The Legislative Recruitment of Immigrant-Origin Candidates in Germany Neutrality, Closure or Openness?

The Legislative Recruitment of Immigrant-Origin Candidates in Germany Neutrality, Closure or Openness? Sara Ceyhan Goethe-University Frankfurt ceyhan@soz.uni-frankfurt.de ECPR Joint Sessions, Pisa 24-28 April 2016 Workshop on The Representation of Citizens of Immigrant Origin in Established Democracies Please Do Not Cite without Author s Permission, Work in progress

Abstract Modern states are characterized by a growing ethnic diversity. Nonetheless, citizens of immigrant origin are still poorly represented in the majority of parliaments. Even countries like Germany which experienced major immigration early in the 1950s and 1960s cannot record an equal representation yet. Despite the pivotal role of party-internal recruitment processes in shaping who runs for election, the question how immigrant-origin candidates are recruited has been scarcely addressed so far. To shed more light on parties recruitment practices towards immigrant-origin candidates, the paper explores how German political parties recruit candidates of immigrant background in comparison to native candidates and asks whether they treat immigrant-origin candidates any differently. By using unique candidate data for the state elections in Hesse and Bavaria in 2013, in Saxony in 2014 and for the German Bundestag election in 2013, the paper provides a snapshot of parties recruitment practices and tries to assess their implications for minority representation.

Introduction As a result of migration processes, Germany is characterized by an increasing ethnic diversity. This growth does not solely date from recent immigration but it is also related to the influx of migrant workers in the 1950s and 1960s and the partial following-on of family members and descendants (Green 2004; Koopmans 1999). Although citizens of immigrant origin make up a large proportion of the German electorate, they are still poorly represented, not only in Germany but in most legislative bodies (e.g. Bird/Saalfeld/Wüst 2011; Ruedin 2009; Ruedin 2013: 39; Schönwälder 2012). While 20 percent of the residents in Germany have immigrant origins (Destatis 2013) and nine percent (Bundeswahlleiter 2013) are even eligible to vote, only 5.9 percent of the MPs in the recent German Bundestag have immigrant origins (Integration 2013). German state parliaments are no exceptions in this context but they are characterized by an even larger bias in part (Donovan 2007; Schönwälder 2013). Consequently, Germany s elected institutions are unrepresentative when it comes to ethnic diversity both, at the federal and at the state level. Faced with today s multicultural societies, political parties are exposed to new challenges. By recruiting political personnel for parliaments, they act as key mediators which are intended to translate social developments into legislative bodies. The growing ethnic diversity exerts heavy adaptational pressure (Soininen 2011: 145) on German party organizations as indicated by a number of exemplary observations. Just recently, the CDU has presented a report called My CDU 2017 which, among other claims, demands a stronger incorporation of immigrants into the party organization. In a similar vein, Nils Schmid, chairman of the SPD in Baden- Wuerttemberg, explicitly highlighted the number of immigrant-origin candidates on promising list positions when presenting the executive board s proposition for the ranking of candidates in the run-up to the Bundestag election 2013. Similarly, the working group Migration, integration and anti-racism in the party Die Linke (2013) has encouraged its regional branches to place candidates of immigrant origin on promising list positions. As indicated by the above-described examples, the underrepresentation of immigrant-origin citizens in parliaments affects political parties and it seems to fall on fertile grounds. In the 2013 Bundestag election, 37 legislators of immigrant origin (5.9 percent) were seated in parliament what was unprecedentedly high, compared to prior elections (Claro da Fonseca 2011; Integration 2013; Wüst 2011). In the previous parliamentary term for example, only 21 immigrant-origin candidates were elected (3.4 percent). Thus, the number of immigrant-origin MPs was nearly doubled. Whether this increase is also reflected in parties recruitment practices is explored more deeply in the following analysis. More specifically, the question is addressed 1

