Does intra-party democracy affect levels of trust in parties? The cases of Belgium and Israel

Previous research has shown a steady decline of citizen’s political trust and growing skepticism towards key institutions of representative democracy. Political parties, which perform the crucial role of linking citizens to the political system, are in the eye of the storm: citizens are generally more distrusting towards parties than other social and political institutions. The relevant literature mentions that parties often implement intra-party democratization to remedy party distrust. This article examines whether democratic candidate selection processes actually affect party trust among voters. The analysis is based on the cases of Belgium and Israel, where politicians made a strong case for intra-party democracy in recent history. The results indicate that, while inclusive selectorates indeed increase trust levels, decentralization decreases trust towards parties in both countries.

The literature on democratic political attitudes is vast. Scholars have examined how country-level determinants, such as economic conditions, cultural features, and institutional structures, shape citizen's satisfaction with democracy, levels of external and internal efficacy, social capital, and political trust. They also test how individual-level characteristics, such as gender, race, or sophistication levels, affect citizens' political attitudes. This article extends the current research and examines the effect of intra-party democracy-a party level characteristic-on the trust levels in political parties. Political trust is a crucial attitude for citizens in representative democracies, as it is often considered as diffuse political support, which enhances the legitimacy and effectiveness of democratic government (Easton 1965;Mishler and Rose 2005). Trust in democratic institutions ''reflects evaluations of whether or not political authorities and institutions are performing in accordance with the normative expectations held by the public'' (Miller and Listhaug 1990, p. 358). In this paper, we adopt an institutional perspective on trust (as opposed to a cultural view) and assert, similar to Mishler and Rose (2001) that institutional trust centers on citizens' evaluation of institutions' performance.
An extensive body of literature shows a steady decline of citizens' trust over the last decades in various democratic institutions such as government, parliament, and parties (Catterberg and Moreno 2005;Klingemann 1999;Norris 1999). Indeed, the severe decrease of trust in political parties is part of a general skepticism and public disenchantment towards key institutions of representative democracy (Dalton and Weldon 2005). The decline of trust in parties was also attributed to the weak linkage between voters and their representatives, which used to be mediated by parties (Norris 1999;Dalton 1999).
Early authors established the common wisdom that political parties are the inevitable basis of democracy (Duverger 1951). Although this consensus has not been called into question by more recent scholars, it seems that parties are one of the least trusted institutions: compared to other social and political institutions, citizens are even more distrusting towards parties as they receive extremely poor ratings in the majority of international public opinion surveys (e.g., ESS, WVS, EVS). Apart from declining party support, the drop in party membership figures and campaign participation rates all point to a general erosion of partisan attachment in advanced industrial democracies (Dalton and Wattenberg 2000).
Parties have experimented with several remedies to overcome the reported partisan decline (Van Biezen et al. 2012). One of the strategies is to democratize intra-party decision-making procedures to mobilize members and citizens into the democratic process (Scarrow et al. 2000;Pilet and Cross 2014). It is often argued that party actors act to bring about the institutional rules (in this case, intra-party candidate selection procedures) that they expect to result in the most desirable outcome (Bawn 1993). Candidate lists and selection procedures could be considered the outcome of intra-party struggles taking place between intra-party factions. Of course, the extent of strategic behavior among intra-party factions varies with the constraining effects of the electoral system and existing laws on the recruitment of political elites (Boix 1999). However, the reported partisan decline functions as an external shock (Harmel and Janda 1994) which urges parties to change (i.e., democratize) their intra-party rules. With these reforms, party elites hope to improve positive attitudes towards political parties. In this context, this paper also contributes to the theoretical discussion on the determinants and consequences of institutional change, as we examine whether intra-party democracy, measured by the inclusiveness of the selectorate and decentralization of the candidate selection process, positively affects levels of trust in parties. The empirical analysis is based on the cases of Belgium and Israel, where politicians and public opinion-makers made a strong case for intra-party democracy. We find support for the hypothesis that inclusive selectorates increase levels of trust in parties. Decentralization, on the contrary, tends to decrease trust levels in both countries.
