing candidacy requirements, opening up the selectorate, decentralizing, and implementing a pure voting system is unlikely to enable selectors to produce a representative outcome. GUARANTEEING REPRESENTATION Representation can be conceived in several ways: in terms of ideas; or in terms of presence and its different types - social or territorial. If any kind of representation is a sine qua non for a particular political party, then the party must take the necessary steps to guarantee that its candidates are indeed representative - at least of its voters if not of the entire population. Producing a more representative group of candidates can be facilitated by candidacy restrictions imposed by the party, because such barriers can be raised for some (overrepresented) groups and lowered for others. Representation is also connected to the inclusiveness of the selectorate, but not in a positive manner, and thus requires the adoption of Competition 125 "If intraparty competition on the fortunes of parties in general elections. The chapter g ends with a short summary of what we know about intraparty competition. Competition In a democracy, we expect to see a free competition of interests, values, and also identities. Parties and candidates present themselves as the representatives of these interests, values, and identities, and from time to time compete with each other for the support of the voters. In the intraparty context, we expect to see competition among candidates for the support of the selectors - be it all voters, party members, party delegates, the party elite, or a single leader. We presume that intraparty competition is important for democracy, especially in those cases where interparty competition is weak - although it cannot serve as a sufficient substitute (Key 1954; Sartori 1976; Turner 1953). Competition - a situation when a plurality of alternatives is presented to the selectors from time to time - is expected to create responsiveness and accountability. That is, incumbents who face competition and who wish to be reselected are expected to be responsive to their selectors, and to be accountable for their actions. While some competition is both necessary and positive, and no competition is bad for democracy, too much competition may also be problematic. If incumbents think that their efforts will not be rewarded by reselection, why should they bother investing any effort? Moreover, heavy competition-might also enhance the importance of money in the campaign and increase the incentives for corrupt practices. High turnover can also affect the ability of the representatives to accumulate experience and thus function better in their legislative and executive posts (Somit et al. 1994). Yet, as a reading of Michels (1915) leads us to expect, it is much easier to identify cases of no and low intraparty competition than cases of excessive competition. In order to explain what competition is in the context of candidate selection, this chapter will start with an examination of the ways that scholars have operationa-lized intraparty competition. We then assess the impact of candidate selection methods on competition, beginning with the influence of candidacy. This is followed by a search for evidence of the linkage between the level of inclusiveness of the selectorate and levels of competition. We then show that territorial decentralization is the main explanation for low competition within parties and also examine the impact of social decentralization. Next, we analyze the influences of appointment/voting systems and of multistage methods on levels of competitiveness. We then look into the impact of changing the candidate selection method itself on intraparty competition. We continue by examining the influence of WHAT IS INTRAPARTY COMPETITION, AND HOW CAN IT BE MEASURED? j j There are several ways to define intraparty competition between candidates. We start with the simplest distinction between those situations where there is no competition compared to those where competition exists. The definition of noncompetitive candidate selection is that the number of candidates is equal to the number of realistic positions (or lower, although we believe such a shortage is rare). For example, if a party has only a single candidate for a realistic seat in a single-member district, then there is no competition in candidate selection. In countries with single-member districts - such as the United Kingdom, the United States, Canada, and Germany - there are often districts in which there is no intraparty competition, and scholars use the very occurrence of a contest - that is, having more than one candidate for the party's single seat candidacy - as one indication of competition (Ansolabehere et al. 2007; Borchert and Golsch 2003; Engstrom and Engstrom 2008; Erickson and Carty 1991; Erickson 1997; Norris and Lovenduski 1995; Porter 1995; Szabo 1977; Turner 1953). This lack of intraparty competition seems to apply also for countries in which single or few candidacies are produced in small multimember districts, such as Ireland (Weeks 2007) and Chile.(Siavelis 2002, 2005). . ........................ When parties produce multiple candidacies for multimember districts, it is more likely that some level of competition exists. We surely have competition when the number of candidates is higher than the number of realistic positions. For example, if a party produces a list of ten candidates for a ten-member district where it can expect five realistic positions, we are likely to see competition - minimally among the incumbents for the higher and safer positions, but also between the weaker incumbents and the stronger challengers for the marginal positions. The case of multimember districts, though, is somewhat more complicated. For example, if a party holds seven seats in a multimember district, and there are only seven contestants, this is low competition rather than no competition. While all contestants will win a realistic position, it is still likely that they will seek to win a higher position on the list, which is safer and can also help them to promote their status within the party and, as a result, in government (Kenig and Barnea 2009). We should still look at the relationship between the number of realistic positions and the number of candidates, but the basic distinction per district will be between high and low competition rather than between competition and no competition, as is the case for single-member districts. 