A Comparative Analysis of Selection Criteria of Candidates in Belgium

The literature on candidate selection has focussed extensively on the degree of inclusiveness and decentralization of the selectorate, as part of the debate on intra-party democracy. However, much less attention has been paid to the degree of openness of candidacies, or selection criteria within parties. Yet parties have a lot of leeway in how they design selection criteria internally. Which guidelines do parties follow when making the crucial choice on which candidates to select for elections? This paper investigates selection criteria from two perspectives: the formal rules set by parties that restrict the candidate’s pool and the (informal) preferences of selectors that shape who gets selected. We aim first at contrasting the degree of party institutionalization and parties’ formal rules in candidate selection and so, we shed light on whether parties formalise their candidacy requirements and candidate selection processes to the same extent as other party activities. Second, the paper investigates the role of the selectorates, and how selectorate’s characteristics matter for the kind of (informal) selection criteria, be they intended at maximizing offices, votes or policies. Drawing on party statutes coded in the Political Party Database (PPDB) and 23 in-depth interviews with selectors, we study three francophone Belgian parties that differ both in terms of inclusiveness of the selectorate who has the final say on candidate selection and in terms of degree of centralisation, and in terms of party institutionalisation: the green party (Ecolo), the socialist party (PS), and the liberal party (MR). Our comparative analysis of parties, selection criteria provides new insights into the secret garden of politics and highlights in particular the major impact of parties, degree of centralization.


INTRODUCTION
Candidate selection processes are not only a matter of internal party life. Political parties are the major gatekeepers impacting who enters politics through their key function of candidate selection (Katz, 2001;Lovenduski, 2016). They decide on the pool of candidates that will be offered to voters on the ballot, and ultimately the personnel and groups represented. It has important political consequences, for instance on party unity in parliament (Close and Nunez, 2017), or on policy decisions, among others regarding issues of relevance to women (Tremblay, 1998).
Scholars have progressively opened the black box of the 'secret garden of politics' (Gallagher and Marsh, 1988). However, the literature has heavily focussed on the first two dimensions: inclusiveness and decentralization. Surprisingly, little research has been conducted on selection criteria set by parties, even though it is one of the foremost predictors of the outcome, above inclusiveness or decentralization. There is a limited literature on selection criteria of candidates that focusses on candidacy requirements, i.e., the formal rules set by parties to be nominated as candidate. Still, we do not know much about the dynamics that lead parties to apply these formal candidacy requirements. We know even less about the mechanisms that lead parties to develop informal selection criteria that further restrict the pool of potential candidates (Bjarnegård and Zetterberg, 2019). This calls for a more qualitative study that would uncover how parties develop their formal and informal selection criteria, which criteria they value, and why. This is what this paper intends to do.
More specifically, we link selection criteria set by parties to their level of institutionalization and the type of selectorate in charge of candidate selection. First, we expect that higher degrees of institutionalization of political parties lead to higher levels of formalization of selection criteria in the candidate selection processes. Second, we expect that different selectorates (on the inclusiveness and decentralization dimensions) have different goals, and hence different views on what makes a 'good' candidate, be it in terms of ideological, political profile or competences.
To investigate these questions, we study three Belgian Frenchspeaking parties: the Green party Ecolo, the Liberal party Reform Movement (MR-Mouvement Réformateur), and the Social Democratic party Socialist Party (PS-Parti socialiste). They share common features in terms of institutional setting but strongly vary on our two key factors: their level of institutionalization and the type of selectorate(s) in charge of candidate selection processes.
Contrarily to the scarce extant research on selection criteria, we do not only rely on official data (Bjarnegård and Zetterberg, 2019;Rehmert, 2020) nor on survey data (Bochel and Denver, 1983;Schindler, 2020;Van Trappen 2021). On top of an in-depth analysis of party statutes coded in the Political Party database (PPDB), we draw on original in-depth interview data with selectors within these three parties. No less than 23 party officials involved in selection choices at different election levels provided some insights on how the selection takes place and which criteria were formally and informally put forward in the selection meetings.
The contribution is structured as follows. We first sketch the main theoretical understandings on candidate selection and selection criteria, before digging deeper into mechanisms of party institutionalisation and types of selectorates as potential factors affecting how parties develop their selection criteria. The following section describes our case selection and data sources-mainly, party rules and interviews with selectors-and outlines our data analysis strategy. We then present our results. This paper demonstrates that party institutionalization is not a prerequisite for formalism in candidate selection, contrarily to normative pressure to follow the existing rules. Our analysis also details how centralization, to a larger extent than inclusiveness, is a major factor impacting the priorities of selectors. We conclude with a discussion of our findings and an outlook for further inquiry drawing on qualitative data retrieved from selectors.

CANDIDATE SELECTION PROCESSES AND THE BLACK BOX OF THE SELECTION CRITERIA
There is a growing literature on candidate selection processes. Scholars have progressively opened the black box of the 'secret garden of politics' (Gallagher and Marsh, 1988). They have emphasized how parties differ on four crucial dimensions: the level of inclusiveness (the size of the so-called 'selectorate', the party body selecting the candidates: a few party elites, delegates, all members or voters?), the level of (de)centralization (the location of the main decision-maker in the party hierarchy, e.g., centralised or at constituency level?), the decision-making method (how decisions are made: acclamation, nomination, vote?), and openness (who can apply, or the selection criteria).
However, the literature has heavily focussed on the first two dimensions: inclusiveness and decentralization. Interestingly, scholars have shown that different candidate selection processes can lead to different outcomes. Especially, the effect of inclusiveness and level of centralization on gender representation has been investigated. (Rahat et al., 2008) emphasize the tensions between intra-party democracy as a process (more inclusiveness and decentralization) and as an outcome (more representativeness). They show that larger, more inclusive, or more decentralized selectorates tend to produce a selection of candidates that is less balanced in terms of gender (Martland and Studlar, 1996;Caul, 1999;Krook, 2010;Kenny and Verge, 2013;. This would be due to atomization and coordination issues in inclusive, decentralized processes, opposed to more centralized processes that can look at equilibrium across electoral districts and be held accountable (Kittilson, 2006;Hazan and Rahat, 2010;Matthews, 2014;Pruysers et al., 2017). However, as Pruysers et al. (2017): 214 note, these two dimensions and their effect on representation 'interplays closely with quotas and formal rules that exist (either at the level of the party, or the state) to facilitate women's representation'. It stresses the importance of selection criteria set by parties, that would be the foremost predictor of the outcome, and not much the process of selection (Hazan and Rahat, 2010).