Thus, this study shows that, even in an era of partisan dealignment, the characteristics of party organizations have far-reaching consequences for the functioning of representative democracies: intra-party decision-making procedures affect citizens' attitudes towards parties as institutions in particular, and towards representative democracy in general.

Determinants of trust in parties: what about party level-predictors?
A significant segment of the literature has focused on identifying the determinants of political trust among citizens. Scholars have found several common individuallevel and contextual macro-level variables to significantly affect citizens' trust in democratic institutions in general and political parties in particular.
At the individual-level demographic variables, subjective well-being, external efficacy, support for coalition parties, and toleration of corruption have robust significant effects on political trust (Catterberg and Moreno 2005). Slomczynski and Janicka (2009) show that individuals' social stratification positions impact prodemocratic attitudes such as trust in democratic institutions, but the effect is mediated by the countries' level of economic development. Mishler and Rose (2001) conclude that individual evaluations of political and economic performance are substantially more important than socialization experience.
Nonetheless, research found that socialization and demographic variables such as age, education, gender, and church attendance often have significant, but rather weak effects in these statistical models. So although some political socialization effects might occur as well (see, for instance : Inglehart 1997), empirical findings mainly support the institutional performance model, which states that political trust is determined by the evaluation of performance by the institutions under consideration (Newton and Norris 2000). This implies that political trust is politically endogenous: institutions are able to generate trust if they perform well.
In addition to the institutional performance predictors, the effects of contextual determinants, such as the age of the democracy, its electoral system and its economic condition on political trust were analyzed (Miller and Listhaug 1990;van der Meer 2010). Van der Meer and Dekker (2011) hypothesize that country-level characteristics such as corruption, economic development, and electoral system affect trust levels, but the effects are mediated by citizens' evaluations of politics. Their model only supports the hypothesis that corruption levels negatively affect political trust, and that this macro effect is explained by micro-evaluations of competence and reliability (Van der Meer and Dekker 2011).
The empirical support for the institutional performance theory implies that even for political parties, one of the most distrusted democratic institution in public opinion surveys, not all is lost. As mentioned, political parties try to remedy party decline through democratizing their decision-making procedures. In this manner, democratizing parties fulfill one of their vital functions in representative democracies: mobilizing people to participate in the electoral process, or in this case to become more involved in intra-party politics (Dalton and Wattenberg 2000).
This suggests that intra-party democracy would generate more trust among voters: democratizing parties more strongly correspond to the expected role of parties linking citizens to the electoral process, and thus perform better than undemocratically organized political parties. Hence, citizens supporting democratically organized parties should exhibit substantially higher trust levels than citizens which support undemocratically organized parties. Yet, to date only few scholars have attempted to examine the effects of intra-party reforms on levels of trust in parties (van Haute and Gauja 2015; Bernardi et al. 2016).
It has been argued that the spreading distrust of political parties probably increases pressures for institutional changes that alter the role of parties in the democratic process (Cain et al. 2003). Dalton (2004) examined whether such electoral reforms in Japan and New Zealand restored public confidence in parties. The data did not support this argument, and it was concluded that as distrust in parties extends across all types of electoral systems and party configurations, institutional reforms will not suffice to solve this problem. We argue that the absence of a significant effect here was because Dalton examined the effect of a general electoral reform on party support. We, on the other hand, propose to study the effects of specific intra-party reforms on the level of trust in parties. In other words, we link the democratic nature of party organizations to the level of trust towards this particular democratic institution. Thus, since only rarely do scholars study the way parties shape citizens' opinion about democracy, the paper's main contribution is to direct scholarly attention to the way parties-a mesolevel characteristic-shape trust in parties.