126 Democracy within Parties Competition 127 Number of contestants One way to estimate the level of competition - which also distinguishes among those cases where there are more candidates than realistic seats or positions - is to focus on the supply side, examining how many actually ran for candidacy (Norris and Lovenduski 1995). This measure was used to estimate the level of competition in US primaries, for example, by looking at differences in competition between winnable and nonwinnable seats (Standing and Robinson 1958); to estimate the influence of incumbency on competition (Ansolabehere et al. 2007); and to compare contests between runoff primaries and contests in primaries where a plurality is sufficient to win (Glaser 2006). Similar measures can be used for multicandidacies, assessing the ratio between the number of candidates and the number of realistic positions or the ratio between competing incumbents and newcomers. Such measures are the two aspirants indices (All and AI2; see Hazan and Rahat 2006; Rahat, Hazan, and Kate 2008). The first aspirants index (All) divides the number of those who competed for a realistic position on the list (Crp) by the number of realistic positions (RP). The higher the value of the index, the more competitive the selection process is, which means that the number of competitors for each position is relatively high. The sensitivity of this index to the relative size of the pie in each selection event enables the accumulation of values for the whole population of selection events. The formula is: All Crp stands for the number of candidates who competed for realistic positions. RP stands for the number of realistic positions. For example, in a single selection event, in party A 30 candidates are competing for 15 realistic positions on the list, and in party B 50 candidates are competing for 20 realistic positions on the list. The value of the first aspirants index for party A is 30/15 = 2, while the value of this index for party B is 50/20 =2.5. Party B is more competitive in this respect. The second aspirants index (AI2) examines how many nonincumbents were motivated to challenge incumbents by measuring the ratio between "insiders" and "outsiders" competing for realistic positions. The higher the index, the more newcomers were motivated to challenge incumbents. The formula is: A12 Cni stands for the number of nonincumbents who competed for a realistic candidacy. Ci stands for the number of incumbents who competed for a realistic candidacy. A higher value for this index reflects greater competition, which means that the number of nonincumbents who challenged incumbent candidates is relatively large. The sensitivity of this index to the number of competing incumbents enables us to sum the values in both parts of the equation and to accumulate data for the entire population of selection events that were conducted under similar rules. For example, in a single selection event, in party A 10 MPs and 20 nonincumbents are competing for positions on the list, and in party B 10 MPs and 40 nonincumbents are competing for positions on the list. The value of the aspirants index for party A is 20/10 = 2, while the value of the aspirants index for party B is 40/10 — 4. Party B is more competitive in this respect Both indices are sensitive to developments that can be (but do not need to be) independent from the influence of the candidate selection method itself. That is, voluntary retirement of incumbents might draw high numbers of newcomers and thus inflate both indices. The accumulation of data from many selection events will balance these "noises." Moreover, with the accumulated data, the fact that there are more "voluntary" retirements in the case of specific candidate selection methods only points out its latent impact on competition. The use of these indices, however, is limited to those cases where we have a clear list of candidates. Such a list is sometimes lacking, especially in cases of appointment by small selectorates in which candidacy is frequently informal and sometimes the party prefers not to make these lists public. Analysis of numerical results We propose a group of measures that use numerical results to estimate levels of competition. Such an analysis is limited to those cases where voting occurs and where the results are accessible.1 Simple distinctions - which are easy to implement when the contest is over a single candidacy - include, first, the definition of those races where the winner won more than a certain percentage of votes as noncompetitive, while defining those where this amount was lower as competitive. For example, in their analysis of primary elections in the United States from 1912 to 2005 for the federal and state houses, Ansolabehere et al. (2007) denned those races where the first-place candidate won 60 percent or less as competitive races. The second distinction is based on definitions of the level of competition on the basis of the margin of victory. Goodliffe and Magleby (2000: 11), for example, defined a competitive primary as one in which "the difference in vote percentage between the first- and second-place candidates was less than 20 percent." Third, there are more sophisticated measures which take into account the number of candidates and their share of the votes. Such measurements are proposed by Kenig (2009&) to 1 For example, when we asked for the results of the candidate selection contests from the Irish Fine Gael, we were told that "usually the results would remain confidential from the candidate." (Personal communication with Vincent Gribbin, Head of Internal Communication, Fine Gael Headquarters, Dublin, Ireland, 26 January 2009.) 