Yet surprisingly little research has been conducted on selection criteria (Rahat and Hazan, 2001;Siavelis and Morgenstern, 2008;Shomer, 2012;Cordero et al., 2016). As King (in Stark, 1996: 124) puts it, 'it is odd that very few writers have addressed themselves, except in passing, to the whole question of criteria, which one might have supposed was central'. When they do, they often work by proxy, by looking at the outcome of the selection process (Put, 2015;Vandeleene, 2016), or by looking at leadership selection processes (Stark, 1996;Kenig, 2009;Pilet and Cross, 2014) or the selection of ministers (Bäck et al., 2016). While theoretical arguments could be retrieved from this literature, one may expect the criteria for leaders and ministers to differ at least slightly from those for regular candidates.
The limited literature on selection criteria of candidates tends to focus on candidacy requirements, i.e., the formal rules set by parties in their internal documents to be nominated as candidate (Krook, 2009;. Increasingly, comparative datasets and studies of political parties allow to compare these requirements across a large set of parties. Indeed, Rahat and Hazan (2001) shows how that, next to legal requirements set in national law (e.g., quotas, citizenship, age-Rehmert, 2020), some parties set additional barriers or restrictions. For instance, Pilet et al. (2015) have compared the criteria set by 145 parties in 27 EU countries. They highlight party-specific requirements that can have either a collective or an individual dimension. The most common collective condition among parties in Europe are gender, ethnic, geographical, linguistic quotas or balance, or for affiliated organizations or civil society candidates. In terms of individual requirements, the most common conditions are party membership, minimum length of membership, age, link to affiliated organizations, sponsorship, and endorsement by elected officials, leaders, factions, or members, but also fee deposit, incumbency, or incompatibility with other professions.
Still we do not know much about the dynamics that lead parties to apply formal candidacy requirements, and even less about the mechanisms leading them to develop informal selection criteria that further restrict the pool of potential candidates (Bjarnegård and Zetterberg, 2019). One early study analyzed the preferences of selectors in terms of characteristics and qualities of candidates based on a large survey among selectors in the Labour party in United Kingdom (Bochel and Denver 1983). However, this has not been followed by many works until very recently. Recent works by Van Trappen (2021) or Schindler (2020) have started to remedy this gap but adopt a quantitative perspective using surveys or experiments. Only Schindler (2021) provides a more qualitative account on the informal selection criteria. This lack of knowledge forms the starting point of this research. Using qualitative data and methods, we uncover how parties develop their formal and informal selection criteria, which criteria they value, and why. We consider in this paper the selection criteria from two complementary perspectives: we posit that criteria to select candidates encompass both formal rules set by parties (among which candidacy requirements) and the informal preferences of those who select.

INSTITUTIONALIZATION, SELECTORATES, AND SELECTION CRITERIA
This paper links (in)formal selection criteria set by parties to their level of institutionalization and the type of selectorate in charge of candidate selection.
First, we expect that the degree of institutionalization of political parties is related to the level of formalization of selection criteria in the candidate selection processes (Reiser, 2014). Party institutionalization refers to 'the process by which organizations and procedures acquire value and stability' (Huntington, 1968: 12), or 'the way the organization "solidifies" (Panebianco, 1988: 49). Authors have developed multiple typologies of dimensions of party institutionalization (Huntington, 1968;Janda, 1980;Panebianco, 1988;Randall, 2006;Bizzaro et al., 2017). These typologies distinguish between internal and external dimensions (Randall and Svasand, 2002). Internal aspects refer to developments within the party itself, such as organizational development (see Mainwaring, 1998;Kuenzi and Lambright, 2001;Webb and White, 2007;Basedau and Stroh, 2008). External aspects have to do with the party's relationship with the society. Scholars also distinguish between organization and value-infusion (Levitsky, 1998). In this paper, we are interested in the internal, organizational dimension of party institutionalization. Following Panebianco's, (1988) view, we define an institutionalized internal organization as characterized by a certain level of regularity, organizational complexity, routinization, and the development of prevalent conventions guiding behaviour. In that line of reasoning, we expect that highly institutionalized parties will also be formal when it comes to candidate selection. Hence they would develop more formal selection criteria in the form of candidacy requirements written in the party statutes and will provide more guidance to (or control over) selectors in the process, leaving less room for maneuver for informality in selection criteria. Conversely, we expect that less institutionalized parties will develop more informal selection criteria, letting some leeway for candidates not meeting them. Our first hypothesis thus reads: H1: Party institutionalization leads to a formalisation of candidate selection and of selection criteria.
Second, we investigate whether different selectorates lead to varying selection criteria. The literature on selectorates has already emphasized that different selectorates 'produce' different outcomes in terms of candidates selected and representativeness (Hazan and Rahat, 2010;Pruysers et al., 2017). These differences are often attributed to coordination issues in large groups. While we acknowledge the role of this factor, this view hides the fact that various selectorates might also have different views and preferences. We want to unpack these mechanisms by looking not at the outcome, but at the process. More specifically, we are interested in the criteria that various selectorates value in the process of selecting candidates. We assume that different selectorates have a different view on what makes a 'good' candidate, be it in terms of social, political profile or competences. Next to the formal criteria, these informal accounts likely play a key role in shaping the candidate's choices.