what recruitment practices towards immigrant-origin candidates are in use on the part of political parties and whether they have treated immigrant-origin candidates preferentially in their recruitment processes to foster minority representation or whether the increase has occurred independently from any specific measures in parties recruitment processes. To map what general recruitment practices towards immigrant-origin candidates can be in use, the paper introduces three different perspectives on parties recruitment processes which build the framework for the empirical analysis. These ideal-typical patterns will help assess the implications of parties recruitment practices for minority representation. The paper will proceed as follows: First, three ideal-typical recruitment patterns towards immigrant-origin candidates are introduced which could be in use on the part of political parties. After four intervention points in parties recruitment processes have been carved out by which political parties can influence candidates success in the candidate selection and a brief summary on the used data set has been given, it is tested empirically what recruitment practices are in use. Finally, the findings and their implications are critically discussed. Recruiting immigrant-origin candidates Neutrality, closure or openness? In the light of a persisting underrepresentation of immigrant-origin citizens in parliaments, parties recruitment processes increasingly came under criticism. Predominantly, they are criticized as too selective and discriminatory to produce an equal representation in parliaments (e.g. Ashe/Stewart 2012; Durose et al. 2013; Kittilson/Tate 2005). However, information on parties recruitment practices towards immigrant-origin candidates is still limited. Data constraints thwart efforts to analyze party-internal recruitment processes wherefore these have remained a secret garden of politics (Marsh/Gallagher 1988). Only a handful of studies have explicitly scrutinized party-internal recruitment practices towards candidates or MPs of immigrant-origin (Durose et al. 2013; Schmitz/Wüst 2011; Sobolewska 2013; Soininen 2011; Thrasher et al. 2013). Durose et al. (2013) for example find for the 2010 British General Election that where candidates from underrepresented groups have been successful, this was because they were acceptably different meaning that they had certain characteristics or experiences which compensated for their electoral disadvantage of being member of a marginalized group. Such candidates are more likely to take pathways into politics which emphasize university education and national politics instead of local politics: If candidates were not white, middle aged, able-bodied, heterosexual and male, then they needed to conform to the other aspects of the archetype, that is, being middle-class, university educated professionals, with previous experience (both paid and unpaid) in national politics (Durose et al. 2013: 258). In a similar vein, Soininen (2011) finds for the local candidate selection in 2

Sweden that formal and informal selection rules have disadvantaged candidates of immigrant origin. More specifically, prejudicial attitudes were assumed to prevail among local party members and voters. This has partly kept party selectorates from nominating immigrant-origin candidates. Against the prevailing view that political parties are one of the main barriers to an equal minority representation, we could in turn argue that their monopoly in the candidate selection (Hazan/Rahat 2006b; Rahat 2007) also makes them one of the main pioneers in paving ways to an equal representation. In line with this argument, Sobolewska (2013) shows for the 2010 British General Election that Labour Party and Conservatives both have employed targeted recruitment strategies to foster minority representation. For instance, they have made use of outside recruitment by fielding more party newcomers. Even though the supply of aspiring candidates of immigrant origin is in all probability below their share in society 1, political parties can use their monopoly in the candidate selection to make this recruitment stage more permeable for candidates from underrepresented groups by treating them preferentially. It implies that political parties act more benevolent towards candidates of immigrant origin by applying less demanding selection criteria and by offering more party support. By these measures political parties can downsize the selectivity of their recruitment processes and make them more permeable for underrepresented groups. As it is the political party and its selectorates which define the selection criteria in the nomination process, we can assume that it is in parties hands to employ a preferential treatment in their candidate selection. This argument is closely linked to the debate on affirmative action (e.g. Norris 1997; 2001). Krook and Norris (2014) have recommended that research on affirmative action should focus more strongly on non-quota strategies to promote descriptive representation. Especially in contexts where no quotas exist alternative strategies have to be taken into account. Otherwise, efforts of equality promotion are systematically overlooked. As no ethnic quotas (on ethnic quotas see Bird 2014) have been introduced in Germany, it is important to take alternative strategies of promotion into account. Consequently, we will put forward the question how immigrant-origin candidates are treated in parties recruitment processes. After having argued that political parties can treat immigrant-origin candidates preferentially in their recruitment processes to increase their descriptive representation, the follow-up question is why political parties should be inclined to do so. In the present paper, political parties are assumed to be rational actors which seek to maximize votes (Downs 1957). Votes have no intrinsic value to political parties but they are prerequisite to pursue underlying policy and 1 By drawing on the German dataset of the European Social Survey 2002-2012 (cumulated waves 1-6, weighted by design weight) we find that fewer immigrant-origin citizens were party members, compared to citizens without any immigrant origin. About 3.8 percent of the respondents without any immigrant origin have indicated to be member of a political party; this applies to only 2.2 percent of the citizens of immigrant origin. These findings suggest that the supply of immigrant-origin aspirants is lower compared to native aspirants. 3