The effect of candidate selection processes on trust in parties
We look at the nature of parties' candidate selection processes and examine whether and how it affects citizens' trust in parties. In the last few decades, we have witnessed an ever growing interest of political scientists in candidate selection procedures. Whether focusing on selection processes' determinants (Lundell 2004;Pennings and Hazan 2001;Shomer 2014) or on selection processes' effects on various political phenomena (De Luca et al. 2002;Hazan and Rahat 2010;Shomer 2009;Sieberer 2006), scholars mainly failed to address whether and how candidate selections affect citizen's political attitudes. We argue that the way parties select their lists-its degree of inclusiveness and centralization-should affect citizens' evaluation of parties themselves, and therefore, as argued by the institutional performance model, their levels of trust in them.

Defining the selectorate and decentralization dimensions
In this paper, we regard two central dimensions the literature identifies as cardinal for the study of selection processes: selectorate and level of centralization. We hypothesize that each dimension affects citizens' trust in parties differently. The first dimension-the selectorate-is the group that selects the candidates and determines who is eligible to represent the party on election-day. Similar to Rahat and Hazan (2001), we conceive the selectorate dimension to range from an exclusive pole, whereby one party leader selects to an inclusive selectorate, in which the entire electorate takes part in a party's selection process. The second dimension-level of centralization-measures the territorial and sectorial level at which the selection takes place. We focus on territorial centralization, which categorizes whether selection is done exclusively at a national level, or whether it ensures regional and/or local participation as well (Hazan 2002).
How, then, does the scope of the selectorate and the selection's centralization level affect citizens' trust in parties? Most current literature does not address this question directly (but see: Amjahad 2015 andBernardi et al. 2016). While hardly testing directly the effect of selection on trust levels, the literature does examine other political consequences of selections, findings that the answer to the question whether democratized selections are beneficial (Cross 1996) or disadvantageous (Kernell 2013) depends on the political phenomena examined and the data used. For example, with regards to citizens' opinions, it has been argued that decentralized inclusive selectorates increase participation levels (Shapira et al. 2010), enhance citizens' satisfaction with democracy (Shomer 2016), and boost legitimacy. On the other hand, some scholars claimed that democratic and decentralized selection processes bring intra-party disagreements to the forefront, consequently leading to greater distrust (Kernell 2013;Greene and Haber 2015). Our paper contributes to the existing literature on intra-party selection processes by trying to empirically resolve this controversy in the literature.
Given the controversy in the literature, we contend that democratic selectorates might either increase or decrease citizens' trust in parties, while decentralized procedures decrease it. We present two mechanisms through which larger selectorates affect citizens' trust in parties, whereby the first-procedural fairness-enhances trust levels, and the second-intra-party conflict-decreases them. We also present one mechanism, which relates selections' centralization levels to trust in parties: intra-party conflict, while differentiating pre-electoral from post-electoral periods.
The first mechanism through which larger selectorates affect trust in parties relates to procedural fairness. Literature on job applicants' perceptions of selection procedures repeatedly emphasizes the importance of fairness perception of the selection procedure itself (Gilliland 1993;Ryan and Ployhart 2000). Indeed Gilliland (1993) maintained that selection information-information provided in advance about the scoring system-affects the perceived overall fairness of the selection. In political science, it has been argued that procedural fairness positively affect citizens' attitudes towards the government (Miller and Listhaug 1999). We contend, by the same token, that procedural fairness should derive partisan trust upwards. As a larger segment of the electorate is entitled to participate in the selection process, it appears to be procedurally fair. No longer do voters perceive the selection procedure to take place at a back room, where party leaders compose the party's list at the expanse of voters' interest (De Luca et al. 2002). The inclusive process is more informative and is, therefore, perceived to be open, transparent, and fair: it allows ordinary citizens to directly participate and affect the outcome. Indeed, this was the reason some parties opted to democratize their selection processes from the get-go (Cross 1996;Scarrow 1999).