Different selectorates might prioritize different goals, between vote-seeking, office-seeking and policy-seeking (Strom, 1990). This might affect how they prioritize between the three main selection criteria developed by Stark (1996) for leadership elections, namely acceptability, electability, and competence. As commonly done in the literature, we distinguish between two analytical dimensions of selectorates: their degree of centralisation and of inclusiveness (Hazan and Rahat, 2010). Along the centralisation axis, we expect that the party in central office might prioritize office-seeking goals, and hence prefer competence as selection criteria given their care for a competent party in public office for the good health of the party as an organization in general. Central party selectors are indeed portrayed in the literature as the ones prioritizing unity the most (Schindler, 2021). Decentralised, constituency organizations might prioritize vote-seeking goals, and therefore electability, because that would be their main evaluation criterion on Election Day: how successful was the constituency. This has been highlighted by Bochel and Denver (1983) based on survey results among delegates participating in selection conferences at the constituency level. He showed how these types of selectors ranked vote-seeking goals the highest. Along the inclusiveness axis, we expect grassroots to prioritize policies and party ideology (Sjoblom, 1968), and therefore acceptability (Quinn, 2016) much more than the less ideologically committed exclusive selectorates. Schindler (2020) has conducted a survey among various types of selectors to investigate whether they differ in terms of selection criteria. He showed how more inclusive selectorates are less guided by vote-seeking goals. We hence develop a threefold hypothesis: H2a: Centralized selectorates tend to favour office-seeking goals, and hence competence as selection criteria. H2b: Decentralized selectorates tend to favour vote-seeking goals, and hence electability as selection criteria. H2c: More inclusive selectorates tend to favour policy-seeking goals, and hence acceptability as selection criteria.

Case Selection
This paper investigates parties, selection criteria in the candidate selection processes among three Belgian French-speaking parties: the Green party Ecolo; the Liberal party Mouvement Réformateur (MR-Reform Movement); and the Social Democratic party Parti socialiste (PS-Socialist Party). Ecolo was founded in 1980. The party has known ups and downs in its electoral success, fluctuating between 4 and 20% of the francophone seats in the federal and regional parliaments (Pilet and Talukder, 2021). The most recent examples are the major defeat at the 2014 regional and federal elections followed up by a large success at the most recent elections in 2019 (Reuchamps et al., 2019). On average, the party ranks 4 th in the French-speaking landscape, after the PS and the MR and close to the Christian Democrats, and more recently to the radical left. It tends to perform better in Brussels than in Wallonia.
The MR traces back to the first political party established in Belgium, the Liberal party, then a nationwide party. It relabelled itself Parti de la liberté et du progrès/Partij voor Vrijheid en Vooruitgang (PLP-PVV-Party for Liberty and Progress) in 1961. Like other mainstream parties in Belgium, it split along the Dutch-French linguistic divide in 1972. After a period of turmoil, the Parti réformateur liberal (PRL-Party for liberal reform) was founded in 1979. In 1993, it enlarged to a federation including the regionalist party FDF, expanded to the MCC (splinter of the Christian Democrats) in 1999. In 2002, this federation relabelled itself MR (Delwit, 2017). The FDF left the federation in 2011 after disagreements on state reform. The party has mainly occupied the second place in the francophone party system, with a short exception of 2007 where it ranked first . The party historically performed better in Brussels than in Wallonia, but the sociological changes among the Brussels electorate has eroded their dominance in the capital city.
The Parti socialiste also dates back to the 19 th century. Its roots lie in the Parti Ouvrier Belge/Belgische Werkliedenpartij (POB-BWP-Belgian Workers' Party) founded in 1885. The party relabelled itself Parti socialiste belge/Belgische Socialistische Partij (PSB-BSP-Belgian Socialist Parti) in 1945. It was the last of the three main parties to split along the linguistic divide in 1978, when it became the Parti socialiste . It has been the dominant parti in French-speaking Belgium since after the war, especially due to its strong local anchorage in Wallonia. It has maintained its status of first party throughout the period, with the exception of 2007. The party has increased its performances in Brussels while its electoral grip on Wallonia has decreased.
The selection of these three parties for our analysis relies on a Most Similar System Design strategy, with the three parties sharing several characteristics, but differing on their level of institutionalization and the selectorate in charge of candidate selection, our two main independent variables.
Indeed, the three parties share similar characteristics. They operate in the same federal multilevel setting with regional and federal elections 1 . They also function in the same party system and under the same set of institutional and electoral rules. For instance, they operate under the list system, where the selection outcome is a group of candidates. More specifically, Belgium applies a flexible list PR system with multiple preference voting (André et al., 2015). However, it has often been labelled a closedlist system in disguise (Crisp et al., 2013) given the difficulty for candidates to break the list order and bypass candidates ranked higher on the list (1,4% of all elected regional and federal candidates from 1995 until 2014 according to Cogels, 2020). Parties have to draft one electoral list for each constituency, and possibly one for each level of election when elections are held simultaneously (which was the case for the last two elections). Moreover, parties draft a so-called substitute list next to the effective list, presenting candidates who will be entitled to sit only if an elected representative renounces her/his mandate during the term (except for the regional election in Brussels). Given these characteristics, parties, selectors remain extremely powerful in determining the future elected representatives (Hazan and Rahat, 2010).
The position of all three parties in the party system implies that they can all count on several realistic positions on most electoral lists and may even hope for some ministerial posts, a factor to keep in mind when the selectorates proceed to the candidate selection. Uncertainty around electability is probably the highest for Ecolo given its history of electoral yoyo.
Even if parties resemble each other in terms of structure (Legein and , they vary in terms of level of institutionalization and selectorate in charge of the candidate selection process. To assess the level of institutionalization of our three parties, we rely on Mainwaring's (1998) operationalization, refined by Basedeau and Stroh (2008). We focus on the internal organization dimension of institutionalization, measured by four indicators: membership strength, regular party congresses, material and personal resources, and nationwide organizational presence and activities beyond election campaigns. To these indicators, we add Baer's (1993) dimension of relatively clear authority structure and division of labor. We retrieve the information from the Political Party Database (PPDB). The dataset focuses on the 'official' story regarding the most important aspects of party organization, based on party statutes and resources (Poguntke et al., 2016;Scarrow et al., 2017). We used the most recent round (Round 2, 2016-19) to retrieve the data on our three selected parties. In Belgium, parties, level of resources is mainly determined by law, as parties are eligible for public funding and private donations are very limited (Gaudin, 2020). Parties are also eligible to staff based on their electoral results (Moens, 2021). Given their respective ranking in the party system, the PS displays the highest level of resources, followed by the MR and Ecolo. There is no party law that provides specifications as to how parties should select their candidates.