office goals. In order to maximize votes, parties have to take voter markets into account and identify relevant sub-groups of voters. The immigrant-origin constituency is a growing voter group in Germany with about 5.8 million in 2013 which corresponds to a share of nine percent in the total electorate (Bundeswahlleiter 2013). As a result of their electoral impact, voters of immigrant origin came increasingly into the focus of political parties (Claro da Fonseca 2011). One major reason was the reform of the Citizenship Act in 2000. After naturalization was facilitated, a tremendous growth of the immigrant-origin constituency was expected. This expectation was reinforced by the novelty that children of foreign residents could now receive German citizenships under certain conditions (ius soli), making them eligible to vote. Before the reforms took place, political parties primarily counted on gaining immigrant-votes by ideological default, i.e. without openly addressing them and thereby risking the loss of native support (Claro da Fonseca 2011: 110). However, after the reforms the electoral weight of immigrant-origin voters has risen and as a consequence, political parties started to compete more actively for their electoral support. Running immigrant-origin candidates for election is one possible strategy to address voters of immigrant background. Political parties utilize candidates descriptive characteristics as information-short-cuts to address specific sub-groups of voters (Cutler 2002; McDermott 1997; 1998; Popkin 1994; Søberg Shugart/Valdini/Suominen 2005). For political parties, candidates characteristics are selling points on voter markets especially for voters without strong party affiliations parties candidate tableau can be decisive for their voting decisions. Previous, primarily US-based research on voting behavior indicates that minority-candidates can be beneficial to mobilize voters of the same background (Barreto 2007; Barreto/Villarreal/Woods 2005; Bergh/Bjørklund 2011; Bobo/Gilliam 1990; Matson/Fine 2006; Tate 2003; Zingher/Farrer 2014). Yet, even if political parties are willing to increase their numbers of immigrant-origin candidates they might refrain from affirmative action due to the screening function of their highly selective recruitment processes. In other cases, political parties aim at selecting the best in terms of merits and experiences, irrespective of demographic characteristics; they do not see any necessity for affirmative action towards underrepresented groups since descriptive representation has simply no relevance for them. For both reasons, political parties can desist from a preferential treatment. Parties highly selective recruitment processes serve as screening instruments to verify the suitability of potential candidates (Best/Jahr/Vogel 2011). Political parties have to find candidates who are professional and qualified enough to gain voters support and who are loyal and reliable (Müller 2000: 327-328). However, party selectorates have no perfect information whether aspiring candidates will meet these demands. By using indicators on aspirants past political performance like their political office experience or their duration of party membership, they try to reduce their information deficit and minimize their risk of making an unfavorably candidate selection. Only in exceptional cases individuals can 4

skip the tedious party-internal recruitment processes and follow the fast track into professional politics (Bailer et al. 2013; Borchert/Stolz 2003: 156; Herzog 1975; Lorenz/Micus 2009: 13; Mögel 2008; Wolf 2007; 2011). In view of this background, political parties can simply put more emphasis on political qualifications than on equal representation either, because they are unwilling to take the risk of recruiting unsuitable candidates or because they dismiss the general notion of descriptive representation. To envision how political parties can behave in this tension field between inducing a social opening of their recruitment processes for underrepresented groups and a social closure, three ideal-typical perspectives on legislative recruitment practices towards immigrant-origin candidates (neutrality, closure, openness) are carved out. These serve as analytic frames which allow drawing conclusions from the found empirical patterns. According to the neutrality perspective, recruitment profiles of immigrant-origin candidates resemble those of native candidates. Both have similar qualifications and receive a similar amount of party support in the candidate selection. Similar recruitment profiles indicate that political parties behave neutrally in their recruitment processes; they neither impede the recruitment of immigrant-origin candidates nor do they make any specific attempts to promote their recruitment. For them, candidates qualifications and experiences are decisive in the candidate selection, irrespective of descriptive characteristics. If empirical evidences point to neutral patterns, parties recruitment processes are no original impediments for immigrant-origin candidates on their pathway to parliament. Not acting in discriminatory ways will automatically result in an equal representation as soon as the supply of potential immigrant-origin aspirants rises. In the case of neutrality, the actual reasons for a poor minority representation do not lie in parties recruitment practices but have to be localized elsewhere. The closure perspective is originally rooted in the research on women representation (e.g. Anzia/Berry 2011; see Black/Erickson 2000; Black/Erickson 2003; Fulton 2012; Lawless/Pearson 2008; Milyo/Schosberg 2000). In this debate, it is argued that female candidates need to outperform their male counterparts in order to run for election. This is mainly due to an assumed gender bias in the electorate which makes party selectorates reluctant to run female candidates. Female candidates therefore must convince party selectorates by having exceptional qualifications and experiences. By the same token, party selectorates can never be fully sure how voters will react to immigrant-origin candidates. Previous research indicates that at least parts of voters discriminate against candidates of immigrant origin (Bieber 2013; Fisher et al. 2015; Stegmaier/Lewis-Beck/Smets 2013; Street 2014). Even if applicants of immigrant origin have similar resources, qualifications and experiences as native aspirants, party selectors still might have reservations when it comes to their selection due to an imputed or a real voter bias (Norris/Lovenduski 1995). As a consequence, political parties might pursue highly 5

