f Í
t \
Part I presents a fourfold classification of candidate selection methods. Each ! chapter is dedicated to one of the four dimensions that delineate candidate \ selection methods. Each of the four dimensions answers a major question: \ Candidacy - Who can be selected as the parry's candidate? The selectorate - [ Who selects the candidates? Decentralization - Where does selection takes place? Appointment and voting - How are the candidates selected? We hope that by dedicating the first part of our book to conceptualization and classification we will contribute to the creation of a common, shared language for the study of candidate selection methods.
Differentiating between candidate selection methods is a precondition for conducting any meaningful studies regarding the origin, preservation, and reform of candidate selection methods - not to mention their political consequences. We present what we believe is the relevant menu for scholars to choose from. In som ■ studies it might suffice to address only the (arguably) two main dimensions: the selectorate and decentralization. Yet, in most studies - especially those that focus on the internal life of parties in general, and their democratization in particular -the other two dimensions will prove to be significant as well........................
We collected data about both new and old democracies around the world, in order to demonstrate how our classification can be used, as well as its applicability. In some cases the data were readily accessible, especially regarding the selectorate ' and decentralization. When it came to candidacy, the data were sketchy. Much to our regret, empirical information on appointment and voting systems is quite rare. Hopefully, shedding light on the analytical potential and the significance of these four dimensions will lead scholars to invest more time and effort in studying them. The systematic accumulation of cross-national data is a necessary precondition in order to make further progress in the study of candidate selection methods.
2
Candidacy
In any attempt to assess candidate selection methods, the first dimension that should be addressed is the question of candidacy: Who can be selected? This is, on the one hand, the simplest of the four dimensions that describe candidate selection methods; but on the other hand, it is the most brutal of dimensions because it has the potential to eliminate an overwhelming majority of the population from the pool of candidates.
Candidacy tells us who can present himself or herself in the candidate selection process of a single party at a particular point in time. Are there any restrictions on presenting candidacy in a given party? If so - how strict are these limitations? How much do they affect the size and nature of the potential candidate pool? The restrictions applied to potential candidates are the defining elements that will allow us to classify candidacy on a continuum according to the level of inclusiveness or exclusiveness, as elaborated later in Figure 2.1.
At one end, the inclusive pole, every voter can stand as a party candidate. Most states in the United States are close to this pole. This phenomenon can be attributed to the fact that state laws, rather than party rules, regulate the candidate selection, process.1. Under such candidacy requirements, or lack thereof, 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, reluctantly or otherwise. Politics in the United States have been described as candidate-centered (Wattenberg 1991), and in the arena of candidacy, possibly the strangest example was the cross-filing system in the state of California where, between 1913 and 1959, a candidate did not need to be a member of any party and could compete in the primaries of more than one party at the same time (Key 1967).
At the exclusive pole, we encounter a series of restrictive conditions. One example is Obleť s (1974: 180) account of the requirements that applied to potential candidates in the Belgian Socialist Parry. Here the party's role as gatekeeper, already at the candidacy stage, was both strong and influential.
1 This fact led Epstein (1986) to liken the American parties to "public utilities," such as water or electric companies.
20
Democracy within Parties
Candidacy
21
While the exact requirements vary from one constituency to another, they generally stipulate that to be placed on the primary ballot aspirants must (1) have been a member of the Socialist parry, trade union, co-operative and insurance association for at least five years prior to the primary; (2) have made annual minimum purchases from the Socialist co-op; (3) have been a regular subscriber to the party's newspaper; (4) have sent his children to state rather than Catholic schools; and (5) have his wife and children enrolled in the appropriate women's and youth organizations. These conditions, in effect, require that a candidate serve as a member of an activist subculture before he becomes eligible to run for Parliament. They involve a form of enforced socialization during which it is assumed (or hoped) that the aspirant will absorb the appropriate values and attitudes as well as a keen commitment to the party.
The more common requirements set by parties for candidacy are less demanding, such as a minimal length of membership prior to the presentation of candidacy and pledges of loyalty to the party. The multitude of real-world examples result in a candidacy continuum, as presented in Figure 2.1, based on the extent of inclusive-ness or exclusiveness imposed by the party - in most cases - or the state on eligibility.
Why would a party adopt more inclusive, or more exclusive, candidacy requirements? Inclusivity might be the result of legal regulations, as is the case in the United States. But it may also be a trait of a party whose main interest is electoral success, and is therefore open to any candidate who could help increase the party's share of the vote. Where little to no membership requirements exist, such as even a minimal membership period, the resulting candidates could be newcomers to the party. For example, in Canada, among the nonincumbent candidates elected to parliament in 1988, 14 percent joined their parties during the year prior to the elections (Erickson and Carry 1991).
Exclusivity, on the other hand, may be due to an attempt by the party to control the supply side of potential candidates, so that those who fulfill the enhanced eligibility requirements, and are subsequently both selected and elected, will behave according to party dictates. In other words, additional requirements can not only remove those candidates with potential personal problems, but also assure certain behavioral patterns once in office. A party with strict candidacy requirements can arrive in office as a cohesive unit, manifesting a patent party culture, thereby removing the need to utilize disciplinary measures in order to keep their
AH Party Party members +
citizens members additional requirements
inclusive %----«&. Exclusive
Figure 2.1. Candidacy Source: Updated from Rabat aad Hazan (2001),
I elected representatives in line (Hazan 2003). Moreover, the party leadership can
I use the more exclusive candidacy requirements to reward loyalists and long-
I time activists, thereby creating a structure of selective incentives for potential
i candidates.
j Overall, the more electorally oriented and catch-all parties might decide to
pursue inclusive candidacy requirements, while the more ideological parties may
'I adopt exclusive restrictions that ensure a candidate's "socialization" into the party culture. The adoption of particular candidacy requirements can thus be a trade-off
.j between different electoral and programmatic goals. Gallagher (1988c: 247)
1 argued that variations among parties in their candidacy requirements cut along
'} ideological lines. Left-wing parties instituted more formal candidacy requirements
. i for selection than right-wing parties; that is, left-wing parties are more exclusive,
i Thiebault (1988; 82) also found that in France the parties of the right, more so than
I those on the left, selected candidates who were not even party members before-
.{ hand. Obler (1970) showed that the very stringent candidacy requirements of
f the Belgian Socialist Party, described above, were quite different from those
" i of the Belgian Christian Social Party, where the relatively inclusive candidacy
i requirements were limited to one year of party membership and (in some districts)
i a maximum age of sixty-five.
At times, parties will ignore their own candidacy regulations, largely due to
t electoral considerations. For example, even the rather exclusive Italian Commu-
J nist Party included nonmembers as candidates. These "prestigious independents
I not previously identified with the party" were "a central feature of the party's
ii electoral (and image-building) strategy" (Wertman 1988: 154). The Irish Labour ? Party required candidates to be party members for at least one year prior to the
elections, but when it came to those unaffiliated potential candidates who the party
; thought were attractive, this requirement was overlooked (Galligan 2003). The
Czech Christian and Democratic Union - Czechoslovak People's Party placed
\ nonparty members on its candidate list for the European Parliament although the
| selection rules did not allow this (Linek and Outly 2006). i Before outlining some of the more common candidacy requirements cross-
: nationally, it is important to distinguish between party-level requirements and
\ state-level candidacy restrictions. This chapter is concerned with the former,
; which is independently decided by each individual party. In other words, there
I is likely to be variation across parties because each can choose the criteria it wants
to focus on. One party can decide to concentrate on general rules such as district
i residency, age, and a monetary deposit, while another might focus on ideological
* loyalty and a history of party activities. Moreover, each party can set the standard
\ for the particular requirements it chooses - residency, age, and a monetary deposit
: might be similar criteria across several parties, but the degree of each could vary
; significantly.
I State-level candidacy restrictions are a different, yet related world. They are
! different because they are typically delineated in a formal document, such as the
22 Democracy within Parties ■
constitution, the election law, laws governing the legislature, etc. Some restrictions f are "hidden" because they are implicit. For example, if the election law states that \ in order to be a candidate one must have the right to vote, then all the limitations [ placed on voting eligibility - from age to residency to serving time in prison, etc. - [ are automatically imposed on candidacy as well.
The most common national-level qualifications on who has the right to be a \ candidate are age, citizenship, residence (country and/or district), monetary deposit, s and incompatibility with other offices. Additional common eligibility restrictions I include the establishment of certain standards for prospective politicians - | candidates are frequently disqualified on the grounds of insanity, criminal convic- I tions, undischarged bankruptcy, and in some underdeveloped countries educational j and literacy disqualifications (Inter-Parliamentary Union 1986).2 [
State-imposed candidacy restrictions are related to party candidacy require- ; ments because they set the lowest common denominator to which all parties : must adhere. For example, if the state decides that all candidates for legislative office must be citizens by birth (Bolivia, Brazil, Ecuador, the Philippines, and Venezuela), or by naturalization (most countries), or that they need not be citizens (Jamaica and St. Vincent) (Massicotte, Blais, and Yoshinaka 2004: 55), then either it is ludicrous for a party to adopt a lesser limit, or parties do not have to address this issue because the state has already taken care of it. State-level restrictions, -whether describing ineligibility criteria - such as incompatibility with other public offices - or eligibility requirements - such as citizenship, number of signatures required on a nomination paper, etc. - create a uniform base which is standard for all parties. No variation across parties in either the criteria or their extent is to be found/
Since our interest lies in how-parties.....select their candidates, we are less ■
concerned with the restrictions imposed by the state, as they do not allow us to distinguish differences between parties nor, as a result, to assess the political consequences of these variations within the state. State-level candidacy restrictions versus party-level requirements are akin to the national electoral law versus the candidate selection method of each party. This book focuses on the latter; we are interested in differences between parties rather than between states. When candidate selection is regulated by law, studying it tells us more about the legal regulations of the political process than about the political parties themselves. s However, we do not ignore the restrictions on candidacy emanating from the state
2 A comprehensive cross-national sample of state-level candidacy requirements is presented by Massicotte, Blais, and Yoshinaka (2004: 42-9).
A curious case is Finland, where the Election Act regulates parties, but the parties are allowed to ::
deviate from many of the stipulations. Most of the provisions of the act apply unless a party enacts rules on candidate selection (Ministry of Justice, Finland). Sundberg (1997) and Kuitunen (2002) point out that the parties have chosen to fol low the legal text almost verbatim, but there are still some differences s between parties, as well as deviations from the Election Act. : (
Candidacy 23
because they define the playing field on which the parties must compete, and thus they are important, yet preliminary, to our discussion.
Common party-level candidacy requirements include criteria that are similar to the state-level restrictions, such as age, but many are quite different. For example, a minimal period of party membership is extremely common, as are pledges of loyalty, usually to the party platform, and a collection of signatures in support of candidacy. Data on party-level candidacy restrictions are quite hard to find. The criteria are delineated either in parry statutes, available usually in the local language only and rarely translated or disseminated in a more general manner -that is, the internet - or they are to be found in party documents that are inaccessible to most. The scholarly literature rarely covers this aspect of candidate selection, if at all. An expansive reading of the academic literature, along with years of searching for these data, allows us to present a preliminary delineation of party-level candidacy requirements along with cross-national examples. Prior to analyzing candidacy requirements at the party level, though, we must delineate the unit of analysis.
Our unit of analysis is the single party, in a particular country, at a specific point in time. Only in cases where several parties in a country impose similar candidacy restrictions (usually due to legal requirements), where a single party implements similar candidate requirements more than once, or when both similarities occur, can we make generalizations over time and across parties.
The tools offered in this chapter, and the ones that follow, can be easily used when analyzing a simple, one-stage, uniform candidate selection method. Such a simple method is one in which all potential candidates face similar restrictions. Empirically, however, we face complex candidate selection methods. Since these largely come into play when we-discuss the selectorate - our second dimension -they will be delineated and analyzed in Chapter 3. When it comes to analyzing candidacy requirements we must be careful, because even in one specific party, different candidacy restrictions might be imposed in different constituencies or regions.
AGE RESTRICTIONS
Age limit is a common requirement placed on potential candidates. For example, the Austrian Socialist Party did not have any age restrictions in 1945, but by 1959, it imposed a detailed restriction: all candidates had to be under the age of sixty-five and those who had never been candidates before, or were nonincumbents, had to be less than sixty years old (the Austrian People's Party copied this ten years later). Circumventing this candidacy restriction was possible, but only with a two-thirds majority in the full party executive (Muller 1992). Most of the Belgian parties
24 Democracy within Parties
imposed an age limit of sixty-five for candidates (De Winter 1988). In the Australian Labour Parry of New South Wales, as well as in the Labour Parry of New Zealand, a candidate's age could not surpass seventy at the end of the prospective term in office the candidate wished to fill (Norris et al. 1990; Sheppard 1998).
The fact that many parties impose an age limit on candidates raises important questions. For example, what leads parties to impose such limits? Are parties disinterested in representing the elderly and in mobilizing their supporters? Are long-term incumbents, many of whom are nearing the age limit, not experienced legislators that the party could use? Age restrictions, it seems, are a reaction to multiple-term incumbents who have become professional politicians with no desire to be replaced. The "young Turks" of the party feel frustrated by these perennial legislators, and conclude that their entrance into politics is stymied by the "old-timers." Moreover, some of the older incumbents can tarnish the party's image, especially if they are perceived to possess health or mental problems that may impair their ability to function. Imposing an age limit thus both creates vacancies and relieves the party from having to deal with problematic personal cases.
It will be interesting to see whether age restrictions will be changed or challenged as life expectancy increases and people are able to be productive well beyond their sixties, as the electorate grows older and pensioners become increasingly successful in their attempts to establish political organizations. An attack on age restrictions could also be a matter of principle in the name of antidiscrimination laws, as the experience of the United States testifies, since these limitations are clearly targeted at a specific segment of the population and are indeed prejudiced.