On the other criteria too, the PS is the most institutionalized of the three parties. With its roots as mass party, it still has the largest membership base, albeit in decline (68,254 in 2018, see van Haute and Paulis, 2017). The party holds regular congresses at all levels (see party statute's articles 15, 20, 28, and 30-34). It has a nationwide organizational presence, with 289 local party branches covering all municipalities in French-speaking Belgium, and 115 lists submitted using the party label at the last 2018 local elections (Legein et al., 2020). Its authority structure and the division of labor is clearly stated in its statutes, including the incompatibilities, motions of distrust, appeals, etc. When it comes to candidate selection, the party's selectorates correspond to the constituency organizations. They attribute the PS logo to a list, have the initiative in proposing a first draft of the lists in accordance with the directives from national executive and congress, and have formal final input. The national party has formal but overarching input: the national executive and the congress set directives for the building of the lists.
Ecolo displays an intermediate level of party institutionalization. Founded on participatory principles as a militant party, Ecolo's membership base is more restricted but stable, around 5-6,000 members in the last decade (5,938 in 2018). While its membership is more limited, the registered members are proportionally more active ( van Haute, 2015). The party holds regular congresses called Assemblées générales, open to all members (Title III, chapter 1 of party statutes). It has a nationwide organizational presence, with 286 local party branches covering all municipalities in French-speaking Belgium and has a strong policy of using the party label at local elections with 144 party lists submitted under the party label in 2018. Its authority structure and the division of labor are clearly stated in its statutes, including incompatibilities and appeal procedures. When it comes to candidate selection, the party's selectorates correspond to constituency member's assemblies depending on the level of election (i.e., all party members registrered on the territory of the electoral constituency). The constituency organization has the initiative to propose a draft list via a list committee set up at the occasion of the elections and gathering both constituency and national leaders (or their delegates). The constituency level has thus formal input (can amend the list) but also the formal final approval. All members analyze candidacies first for all eligible places on the list and later for all other places, and vote in Assembly (quorum of at least 20% of members required) (Vandeleene, 2018). The national party can have formal input besides its involvement in the list committee. The so-called Council of Federation can adopt a procedure of codecision (article 155) and sets the calendar (article 157). The MR displays the lowest level of institutionalization. The party emerged as a cadre party and has transformed into an electoral party. Its membership base is hard to assess given the lack of information provided by the party's headquarters. The leadership elections are therefore the only indirect method to assess its membership figures. In 2019, 24,477 members were listed as potential voters in the leadership race (Vandeleene et al., 2020). The party has a nationwide presence with 282 local party branches. However, only 91 lists were using the party label at the last local elections in 2018. The party statutes are relatively short and outdated (last revision was in 2005; since then, one member has left the alliance, but new statutes have only recently been revised and not approved). The statutes do not specify how frequently a party congress must be held (article 8). Finally, the authority structure is much less clear, flexible, and frequently adapted to the needs of individuals holding the reins of the party at a specific moment in time (Sierens and van Haute, 2017). When it comes to candidate selection, the statutes are relatively vague on the party's selectorate. The national level controls all steps of the process via the electoral commission, from initiative, to formal input to formal final approval. The electoral commission is composed of the Party Leader, the Governmental Leader, and the members of the National executive (article 26). Party members, constituency organizations or affiliated organizations do not have a formal role. This is to ensure an equilibrium between the partners in the alliance. Yet the outdated character of the MR statutes requires to nuance the relevance of these written rules, especially considering the already mentionned flexibility of the party structure (Vandeleene, 2018). Ultimately, our three cases vary in terms of level of party institutionalization (high for PS, intermediate for Ecolo, and low for MR) and selectorates formally in charge (decentralized inclusive members assemblies for Ecolo, decentralized delegates for PS, and centralized executive for MR). Figure 1 summarizes each party's position on both analytical axes.
Accordingly, our expectations regarding our three parties are summarized in Table 1.

Data and Information on Selection Criteria
The formal candidacy requirements as stated in the party statutes are retrieved from PPDB. To go beyond the formal rules, we conducted individual in-depth interviews with party selectors. These interviews aim at collecting information on our dependent variables. First, interviews were used to gather information on the level of formality of selection criteria (H1): how candidates are selected in the party in practice, compared to party statutes? Which role assumed the interviewees, i.e., to what extent they were in the driving seat to select candidates, which kind of candidates this was and for which electoral level(s), and did they experience being an aspirant and/or a candidate themselves? We also asked to what extent candidate selection was steered by the party as an organization or instead whether the selectors felt that they were rather free in their choices of decision-making procedures and criteria, whether they received guidelines prior to the selection phase or whether they had to report to some other party body during and after the selection. The second major layer of the interview tackled selector's preferences in terms of selection criteria (H2). We started with a broad and general question of "what is a good candidate according to you?" before digging deeper in the selection criteria depending on the provided answers. Selectors were then prompted based on a set of vignettes to address three theoretical selection criteria based on parties' strategic goals (Sjoblom, 1968). If not spontaneously mentioned, we proceeded to ask about potential differences between candidates on (un)realistic positions.
We interviewed no less than 23 respondents (seven to nine per party) who were all involved in at least one recent selection process at the regional or federal level. Even though the focus of this research does not lie on local elections, the point of comparison proved to be relevant for many interviewees who could rely on insightful examples from their local experience. Almost all interviewees have also been candidate themselves (with a great variation from head of list to substitute candidate) and some could rely on a parliamentary and even cabinet experience. Being able to understand the other face of the coin by experiencing being the aspirant to a candidate position oneself was extremely useful to encourage respondents' reflection on the critical choices made by selectors. Some interviewees openly declared having been disappointed not having been selected on a particular position at one selection process and reflected on why their profile did not fit the criteria of the selectors on that occasion. The respondents are politically active in various Belgian provinces with varying local contexts (in terms of own party success or population characteristics, e.g., more rural or more urban). For Ecolo, we have about the same number of women and men, but parity was not realistic in the other two parties considering how candidate selection takes place (i.e., most decision-makers are still men). Respondents' level of experience in politics varies, as does their age (from 30 to 60 years old, mean age of 52 years old). Details about the interviewees can be found in the appendix.