defensive recruitment policies towards immigrant-origin candidates. These barriers can only be overcome by being exceptional to make up for negatively valued characteristics (Black 2000: 148). Such defensive recruitment policies do not necessarily originate from direct discrimination against immigrant-origin aspirants but they might express a critical attitude towards aspirants who deviate from the typical nominee, who generally is white, middle aged, able-bodied, heterosexual and male (Durose et al. 2013: 246). This reluctance can be overcome by employing stricter quality checks and by collecting more information on candidates suitability and performance (Soininen 2011: 153). In the case of closure, we would therefore expect that immigrant-origin candidates need to be better qualified than other candidates to run for election and that they receive systematically less party support in the recruitment process. If empirical evidences corroborate this recruitment pattern, parties recruitment processes are genuine impediments for an equal minority representation. In contrast to the previous perspective, the opening model suggests that nominees immigrant origins do not induce a defensive recruitment policy in terms of higher requirements and less party support but trigger a more offensive recruitment policy in terms of a preferential treatment in the candidate selection. Taking up the above-made assumption that political parties have electoral incentives to run candidates of immigrant origin in order to appeal to the growing voter group of immigrant origin (Claro da Fonseca 2011) the critical question is how political parties can recruit more candidates of immigrant origin if their supply is limited (see footnote 1). To this end, political parties can treat minority-candidates preferentially in their recruitment processes by lowering the requirements and by offering extraordinary party support. If empirical evidences reveal an opening, parties recruitment processes are as in the neutrality model no genuine obstacles for an equal minority representation. On the contrary, political parties make targeted efforts to enhance the permeability of their recruitment processes for immigrant-origin candidates. In this case, the actual reason for their parliamentary underrepresentation does not lie in parties recruitment practices but it has to be searched in a different place. In all probability, political parties do not respond in the exact same manner to the underrepresentation of immigrant-origin citizens. Following the dominant reasoning in the literature, we would expect that center-left parties are more inclined to pursue an opening of their recruitment processes than it applies to conservative or right-wing parties. A set of contributions finds that center-left parties more strongly engage in the substantive representation of minority interests than conservative or right-wing parties (Saalfeld 2011; Saalfeld/Bischof 2013; Wüst 2014). Further, center-left parties are more conducive to the descriptive representation of immigrant-origin citizens than conservative or right-wing parties (Dancygier 2013; Wüst 2011; Wüst/Saalfeld 2011). On the other hand, conservative parties are likely to face a lower supply of immigrant-origin applicants. If they strive for candidates of immigrant 6

origin for electoral reasons, they might be more dependent on a preferential treatment to bypass a low supply. For the British case, Sobolewska (2013) finds stronger indications for the first assumption. Both Labour and Conservative Party have employed strategies to increase minority representation but the Labour Party was engaged more effectively. Wüst and Schmitz (2011) find in their study on German city councils that conservative parties were more inclined to actively recruit candidates of immigrant origin, whereas in leftist parties self-recruitment prevails. To take cross-party differences into account, variances in the recruitment practices across political parties are considered in the following analysis. How to promote candidates of immigrant origin Relevant recruitment factors in Germany In the following section, we will briefly discuss relevant recruitment factors which give options for a preferential treatment or a hindrance of immigrant-origin candidates and which define the research design of the empirical analysis. First of all, the duration of party membership is critical to run for election. Entering the parliament without being a longstanding party member is rather unlikely; in most cases, a longstanding involvement as a party member is constitutive. In accordance, Best et al. (2011: 171) find that MPs are party members for 17 years in average before winning a seat in the German Bundestag (for a historical perspective see Best/Cotta 2000). By allowing aspirants of immigrant origin to take the fast track into professional politics, political parties can enlarge their candidate pools of immigrant origin. By diminishing the high requirement of a longstanding party membership and by recruiting more party newcomers from underrepresented groups, political parties can facilitate their access to candidacies (Sobolewska 2013). Second, political parties can take influence on applicants chances of selection by being particularly bolstering. In the German mixed-member-proportional system (Shugart/Wattenberg 2003), the support from state party leaderships is of special importance to run on a party list in one of the multi-member-districts. In contrast, candidates in single-member-districts are particularly dependent on the support from local party branches (Reiser 2011; Schüttemeyer 2002; Schüttemeyer/Sturm 2005). Party support can take various forms, including the recommendation of specific aspirants to party selectors or the provision of practical advices or of access to party-internal networks. To get more aspirants of immigrant origin through the candidate selection, political parties could offer special support for aspirants of immigrant origin. Third, party members generally need to have prior experiences in political offices before they can run for a legislative mandate at the state or at the federal level. While party offices are 7

virtually compulsory, prior legislative or executive offices at lower political levels are certainly supportive for legislative careers (Borchert/Zeiss 2003: 151). Accepting less attractive, mostly unpaid positions at lower political levels is an acid test whether candidates are willing to make sacrifices for their political parties in terms of volunteer time and other resources. What is more, prior political offices can help develop personal profiles within political parties, increase candidates visibility and they allow tapping valuable political resources like party-internal networks. However, having prior office experience constitutes a high hurdle in the legislative recruitment process since offices are very time-consuming. Further, to receive such offices individuals need to be party members for some time and need to be well-integrated into the party organization (Schmitz/Wüst 2011). To increase the number of immigrant-origin candidates, political parties can choose to lower these hurdles by fielding candidates with less previous office experience. Apart from running for election, the question how immigrant-origin candidates are fielded is critical for the level of minority representation. A sheer nomination does not necessarily mean that candidates have any prospects of electoral success. On the contrary, the majority of candidates run for election without any hope of winning. Parties candidate selection has a huge impact on the final composition of parliaments (Hazan/Rahat 2006a; 2006b; Katz 2001; Rahat 2007). Political parties are not only gatekeepers in the sense that they decide who is running for election but they also decide how candidates are running. This makes political parties key players for the demographic composition of parliaments: In marginal seats, who gets into parliament is determined by voters. But in safe seats with a predictable outcome the selector have the de facto power to choose the MP (Norris/Lovenduski 1995: 2). If political parties slate immigrant-origin candidates as pure sacrificial lambs (Darcy/Welch/Clark 1994: 36) to fill up unwinnable seats or to make their candidate tableau more diverse, the selection of immigrant-origin candidates does not change the demographic composition of parliaments. By running immigrant-origin candidates in particularly hopeful races political parties can actively contribute to an equal representation in parliaments. Data and research design Parties recruitment practices are difficult to study since they are not legally defined and therefore, hard to identify (Bjarnegârd 2015). Second, they take place rather veiled within political parties and therefore are described as the secret garden of politics (Marsh/Gallagher 1988). To gain deeper insight into parties recruitment practices survey data on candidates personal recruitment experiences are needed. Alternatively, asking party selectorates about the openness of parties recruitment processes is possible. However, party selectorates probably 8