PARTY MEMBERSHIP REQUIREMENTS
A minimal membership period in the parry is one of the most common political candidacy requirements. There are, however, many parties that do not impose a party membership period as a candidacy requirement. Among these are the Swedish Liberals, who even invited people not associated with the parry to run as candidates in the 2004 European elections (Aylott 2005); the Left Party in Sweden (Pierre and Widfeldt 1992); the Japanese Liberal Democratic Party (Shiratori 1988); the Social Democrats in Iceland (Kristjánsson 2002); and the three main parties in Ireland - Fianna Fail, Fine Gael, and the Progressive Democrats (Gallagher 19886). A study of political parties in Bangladesh, India, Nepal, Pakistan, and Sri Lanka noticed a willingness by parties across south Asia to run candidates who were not party members, but who bad a good chance of winning (Suri 2007).
|
J Candidacy 25
j Our research has shown that membership requirements can be quite minimal. In
| Finland, where parties are somewhat regulated by the state, the law did not require I a candidate to be a party member, although the parties expected this to be the case l (Kuitunen 2002). In New Zealand's National Party, a candidate simply had to be a
f paid member at the time of selection (Jackson 1980). fn the Finnish Social ? Democrats, candidates needed to be members for only four months before the \ primary (Kuitunen 2002), and in the Irish Greens the minimal membership period I        for candidacy stood at six months (Galligan 2003).
| Variation across parties in a single country is common. For example, in Ireland
f while the three main parties (Fianna Fail, Fine Gael, and the Progressive Demo-t crats) did not prescribe any minimum period of party membership, the Labour [ Party had a minimal period of six months, Sinn Fein had a one-year eligibility minimum, and the Workers' Party restricted candidates to those who had been ?        members for at least two years (Gallagher 19886).
Many parties set membership requirements of at least one year, and often two years, before a member could be selected as their candidate. The extreme cases I were some of the Belgian and Italian parties, which required their candidates to be members of the party for at least five years - which effectively meant that they had to be party members for at least two elections and could only attempt to become candidates in the second election. It seems that significant candidacy restrictions are more characteristic of the ideological era of the mass parry. Catch-all and cartel [ parties tend to ease and bypass these requirements, largely due to electoral considerations.
ADDITIONAL REQUIREMENTS
Most parties do not mention criteria such as citizenship and residency in their candidacy requirements, mainly because they are already stipulated in national election law.4 There are, however, a series of requirements that parties impose in addition to those that the state requires of all candidates across parties.
A monetary deposit, which some countries impose at the state level, can also be found within parties. The Canadian Conservatives required a $1,000 deposit (Conservative Parry of Canada 2009). The Christian Democratic Union -Czechoslovak People's Party required payment of a nomination fee of CZK 10,000 (about $500) from their candidate to the European Parliament (Linek and Outly 2006). These sums are not returned to the candidate unless they
J The Christian Democrats in Germany can be seen as a counterexample, when it stipulated that only German natives were eligible for candidacy to parliament (Poguntke and Boll 1992).
26 Democracy within Parties
obtain a certain percentage of the vote. In Kenya, the National Rainbow Coalition required parliamentary aspirants to pay a nomination fee of approximately S3 80 (Ohman 2004).
As cited above, the candidacy requirements of the Belgian Socialist Party in the 1960s seem to be quite excessive. Beyond five years of party membership and membership in the Socialist trade union and health insurance fund also for at least five years, the party required membership and minimal purchases at the socialist cooperative, a subscription to the party's newspaper, and the holding of some party office. All of these pertained mainly to the potential candidate, but there were requirements placed on the candidate's family as well: his wife had to be a party member in the relevant organization, enrolled in the trade union and insurance fund, while his children had to go to state schools and be members in the party youth organization (Obler 1974; De Winter 1988). These stringent eligibility requirements are no longer strictly enforceable, due to a decline in the pillarized structure of Belgian society - for example, the disappearance of party newspapers and cooperatives - resulting from the waning of consociational politics.
Other parties impose their own particular requirements on potential candidates, but they seem to be a far cry from this extreme example. The Irish Workers' Parry, beyond a two-year membership period and a record of party activities, required candidates to take an internal educational course. The parry's National Executive could, however, suspend this requirement. Fine Gael and the Progressive Democrats, on the other hand, required candidates to make a pledge, before being selected by the party convention, that they would "contribute" an amount of money decided by the parliamentary party to the party's election campaign (Farrell 1992).
A common, political requiremeM...is...a..pledge..o£..lo.yalty, usually to the party platform, as practiced by the three main Irish parties (Gallagher 19886). These pledges could include provisions for supporting whichever candidate is eventually selected by the party, as was the case in several of the parties in New Zealand (Milne 1966), or, once elected, promising to vote in parliament based on the party's decision. While such a requirement may influence the potential pool of candidates, its significance compared to other candidacy requirements is secondary.
Some parties require written recommendations by existing party members, regardless of whether the candidate is a member or not. The Labour Party in New Zealand allowed six members to nominate a candidate, while the National Party required that ten paid members should recommend a candidate (Milne 1966). Both the Liberal and Conservative parties in Canada require twenty-five signatures (Liberal Party of Canada 2009; Conservative Party of Canada 2009). In Finland, candidates could be nominated by fifteen members of a local organization, or by thirty members of different local organizations all in the same district (Ministry of Justice, Finland). In Iceland, candidates needed between twenty and fifty endorsements by party members, depending on the size of the constituency (Kristjansson 2002). The Mexican Party of the Democratic Revolution decided, in
| ■
t Candidacy 27
- 1996, to require at least 100 member signatures for candidacy eligibility, but the
" candidate did not have to be a member of the party (Combes 2003). In this
'"• requirement, too, variations across parties in the same country are common.
I In the Netherlands, for example, the Socialist Party called for at least ten members
* to nominate a candidate, the Green Left required fifteen, while the Catholic 1 People's Party insisted on twenty-five (Koole and van de Velde 1992).
1 In some parties, such as the British Labour Party, candidates need to be
i nominated by a party unit in the constituency - a party branch, an affiliated
: union, or another recognized group (Norris and Lovenduski 1995). In the United
"-, States, an interesting requirement for candidacy existed in several of the southern
• states. If a candidate had campaigned against the Democratic Party in the previous ' elections, he could not compete in the party's primary (Key 1967: 392).
The four main Czech political parties required knowledge of at least one world language from their candidates for the European Parliament, and two of these also required membership in the national parliament, or alternatively, experience in municipal or regional boards of representatives (Linek and Outly 2006). The Flemish wing of the Belgian Social Christian Party, prior to the 1968 elections, decided that its candidates were forbidden from holding office in local governments (unless they were composed of 30,000 inhabitants or less) and thus would have to resign before becoming candidates for the national parliament. Such a prohibition was quite significant and obliged candidates to decide which position they preferred to continue holding, as many candidates had backgrounds in local government (Obler 1970).
INCUMBENCY
Incumbents are a special category of potential candidates. In several countries, such as Australia, Great Britain, New Zealand, and The Netherlands, incumbents in one or more parties enjoyed either automatic readoption or almost automatic readoption. In the Irish Fianna Fail, it was common at conventions to pass proposals declaring all incumbents reselected (Gallagher 19886). The Japanese Liberal Democrats, almost without fail, renominated incumbent members (Shir-atori 1988). Even minor parties, such as the Volksunie in Belgium, reassigned the same place on the list to incumbents who wanted to run again, unless their constituency congress voted otherwise by a two-thirds majority (De Winter 1988). Many times, those incumbents with guaranteed candidacy did not even face any of the requirements that nonincumbent candidates had to. In other words, once an aspiring candidate successfully met the candidacy requirements and was both chosen to be the candidate and elected to office, from then on there were no candidacy requirements imposed on that particular person.
28 Democracy within Parties
Why would a party offer its incumbents automatic candidacy for an upcoming election, or at least ease their requirements? The simple answer is that parties want to win elections, and incumbents have already proven themselves in this regard. Incumbents are, therefore, likely to have a strong base of support in their constituency, making reselection especially relevant for parties who function in plurality single-member district electoral systems, and also in electoral systems that employ districts of small magnitude. Incumbents are also strong actors within the party; they know how to take care of their interests and have been working toward the goal of reselection ever since they were first chosen. Incumbent reselection can be seen as a reward given in exchange for loyalty to the party. Parties also want to minimize internal conflict on the eve of the more important general election, and the reselection of incumbents means that they do not have to fight off potential challengers before each election.
Why, then, should a party impose more demanding candidacy requirements for incumbents? The main reason is that if all incumbents are automatically rese-lected, then the party will appear monotonous, which is not a good image. New faces and new blood are perceived as positive projections of an interesting and exciting party prior to the general elections. But, on the other hand, a balance is needed between incumbents and aspirants - wholesale replacement of incumbents is not good for a party, and is also not likely to happen, as explained above. In order to attract new contenders to the parry, the candidacy requirements for incumbents must afford competition, allowing aspirants to enter the game and feel that they have a fair chance, but also giving incumbents the feeling that they are relatively safe and can focus their attention elsewhere. Yet, the party leaders also have their political concerns, and incumbents form the pool of potential challengers for their leadership..Theyjnay..1hus-aeek.Jo..thwart,-leadership-competition by forcing incumbents to focus more on their efforts at being reselected, not to mention the possibility that a potential contender for leadership could be removed when incumbents (but not the leader, or leaders) face special hurdles - such as the support of special majorities.
An interesting study in Britain of thirty-five attempts by the local selecting body to deselect incumbents found that the most common source was ideological, making up one-half of the cases, with other reasons - such as a lack of attention to constituency matters, personal failures, or age - far behind (Dickson 1975).5 Nonetheless, in spite of the conflict, the local organization usually readopts the incumbent candidate because: (a) it fears that alternation could lead to an election defeat, (b) its disappointment with the incumbent can be smoothed over, (c) the Burkean conception holds that the local organization should not control its candidate, and (d) the national party organization gets involved in support of
= Almost all of the ideological conflicts involved incumbents whose ideological leanings were toward the other main party - that is. Conservatives leaning left and Labourites leaning right - and not those incumbents who held more extreme positions.
Candidacy 29
the incumbent. In short, even when readoption is not fully automatic, the incumbency advantage works in favor of the readoption of incumbents.
More recently, some parties have indeed sought to shake up their lists of candidates, mainly to present a fresh face to the public. Automatic readoption seems to be less common than it used to be, and many times incumbents have to face performance reviews. The most well-known example is the "mandatory reselection" of members of parliament adopted by the British Labour Party in the early 1980s.6 Yet, as Gallagher (1988c: 249) stated, "Although few parties appear to have rules specifically protecting incumbents from the full uncertainties of the selection process, the great majority survive nonetheless." Indeed, Labour's adoption of mandatory reselection left few casualties.
Several parties have made it more difficult for incumbents by adding hurdles in the path to reselection. In these cases, the incumbent has different candidacy requirements to new aspirants. For example, in India the Congress Party's Central Elections Committee decided in both 1957 and 1962 that one-third of its incumbents should be replaced (Graham 1986). Moreover, the party recommended that all incumbents who held office for ten years should voluntarily resign. Prior to the 1967 elections, the party eased its requirements but nevertheless decided that incumbents would be reselected only if they had won their district by more than 1,000 votes and the district had remained unchanged. In Italy, the Christian Democrats proposed a four-term limit, but this was not adopted; and the Communists rotated candidates after two terms, unless the party decided that it was crucial for a particular candidate to remain in office (Wertman 1977).
Since 1972, the Austrian People's Party has included a stipulation that the renomination of a deputy for a fourth consecutive term requires a two-thirds majority-in a secret ballot. This rale.does not.apply, to membership in.parliament or to parliamentary terms of less than two years (Muller 1992). In Israel, both Labor and the Liberals required two-term incumbents to win at least 60 percent of the votes in the party's central committee in order to be eligible for reselection (Goldberg and Hoffman 1983). In Argentina, the Radical Civic Union required incumbents to gain two-thirds of the votes in order to be reselected (Field 2006). The German Greens instituted a rotation rule, and in order to remain in office an incumbent had to gain the support of no less than 70 percent of what in the Green Party is known as the "relevant basis" (Ware 1987). In 1986, the Green Party in Sweden decided that MPs may not be elected more than twice in succession, thereby setting a three-term limit (Pierre and Widfeldt 1992).
6 Mandatory reselection as opposed to automatic readoption, is "... a process under which, in place of more or less automatic readoption, they [MPs] would have to be reselected in competition with other contenders before each general election" (Shaw 2001: 36). While automatic readoption was indeed reduced, competition for a position held by an incumbent became open only under specific conditions, rather than in all cases, and changes through the years signified some reversal of this policy - i.e. the number of "nominations" needed in order to challenge an incumbent was increased.
30                               Democracy within Parties rj;
OBSTACLES TO THE CLASSIFICATION OF \
CANDIDACY REQUIREMENTS \
i
There are three main sources for gathering comparative lnlormation on candidacy |
requirements at the party level. The first is a data handbook on political parties. ?
The best-known compendium of this type is Katz and Mair (1992), but many of ?
the country specialists providing the data paid little or no attention to candidacy \
requirements within the parties. \
The second source is volumes devoted to candidate selection, where country I
experts are asked to address this specific issue. The main, and practically sole :■
source of this nature is Gallagher and Marsh (1988). Narud, Pedersen, and Valen \ (2002c) focus on only four Nordic countries, while Ohman (2004) covers Africa alone and concentrates mainly on Ghana. All three volumes pay scant attention to
candidacy requirements at the party level. '
The third source is data compiled by democracy promotion organizations, particularly those who have democracy within parties on their agenda. The 1 National Democratic Institute, for example, presented the best and most up-to- j date information on the subject in its coverage of several parties around the world ] (Ashiagbor 2008). In this report one leams that the British Labour Party, for j example, poses somewhat minimal candidacy requirements: continuous membership in the party for at least one year, membership in a trade union recognized by | the party, and contribution to the fund of that union. The Canadian Liberal Party's j national candidate selection procedures for 2007 spelled out the following specific j eligibility criteria (Ashiagbor 2008: 57): -|
• Current membership (must also be, in goo A standing); •
• Full and truthful completion of relevant forms; f
• Eligibility under the laws of Canada; ! e Satisfaction of any debts to the party and its constituent elements; I
• Compliance with federal and relevant provincial and territorial association j rules; ;|
• The signatures of twenty-five members in good standing; j
• Willingness to undergo background checks or face sanctions; :|
• Approval of the provincial or territorial chair to be a qualified contestant. (This j approval may be revoked by the leader in his/her sole discretion at any time.) 1
Seemingly - but maybe not practically - more demanding eligibility requirements '.j
for selecting candidates to legislative office can be found in the party constitution j
(Article 11(4)) of the New Patriotic Party in Ghana, which states that in order to [ seek the party's nomination an individual must (Ashiagbor 2008: 61):
• Be a known and active member for at least two years; jl
• Be a registered member and voter in the constituency which he or she seeks to i represent; ;i
Candidacy 31
• Be of good character;
• Be of good standing;
» Qualify under the country's electoral laws;
• Pay the fee prescribed by the party's National Executive Council;
• Sign "Undertaking of Parliamentary Candidates."