128 Democracy within Parties analyze leadership contests, but can also be used to analyze any contest for a single candidacy. Analyzing results of a contest for multiple positions requires more sophistication and caution than the straightforward case of competition for a single candidacy. A measure for analyzing the results of a multi-vote selection method is the vote concentration index (VCI; see Goldberg 1992). This index examines the number of votes concentrated on the candidates at the very top of the list. The more these votes are dispersed among the candidates, the more competition there is in the selection process. The formula is: VCI = Vnv stands for the number of votes won by the candidates in the top positions on the party list that is equal to the number of votes allocated to each selector in a selection event. Vt stands for the total number of votes cast. Nv stands for the number of votes allocated to each selector. The formula weights each selection event according to the number of votes that were cast. The higher the value of this index, the less competition there was because the votes were concentrated on fewer candidates. For example, in a single selection event, in party A 100 selectors took part in candidate selection and each was allowed to vote for 10 candidates. The vote total is 100 x 10 — 1,000. The top ten candidates together won 650 of these votes. The vote concentration index is thus 650/1,000 = 0.65. In party B 200 selectors took part in candidate selection and each was allowed to vote for just one candidate. The vote total is 200 x1 - 200. The top candidate won 120 votes. The vote concentration index is thus 120/200 = 0.6. According to this index, competition in party A was slightly greater than competition in party B. The vote concentration index can be used in cases where selectors have a single vote or multiple votes. This measure, however, needs to be developed further since in its current form it is sensitive to influences that can bias results, such as the number of contestants. Incumbents versus nonincumbents Another way to look at competition is to assess the success rate of newcomers in their attempts to challenge incumbents. The selection of newcomers indicates that even after a candidate was selected and elected in the last election, there is a chance that in the next election their position or seat will be lost. Such an analysis can be conducted in voting as well as in appointment systems. There is also no crucial need here for lists of the candidates who took part in the contest - although it can help in Competition 129 distinguishing between retirement and defeat - because the focus is not on the overall pool of candidates, but rather on the party's actual candidates for office in the general election.2 It should also be noted that significant inrraparty turnover - which can serve as a sign of competitiveness - may not reflect genuine democratic competition but rather power that resides in the.hands of local "party bosses" who replace incumbents so that no alternative power center is created (Field 2006). In single-member districts, counting the number of deselections is straightforward, as candidates are either in or out. Indeed, many scholars of US primaries, like Key (1967), look at the rate of incumbent success to compare levels of competition across cases. The analysis of candidate lists is more complicated because incumbents have to compete with each other, not just with newcomers, and as a result may move up or down the list (not off it).3 We suggest two indices to analyze the success of nonincumbents in either single- or multicandidacy contests (Hazan and Rahat 2006; see also Rahat, Kazan, and Katz 2008). The first is the nonincumbents winning index (NIWI), which weights the nonincumbents' share of positions at the top of the list (top is defined as the number of incumbents in the specific contest). The second index is the nonincumbents ranking index (MRI), which weights the relative positions of nonincumbents at the top of the list. A high share of the top positions captured by nonincumbents along with high rankings indicates a high level of competition. These measures can be used in contests with both single and multiple positions, and as such can be useful for cross-national comparisons. The nonincumbents winning index measures the "success" of new candidates according to whether or not they won "an incumbent's position." Thus, if 20 incumbent MPs competed for positions on the party list, the entrance of a new candidate in any of the top 20 positions would be considered a success. The definition of incumbent is any candidate who was elected to the previous legislature. The formula is: NIWI Wni is the number of nonincumbents who won a position on the list that is equal or higher in rank to the number of competing incumbents in each selection event. Ci is the number of incumbents competing in the selection event. For example, let us assume that in a single selection event five MPs compete for a position on the list of candidates for the next parliament. They win the first, second, third, fifth, and seventh positions on the candidate list. This means that one * Moreover, even when there are such lists, it is still possible that several incumbents predicted that they were going to be defeated and retired; and since we do not count them as defeated, the analysis might point to a somewhat lower rate of incumbent defeat 3 The rare occasions when incumbents face each other m selection contests for single-member districts usually occur after redistricting (Herrnson 1997). 