The second mechanism we present-intra-party conflict-explains a negative effect of inclusive selectorate on trust in parties. Democratic selection processes also intensify and highlight intra-party conflict. As opposed to selection via party leadership, which encourages candidates to be loyal to the leadership, selection via democratic processes incentivizes candidates to compete for the support of a large selectorate. Hence, candidates are incentivized to distinguish themselves from their co-partisan candidates (Norris 2004). Kernel (2013) finds that voters are less likely to vote for parties that allow voters, rather than party leaders, to select their candidates. She ascribes these findings to the notion that as a greater segment of the citizenry is involved in selection processes they become more aware to intra-party conflicts and factionalization, and lose support in those parties. Similarly, Greene and Haber (2015) argue that democratic selectorates enhance public's awareness of intra-party disagreements, which in turn negatively affect citizens' evaluations of parties and their tendency to vote for them.
Given the two contradictory plausible projections about selectorate's effect on trust in parties mentioned above, we hypothesize them both and allow the data to empirically determine how selectorate affect trust levels and which mechanism is at play: procedural fairness or intra-party conflict. We hypothesize that: H1a Inclusive candidate selection methods are associated with higher levels of trust in parties.
H1b Inclusive candidate selection methods are associated with lower levels of trust in parties 1 .
Intra-party conflict is also the mechanism that relates selection's centralization levels to citizens' trust in parties. Decentralized selection processes may bring to the political forefront intra-party divisions between the party's local and national levels. For example, it has been argued that there exists a tradeoff between territorial and the nationally desired social decentralization: when selections are conducted at the local level it is harder to optimize social representation, for instance, female representation (Matland and Studlar 1996). Indeed, in the Belgian case, research shows that decentralized selections do lead to more balanced territorial representation, but the latter comes at the expense of female representation (Put 2015).
Decentralized selections also strengthen intra-party divisions in the post-election period. It has been argued that parties, which employ centralized selection processes behave in a unified manner, as the party's representatives tend to toe the party line (Hix 2002;Sieberer 2006), whereas decentralization stimulates conflict since parliament members, who owe their selection to a local group, might defy the national party line and represent their selections' constituents (Benedetto and Hix 2007).
These intra-party divisions, whether during the selection stage or in the postelection period, might give parties a bad image and cause party supporters to be disgusted by the offensive competitive political process. Consequently, they may lose trust in political parties themselves. We therefore hypothesize that: H2 Decentralized candidate selections are associated with lower levels of trust in parties.

Data and method
We use data from Belgium and Israel to empirically test how levels of centralization and the scope of the selectorate affect citizens' trust in political parties. Both countries constitute a most similar research design: both use a proportional electoral system to elect their parliament, which coupled with the countries' social cleavages yield a multiparty legislature. Moreover, in both countries various parties employ divergent mechanisms to select their lists enabling variation on both the selectorate dimension, as well as the level of centralization. We take advantage of this within country cross-party variation in selection processes to examine how they affect citizens' trust in parties. By analyzing each country separately, we hold constant various macro-level determinants that were hypothesized to affect trust levels, for example, a country's electoral system. In addition to the variation in selection methods in both Israel and Belgium politicians and parties themselves publically argue in favor of democratizing selection processes, citing legitimacy crisis, and increasing citizenry trust as crucial reasons. For example, Israeli Knesset Member Ben-Ari said on June 2011 that ''only primaries will restore public's trust'' (Channel 7 2011). In Belgium, especially during the 1990s, several parties have democratized their internal decision-making procedures with the explicit goal of decreasing citizens' distrust towards political institutions (Devos and Verstraete 2002).
We use both Belgium and Israel to increase external validity and verify that the results from one country are not unique. For that reason, while we do not compare directly Israel and Belgium we chose two countries with a similar system: both use PR electoral system in a context of a highly fractionalized party system (most similar research design). While other cases that satisfy these conditions might have been considered, we could not locate public opinion data for them that examine citizens' trust in parties. Therefore, we limit our analysis to Israel and Belgium only. It is important to mention, nonetheless, that the similarity across our cases also hinder generalizability: the extent to which our results are specific to PR countries with a fractionalized party system needs to be tested in future research, by including more countries from divergent settings.