Moreover, the political parties in Ghana routinely require prospective candidates to pay nomination fees. In some cases, there are two separate fees: one for the application form; and a second for filing the application with the party. The African National Congress in South Africa includes even more exceptional requirements. Its candidates must (Ashiagbor 2008: 31):
• Be a members of good standing with a proven track record of commitment to, and involvement in, the democratic movement;
• Have the requisite experience or expertise to make a constructive contribution;
• Have no criminal record, excluding politically related crimes before April 1994;
• Have no history of ill-discipline, corruption, involvement in fostering divisions, or breaching the party code of conduct.
A comprehensive database on candidacy requirements is still unavailable. The proliferation of political party websites makes this undertaking only slightly easier, because many parties do not include the details of candidate selection in general, and candidacy in particular, on their websites, and where this does appear it is usually only in the home language of the country. Hopefully, the recent increase of interest and publications on candidate selection will make such an endeavor possible in the near future. Preliminary research on this topic is thus still extremely difficult, not to mention an analysis of the change in candidacy requirements over time or of the differences across parties.
DEMOCRATIZING CANDIDACY REQUIREMENTS
Democratization is one of the more recent and interesting trends in candidate selection methods (Bille 2001; Hazan and Pennings 2001; Hazan 2002; Kittilson and Scarrow 2003; Scarrow, Webb, and Farrell 2000). Claims concerning the occurrence of this trend are based on data about the selectorate, but not candidacy. However, we can define democratization as a widening of participation in both the supply and the selection process - that is, when parties adopt more inclusive candidacy requirements and selectorates. In order to democratize candidacy requirements, parties have to reduce the restrictions on eligibility, thereby creating a much larger pool of potential candidates, but in order to democratize candidate selection the selectorate must also be more inclusive.
32 Democracy within Parties
Thus, in order to understand the significance of democratizing candidate selection, which will be addressed in the second part of this book, it is important to assess the relationship between candidacy requirements and the party selectorate -the first two dimensions in our framework for analyzing candidate selection methods. A high level of inclusiveness on both has significant political consequences, but inclusiveness on one dimension combined with exclusiveness on the other could preclude or constrain these consequences. For example, if more inclusive candidacy requirements are adopted, yet the same limited selectorate is maintained, control over the final results has not been significantly reduced. The Italian Communists included even nonmembers as candidates, but this was done under the supervision of an exclusive selectorate (Wertman 1988). The same is true for the opposite case of the Belgian Socialist Party that we described above, in which a relatively inclusive selectorate of party members was combined with highly exclusive candidacy requirements. Having described candidacy requirements for legislative office, we now turn our attention to the second and more important dimension in candidate selection methods - the selectorate.
! 3
i __
\ The Selectorate
\
i
S After candidacy requirements, the second dimension in the analysis of candidate \ selection methods is the selectorate: Not who can be selected, but rather who is \ selecting. These two dimensions are akin to the "supply" and "demand" sides of | candidate selection. While candidacy narrows the supply of contestants who can I be selected, the selectorate will further decrease their numbers to those who will eventually face the voters in the general election.
The selectorate as a factor in party politics - as far as we can tell - was first addressed in a book by Paterson (1967), appropriately titled The Selectorate. In it, he describes candidate selection in Britain, which he finds dissatisfying, and argues for the adoption of party primaries. However, despite the focus of the book, there is no systematic treatment or definition of the selectorate. More recently, the term has been adopted by game theorists with a focus on the interplay between domestic political institutions, foreign policy, and the survival of leaders (Bueno de Mesquita et al. 2003). This model provides a definition of the selectorate as "the set of citizens who have a prospect of becoming members of an incumbent's winning coalition" (Bueno de Mesquita et al. 2002: 273), and it is the winning coalition — a subset of the selectorate — that is required for a leader, to achieve or sustain political power (Enterline and Gleditsch 2000). This definition is of little help to scholars of party politics because it equates the selectorate in a democracy to the electorate, suitably adjusted for voter turnout However, it is interesting to note that the rational choice approach emphasizes the size of the selectorate as an important variable.
When speaking of political parties in general, and candidate selection in particular, the selectorate is the body that selects the party's candidates for public office. It is, as Best and Cotta (2000a: 11) described, "an important intermediary actor ... the party organizations, the personal cliques, the groups of dignitaries or state officials involved in the selection of candidates and in their representation to constituencies." Indeed, the selectorate can be composed of one person or several I        people - up to the entire electorate of a given nation. While each criterion used in i        the classification of candidate selection methods has a distinct influence on politics, it is the selectorate that imposes the most significant and far-reaching consequences on politicians, parties, and parliaments more than any other dimen-i        sion of candidate selection.
r
34 Democracy within Parties
Duverger (1954: 353) argued that representatives receive a double mandate. "Before being chosen by his electors the deputy is chosen by the party.... The importance of each varies according to the country and the parties; on the whole the party mandate seems to carry more weight than that of the electors." Particular constraints are placed on the candidates based on the distinctive priorities of each selectorate - different selectorate priorities produce different candidates. In their study of Labour selectors, Bochel and Denver (1983: 45) posit that, "By their choices they effectively determine the range of abilities, social characteristics and ideological viewpoints present in the House of Commons." Best and Cotta (2000a: 11-12) expressed this most appropriately:
Selectorates select candidates according to the result of complex choices considering the probable value of the contender's resources for electoral success, their ideological fit with and their practical function for the selectorates themselves and their likely loyalty, that is, their expected obedience to the implicit and explicit expectations of the selectors after becoming a parliamentary actor. Since selectorates have not only a demand position on the recruitment market but must also make convincing offers to the electorate, the relative weight of factors working in the selective process is variable: for example, in a situation when a selectorate is in secure control of a significant part of the electoral support market, campaign qualities of contenders will be of less importance than their expected loyalty or their ideological fit.
Changes in the selectorate are thus expected to have significant political consequences. These changes do not usually occur before each election, because candidate selection methods, like other political institutions, enjoy a certain level of stability. Yet, significant changes in candidate selection methods indeed occur . among them changes in the selectorate - much more frequently than changes in electoral systems. This makes candidate selection methods in general, and the selectorates as the most important dimension of candidate selection methods in particular, an important source for political renovation and change.
Significant changes and reforms in candidate selection methods result from the interplay between three levels of parry politics. First is the intraparty arena, in which factions, camps, and individuals compete for power. At this level, changes are initiated and promoted to improve the positions of certain individuals and factions within the party. For example, the young guard might promote the adoption of party primaries, believing that they might improve its chances to advance within the party and push out the old guard. Second is the interparty arena, in which parties compete with each other for power, mainly through elections. Here, many times, changes are initiated and promoted to improve the position of one party vis-a-vis its competitors, the other parties. For example, a party that suffered defeat in the elections may reform its candidate selection method in order to improve the party's image vis-a-vis the other parties. Finally, there is the political system level - the general environment in which the parties
t The Selectorate 35
act. Here we would expect general social, technological, and cultural develop-l ments to direct and constrain parties when they decide to preserve or reform their
; candidate selection methods (Barnea and Rahat 2007). The adoption of more
inclusive selectorates is explained, at this level, as a reaction to general long-1 term developments that led to the decline or the adaptation of the parties (Scarrow
\ 1999a; Katz 2001).
I This chapter focuses on a delineation of the selectorate, so that researchers
across political systems will have a common language when trying to classify and ; analyze selectorates both of their own country and of other countries. Adopting
I tiiis classification allows for comparison within countries, across parties and time,
> and between countries. Subsequent sections of this book will address the political
\ consequences of a democratic reform in a party's candidate selection procedure as
a result of expanding the size of its selectorate.
SIMPLE, ASSORTED, MULTISTAGE, AND WEIGHTED CANDIDATE SELECTION METHODS
A simple selection method is one in which a single selectorate selects all of the candidates. Such methods are easy to classify according to their levels of inclu-siveness. As Figure 3.1 suggests, we can distinguish five archetypical kinds of selectorates:
1. The most inclusive selectorate: voters. This selectorate includes the entire electorate thai has the right to vole in general elections.1
2. The highly inclusive selectorate: party members. Here we include party membership in its European sense; that is, not simply registration as a party affiliate
Voters Party Party Party Single
members delegates elite leader
Inclusive «------ -............-...........-—---• Exclusive
Figure 3.1. Party selectorates
1 In some countries, tbe rules for this category can be even more inclusive than those for the electorate. For example, in Canada, immigrants who are not yet citizens can become party members and take part in candidate selection (Cross 2004). In Israel, an amendment to the Parties Law of 1992 (article 20A) lowered the minimum age of party members to seventeen, which allows a minor to vote in those parties that hold primaries, while the minimum age for voting in general elections is eighteen (Israel Parties Law). In Finland, the Center Party allowed its members to take part in party primaries from the age of fifteen (Kuitunen 2002).
36
Democracy within Parties
administered by the states as in the United States - this belongs to the above category - but registration that is controlled by the party itself.2
3. The in-between selectorate: party delegates. This selectorate is composed of representatives selected by the party members. They can be members of party agencies (e.g. conventions, central committees, or congresses) or delegate bodies that were especially selected for this purpose alone.
4. The highly exclusive selectorate: the party elite. Here we include small party agencies and committees that were indirectly selected, or whose composition was ratified by wider party agencies, and also other less formal groupings.
5. The most exclusive selectorate: a nonrinating entity of a single leader.
The categories proposed above can be easily used when analyzing a simple, one-stage, and uniform candidate selection method. Such a simple method is one in which all potential candidates simultaneously face a similar selectorate. Empirically, however, we often face complex candidate selection methods - methods in which different candidates face selectorates with different levels of inclusiveness, or the same candidates face several selectorates with differing levels of inclusiveness.
We distinguish between three kinds of such complexities: the assorted, the multistage, and the weighted candidate selection methods. In an assorted candidate selection method, as shown in Figure 3.2, different candidates face selectorates that differ in their levels of inclusiveness. The Belgian parties, from the 1960s until the 1990s, serve as an example of an assorted system, one that used different selectorates for selecting candidates of the same party. Inside the main parties, some candidates were selected by party members while others were selected by delegates or members of local and central party agencies, and still others were appointed by local elites (De Winter 1988; De Winter and Brans 2003; Deschouwer 1994; Obler 1970). In Australia, candidates from the same party face substantially diflcrent selectorates across the Australian states (Norris et al. 1990).
In the multistage candidate selection method, the same candidates have to face more than one selectorate during the selection process. For example, in the British Conservative and Labour parties, special national party committees screened aspirants and created a list of eligible candidates. Then, a small local executive party agency (about twenty to twenty-five people) filtered candidates and compiled a "short list" from the dozens or even hundreds of aspirants, which was subsequently presented to a more inclusive party agency for selection - and in recent decades, to the even more inclusive selectorate of party members. The first selectorate (sometimes even selectorates) filters, or screens, the candidates, further minimizing the overall pool that was previously narrowed by candidacy requirements, yet the last selectorate - be it party delegates or party members - still has the last word (Norris and Lovenduski 1995).
On the differences between US party registration and party membership, see Katz and Kolodny (1994).
The Selectorate
37
Assorted method Intraparty Intraparty candidates A,, A2, A3...
Selectorate A
candidates B,, B2, B3..
J}
©
Multistage method
intra] mdida
All intraparty candidates
Weighted method
Ail intraparty candidates
Selectorate A
Party candidates
trapai
' Selectorate \
B
J
Remaining intraparty candidates
Party candidates
Selectorate B
Party candidates Figure 3.2. Complex candidate selection methods
The multistage method can also place a selected party agency after the party members. For example, candidate selection can be entrusted to the party agency after conferring with the party membership, or a party agency can exercise its veto over candidates chosen by the members. In the French Socialist Party, the executive committee at the level of the department chooses the candidates, but often after consulting the constituency party membership (Depauw 2003). The experience of British (Norris and Lovenduski 1995), Canadian (Cross 2002; Erick-son 1997), New Zealand (Vowels 2002), and Irish (Gallagher 2003) parties tells us that such veto is activated only on rare occasions. The party agency still has an impact, though, because selectorates are usually sensitive enough to refrain from selecting someone who may be vetoed by the executive agency.
Weighted candidate selection methods are those in which the result is determined by weighting together the votes of two or more selectorates for the same candidate or candidates. The British Labour Party used a weighted method for selecting its candidates for the 1992 elections (Criddle 1992) in the last stage of the selection process (it was also a multistage method). The result that determined which candidate would stand as the party candidate in the constituency was based on weighting the choices of the affiliated unions (up to 40 percent) and of the party members (60 percent or more). The Labour Party of New Zealand also used a weighted method to select its candidates in the single-member districts (Mulgan 2004; Sheppard 1998). It weighted the votes of delegates nominated by the national party agency (three delegates), delegates nominated by the constituency party agency (one to two
38
Democracy within Parties
The Selectorate
39
delegates), a delegate selected by party members, and the vote of the party members (as an additional, single delegate vote). In Taiwan, the Kuomintang and the Democratic Progressive Parry used various weighted methods over the years. These included weighting the votes of party cadres and party members and later, in a more inclusive manner, weighting the votes of party members and the results of opinion polls (Baum and Robinson 1999; Fell 2005). Similar to the assorted and the multi-stage methods, such weighted methods spread power among several forces within the party. These may result from a compromise between forces within the party, but may also be seen as an attempt to create the optimal blend in order to produce the best possible candidates (Rahat 2009).
Dealing with the complexities
In the case of complex candidate selection methods, classification becomes difficult. As the goal here is to offer a cross-party and cross-national analytical framework, one must integrate one-stage, uniform candidate selection methods and different complex cases, such as assorted, multistage, and weighted methods, into the same framework.
The difficulty with assorted candidate selection methods can be addressed in two steps. First, there should be a separate analysis of each distinct selectorate within the party. Second, the relative impact of each selectorate should be weighted by calculating the ratio of realistic candidacies that are filled by the particular selectorate, with the goal of "sununing up" the parry's candidate selection method. If one-half of the candidates are selected by party delegates and one-half by party members, then - when summing up for comparative needs - one can locate the selectorate between these zones.
The complexity of a multistage candidate selection method should be approached in a slightly different two-step method. First, there should be a separate analysis of each stage. Second, the relative importance of each stage should be estimated. If certain stages are found to be mere formalities, then they should be removed from consideration. When more than one stage has a real impact on the composition of the candidate list - in terms of the realistic candidacies - then these stages should be weighted to produce results that will enable us to locate the system along the continuum suggested.