The interviews took place in two phases. For Ecolo, interviews were conducted in February and March 2020, at various places (respondents' office, home or in a coffee shop). Due to the COVID-19 pandemic, the interviews for PS and MR took place online via the Zoom or Teams platforms, in May-July 2021. Interviews lasted on average 47 min. All interviews, conducted in French, were recorded and entirely transcribed, either by the authors or by job students. The analysis of the interviews' transcribed texts has been undertaken according to a cross-sectional code and retrieve method in NVivo, starting with a categorization of chunks of text into large categories (chiefly, types of selection criteria, role and kind of selectors, features of selection process) combined with a later refining of categories, both during the coding process and afterwards (Ritchie and Lewis, 2003). A systematic comparison of categories emerging from the data helped uncover the relative importance of more or less formalized practices of candidate selection and the related criteria put forward by selectors.

Party Institutionalization and Formal Rules in Candidacy Requirements
Our first goal is to uncover whether more institutionalised parties develop more formalism in their candidate selection process, in particular regarding the establishment of formal selection criteria, of guidance to people in charge of the selection decisions and of a reduced room for maneuver for informality in candidate selection. We selected three cases varying on their level of party institutionalization. Based on these levels, we expect the degree of formalism in candidate selection to be high for PS, intermediate for Ecolo, and low for MR. We first sketch the formal candidacy requirements at the legislative level and partyspecific requirements based on the PPDB. Next, we rely on our interview data to investigate the extent to which respondents reported the process to be formal, before digging deeper into the formalism of the selection criteria as such and in particular the control by the party over selectors.
In Belgium, the electoral law sets individual and collective formal candidacy requirements. Individual candidates for the Lower House must have civil and political rights, be 18 years old or more, and be Belgian and residing in Belgium. Collectively, they must submit a list with gender parity, with a maximum difference of one for odd lists. Moreover, the first two positions on the list shall be occupied by a man and a woman (note that for regional elections in Wallonia, full gender alternance on the list is applied, i.e., the so-called zipper system).
Parties sometimes develop additional formal candidacy requirements in their statutes. It is the case for the three parties under study, but to different extents. The PS applies collective selection rules. It reiterates the national rules regarding gender parity whatever the level of election (article 7). It introduces an additional age criterion, stating that a list cannot have more than 15% of candidates older than 65 years old on the day of the election (article 68). In terms of individual rules, candidates must be member of the party and sign a loyalty pledge committing to party group discipline in the legislature (the 'Charter of the candidate', see below). There are no formal rules regarding links to groups, no endorsement or fee deposit required.
Ecolo also reiterates the gender parity rule in its statutes. At the individual level, to be allowed to be listed as candidate on an Ecolo list, one should 1) be a member of the party (or of the sister party Groen or another party with which Ecolo has an agreement), 2) be a non-member but be approved by a 2/3 majority by the selectorate (article 152). Candidates must also sign a loyalty pledge. There are no formal rules regarding links to groups, no endorsement or fee deposit required. The Council of Federation can impose additional conditions, specific to the context of the elections (article 153).
The MR is much more unclear regarding candidacy requirements in its statutes. There is no mention of collective rules regarding gender or ethnic background. In terms of individual requirements, there is no need to be a party member, but candidates must sign a loyalty pledge. There are no formal rules regarding links to groups, no endorsement or fee deposit required.
Yet one knows that candidate selection goes beyond formal rules and candidacy requirements. De jure and de facto procedures do sometimes not match, and scholars should at least consider the divergences between both (Meserve et al., 2018;Kelbel, 2020). The accounts of selectors retrieved from in-depth interviews allows to assess how much room for maneuver for informality is left for selectors.
Our interviews confirm the high degree of formalism in candidate selection for the Socialist party (PS). However, some degree of informality prevails in the early and most relevant stages of the list drafting. What is striking from the interviews is the importance granted by most respondents to the statutes together with the acknowledgment that the most important decisions are taken informally by one or some party elites. The candidate selection core leverage is clearly situated at the decentralised level with the Federation presidents steering the processes. The federations, one per arrondissement (a sub-territory of the province, corresponding to the regional constituencies), form the backbone of the party structure. These local leaders are responsible for the list drafting and are rather free to organise their own selection process like they wish, resulting in a potential variety of processes across the different federations. Some take advantage of the formal party bodies and rely for instance on the federation board to validate their choices while some only informally consult the main local party sages (e.g., former top politicians, powerful mayors or local party chairs) or the head of list when they do not self-designate.
Most processes rely on an open call for candidacies, which sometimes results in a formal endorsement of some candidates by their local party branches (i.e., an even more decentralised selection process). Besides, most lists are at the end of the day formally validated by the federation congress gathering delegates or rank-and-file members. These large gatherings seemingly never hamper the decisions made by the federation leader(s)-one may rather talk about rubber-stamping. The congress votes 'en bloc' on the list either by secret ballot or more informally by a show of hands.
The national party level does not seem to strongly interfere in the process even though respondents refer to a validation by the national headquarters. A continuous coordination throughout the process guarantees selectors that their proposal will be accepted. This coordination is likely to happen rather formally, for instance during the meeting of all Federation's presidents with the party General Secretary. At this meeting, "there is of course inevitably a progress report on the constitution of the list on the agenda and therefore each federation can report on the difficulties it encounters or not in the framework of the drafting of the list" (P2). But the national and the constituency levels also informally come together, primarily to determine the candidates on the most realistic positions, and this starts long before Election Day.