For Belgium we examine five consecutive legislative sessions from 1995 until 2014, and in Israel we examine three consecutive legislative terms from 2003 until 2013. 2 We use individual-level data to measure respondent's trust levels and other independent variables, and utilize a two-level hierarchical linear model, whereby individuals constitute the first level, and parties within a given legislative session constitute the second level. We nest individuals within parties (in a given legislative session) using their answer to a vote question. In the Israeli case we use the Israeli Democratic Institute's Democracy Index, and utilize the question: ''If the elections were held today, which party would you vote for?'' to connect respondents to parties 3 . In Belgium we use the Belgium General Election Studies for the 1995, 1999, and 2003 elections, and the European Social Survey for the 2007 and 2010 elections 4 . In both studies, the respondents were asked to answer the question: ''What party did you vote for in the last national election?'' These survey items allow us to nest respondents in political parties and link their levels of trust with intra-party candidate selection dimensions. Data on parties' selection processes were collected by the authors.

Variable operationalization
The outcome variable is a categorical variable which measures citizens' trust levels towards political parties. 5 Note that the question wording and the number of categories in the trust questions' answer differ between Israel and Belgium. While in Israel respondents were presented with four possible answers, in Belgium they were presented with five. 6 Since we do not pool the two cases together, these wording differences neither pose a serious challenge for the analysis nor do they constitute a validity concern.
The main predictors of this research refer to candidate selection processes' selectorate scope and their levels of centralization. We operationalize the selectorate using three categories, whereby 0 represents selection via a small group of party leaders, 1 represents selection via party delegates, and 2 selection via primaries (O'brien and Shomer 2013). Centralization is also a three-category variable, whereby 0 was assigned to parties that use solely nationalized procedures, 1 represent parties whose selections involved both the national as well as the local level, and 2 was assigned to parties with exclusively localized procedures (Tavits 2012). For each country we present two models, whereby the first treats each of the three-category variables as an index, and the second validates the index by treating each variable as a categorical indicator, and using treatment contrasts. We use selection via party leaders as a reference group in the selectorate treatment contrast, and selection solely at the national level as the reference group for centralization 7 .
At the party level, we control for two additional variables. First, a dummy variable indicating whether the party which the respondent voted for is a coalition (1) or an opposition party (0). Norris (1999) finds that people who support governmental parties have higher levels of institutional trust. Second, we control for the party's ideology, whereby -1 represents right parties, 0 center parties, and ?1 left parties.
The literature review demonstrated how demographic predictors have modest effects on political trust, while emphasizing that performance evaluations are more consequential. Nonetheless, we control for respondents' education levels; income levels (Slomczynski and Janicka 2009); age (while also squaring it to allow for curvilinear relationships); and gender. Data availability concerns prohibit us from including direct measures of economic and political performance. To overcome this drawback we use a proxy: satisfaction with democracy, which has often been regarded as a performance measure (Huang et al. 2008;Linde and Ekman 2003). Indeed, Klingemann (1999) used confirmatory factor analysis to find institutional trust and satisfaction with democracy to load on the same dimension he terms 5 Since the outcome variable is ordinal we also estimated a two-level ordered logit hierarchical model, for each country (see the on-line appendix). Substantive results are similar to the ones presented in the paper, especially concerning the effect of selection processes, and we therefore chose to present the more easily interpretable results. 6 See online appendix for details. 7 In Israel, we could not include treatment contrast for both selectorate and centralization, as this led to perfect multicollinearity. Therefore, we only include contrast for the selectorate variable. performance of the regime. Thus, we expect high levels of satisfaction with democracy to positively correlate with high levels of trust in parties 8 . Table 1 presents four hierarchical models, whereby the first two columns present the results for Israel and the last two present them for Belgium. As mentioned, we use both an indexed version and a categorical version of the two main predictors: selectorate and centralization. The results we obtain support the hypothesis that using inclusive selectorate positively relates to citizens trust levels in parties: citizens who vote for parties that employ primaries tend to trust them more. The positive effect of democratic selections' procedural fairness, therefore, outweighs intra-party conflict's negative effect. Intra-party divisions play, though, an important role in explaining the findings that voters, who vote for parties that employ decentralized processes, tend to exhibit lower level of trust towards parties.