Weighted candidate selection methods should be treated according to the relative importance that is allotted to each selectorate. When, for example, the votes of the party members and the party delegates are equally weighted, then the level of inclusiveness is between those two categories. When the vote of one selectorate carries more weight - for example, 70 percent for party members versus 30 percent for party delegates - then we are dealing with a case that is closer to party primaries in its level of inclusiveness. The following sections offer examples of the operationalization of such solutions, alongside the simpler methods, by delineating the continuum of party selectorates.
A JOURNEY ALONG THE rNCLUSIVENESS-EXCLUSiVENESS CONTINUUM
In this section we take a journey along the inclusiveness-exclusiveness continuum, starting from the most inclusive selectorate - all voters - and ending with the most exclusive one - the single leader. Our journey will pass through empirical examples from the democratic world that will refer to parties from established and new democracies. It will relate not only to the simple methods, but also to the complex ones: the assorted, multistage, and weighted selection methods.
Voters
The primaries of the fifty states in the United States provide us with most of the examples at the inclusive end of the continuum. The exact location of American primaries depends on the conditions set for participation in the primaries that are defined by the different state laws (Gerber and Morton 1998; Kanthak and Morton 2001; Merriam and Overacker 1928; Ranney 1981).3 Duverger (1954: 363) noted that in some US states there is no party affiliation beside the names of the candidates, and thus, "Really this is no longer a primary but the first ballot of an election." Indeed, at the extreme end are the American nonpartisan primaries, used in Louisiana from 1978 to 2006 to select candidates for Congress (Engstrom and Engstrom 2008; Maisel and Brewer 2007).4 These primaries, where every registered voter could vote for candidates from any party, are located at the inclusive end of the continuum (Ranney 1981).
Blanket primaries used in Washington (since 1938), Alaska (since 1968), and California (1998,2000)5 (Engstrom and Engstrom 2008) - are also at the extreme inclusiveness pole, as shown in Figure 3.3. Here voters receive a single ballot listing all the candidates from all the parties and decide, for each post separately, which party candidate to vote for. In both the nonpartisan and blanket primaries, participants do not need to declare their party affiliation in order to take part in candidate selection.
Open primaries, slightly less inclusive than the two previous kinds, are used in several states in the United States.6 As in nonpartisan and blanket primaries, the
3 Endorsements in preselection delegate conventions can also affect the location of the specific method, turning it into a more exclusive two-stage process. We ignore this element because it seems to have only a marginal effect on the selection process (Galderisi and Ezra 2001).
1 Nonpartisan primaries are much more common in the lower levels of government in the United States, especially the local level.
5 On the abolition of blanket primaries in California, as a result of a Supreme Court decision, see Persily(2001).
6 The Democratic and Republican parties of the following states used open primaries (in parentheses are the time periods for which we have reliable data): Michigan and Montana (1960-90), Minnesota,
40                               Democracy within Parties ]A
voters are allowed to decide in which primaries they want to take part without the :: ■ fj
need to announce their partisan preference. Yet, unlike these types, voters can take | part hi the primaries of only one party.
Semi-closed primaries are used in the United States by the Republican and ; -if
Democratic parties of several states.7 These primaries require participants to :!
declare their party affiliation only on the selection day, and/or allow independents.....;.|
to take part in candidate selection of the party they announce their wish to vote in.8 : Si
Here we are slightly away from the inclusive end of the selectorate continuum :t|j because voters need publicly to affiliate with a political party. We also have
examples from Iceland, Taiwan, Mexico, and Spain.9 According to Kristjansson JJ (2002), from 1971 on, parties in Iceland (Social Democrats, Progressives, and
Independence Party) adopted primaries in some, and sometimes all, electoral \
districts, where every citizen in a particular electoral district could participate. In .
Taiwan, the National Party adopted primaries in which all voters could participate ,:%
in 1998 (Fell 2005), and so did the Mexican Party of the Democratic Revolution in ::;] 2003 (Wuhs 2006). The Catalan Socialist Party in Spain opened its candidate
selection to "registered 'sympathizers'" - non-members who could register as vd
party supporters without paying any membership fee (Hopkin 2001). J:j
American closed primaries, which demand that voters register according to their.....: a
party affiliation before the day of the primaries, are located somewhat further away ;:
from the inclusive end. Closed primaries, which are used in several states of the : ?j United States,10 are in the middle - between the category of voters and that of . "izi
Utah, and Wisconsin (all three 1960-96), Alaska (1960-6), North Dakota (1968-86), Vermont (1972- .£]
96), Hawaii (at least since the 1960s), and Idaho (since 1976) (Kolodny and Katz 1992; Goodliffe and :Z\
Maelehv 2000: State ofHawaii; Idahe-Seer^rv nf-State)™................................................................................
In the research literature, this type is sometimes labeled "semi-open." We prefer the semi-closed : "1
label because the main difference between dosed and open primaries is that in the former the voter ■.;■)
needs publicly to affiliate with a party. In semi-closed primaries, the voters still need to publicly ifl
announce in which party primaries they will participate, even if at the last moment ::':->
8 The Democratic and Republican parties of the following states used semi-closed primaries (in parentheses are the time periods for which we have reliable data); Alabama, Arkansas, Georgia, Illinois, Indiana, Mississippi, Missouri, South Carolina, Tennessee, Texas, and Virginia (all in the 1980s and 1990s), Arizona (since 2000), California (since 2002) (Goodliffe and Magleby 2000; Arizona Constitution; California Secretary of State).
9 It should be noted that what are called "open" primaries outside the United States is different from : :j what the Americans call open primaries. Outside the United States, open primaries are a candidate -51 selection method in which nonmembers can participate. In the US open primaries, voters can participate ■■<; '< without publicly exposing their party affiliation. We thus place the non-American open kind together
with the semi-closed American primaries. ':!
10 The Democratic and Republican parties of the following states used closed primaries (in :}
parentheses are the time periods on which we have reliable data): Florida (1960-2008), Arizona :; (1960-98), California, Connecticut,  Delaware, Kentucky,  Maine, Maryland, Massachusetts, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, Ohio,
Oklahoma, Oregon, Pennsylvania, South Dakota, West Virginia, and Wyoming (all from 1960 to 1996) :
(Kolodny and Katz 1992; Goodliffe and Magleby 2000; Florida Department of State). Some of these S|
states allow the parries to conduct semi-closed primaries if they wish; that is, to allow unaffiliated voters . [
to take part in their primaries. Yet, the default alternative is still a closed primary method. it
The Selectorate 41
party members. The level of inclusiveness of the selectorate of the Democratic Progressive Party in Taiwan (1998-2001) also places it at this middle point. It used a unique weighted selectorate that combined the results of both a party members' vote and a public opinion poll (Baum and Robinson 1999; Fell 2005).
The open convention, which allows any voter to take part in a candidate selection meeting, is probably the less inclusive version of the most inclusive family of selectorates. While it allows any voter to take part in a selection meeting without any need to prove party affiliation or even to preregister, it is still quite a demanding system, as it requires the voter to attend a meeting at a certain date, time, and location. This kind of selection - which is similar to some of the American caucus systems (Marshall 1978) - was used in the past in Canada, especially in the 1920s-1950s period, before the parties started to institutionalize their membership (Engelmann and Schwartz 1975; O'Brien 1993).11
Party members
We now move into the party members' zone, but at its inclusive end, closer to the general electorate. Here we have the case of the Dutch Democrats 66 in the 2004 European elections, which held meetings where both party members and all voters could participate, but then allowed only party members to make the final decision through voting via postal ballots (Depauw and Van Hecke 2005; Hazan and Voennan 2006). We also have the case of the Taiwanese Kuomintang (2001^) that weighted the voting of party members with public opinion polls (Fell 2005). Another example is the Argentine Peronist (Justicialista) Party and the Radical Civic Union, who in some districts (1983-2001) allowed party members and Independents to participate in their primaries (De i.uca. Jones, and Tula 2002; Jones 2008).
In the middle of the party members' zone we find the typical European closed primary (Newman and Cranshaw 1973), which - as opposed to American closed primaries - usually means "party primaries" (Gallagher 1988c: 239-40) in which the selectors are party members, not merely registered adherents. From this point on in the selectorate continuum we exclude the party supporters. Over the years, usually in an incremental fashion, more and more Western democracies allotted their members a significant role in candidate selection (Scarrow, Webb, and Farrell 2000; Bille 2001; Kittilson and Scarrow 2003). The "purest" type of party primary is where the party members' votes alone decide the composition and rank of the candidates. Several parties, across a wide spectrum of time and space, have used party primaries, albeit not consistently over time and not necessarily in all districts. Among them are the following examples: the Australian Labor Party
11 Scarrow (1964) described the dynamics of such a selection process taking place in 1962 in the Liberal Party in Ontario.
42
Democracy within Parties
The Selectorate
43
(Epstein 19776); the Belgian ECOLO (Deschouwer 1994); the German Social Democrats, Christian Democrats, and Greens at the single-member district level (Borchert and Golsch 2003; Schuttermeyer and Strum 2005); the major Icelandic parties - Independence Party, Social Democratic Party, Progressive Party, People's Alliance (Hardarson 1995; Kristjansson 1998, 2002); the major Israeli parties -Labor, Likud, and Kadima (Hazan 1997a, 19976; Rahat and Sher Hadar 1999a, 19996; Rahat 2008«; Rahat Forthcoming); major Mexican parties - the Institutional Revolutionary Party and the Party of the Democratic Revolution (Baldez 2007; Langston 2006, 2008; Wuhs 2006).
When party members have a dominant or significant role in candidate selection -but they are not the sole selectors and other, more exclusive party actors take part in the selection of candidates - we are still in the party members' zone but are moving toward the party delegates' area. The cases here involve multistage methods. For example, in several of the Danish parties, from the 1970s until the twenty-first century, central party agencies could veto or change the selection made by the party members (Bille 1994, 2001; Pedersen 2002). In the Kuomin-tang in Taiwan in 1988-9, the party's executive could ignore the members' vote, yet chose to respect their verdicts in 90 percent of cases (Baum and Robinson 1999). In Finland, the election law states that party organizations have the right to change up to one-fourth of the candidates selected by the members' vote.12 In Canada, national party leaders have veto power over the party members' selection, although they usually refrain from exercising it (Cross 2002, 2004),13 while in Ireland the national party leadership kept and even enhanced its veto power in those parties that adopted a membership vote (Galligan 2003; Weeks 2007).
The right that party agencies (typically-national executives and/or party leaders) possess is rarely used because it can cause conflicts within the party, with allegations that the oligarchy does not respect the more popular democratic will. This is why party agencies are more influential vis-a-vis the members when the order is reversed - when the party delegates (or even the party elite) screen the potential candidates, and the party members make the final decision.
12 Section 117 of the Election Act of Finland states:
On the recommendation of the party board, the result of the vote by the members can differ no more than one fourth from the number of candidates nominated by the party (right of change). Even then at least half of the candidates of the party must be persons who have received most votes in the vote by the members. (Ministry of Justice, Finland [Italics in original])
13 In 1970, with the law that required the printing of party labels beside candidate names in federal ballots, party leaders were given the power to veto candidacies. In 1992, the Liberal Party granted its leader the right to appoint candidates, thus enhancing his powers as a selector, though most candidates at most times were still selected by party members at the constituency level. (Carry and Eagles 20031.
When party agencies can filter the candidates, who are then put to a membership vote, we are getting close to the middle between the party members and the selected party delegates zones. If the screening process still leaves a large and viable pool of candidates from whom party members can make the final decision, then we are still on the party members' side. In Israel, the Meretz Party in 1996 produced a sizeable "panel" of candidates from which the members chose the final list (Hazan 1997a; Rahat and Sher Hadar 1999a). The Social Democrats and the Liberals in Britain did much the same (Criddle 1984, 1988; Norris and Lovenduski 1995; Rush 1988), as did several of the parties in Belgium in some of the districts, especially the Belgian Socialist Party during the 1960s (Obler 1970,1974).
When equal weight is given to the party delegates and to the party members, then we are in the middle between these two categories. In Taiwan in 1995-6, the Democratic Progressive Party used a weighted method in which the vote of the members was equally weighted to the vote of the party representatives (Baum and Robinson 1999). British Labour's use of a multistage method since 1997 -where candidates were screened by party agencies, selected by party members, and could still be vetoed by the National Executive Committee - can also be seen as a middle-of-the-road example (Quinn 2004). The same seems to be the case for the Democratic Party of Botswana in 2002, where party members selected candidates after a national party agency screening (Ohman 2004), and for both Dutch Labor (1960-4) and the Pacifist Socialists (1957-73) that allowed their national executives to propose the list of candidates, yet let the members then vote and alter both the rank and the composition of these lists (Koole and van de Velde 1992).
The "party members" selectorate can be further distinguished according to the restrictions on party membership; the additional requirements that are placed on members with a conditional right to take part in the party selectorate, and the level of accessibility of the selector to the selection procedure. For example, one rule that could restrict membership, or just the right to participate in candidate selection, is the rate of membership dues. Members' participation may also be restricted by the requirement of a ininimal party membership period prior to candidate selection, proof of party activity, etc. The Mexican Party of National Action is a good example of a party that placed banters on membership participation. A member could take part in candidate selection only after six months of membership, and in order to keep this right had to pay party dues and take part in party gatherings (Langston 2008).
Accessibility may also be an important factor in distinguisliing between such selectorates. Levels of accessibility and inclusiveness are higher if a party adopts postal ballots14 or e-voting.'5 Spreading polling stations all over the country is less
14 Postal ballots were used by the Dutch D66, the British Labour Party, and by several Danish parties.
15 In 2000 in Arizona, the Democratic Party conducted its primaries online (Alvarez and Nagler 2001).
44 Democracy within Parties
accessible.16 A much less accessible yet inclusive method is an all-member party convention, as is usually the case with candidate selection in the Canadian parties and some of the Irish parties over the last decade. While all members can attend such a meeting, it requires more effort on their part.
Selected party delegates
When party members have less of an impact than selected party delegates, the selectorate is still located between these two zones, but is closer to the latter. Here, for example, the members can ratify or reject a list of candidates drawn up by the party agency, as was done by the French Socialists in 1986 (Thiebault 1988). Also in this category is a multistage process where the members are only one stage in a process that is more than two stages. Since the 1980s, the British Conservatives used a multistage method, which started with a screening by a nonselected national party agency, followed by a local selected party agency screening, and ended with a party members selection meeting (Norris and Lovenduski 1995).