In terms of selection criteria as such, PS respondents claim not being given instructions by the national headquarters. There is however a 'Charter of the candidate', mentioned by several interviewees, originating from the national party but that can be fine-tuned by each Federation. The very usage of this loyalty pledge and its respect also seem to vary from constituency to constituency. This written document to be signed by all candidates theoretically compells them to follow the party rules, but respondents admit that the party is sometimes powerless when a candidate deviates from the party line at campaign time or once elected. Apart from excluding the Frontiers in Political Science | www.frontiersin.org November 2021 | Volume 3 | Article 777747 freerider, there is not much the party can do. Next to the formal Charter, some selectors report the existence of some guidelines set by the national party board: "These are not injunctions, instructions or pre-established grids by the big war party machine which has a research department that almost sends you the age and sex of the third candidate of the second list." (P4). Examples of these guidelines are paying attention to the socalled opening candidates, i.e., non-member candidates, to prevent disorder during the campaign. A respondent recalls the received instructions from the party: "If you appoint opening candidates, pay attention to the way they will communicate, assist them. We don't say no to you, but we don't say yes." (P9). The representation of some population categories would also be encouraged, be it to remind selectors to select enough young but also senior candidates. Beyond these recommandations, selectors feel rather free in their decisions, what -they report-is for the good of the list quality given their knowledge of the constituency, counter to national leaders. We expect an intermediate level of formalization of the candidate selection processes for our second case, Ecolo. Based on the interviews, it appears that the written rules are narrowly followed in the party: the constituency organization holds the power via the members' General Assembly, entitled with the nomination of who might sit in the list committee on behalf of the constituency organisation (and will decide together with representatives from the party national leadership and from the Council of Federation-the party national delegate's assembly) and after the list committee has drafted a list proposal, is charged with the final approval of the draft list. The selection process starts with a formal call for candidacies. Lists are constituted according to an assorted process: first the eligible positions and later the other list positions, determined by another committee in which the heads of list hold a key role. Respondents from various constituencies reported very similar decision-making procedures and they all highlighted the importance of respecting the rules, and in particular the extent to which party members grant importance to these rules ensuring fair decisions as well as a smooth intraparty competition (as opposed to the poll system formerly in place in the party that proved to be harmful for party cohesion, see Vandeleene and De Winter (2018)). There is still some room for informalism in the working of the list committees that are free to organise their decision-making how they want: some foresee (several rounds of) individual interviews with aspirants and others gather all aspirants who have to defend their candidacy before the group.
Ecolo formalises the establishment of selection criteria in two respects. First, the Council of Federation sets general guidelines for candidate selection, such as gender parity among heads of list in the same constituency or trying to avoid local office-holders on the list. Second, beyond these national instructions, each members general assembly gathers before the list committee starts the selection process and establishes a set of selection criteria: "We made working groups of five-six people. . . And we gave them half an hour to create the criteria. And then the working groups came back. They said: 'We saw the similarities in the working groups, the differences'. Then we had another informal discussion." (E1). The result is a non-constraining list of criteria that the selectors have to follow when drafting the list: "It makes the General Assembly accountable to criteria. It's as if they gave us a mission statement." (E6) The interviewed selectors acknowledge the necessary degree of informalism in the selection process, in particular to stimulate some candidacies or to decide between two very similar profiles, but also emphasize the importance of the final vote by rank-and-file members: "We are required, as we know, to respect the balances because if we don't respect the equilibrium that has been decided with the members, it automatically won't be accepted because there is a vote." (E2) All in all, both the process and the criteria are rather formalised. It is not so much the national party who controls selectors, but selectors themselves feeling compelled to abide the rules.
We expected our third case, the liberal party MR, to have the least formalised selection procedures and criteria. Respondents consistently report that the formal rule is that the party national leadership holds the power to designate the head of list (possibly with the provincial party leaders), after which the head of list becomes the main decision-maker to select the remaining candidates. Some heads of list let a party constituency body (the board or even the members' assembly) formally validate their proposal, but most acknowledge that the validation is rather a formal approval that denotes the launch of the electoral campaign rather than a moment when the list composition is discussed. There is no systematic call for candidacies; this is let at the discretion of the main selector/head of list. A large degree of informalism prevails in the list drafting process, and the number of decision-makers varies depending on how much the head of list is willing to share the power with other constituency figureheads or with the national party leaders.
No MR respondent reported formal selection criteria. On the contrary, they rather highlighted the subjective character of their decisions. The decisions fall on the shoulders of the heads of list (often the constituency party chairs) who choose candidates on their own: "There are no directives from above, we do it ourselves and we have the wisdom to know our territory well enough to represent it at best." (M6). This results in the extreme importance of individuals and their own preferences. Moreover, "there is also a courting logic that takes hold with the president, especially in the months before the lists are drafted, when everyone is nice to him or her." (M2). Yet some informal coordination takes place, between the heads of list for regional and federal elections running in the same constituency, and with the national and/or provincial leadership who might interfere especially when conflicts arise. For instance, when there are only a few realistic positions, "there is a need for arbitration by the party leader, who must not only arbitrate on the human dimension, but who must also arbitrate on the dimension of the party's interest, and the human dimension is not always in line with the interest of the party" (M4). This informal logic is also emphasized to be conform to this cadre party centred around individualities and office-holders' relative freedom.
Our first hypothesis stated that highly institutionalized parties will develop more formal selection procedures and criteria, and will closely control the selectors, resulting in a narrow room for maneuver for informality. Our findings slighlty nuance this assertion. It appears first that party institutionalization interacts with the degree of centralization and inclusiveness of candidate selection procedures. The degree of centralisation of the selection processes affects the possibilities for parties to have homogeneous list drafting procedures, i.e., following the same rules. The highly institutionalised PS knows rather heterogeneous selection processes due to the pivotal role of the constituency organizations and the prevalence of their own rules. Yet the national party strives to coordinate and influence the decisions via the meeting of the federation presidents. The story is different for Ecolo, which is less institutionalised as a party but manages to have homogeneous selection processes thanks to a deep concern of party members, and so of selectors, for a respect of the rules. National leaders (or their representatives) are part of the selectorate in charge of the main list draft, what allows them to possibly steer the decisions. The least institutionalised party of our research, MR, follows our expectations by acknowledging the occurence of informalism in the selection processes, apart from a formal designation of the heads of list who later select the other candidates in a rather informal way. The national level might informally provide guidance, in case of problems only. Second, we show that the degree of institutionalisation can also affect the establishment of formal selection criteria. Again, the most institutionalised party in our analysis is not the party relying on the most formal list of criteria for candidates. Although a so-called PS Charter of the candidate does exist, this document is nor widespreadly used nor is it similar across constituencies given the decentralised authority. The party board rather sets guidelines but these are not seen as formal requirements for selectors. On the contrary, the intermediate party in terms of party institutionalisation, Ecolo, relies on a very formal set of selection criteria systematically established by the members' assemblies prior to the selection of the candidates. The list committee's members consider this criteria list as their contract to which they feel accountable. Besides, similarly to PS, the national Council of Federation sets general guidelines for candidate's preferred profile. The weakly institutionalised MR is in line with our hypothesis as no formal selection criteria, nor from the central or the decentralised level seem to exist.