Results
With regards to selectorate's effect on respondent's partisan trust levels, we see in the second column of Table 1 that Israeli voters who vote for parties that use primaries have higher trust levels by 0.123 (on a 4 point scale), than voters who vote for parties that select their lists via party leaders, all else held constant. Similarly, voters who support parties that use delegates exhibit trust levels that are higher by 0.181-a 1/5 standard deviation-than those who support parties that select in an exclusive manner. Both results are statistically significant at a conventional level. The results for the Belgian case are comparable. Voters who vote for parties that use delegates exhibit higher trust levels by 0.340 (on a 5 point scale), which are about 0.4 standard deviations. Similarly, citizens who support parties that use primaries trust them more by a magnitude of 0.354 compared to citizens who support parties that select via leaders. Once again, these results are statistically significant at the 0.05 level.
Theoretically we proposed two opposing explanatory mechanisms that connect parties' selectorates to citizens' trust-procedural fairness and intra-party conflict. We allowed the data to determine which mechanism is at work. The analysis clearly supports H1a, revealing that voters who support parties that use primaries have higher levels of partisan trust. Therefore, we conclude that either procedural fairness is the mechanism at work, or at a minimum its effect is stronger so it outweighs the negative effect of intra-party conflict.
Despite the correlation found in the table between more inclusive selectorates and trust levels, we may still wonder whether a simultaneity effect is at work: could it be that parties with voters that have high trust levels delegate decision-making powers to a wider selectorate (and not vice versa)? Accounting for the issue of timing is, therefore, important. However, since most parties in our datasets do not alter their selection processes from one election cycle to the next, we cannot use panel data to account for the timing in a systematic manner. Nonetheless we use anecdotic evidence from parties that altered their selections over the years to provide support for our argument that it is the increase in the scope of the selectorate that led to higher trust levels, and not the other way around.
In Israel, the Likud party selected its candidates for the 2006-2009 session via party's delegates, whereas prior to the 2009 elections, Likud used primaries. According to expectations, the average trust levels of Likud supporters was 1.97 in 2006, and rose to 2.14 in 2009 after the party expanded its selectorate. Conversely, when Likud decided after the 1996 elections to abandon the primaries and select in a more exclusive manner, Minister Michael Eitan from the Likud party said to Haaretz newspaper that Netanyahu should not cancel the primaries as ''it will be a no-confidence in democracy, and Likud's chairperson cannot express no-confidence in democracy'' (Verter 1997). We unfortunately do not have public opinion data from that time period.
Similarly, during 2015, the Likud contemplated replacing again its primary system. However, a group of Likud Knesset Members objected to the move, arguing that selection via primaries is a democratic act, which Likud should be proud of: ''we cannot ignore the public damage that Likud might suffer from such a move…A  Does intra-party democracy affect levels of trust in parties… decision to transform the selection process into a ''closed club'' business conveys the opposite message, saying that the party is caving into an old, non-transparent and undemocratic selection'' (Lev 2015). Trust levels of Kadima's supporters further support our argument: Kadima selected its candidates for the 2006 Knesset via the most restrictive manner, but adopted primaries prior to the 2009 election. Kadima's voters' trust level rose accordingly from an average of 1.92 in 2006 to 2.07 in 2009. Belgian parties that altered their selection also support our conclusions. Indeed in 1993, the Flemish Liberal Party adopted a new name and democratized its internal decision-making procedures, improving the party's image and increasing its popular support according to the polls (Verleden 2013). This positive effect of democratizing intra-party procedures did not go unnoticed: the Flemish Christian Democratic Party followed the example of the Liberal counterparts and democratized their candidate selections (Verleden 2013). And indeed while the CVP selected prior the 1995 elections via party delegates 9 and its supporters' average trust levels was 2.49, prior to the 1999 elections it selected via a more inclusive manner 10 and ' trust levels rose to a mean of 2.65. Likewise, when the Francophone Liberal Party (PRL) expanded its selection from an exclusive selection in 1995 to a system, in which member delegates ratify the resulting lists of candidates in 1999, its supporters' trust level increased from 2.01 in 1995 to 2.56 in 1999.