We move slightly further toward exclusivity when the candidates are produced by a wide delegate convention - when the ratio of members to delegates is in the low range of one delegate for each three to four members - as was the case in Ireland with the Fianna Fail, Fine Gael (until it adopted membership ballots in the 1990s), and Labour parties (Gallagher 1980,1988ft). Another example is selection by a party agency that might be followed (or preceded) by a membership ballot, as was the case in the Swedish Communist/Left Party (Pierre and Widfeldt 1992).
When the selectorate is an agency of the party, we are in the middle of the continuum. Inside the party the relative size of each agency is a sign of its inclusiveness: conventions arc usually larger than central committees, which in turn are usually larger than executive bodies, such as bureaus. As the size of the particular party agency gets smaller, we move closer to the exclusive pole of the continuum. The terminology used in each country is not necessarily equivalent, and hence one must be cautious when inferring the extent of inclusiveness based solely on what a particular party calls a specific agency. In addition, the more inclusive party agencies contain delegates selected by party members, while the more exclusive ones include representatives who were selected by such delegates. The use of party delegates is widespread. Since the 1950s, the major German parties used delegate conventions at the single-member district level, preferring them in most cases over the alternatively more inclusive selectorate prescribed by the party law - selection by the members themselves (Borchert and Golsch 2003; Roberts 1988). This was also the typical selectorate in Australia (Epstein 1977ft; Norris et al. 1990). Several Israeli parties (National Religious Party 1996-2006,
16 Primaries conducted in the United States involve the spreading of polling stations across the particular state.
The Selectorate 45
Shinui 2003, Herat 1977-88, Likud 1992,1999-2006) used their central committees to select candidates (Barnea and Rahat 2007; Rahat 2008a). From the 1920s to the 1950s, this was also one of the methods that was used by the Canadian parties (O'Brien 1993). Even in those countries where more inclusive selectorates are now the norm, or already were before, we still find the use of delegates to select candidates from time to time in certain parties, or in certain constituency organizations. Such cases include the occasional use of delegate conventions by both Belgian and Argentine parties (De Luca, Jones, and Tula 2002; De Winter and Brans 2003; Jones 2008; Obler 1970).
When a nonselected party agency, such as a nomination committee, has an influence on the selection of candidates, alongside the selected party agency, we move toward less inclusiveness (or more exclusiveness), but remain within the selected agency zone. One example is the assorted system used by both the Parry of National Liberation and United Social Christian Party in Costa Rica, where most candidates were chosen by a selected party agency, but several were nominated by the party's president (Taylor-Robinson 2001). There are many examples of multistage methods when both directly and indirectly selected party agencies take part in candidate selection, as in the Austrian Socialist Party 1945-90 (MuTler 1992); the Dutch Christian Democrats 1986, Christian Historical Union 1960-79, and Radical Political Party 1973-89 (Koole and van de Velde 1992); the British Conservatives from the 1950s to the 1970s and Labour from the 1950s until 1987 (Denver 1988; Lovenduski and Norris 1994; Ranney 1965; Rush 1969). Yet another example is a weighted method where both nominated and selected delegates choose the candidates, such as in the New Zealand Labour Party since the 1950s (Catt 1997; Milne 1966; Mulgan 2004; Vowels 2002).
We are in the middle, between, the. selected party delegates, and. the party elite, when there is a relative balance of power between the selected and the nonselected party agencies. This was the case with the multistage methods used in the 1980s in the Union for French Democracy (Thiebault 1988), the Dutch People's Party for Freedom and Democracy from the 1960s to the 1990s (Koole and van de Velde 1992), and the Spanish Socialists from 1979 to 1998 (Field 2006). The Norwegian parties, in most cases, from the 1920s and until 2002, followed such a system, according to a law that allocated funding to political parties that selected their candidates in a (selected) delegate convention at the level of the multimember constituency. In practice, a nomination committee submitted a list of candidates, and the selected delegates either ratified or changed the list, position by position (Valen 1988; Valen, Narud, and Skare 2002). Candidate selection at the Land level in Germany also belongs here. Although party delegates are the final decision makers at this level, the selection is made on the basis of a recommended list that is designed by the Land party elite (Borchert and Golsch 2003; Porter 1995).
46
Democracy within Parties
The Selectorate
47
Party elite
This category includes nomination committees that are formed for the sole purpose of selecting the party's candidates, as well as nonselected party agencies (typically small executive boards) that are entrusted with several different tasks including the selection of candidates.17 Nomination committees are usually composed of a few party leaders, their representatives, or aficionados. Their composition, as well as their decisions, is many times ratified en bloc by more inclusive party agencies. The composition of both nomination committees and nonselected party agencies can be regarded as slightly more inclusive if it is indirectly selected, or somewhat more exclusive if it is not.
When the selection power of the party elite is stronger than that of the selected party agency, we move into the party elite zone and are squarely within the exclusive part of the continuum. Here we have cases of multistage methods that involve various selected and nonselected (or highly indirectly selected) party agencies, with more influence to the latter. For example, the French Rally for the Republic in the 1980s allowed a nominating committee to select its candidates, but afforded some influence to other directly and indirectly selected party agencies (Thiebault 1988). The Italian Communists from 1956 to 1986 allowed for the involvement of a selected party agency, but the final word over candidate selection was given to an indirectly selected party agency (Bardi and Morlino 1992). The Chilean Party for Democracy and National Renovation allowed their national board and national council, respectively, to be involved, but the final decision was made - because of the constraints set by the binominal electoral system - by negotiations between the party leaders (Navia 2008). In the Japanese Liberal Democratic Party during the 1950s-1990s period, the last stage of candidate selection was also in the hands of an exclusive national committee that "consisted'" of 1fflem"sehjor leaders (Fukui 1997). When, for example, the selected party agency is only asked to ratify a decision made by the nomination committee, we are still leaning slightly toward the inclusive side of the party elite zone. This was the case in the Taiwanese Kuomintang in 1995 (Baum and Robinson 1999), and in the Israeli Mapai from 1949 to 1955 (Brichta 1977). In the Greek Pan-Hellenic Socialist Movement, local and relatively inclusive agencies were consulted in the early stages of the selection process, but the final list of candidates was determined by a committee that was appointed by the party leader. In 2006, the process was democratized when it was decided that the selection committee would be chosen by the members of the party's National Council. This is clearly a more inclusive selectorate, which moves the party to the "selected delegates zone" yet leaves it in an exclusive position within that zone because the delegates are not selected directly by party members (Ashiagbor 2008).
1' We inctude in this category small party agencies (not more than a few dozen members) that are composed of people who were selected indirectly or nominated by other (possibly selected or indirectly selected) party agencies.
When only a nonselected party agency/group is involved in candidate selection, then we are squarely in the middle of the party elite area. This is the kind of selectorate that seems to reflect Michels' (1915) notion of party politics. That was the case in the Venezuelan Democratic Action Party in the 1990s (Coppedge 1994); the Italian Christian Democrats from 1957 to 1984 (Bardi and Morlino 1992); the Mexican Institutional Revolutionary Party before 2000 (Langston 2001); the Chilean Independent Democratic Union from 1989 to 2001 (Navia 2008); in a significant number of constituencies in several of the main Belgian parties, especially since 1968 (De Winter 1988; Deschouwer 1994); the Indian Congress Parry in the 1950s and 1960s (Graham 1986; Kochanek 1967); the Danish Peoples Party in 1998 (Pedersen 2002); and in the Argentine Peronist and Radical parties in some districts in Argentina (De Luca, Jones, and Tula 2002; Jones 2008).
We move toward the single-leader pole with such exclusive selectorates as a gathering of the party founders in a new party, or an informal group of factional leaders in older parties. Israel's ultra-religious parties serve as an example of such highly exclusive selectorates. In one party, Shas, a Council of Sages - a small body of Rabbis headed by a highly influential spiritual leader - formed the candidate list (Rahat and Hazan 2001).
A single leader
The extreme end of the exclusive pole is defined by a selectorate of a single individual. If the leader does not have complete control over candidate selection, then we are close to the exclusive end of the selectorate continuum, but not at its pole, fn Forza Italia, in the 1990s, founding leader Silvio Berlusconi chose the candidates in cooperation with the parry's regional coordinators (Hopkin and Paolucci 1999). Jean-Marie Le Pen, the French National Front leader, together with its general secretary, chose the party's candidates, with some influence given to nonselected forces (Thiebault 1988). Winston Peters, New Zealand First's founder, was also given almost complete control over candidate selection (Cart 1997; Miller 1999; Mulgan 2004).
Similar to the party elite, a single leader will lean more toward the inclusive side if me leader is selected, and more to the exclusive pole if it is a nonselected leader. Once again, some of the Israeli parties serve as examples of such an extremely exclusive selectorate. In 1988-96, in one ultra-religious party in Israel, Degel HaTorah, a single rabbi was authorized to decide the composition and order of the party list (Rahat and Sher-Hadar 1999a). In 2005, Prime Minister Ariel Sharon quit his party (Likud), formed a new one (Kadima), and also called for new elections. He alone was to choose its list of candidates (Hazan 2007).18
18 After suffering a stroke in the middie of the election campaign, the leadership of the new party was taken over by his deputy, Ehud Olmert, who singlehandedly completed the list of candidates.
48
Democracy within Parties
The Selectorate
CLASSIFYING PARTY SELECTORATES ALONG THE SELECTORATE CONTINUUM
Figure 3.3 illustrates the levels of inclusiveness of the selectorates using a twenty-five-point scale (0-24). It is based on the continuum in Figure 3.1. Indeed, every six points along the scale relate to one selectorate, or category, along the continuum that appears in Figure 3.1: single leader (0), parry elite (6), party delegates (12), party members (18), or the voters (24). Its additional property - designed to deal systematically with the challenge of complexity - is a distance of six points between each category. This distance is needed in order to allow the categorization of those cases where we deal with more than a single selectorate (the assorted, multistage, and weighted candidate selection methods). To demonstrate the usefulness of this continuum we located examples of cases that were mentioned in the previous section at most of its points.
The six-point distance allows us to illustrate four possible scenarios:
1. A single selectorate is responsible for candidate selection, in which case the classification falls clearly into one distinct category (0, 6, 12, 18, or 24).
2. Two selectorates (located near each other) have equal weight in the selection. Each selectorate may choose only one-half of the candidates, or their influence is equal in a weighted or multistage method. In these cases classification falls in the middle, between the two selectorate categories (3, 9, 15, or 21).
3. Two selectorates (located near each other) are involved in candidate selection, both have a significant role, yet one is either somewhat more important than the other or clearly dominant compared to the other. For example:
(a.) If in an assorted systemrone selectorate seleets~two=thirds of the candidates and the other selectorate the remaining third; if in a multistage method, we estimate the influence of one selectorate to be higher than the other; or if in a weighted system, the weight of one selectorate is around two-thirds, while the other is about one-third. In these cases we are no longer in the middle, between two categories, but leaning closer to one category (2,4,8. 10, 14, 16, 20, or 22).
(b) If in an assorted system, one selectorate selects 80 percent of the candidates and the other the remaining 20 percent; if in a multistage method, we estimate the influence of one selectorate to be much higher than the other; or if in a weighted system, the weight of one selectorate is around 80 percent, while tire other is about 20 percent. In these cases we are much closer to one category than the other, but not clearly in that category alone (1, 5, 7, 11, 13, 17, 19, or 23).
After considering several options - using a continuum with less or with more resolution - we concluded that the twenty-five-point continuum is the most appropriate. Beyond the cross-national comparison exhibited in Figure 3.3, it
5 2
1 e: .s °
w. rj u * w
O 85
.2 '8 1 as *j ■
i a
g a '
£3.
O .-I
£ 2
■3   g 01
5 <- c
2 a S P.g St
< g.-s
£> BJ £
« -5
i—i rM
CI
a 1
_ o
o v>
Z =6
s    -S <=
™ o\ 1-1 0\
[
o M —
<u-N
.2 S
P P5
R .5 Q g
| §
^ bfl
§ °
o
3 -c
8 3
o Q S
P3 £>
T-h GO
O 5
& is
fe a ' 2.2 L
^ P ...
» s ■ ^ o
«.2 E c *a .3
S ^ g < a g
50
Democracy within Parties
i
The Selectorate
Voters Parry Party Party Single
members delegates elite leader
24 23 22 21 20 19 18 17 16 15  14 13 12  11   10   9   8  7 6   5  4 3  2   1 0
Inclusive „
Labor
NRP
Meretz
Likud Hadash
AI
Gil
Shas Kadima loH
t Exclusive
NRP = National Religious Party AI - Agudat Israel loH - Israel Our Home
Figure 3.4. Party selectorates in Israel 2006 - high variance within the same party system
affords both synchronic and diachronic comparisons. First, we can get a clearer perspective on the difference in candidate selection methods across parties in one country at a specific time, as shown in Figure 3.4, which exhibits the variance in the inclusiveness of party selectorates prior to the 2006 general election in Israel.
Second, such a continuum also allows an examination over time of shifts -both large and small - as exemplified in Figure 3.5. Israel's Labor Party and its predecessor, Mapai, show a linear pattern of democratization, one that starts with small steps taking place over three decades (moving from 7 to 10), and ends with a great leap (from 10 to 18) in 1992, Likud and its precursor, Herut, also democratized, but at a different pace of greater leaps up to 1996, then returned to a less inclusive selectorate from 1999 to 2006, and finally redemocratized in 2008.