Selectorates and Selection Criteria
Our second research objective was to investigate whether different selectorates favour different selection criteria. Centralized selectorates would prioritize office-seeking goals and competence as selection criteria, while decentralised selectorates would prioritize vote-seeking goals and electability. More inclusive selectorates would prioritize policy-seeking goals and acceptability. Drawing on the insights from our interviewees, we sketch in the following paragraphs the main priorities and criteria respondents reported, depending on the selectorates' characteristics.
Leaders at the central party level are expected to prioritize office-seeking goals and competence as selection criteria (H2a). This is confirmed in the interviews. Interviewees refer to the importance of a balance of competences among the would-be MPs so they can cover as much portfolios as possible once in Parliament. "We wanted to have a group that was both diverse and coherent [. . .] in order to be functional and effective." (E6). We argue that the prioritization of office and competences is related to the centrality of the selectorate. Centralized selectorates enjoy the helicopter view on the lists' drafting processes and can steer the selection of candidates on realistic positions to ensure some balance of profiles within the parliamentary group. Their ability to achieve these goals is linked to their size and ability to coordinate the selection on the eligible positions (Hazan and Rahat, 2010). Making wise choices in terms of candidates' competence would be easier: "The list committee, it's what they would have chosen. 'We know him, we know him well.' But, 'we know him', that means: he is able to handle a project." (E1). They also have the power to allocate resources to achieve these goals. To boost a candidate's chances to be elected, the party in central office can increase her visibility, with the expectation that voters will follow suit and cast preference votes accordingly: "At some point you will even have to tell yourself: 'I want this one to be elected for my political work'. And we're going to make videos, we're going to make things and posters and for others not, because this one must be elected." (E1).
However, centralized selectorates are not only about officeseeking. Interestingly, interviewees also emphasize vote-seeking goals, especially in highly competitive constituencies, and their core message is that electoral lists have to succeed in getting candidates elected: "It's a like the player who gets on the pitch: he can get on, but he has to win." (P9). To achieve this goal, centralized selectors focus on balance and diversity of profiles of candidates. Central selectors can identify and recruit votesboosters like celebrity candidates, to maximize votes. The recruitment of these categories of candidates is easier for party leaders and central elites. Besides, they also favour balance in the socio-demographic composition of the list: "Each regional selectorate could select a male head of list. And then it would be nothing but a group of male MPs. So, the federal level intervenes." (E1). This is often done in coordination with more decentralized bodies, who are better positioned to draft lists that nicely meet the needs of their own constituency, and to avoid the potential drawbacks of centrally chosen candidates who would not be supported by constituency elites: "If at some point you impose on the heads of lists candidates that they don't necessarily want to have on board, that can result in problems." (M5). Finally, centralized selectorates can also lead to selection criteria that prioritize personal interests: "I think the people who were on that committee were mostly looking out for their own personal interests and not developing a collective synergy." (M7).
We expect decentralized selectorates situated at the constituency level, typically the head of list or local leaders in a committee, to prioritize vote-seeking goals and electability (H2b). These selectors might indeed above all seek to win elections and maximize their constituency party's strength. Some interviewees point to the importance of vote-attracting candidates and the added-value of decentralization. They put forward their knowledge of the local context to recruit candidates with eligible profiles: "It also allows, with successes and failures, eh, but to each one to be responsible actually, but of a responsibility, I find, which is appropriate because it is integrated in a sociology, in a geography which sometimes is deeply different [from one constituency to the other]." (P1). Eligibility also means to come up with geographically balanced lists representative of the various sublevels in the party. To assess eligibility, selectors report evaluating candidates retrospectively based on their individual electoral results. Yet many acknowledge the trade-off to be made between popular and competent candidates. While constituency organizations would like to recruit candidates meeting both requirements, the lack of ideal candidates might force them to take a popular candidate on board despite poor skills potentially harming the constituency party's reputation as well as the substantive representation of the constituency best interests by the future MP: "It's not about being a potentially excellent representative. It's not about being a certified jurist because we're going to do legislation, no. The first quality of a candidate is to be known, so sometimes you have a fool who is, who is well-known in the area where he lives." (M5). In other cases, respondents refer to the opposite choice: they selected a competent candidate despite a low electoral popularity. "Someone who was very, very good, but he doesn't get any votes. And we told ourselves: 'Well, we'll put him there, so that if we have prospects of entering the majority, we know that. . . he'll go straight up [to the Parliament].'" (E6).
Finally, we assume that larger, more inclusive selectorates prioritize policy-seeking goals and acceptability as selection criteria (H2c). We found this assertion directly in some respondents' accounts. Our interviews suggest that larger selectorates are less office-and vote-seeking than more exclusive selectorates, what could play in favour of policy-seeking objectives and prioritize aspirants' involvement in the party, and thus acceptability: "The activists, their first argument, it is always difficult to make them understand other arguments, it is the loyalty to the party. And so, we need people [candidates] who are committed, who come to the General Assembly, who are present in the party, . . ." (E7). Contrarily to more exclusive selectorates, inclusive ones are portrayed by respondents as lacking the strategic skills to help them make strategic informed choices to pursue vote-or officeseeking goals-as already reported in the literature (Kittilson, 2006;Pruysers et al., 2017). First, inclusive assemblies are said to be less able to grasp what makes a popular or competent candidate: "They managed to designate candidates who were not at all, uh, who were not made for it and where the incumbents were well aware of it but, as it is the base who decides, it is very democratic [but] we arrive at, sometimes, also bad casting." (M4). Furthermore, the coordination issue in larger selectorates is underlined: "The general result is not guaranteed, because obviously as we do, we focus on the individuals and not on the collective or collectives: it necessarily has an impact." (E5).
Yet, even inclusive selectorates are not fully exempt of focus on candidates' competence: "I think that the members have a clear and mature view, and very. . . I would say full of competence. That's not unreasonable actually. When they, in general [. . .] come with questions 'Why this person and not that one?' [. . .], these are questions we were prepared for, because we have often been faced with them in the list committee." (E2). Similarly, they also tend to focus on some aspects of profile and electability, especially when it comes to territorial balance and representation of their own local chapter on the list: "This municipal mechanism was stronger than we imagined, and we became aware of it as time went by. And so, in the end, we realized that. . .it was stuck on that side. And so you say: 'The candidate from X that we put [. . .] on place thirteen, we're going to put her eleventh.'" (E6). Inclusive assemblies often also select candidates based on their personal (lack of) acquaintances with them: "We often get reactions from an angry guy because he had an argument with someone from his chapter and says 'I don't want her, she did this, she didn't do that in her municipality, that's shameful, you are taking her!?'" (P3). This personal, proximity dimension is even a mobilizer for selectorates: "There's a bus coming to the general assembly, whose mission will be: I have to vote for candidate X, because he's my friend and that's it." (E4).