hesitate to fully disclose their strategies in the candidate selection. What is more, the paper focuses on the question how immigrant-origin candidates are recruited. Parties recruitment practices reflected in candidates recruitment profiles are harder measures than selectorates perceptions. For this reason, candidate surveys are seen as the appropriate way of measuring. Researchers in the field of minority representation often struggle with a small case number. Studying the recruitment profiles of immigrant-origin candidates at the national and at the state level is a possible research design strategy to address the small n-problem. It allows pooling candidate data to end with a useful sample size. In this light, Germany offers an optimal testing ground due to its federal structure. Although it generally allows for a high permeability between the state and the national level rather limited movement occurs (Borchert 2011; Borchert/Stolz 2011a; 2011b; Stolz 2003). Instead, state parliaments have turned into a career arena in its own right (Borchert/Stolz 2011a) with similar career patterns and requirements as at the federal level. By pooling candidate survey data from the state and the federal level we can end up with a reasonable sample size of immigrant-origin candidates. We will use candidate survey data from the Hessian state election in 2013, the Bavarian state election in 2013, the Saxon state election in 2014 and the Bundestag election in 2013. The data for the Bundestag election are available from the German Candidate Study 2013 (Rattinger et al. 2014) which is part of the German Longitudinal Election Study. The state elections were mainly selected because of their timely proximity to the Bundestag election. This is important since parties recruitment practices towards immigrant-origin candidates have changed over time due to altered electoral incentives which stem from the increasing electoral power of immigrantorigin voters (Claro da Fonseca 2011). The empirical results thus provide a conclusive snapshot of parties recruitment practices. To this end, survey data from candidates of the Social Democrats (SPD), Christian Democrats and Christian Social Union (CDU/CSU), Liberals (FDP), Green Party (Die Grünen) and Left Party (Die Linke) are analyzed. While SPD, Die Grünen and Die Linke are leftist parties, FDP and CDU/CSU both are liberal conservative parties. In total, data on 1.546 candidates are available. In Hesse, 224 of 426 candidates took part in the survey (52.6 percent). In Bavaria, 339 of 842 candidates answered the survey (40.3 percent). With respect to Saxony, questionnaires from 157 of 344 candidates are available (45.6 percent). Finally, the candidate survey at the federal level was answered by 826 of overall 2.031 candidates (40.7 percent). As extensively discussed in the literature, a variety of criteria exist to define immigrant origins, ranging from biographical to visibility to name-based approaches (Bloemraad/Schönwälder 2013). In the present paper, a biographical approach is adopted, based on the German micro census (Destatis 2013). It includes citizens of immigrant-origin, irrespective of the visibility of their immigrant origins. Further, it takes into account children of immigrants. Following this 9

approach, the first migrant generation comprises persons who were born in a foreign country and immigrated to Germany within its present territorial boundaries after 1949 2. The second generation includes persons who were already born in Germany but have at least one parent belonging to the first generation. As we scrutinize whether immigrant-origin candidates are treated any differently in parties recruitment processes and follow deviant career tracks, native candidates are the reference group. They were born in Germany with a German citizenship; further, their parents were both born in Germany with German citizenships. Due to the overall small case number of respondents with immigrant origins, we cannot account for different countries of origin. Based upon the above presented criteria, 99 of the overall 1.489 3 candidates in the dataset have immigrant origins (6.7 percent). Empirical results What recruitment practices are in use? This section explores how candidates of immigrant origin have been recruited and asks whether political parties have employed any deviant recruitment strategies vis-a-vis immigrant-origin candidates. It is an attempt to map in an explorative manner what recruitment strategies towards immigrant-origin candidates are in use. The analysis builds on the above introduced recruitment factors which give options for a preferential treatment or a hindrance of immigrant-origin candidates. The general starting point for legislative careers is the long-standing engagement as a party member (Edinger 2009: 194). To lower this requirement for immigrant-origin candidates, political parties may choose to recruit more party newcomers and in this way, enlarge the pool of immigrant-origin candidates. Table 1 presents evidences on the duration of party membership in group comparison. The sample is restricted to candidates who stood for election for the very first time. Model 1 is the baseline model which tests the basic relationship between candidates immigrant origin and their duration of party membership in years. In model 2, individual-level variables are added which might confound candidates duration of party membership at the first candidacy. Since a broad literature works on party discrimination against women (Carroll/Sanbonmatsu 2013; Sanbonmatsu 2006) we will include gender as one confounding factor in our models. Further, it is controlled for age since younger candidates are more likely to be seen as inexperienced and therefore, have a lower probability to run for election. Further, candidates educational level is included since highly educated are more inclined to pursue political careers (Best/Cotta 2000). By controlling for these factors, we aim to establish to what extent candidates immigrant origins have direct impact on their duration of party membership at the first candidacy. In model 3, state and party dummies are introduced 2 This criterion is intended to rule out war-related migration flows. 3 In the case of 57 candidates the biographical data were incomplete. They are not included in the following analysis. 10