Using a scale with lower resolution, for example a thirteen-point scale, would place two examples next to each other, leading one to conclude that the differences are minor, when their differences are actually critical. For example, a multistage method in which the party delegates select a short list while the party members make the final selection (located at 8 on a thirteen-point scale, such as the Meretz Party in Israel in 1996), as opposed to a multistage method in which the order is reversed - the party members select the shortlist while the party delegates make the final selection (located at 7 on the thirteen-point scale, such as the Tsomet Party in Israel in 1996), are too distinct to be placed right next to each other. Clearly, ceteris paribus, the selectorate with the final say is the more important one (if the list of candidates is long enough, e.g., it exceeds the number of realistic positions) because it can move candidates into, or away from, the realistic positions. The Meretz Party's selected agency (a central committee) produced a shortlist of thirty candidates, at a time when the party had only twelve representatives in the parliament - the list thus contained more than twice the number of realistic positions. The party members then ranked the candidates, which gave them the more important role. Tsomet, on the other hand, allowed its members to pick twenty-three candidates out of thirty-eight - the remainder was comprised of the four sitting representatives and nine picked by a small committee. Ostensibly, the membership played an important role, picking almost two-thirds of the candidates.
vc
S 3
o >
-a
tu <ü
S 5 j2 o
3
52 Democracy within Parties
However, when the selected party agency ranked the candidates, the four sitting representatives came first, followed by two nominated by the small committee, and only then did it rank the first candidate chosen by the members. However, since the parry had only five seats in the outgoing parliament, this was not only a completely unrealistic position, but all the realistic positions were given to the incumbent representatives. Placing these two at 8 and 7 on a thirteen-point scale would locate them too close to each other, while the gap between them on a twenty-five-point scale (16 and 13, respectively) is both more conceptually appropriate and empirically necessary.
The twenty-five-point scale is also sensitive enough to identify small yet significant changes. For example, in 1949-55, Israel's Mapai used a highly exclusive nomination committee composed of a few party leaders to decide both the composition and the ranking of its candidate list, thus supposedly placing it at 6 on tire scale. The only reason that its final location is slightly different, position 7, is the fact that the composition of the nomination committee, and also its decisions, were usually ratified by wider party agencies. Subsequently, prior to the 1959 elections, Mapai decided to select its candidates through a two-stage process that involved additional selectorates. In the first stage, twenty-five candidates were named by a nominating committee and twenty-five by the party's eleven regional councils. In the second stage, the nominating committee ranked these candidates in positions 1-50 on the candidate list. While the nominating committee remained the dominant selectorate, the regional councils effectively influenced the composition of the list, because Mapai always had more than twenty-five realistic positions (Bamea and Rahat 2007). This is reflected by placing it in position 8, which clearly signifies the dominance of the party elite yet alludes to the.. secondary -but., significant-role, of the- regional selected party agencies.
A continuum with a higher resolution would be too demanding, and one with lower can always be produced by simply collapsing the points into fewer categories. We, therefore, believe that the twenty-five-point scale is optimal, but not necessarily ideal. There are still problematic cases that require making rough estimations because the simple rales given above do not apply, for example in cases where three or more selectorates are effectively involved in the selection process, or even those cases in which two selectorates are involved yet they are not close to each other in terms of inclusiveness (or party elites and party members). We may end up with an estimation of inclusiveness at 12 on the selectorate continuum, not because party delegates chose the candidates but rather because the weight of both the party members and the party elite in the selection was equal. Nevertheless, this price is unavoidable if we need to operationalize inclusiveness in order to conduct a large N analysis, and want to avoid creating multidimensional models that produce almost as many categories as empirical cases.
The Selectorate 53 DEMOCRATIZING THE SELECTORATE
As we conclude our discussion of the selectorate, there are two issues that need to 1        be addressed. The first looks back at candidacy, which was discussed in the
previous chapter, while the second looks forward at two dimensions that we
discuss in the next chapters - decentralization and voting.
Table 3.1 integrates the two dimensions of candidacy and the selectorate,
presenting each party's candidate selection method according to its level of '        inclusiveness or exclusiveness. In the American case, when candidacy and the
selectorate are inclusive, the party hardly has a say when it comes to candidacies
under its label. Scarrow (2005: 9) describes one of the more well-known cases:
A notorious instance of what happens in a party without such a safeguard occurred in the U.S. state of Louisiana in 1991, when voters in a Republican primary nominated David Duke as the party's gubernatorial candidate. Party
i leaders could personally repudiate the candidate, an outspoken white suprem-
acist and former Ku Klux Klan member, but they had no way of denying him
: the use of the party label.
i We find no example for the opposite case, where both candidacy and the selectorate are exclusive, and while this is no evidence that there is indeed no such case, its rarity is quite logical: Why would a party elite limit its pool of candidates if it is
I        in full control by being the exclusive selectorate?
A high level of inclusiveness on one dimension combined with a high level of
' exclusiveness in the other would mean that the party leadership, or the party apparatus, retains control over the process. For example, the Italian Communist
; Party included nonmembers as candidates, but this was done under the supervision of an exclusive selectorate (Wertman 1988). In a similar way, the Israeli ultra-religious parties have no formal rules regarding candidacy, yet the highly exclu-
i sive selectorates ensure that all selected candidates will be ultra-religious men. This seems to be a pattern also in the leader-dominated extreme right and populist right European parties, where anyone can be a candidate - their selection is based
1        on the leader's predisposition toward them. An opposite case, with the same logic
! i
■ i Table 3.1. Inclusiveness in candidacy and in the selectorate
I -
! Inclusiveness in the selectorate
i___
I Low High
Inclusiveness in Low    — Belgian Socialist Party (1960s)
candidacy
| High    Italian Communist US Republican and Democratic parties
Party (1976) (since the 1920s)
54
Democracy within Parties
of retaining control over those selected, is the Belgian Socialist Party in the 1960s, j which frequently used the most inclusive selectorate among the Belgian parties to :::hM select its candidates but ensured their cohesion through very exclusive candidacy (Obler 1970). These days, however, the significance of candidacy requirements is low as politics becomes less partisan and ideological, and more personalized and -; electoral. When candidacy is highly inclusive, dejure and especially defacto, the significance of the level of inclusiveness of the selectorate increases.
Democratization of the candidate selection process is expressed as a widening \ of participation in the process; that is, when the selectorate following a refonn of the candidate selection method is more inclusive than previously. Adopting only more inclusive candidacy requirements (the first dimension in the analytical framework), implementing decentralization (the third dimension), or shifting from appointments to a voting system (the fourth dimension) may be labeled democratization, but they are not. They are facilitating variables only and neither i define nor exhibit democratization. More inclusive candidacy requirements may i be adopted, yet the same limited selectorate could still have control over the final results, thereby curtailing the impact of democratization. Decentralization could mean only that control of candidate selection has passed from the national to a local oligarchy. Indeed, if the selectorate is decentralized from a national party congress of several thousand participants to a handful of local executive committees each consisting of a dozen party notables, the overall selectorate may have ~-\ actually become more exclusive. Voting procedures may replace appointments, t but the vote itself could be restricted to a very exclusive body. In other words, it is the inclusiveness of the selectorate that is the necessary variable for democratizing candidate selection methods. Sartori (1973: 19-20) appropriately equated democratization with the "massification" -of polities,- because the hitherto excluded masses are now allowed to enter. The consequences of such intraparty democrati-zation can be curtailed if the party can still exert power through exclusive candidacy requirements. On the other hand, if both the selectorate and candidacy become more inclusive, then the party will experience more of tire political ramifications associated with democracy within parties.
The nature of the selectorate, along the inclusiveness-exclusiveness continuum, ; is important because it allows us not only to classify candidate selection methods, ':M to assess the political consequences of each selectorate, and to analyze differences :-|| along the continuum and their ramifications, but also to uncover trends over time ^fl in one country and across countries. Such trends are fairly evident, and their : consequences are the focus of the second part of this book. However, before i moving to the political consequences of democracy within parties, there are two :':';-\ more dimensions which need to be discussed - decentralization and voting versus appointments. ^jj
Decentralization
Several of the most prominent scholars who pioneered the research on candidate selection chose to devote substantial consideration to the degree of centralization of the candidate selection method. Ranney (1981) proposed centralization as the first of three dimensions for measuring the variation among candidate selection methods; only afterward does he mention inclusiveness and finally direct or indirect participation. Gallagher (1988a) chose three main aspects in order to describe candidate selection: the first was centralization, followed by participation (i.e. what we label inclusiveness), and then the qualities required of aspirants. Marsh (2000) picked centralization as the first dimension of two, the second being participation, as do Narud, Pedersen, and Valen (2002a, 20026, 2002c). Centralization, these scholars argue, is the extent to which the national level influences candidate selection, as opposed to the weight of the regional and/or the local levels. This is, we will argue, only one aspect of centralization - centralization need not be only territorial, it can also refer to nonterritorial aspects such as gender and minorities.
Marsh (2000) was correct when he stated that his two chosen dimensions -centralization and participation are not entirely independent of each other, indeed, more centralized selection methods are usually more exclusive, and vice versa. The national party's involvement is typically that of relatively small executive agencies, while regional and local involvement is typically that of selected delegates or even party members. However, as Gallagher (1988a: 4-5) stated, "Even knowing that selection is made by constituency agencies, of course, still leaves open the question of how widely party members and voters are involved." We argue that most scholars do not make this distinction as clear as it needs to be, even those who recognize that inclusiveness of the selectorate and centralization can vary widely. Generally the more decentralized candidate selection is, the greater the possibility for individual party members to play a role - but this is an inclination and not a rule. Decentralization, therefore, deserves to be addressed on its own.
Candidate selection methods may be seen as decentralized in two senses, territorial and social, parallel to the concepts Lijphart (1999) proposed when he dealt with the division of power in federal and unitary democratic regimes. It is territorial when local or regional party selectorates nominate party candidates, as opposed to national party selectorates. However, regardless of the extent of
56
Democracy within Parties
1
decentralization, each territorial category can vary in its level of inclusiveness. For example, a decentralized selection method would be one in which the local level has the power to decide, but this power can be held by a local leader, a party branch committee, all party members, or even all the voters in an electoral district
Decentralization based on territorial mechanisms, in order to ensure regional and/or local representation, is rather straightforward. In many European cases, the selectorate at the district level plays the crucial role in candidate selection. The Norwegian case, where the national party agencies cannot veto candidates who are determined at the district level,1 and territorial representation is taken into account inside each district, is an example of territorial decentralization (Valen 1988; Valen, Narud, and Skare 2002).
Decentralization of the selection method can also be social or corporate; for example, it ensures representation for representatives of groups that are not defined territorially, such as trade unions, women, minorities, or even subgroups within these groups. Nonterritorial decentralization can be found in many parties that are closely connected to interest groups. In many socialist parties there is significant representation for union representatives while in right-wing parties there is, many times, representation for business and farmers' associations. Probably one of the most prominent examples is the role of the trade unions in the British Labour Party. In the 1950s-1960s, the trade unions controlled approximately one-fifth of the candidacies and one-third of its legislators (Ranney 1965; Rush 1969). In Belgium, legislators not only kept close ties with interest groups, but maintained their interest group positions after being elected because it was via these organized interests that they were selected. De Winter (1997) states that interest groups in Belgium had an important role in the selection or removal of candidates, and at times even a monopoly over candidate selection, ttese-decentralizations are not-likely to be found in the parties' regulations or their constitutions, but are rather phenomena that express the power of these groups within the parties.
When candidates are selected exclusively by a national party selectorate, with no procedure that allows for territorial and/or social representation - be it a nonselected leader, a national party agency, the entire party membership, or even the national electorate that selects all candidates from the whole nation then we have a method that is located at the centralized pole (Figure 4.1). At the decentralized poles, candidates are selected exclusively by local party selectorates and/or intraparty social groups.
Once again, we have to determine and weigh the impact of different selectorates at different levels in the case of a mixed selection method. A case in point is that of
1 The Norwegian Nomination Act of 1920, which was revoked in 2002, stated that candidates were to be selected at the level of the constituency, and that die center did not have the power to atter this decision (Narud 2003).
Subsectors/ social subgroups
S
0 c
1
A L
Large sectors/ social groups
National
Decentralization D
57
D
Centralized
National Regional Local
TERRITORIAL
Figure 4.1. Centralization and decentralization of candidate selection
the Italian parties in the 1980s, in which central} .provincial, and local selectorates took part in a multistage candidate selection process. According to Wertman (1988), the provincial level party agencies played the main role in candidate selection vis-a-vis the center and the district level selectorates. Thus, the Italian parties of the 1980s are in the middle area of the territorial continuum. Still, there were differences between the parties, which placed the Italian Communists, for example, closer to the centralized pole than their Socialist counterparts. Similar calculations might have to be made for weighted and assorted candidate selection methods.
Territorial centralization is not a direct consequence of the national electoral system, although the latter does influence the former. Federal systems do have an inclination toward decentralized candidate selection methods, but there are many exceptions - candidate selection in Austria, a federation, is more centralized than that of a unitary state like the United Kingdom. Both India and the United Kingdom use single-member district plurality electoral systems, but candidate selection in India - at least in the 1950s and 1960s - was highly centralized, and in the UK it was decentralized. While both Israel and the Netherlands use a single, national constituency in their general elections, local branches and regional
58
Democracy within Parties
Decen tralization
59
agencies played a significant role in the selection process in many of the Dutch parties, but much less so in Israel where the process is very centralized.2 Moreover, the fact that various parties in one country, at a specific point in time, use different candidate selection methods underlines this distinction, which is buttressed by the more frequent changes to candidate selection methods by parties while the electoral system remains intact. Other, party-specific factors - such as party age, size, and ideology - may be associated with the extent of centralization, but they are far from supplying general, universally applicable rules. For example, ideology may have an impact, yet both centralized and decentralized candidate selection methods can be found in parties of the left as well as those of the right.
If we have to generalize, then it would be correct to say that in most parties the selectorate at the electoral district level plays a significant, and even a dominant role in candidate selection, while other nonnational levels (local or regional) play secondary role and the central party leadership usually supervises the process to some degree (Ranney 1981: 82-3). It is difficult to point to any recent trends regarding the balance of power between the center and the lower regional and local levels. Marsh (2000) argues that in recent years political developments that have strengthened the party leadership have also led many parties to try to increase the degree of centralization. This could be tied to developments such as the "pre-sidentialization" of politics (Poguntke and Webb 2005), the increase of public financing and "cartelization" of parties (Katz and Mair 1995), and the profession-alization of ever more capital-intense election campaigns. Krouwel's (1999) study of parties from 1945 to 1990 concluded that the dominant trend in the selection of parliamentary candidates in Western Europe was one of increased centralization. However, a study of candidate selection procedures across Western Europe from
1960 to the 1990s.by Bilk. .(2.0.01)..reachedthe^pposite conclusion: There is no.....
movement in the direction of more centralization, and if there is any evident trend it is toward increased decentralization.