Our second hypothesis expected different selectorates to be driven by different goals and selection criteria. Our findings corroborate our expectations in that selectors point to varying preferences depending on the party body in charge of candidate selection. Our interview data allow to qualify and nuance our explorative hypotheses. More centralized selectorate do indeed benefit from more coordination power. This allows them to focus more on office and competence of candidates. Yet winning seats and office requires winning votes, and centralized selectorates often take into account electability, in coordination with decentralized constituency party bodies. They are also prone to the influence of personal interests of selectors. Decentralised selectorates prioritize winning lists in terms of the local specificities, seeking a balance between profiles, groups, and territories. Yet they cannot fully ignore competence as selection criteria, and sometimes face a tradeoff. Finally, more inclusive selectorates focus less on votes and office, partly due to the coordination issue linked to larger groups, and more on policy. They also focus more on candidates' acceptibility and favour aspirant candidates who can demonstrate an involvement toward the party.

DISCUSSION AND CONCLUSION
When parties select candidates, they have to abide the legal eligibility rules but they also set their own rules and hereby restrict the pool of potential candidates running for elections. Building on the literature on candidate selection processes and party institutionalization, this paper intended to shed light on this largely understudied aspect of selection criteria. More specifically, we linked selection criteria set by parties to their level of institutionalization (Panebianco, 1988) and the type of selectorate in charge of candidate selection. First, we expected that higher degrees of institutionalization of political parties lead to higher levels of formalization of selection criteria in the candidate selection processes. Second, we expected that different selectorates (on the inclusiveness and decentralization dimensions) have different goals (Strom, 1990), and hence different views on what makes a 'good' candidate, be it in terms of ideological, political profile or competences. We tested these expectations in a qualitative analysis of three Belgian political parties (Ecolo, PS, and MR), using party statutes retrieved in the PPDB and original interview data among 23 selectors. These three parties display rather common features of party organizational models in terms of level of institutionalization and types of selectorates . Hence, we expect our findings to travel to other contexts.
Interestingly, and contrary to our expectations, our results show that, in the three parties under study, formalism in candidate selection is not per se linked to party institutionalization, as the most institutionalized parties do not automatically strongly formalize their selection processes and criteria. What our data reveal is rather that party institutionalization interplays with the degree of centralization and inclusiveness of the selection procedure. The authority of central party structures can trigger or hinder formalism in candidate selection. A loyalty pledge is for instance officially in use in all three parties under study in this research but its usage has been unevenly mentioned by selectors, emphasizing not only the obvious relevance of examining party practices next to written rules, but the pivotal role of the party in central office in guaranteeing homogeneous selection procedures following the party official rules. When the power is decentralized, the degree of formalism weakens at least in some of the decentralized party entities caring less for the existing rules. Another major finding from our analysis is that party culture matters, and the feeling of being or not compelled by formal rules appears to be more relevant to predict the degree of formalism in candidate selection than the extent of party institutionalization. When selectors feel concerned by the existing rules, they are likely to respect them, whereas the absence of central rules might even be a source of pride for some selectors, highlighting their high degree of freedom and even self-rule. A case in point is the set of selection criteria established by the member's assembly of the Belgian francophone green party and serving as a mandate given to a list committee charged with the draft list. This list committee's members feel accountable towards the assembly to abide the criteria they collectively determined. In contrast, the Socialist party officially provides a Charter of the candidate (the loyalty pledge) to be signed by all aspirants but irregularly in use in the various decentralized party structures who hold the real selection power. We encourage party researchers to test this link between centralism and formalism in other settings to strenghten our knowledge on what stimulates or hinders the formalization of selection processes and criteria.
Regarding our second set of expectations, our results corroborate the idea that different selectorates value different goals and hence different priorities in terms of selection criteria. Centralized selectorates care primarily about offices and value competent candidates, while decentralized selectorates are more concerned with winning votes for their constituency's candidates and value electability, while more inclusive selectorates prioritize policy and acceptability. Yet these are trends rather than hermetic categories. Centralized selectorates also care about votes, and seek out the expertise of constituency bodies. Decentralized selectorates also care about competence, and inclusive selectorates also seek out balanced lists in terms of competence or profiles.
More generally, our interview data confirm the coordination issue in larger, more inclusive selectorates who are often described as lacking the strategic skills to efficiently prioritize office and votes goals, and the capacity of centralized bodies to benefit from more coordination to design and implement an informed strategy. Yet we dit not interview ordinary rank-and-file members directly. Rather, we collected information from party figures who took on responsibilities in the selection processes. Our report of the priorities of inclusive selectorates thus relies on how these party figures perceive inclusive selectorates, their priorities, and the criteria that guide their decisions in terms of candidate selection. It limits our conclusions on the preferences of inclusive selectorates. Our findings also point that centralization and inclusiveness matter more than institutionalization when it comes to selection criteria.
Lastly, our findings emphasize that the 'secret garden of politics' (Gallagher and Marsh, 1988) is definitely a black box. Even in highly institutionalized parties, a large degree of informalism prevails in the implementation of candidate selection processes. Much has yet to be uncovered in this secret garden. This study has advocated for the added value of an in-depth analysis of selectors' views on the process, beyond the formal story of candidate selection and candidacy requirements. We hope this study can inspire other works investigating different combinations of degrees of party institutionalisation and selectorates to disentangle in particular the role of central party bodies in designing the formal processes and the preferred selection criteria. Extended analyses of selector's insights in various contexts will certainly prove valuable to understand who our political elites are and how they eventually reach office.

DATA AVAILABILITY STATEMENT
The PPDB dataset is available on https://dataverse.harvard. edu/dataverse/politicalparties. The qualitative dataset presented in this article is not readily available because the interview data contain information that is hardly made anonymous. Therefore we do not publicly disclose our dataset. Requests to access the dataset should be directed to audrey.vandeleene@ugent.be.