which account for the pooled character of the data set to rule out any systematic cross-state and cross-party variations. Table 1: OLS regression Duration of party membership at the first candidacy Duration of party membership (in years) (1) (2) (3) Immigrant origin -4.356 (1.42)*** -3.124 (1.31)** -2.690 (1.18)** Male 0.657 (0.71) 0.975 (0.64) Age 0.328 (0.03)*** 0.376 (0.02)*** Education 0.265 (0.28) 0.506 (0.26)* State: Hesse -0.289 (0.92) Bavaria -1.234 (0.78) Saxony -0.879 (1.07) Political party: CDU/CSU -0.062 (0.92) FDP -6.226 (0.97)*** Die Grünen -6.888 (0.89)*** Die Linke -11.625 (0.97)*** Intercept 13.009 (0.37)*** 12.629 (0.86)*** 17.035 (0.95)*** N 832 832 832 R 2 0.011 0.166 0.340 Adj. R 2 0.001 0.162 0.331 Log likelihood -3129-3058 -2961 Notes: Cell entries represent unstandardized OLS regression coefficients, with standard errors in parentheses. Dependent variable coding is continuous. References: no immigrant origin, female, mean age, low education, federal level, SPD; * p<0.1, ** p<0.05, *** p<0.01. Source: GCS 2013 and state-level candidate surveys. The estimated coefficient for candidates immigrant origin is significantly negative in the baseline model. Without the inclusion of control factors, we find strong empirical evidence that candidates of immigrant origin are party members for a shorter period of time when running for election for the very first time, compared to native candidates. The difference amounts more than four years. However, the statistical relationship in model 1 might be inflated by confounding factors. We therefore control for candidates gender, age and educational level in model 2 and in addition, introduce state and party dummies in model 3. The coefficient for candidates immigrant background still points into a negative direction although its size has declined. Immigrant-origin candidates can take a somewhat faster track to their first candidacy although the difference in years is not large enough to speak of newcomer recruitment as indicated by the high intercept. Age and education have a small, albeit significantly positive effect on the duration of party membership. Strikingly, no cross-state differences exist with respect to the length of party membership at the first candidacy. Yet, strong inter-party differences become evident. Compared to the SPD, the duration of party membership is substantially lower in the smaller political parties FDP, Die Grünen and Die Linke. This is related to their smaller recruitment pools; small political parties often find it difficult to recruit enough political personnel and therefore, are more dependent on recruiting party newcomers. Further, the party Die Linke has been founded relatively late in 2007. In contrast, no systematic difference between the two large Volksparteien SPD and CDU/CSU becomes evident. 11

0 5 Predicted probability 10 15 20 Figure 1: Predicted duration of party membership across political parties SPD CDU/CSU FDP Die Grünen Die Linke No immigrant origin Immigrant origin Notes: The figure displays predicted probabilities. The vertical lines represent 90-percent confidence intervals. Source: GCS 2013 and state-level candidate surveys. As political parties are rational actors which seek to maximize votes, they are assumed to adapt their recruitment strategies towards immigrant-origin candidates, depending on their expected electoral gains and losses. Due to the fact that conservative parties face a higher electoral risk of losing votes when selecting immigrant-origin candidates (Dancygier 2014), their recruitment channels are expected to be more closed than it is the case in center-left parties. For them, the stakes to reach out to voters of immigrant background are much higher. Building upon this reasoning, we expect to find systematic cross-party differences in the recruitment practices towards immigrant-origin candidates. Figure 1 presents group differences in the predicted duration of party membership at the first candidacy. More specifically, the predicted probabilities are based upon interaction terms between candidates immigrant origin and party dummies. By a visual inspection, we find that immigrant-origin candidates were party members for a remarkably shorter period of time when standing on the ballot paper for the SPD. The difference is significant. But still, they are by far no party newcomers. By a majority, they were party members for about 11 years before running for election for the first time. Surprisingly and against the theoretical expectations, we find that immigrant-origin candidates also seem to take a faster track to their first candidacy in the conservative party CDU/CSU. While native candidates were party members for about 17 years it was only 13 years for immigrant-origin candidates. However, the group difference fails statistical significance wherefore the statistical relationship is highly uncertain. Turning to the remaining parties, no strong differences in the duration of party membership become evident. Against the theoretical expectations, no clear party divide exists between conservative and center-left parties. Although the SPD seems to lower its membership requirements for candidates of immigrant background, no strong pattern emerges in the other center-left parties Die Grünen and Die Linke. 12