The centralization of candidate selection also has its own contentious political consequences. One school argues that if the candidates have to appeal to the central party leadership in order to be selected, then the party's representatives in parliament are more likely to toe the party line. Alternatively, candidates whose selection is decided within the constituency will respond to the demands of their local base and might be willing to rebel against the national party leadership more often. Another school argues that it is possible to give the local level significant or even exclusive control of candidate selection while maintaining high cohesion and discipline, either because the local agencies see these qualities as important, such as in the United Kingdom (Ranney 1968), or because there is a division of labor between the local
2 The large parties in Israel employed a mechanism for territorial representation, but a clear advantage was given to the center. For example, between 1959 and 198 8, Mapai, and its successor Labor, allowed regional and local party agencies to select about one-half of its candidates for approximately the first fifty positions on the list, yet the national party organs decided their final positions on the list.
and national levels - the former selects while the latter governs - such as in Canada (Carty 2004). This issue will be elaborated and developed in Chapter 9.
DEMOCRATIZING VERSUS DECENTRALIZING
While an inclusive selectorate and a constituency-based selection method seem to go hand in hand, they must be theoretically and analytically distinguished from each other, as shown in Figure 4.2, and may be divorced in practice. In Denmark, for example, the selectorate became more inclusive, marked by the adoption of members' postal ballots in four parties, but the level of decentralization remained unchanged (Bille 1992). In the United Kingdom, the selectorate became more inclusive over the years, with the adoption of a membership vote for selecting candidates who were screened by party agencies at the national and constituency level. At the same time, the role of the central party agency in screening candidates, and its general involvement in the process, increased (Norris and Love-nduski 1995). Many scholars have not made this distinction, while others who have done so still mix both dimensions.
Bille (2001: 365-6) correctly states, "The phenomenon of decentralization is related to democratization... although in addition to decentralization, true democratization requires reforms that make both the candidacy requirements and the selectorate more inclusive at the local level" [emphasis added]. In other words, decentralization does not necessarily entail the transfer of power from an exclusive national oligarchy to a more inclusive local leadership. If the centralized process was based on a membership ballot^ and decentralization shifted power to 5, 50, or even 500 exclusive regional or local nomination committees, fewer people might end up being involved in the process rather than more. Gallagher's (1980: 500) study of the Irish case leads him to conclude, "The important decisions, then, are taken at local level, but this does not of itself mean that the candidate selection process is any more democratic in Ireland than elsewhere. Local parties can be controlled by elites in the same way as national parties." Thus, decentralization may be seen as a contingent step in the direction of democratization, but only if the decentralized selectorate is more inclusive than the earlier centralized selectorate.
Bille himself, after making the distinction between decentralization and democratization, then mixes up the two dimensions when he attempts to measure democratization based on six categories, the first five of which focus on national versus subnational control of candidate selection rather than on a more inclusive selectorate - his sixth category.3 A similar mix can be found in Kittilson and
3 BiUe's (2001: 367) first five categories include: complete national control over the selection of party candidates, subnational organs propose and national organs decide, national organs provide a list from which subnational organs decide, subnational organs decide but subject to the approval of the
60
Democracy within Parties
Decentralization
61
C E N T R A L I
Z A T I O N
National
Regional
Local
Selection of all candidates by all voters
Selection of all candidates by single national leader
Selection of candidates by regional delegate convention
Selection of constituency candidate by constituency voters
Selection of constituency candidate by
single local leader
Voters
Party members
Party delegates
Party elite
Single leader
EXCLUSIVENESS
F ] g u r e 4.2. Excfusiveness and centralization as separate dimensions
Scarrow (2003), whose first three categories of inclusiveness (out of five) also focus on national versus local selectorates 4 Gallagher (1988c: 236) also presents aspects of inclusiveness alongside those of centralization - under a category that clearly lumps them together ("The selection process: centralization and participation"). Janda (1980) devotes a chapter in his book on political parties to the centralization of power, defining it as the number of participants in the decision and their location in the organizational hierarchy, which brings together two
national organs, subnalional organs control the selection of candidates. Only his last category alludes to what he earlier called "real" democratization - the introduction of a membership ballot.
4 (CittJlson and Scarrow's (2003: 70) categories are: national leadership, regional delegates, local party selectorate, members can vote, nonmembers can vote.
Gallagher's (J 988c: 237) categories are: party voters, party primaries, subset of constituency party members, national executive, interest groups, national faction leaders, party leader.
dimensions that need to be separated.6 Hence, as Janda (1980: 111) wrote,"... the more restricted the privilege to participate in candidate selection, the more highly centralized is the party." Norris (2004: 27) divides between the degree of centralization and the breadth of participation in her analysis, but then merges them when she concludes, "In the most decentralized processes, nomination decisions in each local area rest in the hands of all grassroots party members who cast votes in closed primaries, or even the mass public in open primaries." Krouwel's (1999) study assesses internal party democracy using a scale of centralization that combines these two dimensions.7 An incomplete overlap between inclusiveness and decentralization is to be found in LundelPs (2004) study of the determinants of candidate selection. Lundell's scale of centralization combines these two dimensions, but does not allow inclusiveness to vary independently within many of the centralization categories. For example, the centralized end of the scale allows inclusiveness to fluctuate widely from selection by the party leader to primaries at the national level, whereas selection at the more decentralized district level covers only a selection committee, the executive district organ, or a delegate convention.8 The overlap is almost complete with Ohman's (2004) scale of the centralization of candidate selection. For example, the only difference between the two most decentralized categories is that the extreme one allows all or most party members to vote while the next-most extreme is a delegate conference.9 In other words, both
6 Janda's (1980: 111) categories are: nominations are determined locally by vote of party supporters, for example in a direct primary; selection is made by local party leaders whose selection must be ratified by party members; selection is made by local leaders with little participation by members; selection is made locally, but the selections must be approved by the national organization; selection is made by associations affiliated with the party or regional associations, but the selection muse be approved by the national organization; selection is done by the national organization, but the selection must be approved by local or affiliated organizations; selection is determined by a national party congress or caucus; selection is determined by a national committee or party council.
7 Krouwel's (1999: 94) categories are: incumbent national leader, party central office or national executive, interest or other external groups, parliamentary delegates, national convention or congress, select group of local party members, party members.
8 Luudell's (2004:31) categories are: selection at local party meetings, by local selection committees or by primaries open for all party members; selection at the district level by a selection committee, by the executive district organ, or at a convention by delegates from the local parties; the same as the first two but regional or national organs exercise influence over the selection; the same as the previous categories but local, district, or regional organs exercise influence over the selection; selection by the party leader, by the national executive organ, by a national selection committee, or by primaries at the national level.
9 Ohman's (2004: 48) categories are: selection at the local level by all or most party members through a direct vote, with no or only nominal confirmation by other party agencies; selection by a constituency-level delegate conference, with no or only nominal confirmation by other party agencies; the same as the previous category, but with confirmation from central party agencies; selection by regional bodies, or by a national conference, where delegates from all parts of the country decide on all candidates without confirmation by higher levels; selection by national party leadership subject to confirmation by lower levels; selection by the party leadership without corrfirmation by lower levels.
62
Democracy within Parties
Decentralization
63
are constituency ievel decisions, with no or only nominal influence by other party agencies, and what varies is the extent of inclusiveness. Ohman (2004: 12) clearly combines the two dimensions when he states that, "The most decentralized method would allow all eligible voters in the country to take part in the selection of the candidates, even if they are not members of the party in question. The other extreme would be when a single individual chooses all candidates for the parliamentary elections."
Contrary to what the aforementioned scholars contend, inclusiveness, as it was defined in Chapter 3, focuses on the extent of participation in the process and should be distinct from centralization. Tin's may be what Denver (1988) alluded to in describing the British political parties as being centralized but with decentralized selection; that is, the selection process was characterized by high territorial decentralization yet also by small selectorates. Ranney (1981), for example, proposed three dimensions for analyzing candidate selection - centralization, inclusiveness, and direct or indirect participation - and clearly distinguished between the first two. Centralization is described along a six-point scale made up of the following purely territorial elements: selection by national agencies with occasional suggestions by subnational agencies; selection by national agencies after serious consideration of suggestions by subnational agencies; regional selection with national supervision; regional selection with no national supervision; constituency selection with national supervision; constituency selection with regional supervision; constituency selection with no supervision (Ranney 1981: 82).
Ware's (1996) five dimensions of candidate selection also distinguish between the extent of centralization (his second variable) and the degree of participation (his third). He acknowledges that they are not independent of each other, but he does not mix between them.. An interestingsteptoward-differentiatingthe democratization of candidate selection from decentralization is given by Scarrow, Webb, and Farrell (2000). They raise two separate hypotheses concerning candidate selection: the first concerns its transfer over time to the party members; and the second is that the national party elite maintains or obtains the right to veto the decision made by local party members. In other words, what we see here is first democratization, and second centralization. The local party elites are weakened by democratization, while the party's national elite does not lose power because candidate selection is simultaneously being centralized. In their words,"... parties' decision-making processes will display movement toward... greater centralization and greater inclusiveness" (Scarrow, Webb, and Farrell 2000: 137, emphasis in original). The two hypotheses are validated by the data, but not equally. The democratization of candidate selection received limited empirical support; that is, more countries have expanded their selectorates than have curtailed it. Clearer support is given to the centralization hypothesis, yet this does not display a trend but rather an existing situation. While there are cases where the party center has gained power - such as Australia, the two main parties in Ireland, the Liberal Party in Canada, and British Labour - the interesting element is that in most cases the
existing power has not been cut. This shows that an expansion of the selectorate does not need to go hand in hand with decentralization, and that these two could also move in opposite directions at the same time.
Narud, Pedersen, and Valen (2002a: 13) clearly delineate between their two main dimensions of candidate selection - centralization and inclusiveness - and show that mere does not have to be an apparent connection between them:
In combination, these two dimensions include four distinct varieties of nomination systems. At one extreme is found an inclusive, yet cental, process: Participation is open to many individuals, though the outcome of the process is decided at the central level - not in the local branches of the party. An alternatíve nomination system is one that is decentralized and open at the same time. But we may also envisage decentralized and closed systems, where the decisions are made in the local party branches by a small group of "gatekeepers." A fourth type would be one with centralized nominations restricted to a few party members.
DEFINING AND MEASURING TERRITORIAL DECENTRALIZATION
Territorial decentralization, which focuses on the local-regional-national dimension, is quite a clear concept. If the party's local base has the full power to decide who its candidate will be, then we are at one end of the spectrum. If the national party is in full control of candidate selection, then we are at the opposite end. In the middle we find examples where candidates are decided at a level higher.than, the local, but lower than the national. Or, alternatively, if both the local and the national levels share power in the selection of candidates, then we are also near the middle, the exact balance of power determining how close to which end.
There is, however, another way to look at territorial decentralization, and that is from the perspective of the candidates. Candidates are selected either by a selectorate that is theirs and theirs alone (defined territorially or socially), in which case we are at the decentralized pole, or by the same selectorate as all of the other candidates, in which case we are at the centralized end. If some, but not all, of the candidates share the same selectorate, then we are in the middle of the dimension. This new and different, not to mention unconventional, perspective on the decentralization of candidate selection requires further explanation.
While some researchers address candidate selection at the national level, and most scholars - including the authors of this book - prefer to assess this topic at the level of each separate party at a specific point in time, when it comes to decentralization we can empirically design a framework for classification by looking at candidate selection from the perspective of the individual candidate. There are two questions raised by the individual candidate:
64 Democracy within Parties
1. How distinct is/are the selectorate(s) involved in my selection?
2. If there are two or more selectorates involved in my selection, what is their relative weight in the process?
The answer to the first question shows us the degree of decentralization for each selectorate. The answer to the second question tells us where exactly to classify candidate selection along the subsections of the dimension.
Consider a country divided into single-member districts where candidate selection is decided by the party at the electoral district, without any influence exercised at any other level. This means that each individual candidate will be selected by a selectorate - of the particular party in a single constituency - that is different from all the other selectorates. That is, in each constituency the selectorate is different, for each party. The United States is an example where voters in the primaries of each constituency cannot participate in selecting candidates in any constituency other than their own. In other words, each of the 435 Congressional districts produces different selectorates for each party, and hence for each of the party candidates in the constituency. From the candidates' perspective, their particular selectorate is made up of people who can only select a candidate in his or her own constituency, and are not involved in the selection of candidates in any another constituency. Two adjacent constituencies will thus have no overlap in their selectorates; they will be mutually exclusive. This is a picture of complete decentralization: a distinct selectorate makes the decision over who will be the party's one and only candidate.
Now consider a country that has a single, national electoral district, and where candidate selection for the party lists is decided at the national level, without any interference from lower levels. In this case, each individual candidate will face exactly the same selectorate - of the national party as each of the other candidates. Israel is an example of a country where the parties produce national lists for which all of the candidates must compete. In most cases the selectorate (and even the selectorates) for each and every candidate within a particular party is identical. From the candidates' perspective, they must face a selectorate that will choose all of the party's candidates. The exact same selectorate is involved in the decision concerning each and every one of the candidates. This is a situation of complete centralization: a single unified national selectorate makes the decision for all of the party candidates.
The interim case is a country with multimember districts, where each party produces a regional list of candidates. The candidates here must face a selectorate that is similar to that of all the other candidates in the district, but different from that of all the other districts. Iceland's six electoral districts each send nine representatives to the legislature.10 Each candidate in a multimember district,
10 Nine additional representatives, for a total of sixty-three, are elected in a second tier based on the results in the six districts.
Decentralization 65
Plural "Limited" Single
selectorates selectorates selectorate(s)*
Decentralized <t-————■--► Centralized
Local Regional National
decision decision decision
Figure 4.3. Decentralization and the number of selectorates per candidate * Complex candidate selection methods can produce several national-level selectorates that are involved in the selection of all the candidates.
will face a selectorate that chooses eight other candidates, and there will be six such separate selectorates. This is a case of neither complete decentralization nor absolute centralization, but somewhere in the middle. Between the two poles of singular versus plural selectorates, as shown in Figure 4.3, the middle can be labeled as limited in the number of selectorates. In the most decentralized case, the number of selectorates would be equal to the number of legislative seats (or two and even three times that number, if several selectorates at the same level are involved in the selection). The most centralized case will involve a single selectorate, or possibly a few selectorates as long as they are all at the national level.
What must be emphasized is that the question of how many selectorates exist in general - whether each candidate will be selected by his or her own distinct
i selectorate - divorces the issue of decentralization from that of inclusiveness. A singular selectorate can be either inclusive (national primaries) or exclusive (the national party leader). Exactly the same holds true for plural selectorates: they may involve the entire party membership within each constituency, or can be limited to the party oligarchy in the districts. This also follows for the interim case of a limited number of selectorates, and for all the possibilities along this dimension.