These anecdotes refute the argument that parties with high levels of partisan trust delegate selection processes to a wider selectorate, and support H1a which asserts that the expansion of the selectorate increased supporters' trust levels.
The analysis 1 also supports H2. For both Israel and Belgium, we find a negative effect of decentralization on partisan trust levels. In Israel, holding all else constant, voters who support parties that use more decentralized processes have lower trust levels by a magnitude of 0.238 than voters who support parties with centralized candidate selection. This result is statistically significant at the 0.01 level. In the Belgian case, where we can use the treatment contrast, we clearly see that the negative effect of decentralization stems from citizens who support parties that used solely localized procedures: the coefficient of local selections is -0.246, and is significant at the 0.05 level.
Our analysis also reveals that respondents who support governmental parties exhibit higher trust levels compared to respondents who support opposition parties. In three of the four models, the coalition variable's coefficient is positive and statistically significant at least at the 0.1 level.
As many scholars argue, the effect of demographic predictors on trust levels is limited, as the more important determinants of partisan trust relate to citizens' 9 The initiative to draft a model list was assigned to the bureau of the district party. Afterwards the national party board had the possibility to make a number of modifications to this list. Finally the assemblies of member delegates at the district level were expected to approve the model lists. 10 The initiative was still taken at the district level, where the majority of the district parties created an informal and highly exclusive list formation committee to coordinate the process. The first draft of the candidate list needed to be ratified by the bureau of the district party. Afterwards it was passed to the national level where the general assembly ratified the lists of all of the districts. In the final step, all members had the possibility to ratify or reject the proposed list through member polls at the district level. performance evaluations. Indeed, controlling for all other variables, and especially the satisfaction with democracy variable-our proxy for performance evaluationwe find no demographic variable with a consistent significant effect across all models.
While the effect of the demographic covariates is weak and inconclusive, the impact of satisfaction with democracy-our proxy for citizens' performance evaluation-is strong and steady. Higher levels of satisfaction with democracy are positively correlated with high levels of partisan trust. In Israel, a one unit increase in respondent's satisfaction levels increases his/her trust in parties by a magnitude of 0.169. Indeed movement on the full range of satisfaction with democracy, from respondents who are at not at all satisfied with democracy to those who are very satisfied, increases partisan trust by a magnitude of about 0.7. Belgium presents a similar picture. A one unit increase in satisfaction with democracy is associated with an increase of 0.426 units in the 5 point trust in parties scale. This means that movement on the full range of satisfaction with democracy is associated with an increase of almost two units (out of a possible five) in trust for parties.

Conclusion
Ample scholarly work has been devoted to identifying the trends in citizens' trust in political parties and the reasons for its apparent decline. Scholars often look at macro-country level-factors and individual-level determinants of partisan trust, debating whether and to what extent demographic characteristics affect citizens' trust in parties. While some found support for these measures, others argued that it is the respondent's evaluations of political and economic performance that mainly shape their rust towards political institutions. In this paper, we add to our scholarly understanding about trust in parties by focusing attention on the meso party level. We specifically examine whether and how the way parties select their candidates affects citizens' trust levels.
Parties themselves, as part of their struggle to revamp their glory days of strong partisan attachments and support, adopted democratic candidate selection processes. This was done in an effort to facilitate greater citizen's involvement in and mobilization to the democratic process, hoping it will increase citizen's efficacy and sense of belonging, and restore their trust in political parties, as they will perceive intra-party decision-making processes to be more open, transparent, and reliable. To what extent, then, do selection processes affect citizens' trust in parties? To what degree might selection procedures indeed be considered a tool parties can utilize to re-establish the electorate's confidence in them?