Table 2: Ordered logistic regression Importance of party support Support from state party leaderships Support from local party branches (1) (2) (1) (2) Immigrant origin -0.078 (0.31) 0.012 (0.33) -0.515 (0.29)* -0.479 (0.30) Male -0.875 (0.17)*** -0.287 (0.17)* Age 0.013 (0.01)* -0.003 (0.01) Education -0.077 (0.05) -0.024 (0.05) Incumbent -0.266 (0.25) -0.186 (0.24) Mode of candidacy Party list 0.488 (0.30) -0.251 (0.30) Dual 0.638 (0.27)** 0.180 (0.26) Duration party membership -0.000 (0.01) -0.005 (0.01) State: Hesse -0.600 (0.22)*** 0.030 (0.22) Bavaria -1.169 (0.23)*** -0.005 (0.22) Political party: CDU/CSU 0.504 (0.22)** 0.460 (0.22)** FDP 0.211 (0.25) 0.191 (0.24) Die Grünen -0.352 (0.23) -0.320 (0.22) Die Linke 0.444 (0.26)* 0.065 (0.25) Intercept 1-0.457 (0.08)*** -1.469 (0.36)*** 2.519 (0.15)*** 2.757 (0.38)*** Intercept 2 0.461 (0.08)*** -0.463 (0.36) 1.657 (0.11)*** 1.898 (0.36)*** Intercept 3 1.417 (0.10)*** 0.585 (0.36) 0.636 (0.08)*** 0.858 (0.36)** Intercept 4 2.365 (0.14)*** 1.597 (0.37)*** -0.754 (0.09)*** -0.566 (0.36) N 643 643 647 647 McFadden s pseudo R 2 0.000 0.043 0.002 0.016 Log likelihood -976-906 -941-927 Notes: Cell entries represent unstandardized regression coefficients, with standard errors in parentheses. Dependent variable coding is ordinal: not important at all (1), very important (5). References: no immigrant origin, female, mean age, low education, no incumbent, nominal, mean duration of party membership, Saxony, SPD; * p<0.1, ** p<0.05, *** p<0.01. Source: State-level candidate surveys. Another point of intervention is the provision of party support in the candidate selection. It can emanate from two party bodies which both play key roles in the candidate selection these are state party leaderships and local party branches. For the selection to party lists, the support from state leaderships is of particular importance (Schüttemeyer 2002; Schüttemeyer/Sturm 2005). In single-member-districts, the candidate selection is largely under the control of local party organizations (Reiser 2011). The dependent variable measures the reported importance of party support for the own nomination on a 5-point scale separately for state party leaderships and local party branches (1=not important at all; 5=very important). Due to its ordered nature, ordered logistic regression models are estimated. The parallel regression assumption is not violated (Williams 2006). Unfortunately, the findings only refer to state candidates since the item was not included in the national survey. In addition to the previous control factors, we will further control for incumbency status, the mode of candidacy and the duration of party membership. Incumbents are expected to receive more party support in the candidate selection than novices; their re-selection is virtually guaranteed (Hainmueller/Lutz Kern 2008; Reiser 2014). Since systematic patterns of party support across the mode of candidacy are possible, we will further introduce dummies for nominal, party list and dual nominations. The duration of party membership (centered) is included as a proxy for candidates party involvement. Candidates with a long-standing party involvement are expected to receive more party support in the candidate selection. As revealed by model 1 in the left panel of table 2 which measures the provided support by state party leaderships, no significant differences between native and 13