The second question focuses on the number of selectorates involved in the selection of a particular candidate, rather than in general. This question assesses
i the complex cases of candidate selection, specifically where a selectorate at one
level can influence the choice made at another level. The answer to this question will influence the location of each party along the continuum in Figure 4.3. For example, if the national party leadership is allowed to create a shortlist of
| potential candidates from which the local party at the constituency level can choose its candidate (such as in the United Kingdom), then selectorates at two levels are involved in the process.11 This is, therefore, not a clear case of centrali-
i        zation or decentralization, but somewhere in-between. The exact placement of the
! party along the decentralization dimension depends on the relative strength of the central party vis-a-vis the constituency party. Alternatively, the availability of a
I i
11 It should be noted that in the UK several selectorates are involved in candidate selection at the local level. They are all local, yet they vary in their levels of inclusiveness, from small exclusive party i agencies to larger more inclusive party agencies to party members.
66 Democracy within Parties
veto for the national party, after a candidate is chosen by the constituency {such as in Canada), is a similar case of two selectorates influencing the decision. In-between cases can also exist, such as a combination of the interim case of a '■" limited number of selectorates and selectorates from either of the two poles. If a | regional selectorate has to take into account the demands of the separate territorial i elements within the region (such as in Norway), or a regional selectorate at the multimember district level can be influenced by the national party (such as in |!| Austria), then we are close to one of the two ends of the dimension, but we are not 1 at the end. ■. j
This framework for classifying territorial decentralization has its drawbacks, 3} particularly when it comes to parties that employ more than two stages of candidate selection or when we encounter either assorted or weighted candidate selection methods. However, regardless of these complexities, the attempt to classify decentralization by looking at the individual candidate and his or her selectorate(s) achieves several goals. First, it clearly delineates the extent of decentralization using empirical data that should be relatively accessible, resulting     : t:j in a simple dimension that can provide researchers with an easy and valid way to 1 categorize and operationalize decentralization. Second, it unmistakably separates \\H between decentralization and democratization of candidate selection. There is no .::!:,4 mixing or combining of the centralization and the selectorate dimensions, and one ] does not vary from the other interdependently. This should, we hope, result in a   .' :K| dimension that scholars will no longer confuse with inclusiveness. Third, it is a j flexible dimension, allowing parties to shift based on a change in their method of candidate selection, or in the relative power of one selectorate versus another visa-vis the individual candidate.
MECHANISMS OF DECENTRALIZATION
Two mechanisms can be used for ensuring territorial and social representation via decentralization. The first mechanism used to ensure territorial and social representation is the establishment of separate territorial, sectarian, or social group districts, where the candidates and the selectors live in the same region, or are identified as or are members of the same sector or social group. The second is the reserved place mechanism, which guarantees a minimum number of positions on the list (or a minimum number of realistic seats in the case of single-member districts) for candidates representing specific localities or regions, or belonging to a distinct sector or social group. Establishing quotas for women, a practice adopted by many parties, is one example. If decentralization based on territorial mechanisms is usually ensured by using districts, reserved places are often used for ensuring social decentralization.
Decentralization 67
We have already discussed the issue of territorial districts, so here we will address the sectarian or social group district, where the candidates and the selectors are members of the same sector or social group. Unlike the territorial district, it is not built on a parallel electoral district, and it is thus no wonder that it is quite a rarely used mechanism. This mechanism decentralizes both candidacy and the selectorate, because the candidates not only need to be different from other candidates, but they also compete only against similar candidates and are selected by a distinct selectorate. A social group district in the candidate selection process is similar to the kind of social group constituencies found in the general elections in New Zealand, where members of the Maori minority are asked to decide if they wish to vote in the general district where they reside or in specific Maori districts. Israel's two main parties adopted such mechanisms of social decentralization in the mid-1990s. Both Labor and Likud allowed all their dues-paying members to participate in the selection of candidates, and each member was asked to choose from two lists: one composed of candidates who ran nationally and one made up of constituency candidates. The latter were somewhat fictitious, since Israel has no districts and employs a single, nationwide electoral constituency. The parties, however, established internal districts, both territorial and social. Labor, for example, had four social districts: kibbutz members (collective settlements), moshav residents (cooperative settlements), minorities (i.e. Arabs), and Druze.'2 Party members had to belong to a social district, a specific social group, in order to run and in order to vote in the district. Specific positions were reserved on the party list for these social districts, alongside reserved places for women, immigrants, and the territorial districts (Hazan 1999a). In those cases where parties allocate candidacies to affiliated social groups, and let them decide who will be their representative, we see the de-facto, creation of a social district in which.the.selectorate and the candidate are denned as belonging to the same group. For example, the allocation of specific candidacies to union representatives in the Japanese Socialist Party can be seen as the de facto creation of a district in which the union is the selectorate (Shiratori 1988; Youn 1977).
Belgium also supplies us with examples of both of the representation mechanisms, which were used at the (territorial) district level. In the Belgian Christian Social Party in 1961, the reserved place mechanism was used when it was decided that in some Brussels districts, Flemish and Francophone candidates would alternate seats on the party list.13 In 1965, separate intraparty subdistricts were actually established when Francophone and Flemish party members in these districts selected, separately, Francophone and Flemish candidates for parliament (Obler 1974).
12 A religion that split from Islam and is found mainly in Israel, Lebanon, and Syria. 1J In 196! and 1965, reserved places were also allocated to German candidates in a constituency with a large German-speaking minority.
68
Democracy within Parties
Decentralization
69
The reserved place mechanism - securing a rrhnimal number of realistic positions on the party list, or a minimal number of realistic seats in the case of single-member districts, for candidates belonging to a distinct sector or social group -implies the decentralization of candidacy alone. That is, unlike the case of districts, the selectorate is not defined along territorial or social lines. In some cases, candidates who are eligible for reserved places compete for their place on the list against all of the other candidates. The reserved representation mechanism is implemented in these cases only if the candidates do not attain the reserved position or higher ones. In other cases, the candidates compete against candidates like themselves for the predetermined positions only. Establishing quotas for women, a practice adopted by many parties, is one example.14 Here, women are allotted a certain percentage of seats on the party's list, or women are the party's candidates in a specific percentage of constituencies, yet they are selected by men and women alike.15
In order to make this kind of social decentralization more actual than virtual, women have to be given positions high enough on the party list so that a substantial number can be elected. In other words, the parties should seek a specific number of women representatives, not a symbolic percentage in nonreal-istic positions low on the party list. One way of doing this in a list system is "zipping," creating a zipper on the party list where every position is alternatively a man or a woman, such as in the German Green Party since 1990 (Davidson-Schmich 2006), the South African National Congress Party (Ashiagbor 2008), and the Dutch Labor Party in 2002 (Andeweg and Irwin 2002), or in which one member of each gender must appear in every set of three candidates, as is the case of the party lists of the Mexican National Action Party (Ashiagbor 2008).
The challenge in coimtries .with..smgle-^embe^ It is not
about locating women in realistic positions on the list but rather about ensuring that women are the party's candidates in realistic seats - safe, marginal, and targeted seats - rather than in those that the party expects to lose. This is quite difficult to achieve, as it means the active involvement of the center in coordinating the use of these mechanisms in the face of likely resistance at the district level. Such involvement is easier when the party is in opposition for a long time and when the adoption of such mechanisms is accepted as part of the recovery strategy. Indeed, prior to the British election of 1997, the Labour Party (after spending eighteen years in opposition) implemented all-women short lists, from which women candidates were selected in about one-half of the vacant Labour districts
14 There are several online databases that show the proliferation of quotas for women, avowing scholars to distinguish between those set at the state level and those established by the political parties. See, for example, Global Database of Quotas for Women (2009).
15 In many cases the quota is gender neutral, in the sense that it determines a minimum or maximum share of the candidacies for each gender without explicitly relating to women, although the aim is to ensure their representation.
and in marginal ones where the gap in the previous elections was under 6 percent (Criddle 1997).16 It is also easier to adopt such mechanisms when incumbents are not involved. In the first Scottish and Welsh legislative elections in 1999, the Labour Party used a mechanism called "twinning" in which every two constituencies had to present a male candidate in one and a female candidate in the other (Bradbury et al. 2000; Edwards and McAllister 2002; Squires 2005).
The goal of enhancing representation through the use of mechanisms of social decentralization may be achieved through adherence to quotas established via legislation, thereby circumventing the parties. This is the case in about fifty countries (Htun 2004). In other cases, parties voluntarily establish quotas (Kittil-son 2006). The more effective are the quotas with clear rules concerning ranking or district placements, and sanctions for noncompliance (Dahlerup 2006; Krook 2009; Mateo-Diaz 2005; Tremblay 2008).
Beyond the most outstanding and well-researched mechanism of social decentralization - women's quotas - these mechanisms can be, and have been, applied to regional, ideological, age, and occupational groups, among others. For example, in one German Land, the Social Democrats allocated 10 percent of its list positions to those under thirty-five years old (Scarrow 19996). British Labour comes somewhat close to this in ensuring the representation of minorities. If one or more minority candidates (black, Asian, or other minority) are nominated by the branches or another local party group, then the resulting shortlist must include one minority candidate (Ashiagbor 2008).
There is a connection between territorial and social decentralizations, in that the former has an impact on the latter. If candidate selection is decentralized along territorial lines, it becomes more difficult for the party leadership to implement the mechanisms of social decentralization. Matland and Studlar (1996) argued that more centralized candidate selection methods allow the party leadership to respond to pressures for increasing the diversity of representation, such as in the adoption of gender quotas. Moreover, centralized methods enable the party to intervene in order to ensure that quotas are applied. In contrast, more decentralized candidate selection lowers the influence exercised by the central party organs. Therefore, a party might need to centralize in order to increase the influence of the party center in candidate selection, and thus to decentralize socially. In other words, without central control, candidate selection could prove to be a barrier for social groups that is as difficult to overcome as the obstacles raised by majoritarian electoral systems.
The main obstacle to increasing the representation of women in the selection of candidates in single-member districts is the incumbents - most of them men whose political future is threatened when quotas are used. That is, they have to give up
16 The British Conservative Party, after ten years in opposition, adopted mechanisms for enhancing women's representation, after it had strongly objected to these for decades (Ashiagbor 2008).
70 Democracy within Parties --^i
their seat (rather than moving to a lower position in a candidate list, which can still leave some hope.for reelection). Indeed, in the United Kingdom, the greatest ■ ■■3; breakthroughs in women's representation were achieved when there were no incumbents - when the new parliament of Scotland and the new Welsh Assembly    : 3 ■; were elected (Bradbury et al. 2000; Mitchell and Bradbury 2004; Squires 2005). i Another opportunity is when the party is in opposition and the number of incumbents shrinks. .^ii
CONNECTING BETWEEN DECENTRALIZATION AND ELECTORAL SYSTEMS
It is intuitive to say that candidate selection is influenced by the electoral system. Aspects of the electoral system, such as the electoral formula, the district magnitude, and the availability of preferential voting have significant consequences on ';■■■} the political parties' choice of candidate selection methods. At the most basic : level, if a country has a single nationwide district, the parties need to choose only a J list of candidates. If, on the other hand, the country is divided into single-member      ■:] districts, the parties must choose individual candidates to stand in each district. Mixed-member electoral systems force the parties to produce individual district candidates alongside a party list. Here the parties must employ different criteria in each selection process, yet the two processes are not isolated, especially if some of the same candidates are allowed to run in both the districts and the list. The more
intricate the electoral systemrthe-more.....constraints.....are-placed onthe parties'
candidate selection methods.'7 In short, the connection between the electoral system and the candidate selection process is not only intuitive but also rather apparent.
Moreover, the impact of a change in the electoral system on candidate selection might be substantial, forcing the parties to adapt their internal nomination procedures. For example, in the 1990s, both New Zealand and Japan reformed their electoral systems and adopted mixed-member systems, and the parties, for the first time, had to walk through unfamiliar territory in selecting candidates for a list that in the case of New Zealand reached fifty-five places. Adjusting to these changes is not only a major organizational undertaking for the political parties, it also undoubtedly involves political consequences - district nominations in both
17 For example, the unique Chilean electoral system, based on two-member proportional districts, coerces the party alliances to make complex strategic calculations of their candidacies in the general elections. This makes the involvement of exclusive national party elites a necessity, and limits the significance of the selection of either more inclusive or decentralized party selectorates (Siavelis 2002, 2005).
Decentralization 71
countries were more prized than list places, especially by incumbents, which led to fierce battles within some of the parties (Hazan and Voerman 2006). Also in the 1990s, when Italy shifted from proportional representation to a mixed system, the parties there had to produce individual constituency candidates for the first time -a no less daunting or politically significant challenge. In 1917, when the Netherlands adopted proportional representation, the parties' candidate selection procedures faced similar upheavals that included intraparty political ramifications.
Decentralization of candidate selection, largely in its territorial sense, can and should be assessed in relation to the electoral system of a particular country.18 For example, if we are dealing with a case of single-member district elections, it is both interesting and important to assess whether the candidate selection methods of the particular parties are less decentralized or the same as the electoral system. In countries with somewhat centralized electoral systems, such as multimember regional districts, we should gauge decentralization in the parties' candidate selection methods in regard to whether they are more decentralized, more centralized, or similar to the national electoral system.
It is quite likely that decentralized electoral systems, such as single-member districts, will influence the parties to adopt more decentralized candidate selection methods. However, as with other aspects of decentralization, this is not always the case. The British electoral system is more decentralized than that of Norway, but the candidate selection method is more influenced by the central party organs -such as in the production of approved candidates lists - in Britain than it is in Norway. What does seem to be clear is that the more inclusive a candidate selection method is in terms of the selectorate, and the more territorially decentralized it is, the more there is a need for representation correction mechanisms to ensure social representation.
As long as we are able to distinguish clearly between the extent of candidacy requirements, the level of inclusiveness of the selectorate, and the degree of decentralization - both territorial and social - we are closer to having a more comprehensive and precise picture of candidate selection methods. What remains is to assess the appointment or voting method (or a mix of the two) used in the process of candidate selection, in order to complete our framework for assessing candidate selection methods. Only after this last dimension is explained can we pursue our inquiry into the political consequences of different candidate selection methods.
1 s Lundell (2004) found no evidence of a relationship between the electoral system and the degree of decentralization of candidate selection methods. However, his findings are questionable because he mixes centralization and exclusiveness.