We theorized about the effect of candidate selections on partisan trust, while distinguishing between selections' centralization levels and the scope of the selectorate. The theoretical mechanism that links centralization levels to trustintra-party conflict-pointed to a negative relationship between decentralized selection processes and citizens' trust. To mimic the division in the literature about selection processes' consequences, we present two theoretical mechanisms that relate the scope of the selectorate to trust in parties. Procedural fairness links democratized selections to high levels of trust. As citizens view democratic selection processes as more transparent and fair, their confidence in the selection process is projected to the party as a whole. The second mechanism-intra party conflict-explains why democratic selection processes might negatively affect citizens' partisan trust. The increased size of the selectorate incentivizes candidates to compete against one another in a personalized campaign frequently besmirching and discrediting their opponents even at the expense of the collective party's reputation. Thus, democratic selectorates are associated with more intensified intraparty conflict, which should derive downwards citizens' trust in political parties. Since both explanations are plausible, we let the data determine between these two explanatory mechanisms.
We test the hypotheses using individual and party level data from Israel and Belgium, two countries which use PR systems in a highly fractionalized party system and which present a great deal of variation in their parties' selection processes. Moreover, in both countries debate over selection methods revolved partially around arguments about partisan trust. The analysis supports a positive relationship between selectorate and trust levels, and a negative relationship between decentralized selection and trust levels.
While the literature is divided concerning selection processes' consequences in general, and their effect on citizens' opinions in particular, our analysis provides clear findings in this regard: we find that inclusive selectorates increase partisan trust levels. From a theoretical perspective, these results suggest that the positive impact of democratic selections' procedural fairness outweighs intra-party conflicts' negative effect. Our results, thus, seem to extend Scarrow's argument that some procedures of intra-party democracy ''are better suited to some circumstances than to others'' (Scarrow 2005, p. 3). Our analysis, in conjuncture with the broader literature on intra-party democracy, suggests that while some consequences of intraparty democracy are negative others are positive, benefiting citizens, parties, and democracy. Our study demonstrates the latter, as it shows how democratized intraparty candidate selections enhance citizen's trust in parties. Consequently, these results challenge Katz' argument that partisan organizational re-structuring should not affect declining party membership and citizens' satisfaction with party politics (Katz 2013).
The analysis presented above is cross-sectional, and does not directly test the mechanisms through which selectorate and centralization affect trust in parties. Future research will have to directly examine, possibly through experimental manipulation, whether and to what degree respondents perceive democratic selection procedures to be fairer and more transparent compared to exclusive selections, and whether this perceived procedural fairness translates to greater trust levels. A carefully designed lab experiment might enable us to also ascertain how voters weigh the two competing mechanisms of procedural fairness and intra-party conflict as they form evaluations about parties. Furthermore, future research will have to ascertain whether our findings hold in other settings (like PR with nonfractionalized party system).
Moreover, to date, to the best of our knowledge no cross-national survey and almost no country-specific survey asked respondents directly about their opinions regarding candidate selection processes. In fact, the only survey we know that asked the general public is the Israeli National Election Study from 1996 11 . Other surveys asked party members/activists of specific parties about their opinion on democratic selection processes, for example, Bernardi et al. (2016), who study the Italian Democratic Party or Sandri and Amjahad (2015) who study the Belgian Socialist Party. It would be of great importance to systematically study citizen's views about intra-party selection processes so that we can progress our understanding about the potential multifaceted ability of primaries to amend and overcome parties' decline.
Our analysis is one of the first attempts that focuses on party level characteristics-namely intra-party candidate selection processes-and examines how they affect citizens' trust levels in parties. Future research will have to ascertain the causal mechanism and generalizability of our findings. Yet, our results suggest that by adopting inclusive candidate selection processes parties could perform the role of democratic attitude-promoters.