130 Democracy within Parties Competition 131 nonincumbent won (the one who won the fourth position on the list). The non-incumbents winning index is therefore 1/5 = 0.2. Where selection is by party delegates or party members, those incumbents who appeared on the intraparty ballot - from which either party delegates or party members select their candidates - are considered as trying to win a realistic position. Where there is a more exclusive selectorate, such as a nominating committee, usually there are no such ballots or formal lists of candidates. The task of distinguishing the losers from those who willingly retired requires the construction of a small biographical profile for each MP who either disappeared from the official candidate list produced by the party or appeared in an unrealistic position. Newspaper accounts from the time of the selection, biographies and autobiographies, intraparty material, and historical studies can help to accomplish this complex task. The second index, the nonincumbents ranking index, counts the number of nonincumbents who won "an incumbent's position," taking into consideration the relative position that they won and giving higher values to winning higher positions on the list. That is, each position on the list, up to the position that is equal to the number of competing incumbents, is given a value in descending order - the last position is given one point and each higher position is "worth" an additional point. The formula is: NIRI = Vpni stands for the value of the positions won by nonincumbents in the selection event. Vpi stands for the total value of the "incumbenf s positions" in the specific selection event. The value of the victory in each selection event {Vpni! Vpi) is then multiplied by the number of incumbents competing in the selection event (Ci). We then sum all the selection events and divide the product by the sum of the number of incumbents competing in all selection events. For example, let us assume that in a single selection event, five MPs compete for a position on the list of candidates for the next parliament. They win positions one, two, three, five, and seven on the list. The total value of the list is five (for the first position) plus four (for the second position) plus three (for the third position) plus two (for the fourth position) plus one (for the fifth position) = 15. The single nonincumbent won position number four on the list which has a value of two. The value of the index (for an example with a single selection event) would thus be 2/15 =0.133. The "sophomore surge" is another type of measure that focuses on incumbency. It is calculated as the gap between the percentage of votes a candidate gets when first selected and his or her second bid (this time for reselection) as an incumbent (Ansolabehere et al. 2007). This measure can also be used in contests with multi-candidacies (such as a contest for several positions on the list), although its sensitivity to the influence of a change in the rules, such as the number of votes allotted to each selector, calls for caution and improvement. Summary of the indices Each index of competitiveness has its advantages and weaknesses. Each stresses a specific aspect of competition and ignores others. Some may prefer specific indices, pomting to their strengths over others from a theoretical point of view. All scholars, however, will be constrained by the type of data they have and by the ability of the measures to be adapted to a wide variety of selection methods. It should also be noted that the results of these measures are not independent from each other. For example, a high supply of candidates could point to a particular interpretation of competition - indicating that the candidates estimate that there is a chance for success - but this might actually reduce the values of other measures of competition. This is the case, for example, if we use turnover, or incumbency defeat, as a measure of competition. That is, a high number of competitors might benefit incumbents in their attempt to be reselected. When there is an incumbent competing, he or she is likely to be well known, and thus voting will largely be for or against the incumbent. The more challengers there are, the more these "negative" votes could be spread out, and the lower the plurality needed in order to win. Thus, being more competitive in one aspect (number of candidates) may lead to being less competitive in another one (defeat of incumbents). The optimal strategy for analyzing competition is to employ several indices, taking into consideration the possible covariance between them. The menu of indices for analyzing single candidacy contests is quite long. A comparative study of multiple candidacy contests requires an improvement in the existing measures and the development of additional ones. THE IMPACT OF SELECTION METHODS ON COMPETITrVENESS The high success rates of incumbents in their bids for reselection and reelection are well documented (Somit et al. 1994). Indeed, "deselection appears to be a relatively rare phenomenon, the norm being that incumbents desiring to run for reelection are renominated" (Maffand and Studler 2004: 97, emphasis in original). The advantage of incumbents results from their being recognized and known; the resources they have as members of the legislature to sustain visibility and their link with their constituency; and their ability to present achievements rather than promises, which gives them an edge in recruiting votes and other resources for their campaign (Gallagher 1988c). Yet, we also know that beyond the universal