immigrant-origin candidates emerge. This finding remains unchanged after introducing a set of possible confounders. For immigrant-origin candidates, the support from state party leaderships was not more important than for native candidates. Apparently, state party leaderships did not play a remarkable role in the promotion of immigrant-origin candidates. Interestingly, male candidates viewed the support from state party leaderships as significantly less important for their selection, compared to female nominees. One possible explanation is that party leaderships have a keener interest to balance their party tickets and for this reason, offer more support for female applicants who still are underrepresented. Further, in comparison to pure district nominations the support from state party leaderships was significantly more important in the case of dual nominations. State party leaderships have less influence on the candidate selection in single-member-districts; this is mainly incumbent on local party organizations (Reiser 2011; Schüttemeyer 2002; Schüttemeyer/Sturm 2005). Moreover, compared to Saxony state party leaderships seem to play a less significant role in Hesse and Bavaria. This pattern can be traced back to the fact that in East Germany interventions by state party leaderships in the local candidate selection are more common than in West Germany (Reiser 2011: 251). The question arises whether the same pattern of neutrality also holds true for the support from local party branches. In the existing research, evidence is provided that local party organizations are more closed and harbor stronger reservations when it comes to the recruitment of immigrant-origin candidates (Durose et al. 2013; Soininen 2011). In line with these previous findings, candidates of immigrant origin receive significantly less support from local party branches, compared to native candidates as revealed by model 1 in the right-panel of table 2. It appears to be more difficult for immigrant-origin candidates to cultivate a support base in the local party organization than it holds true for native candidates. It might be the case that local party organizations resemble closed clubs which are less open to candidates who deviate from the average candidate type. Turning to model 2 where it is controlled for various confounding factors the estimated coefficient remains nearly unchanged but it now fails statistical significance. Although the findings suggest that candidates of immigrant origin have received less support from local party branches we cannot draw any definite conclusion here. 14

0 0 Predicted probability.2.4.6.8 Predicted probability.2.4.6.8 1 1 Figure 2: Predicted party support across political parties SPD CDU/CSU FDP Die Grünen Die Linke State party leadership No immigrant origin Immigrant origin SPD CDU/CSU FDP Die Grünen Die Linke Local party branches No immigrant origin Immigrant origin Notes: The figure displays the predicted probabilities. The vertical lines represent 90-percent confidence interval around point predictions. Source: State-level candidate surveys. This section goes on to analyze whether systematic party variances in the amount of provided support exist. Due to the small case number of immigrant-origin candidates and separation in the data set ordered logistic regression with party interaction terms cannot be conducted. As a solution, the ordinal dependent variable is transformed into a binary variable, taking on the value of one if a candidate has indicated that party support was important or even very important, whereas it takes on the value of zero if it was not important, not so important or a respondent has ticked the middle category. The left panel of figure 2 illustrates party differences in the provided support from state party leaderships, separated for immigrant-origin candidates and candidates without any immigrant background. Due to the fact that no single immigrantorigin candidate of the FDP and Die Grünen has reported that the support from state party leaderships was important for their selection, no predicted probabilities can be estimated for these cases. In the SPD, immigrant-origin candidates have a somewhat higher probability to receive specific support from state party leaderships but the group difference fails statistical significance. With regard to the conservative party CDU/CSU virtually no difference in the predicted support emerges. The support from state party leaderships seems to play a similar role for the selection of immigrant-origin and native candidates. When we take a look at Die Linke we find that immigrant-origin candidates have a higher probability to receive support from state party leaderships, compared to native candidates. But again, the predicted group difference is not significant. Overall, the group differences in predicted probabilities are not very strong. Apparently, providing more party leadership support for immigrant-origin candidates is not a prevalent party strategy to foster minority representation irrespective from the ideological orientation. A glance at the right panel of figure 2 reveals that throughout, candidates of immigrant origin have a lower predicted probability to receive specific support from local party 15

branches except of CDU/CSU where virtually no difference emerges. However, the differences in predicted probabilities are small and therefore, turn out to be insignificant. Further, no systematic cross-party variance in terms of a clear divide between center-left and liberal-conservative parties exists. Party ideology does not seem to be a critical factor which conditions the amount of party support for immigrant-origin candidates. Prior office experiences serve as acid tests to check candidates qualification to maximize votes and their reliability in terms of representing party interests (Burmeister 1993: 68-69). But prior office experiences also constitute high hurdles in the legislative recruitment process since offices are very time-consuming. To increase the number of immigrant-origin candidates, political parties can lower these hurdles by running immigrant-origin candidates with less previous office experience. To test this relationship, a political experience index is generated which measures the proportion of offices in relation to the total number of items (Giebler/Haus/Weßels 2010). 4 All offices enter into the index by equal weight although some offices might count more in the candidate selection than others. However, as weighting will remain arbitrary to some extent no weighting is applied. Due to the fact that only few candidates have held six or more offices these enter the highest category of five or more offices. The index thus is 6-point scaled and ranges from 0 to 1. To account for its bounded nature, generalized linear regression models are estimated (Papke/Wooldrid 1996). The sample is restricted to candidates who ran for election the very first time to measure their office experiences at the first candidacy. Table 3: Generalized linear regression Number of prior offices at the first candidacy Index of political offices (1) (2) (3) Immigrant origin -0.290 (0.11)*** -0.194 (0.11)* -0.194 (0.11)* Male -0.125 (0.06)** -0.078 (0.06) Age -0.014 (0.00)*** -0.011 (0.00)*** Education 0.039 (0.02)* 0.049 (0.02)** Duration of party membership 0.033 (0.00)*** 0.030 (0.00)*** State: Hesse 0.241 (0.08)*** Bavaria -0.232 (0.08)*** Saxony 0.095 (0.10) Political party: CDU/CSU 0.108 (0.08) FDP -0.218 (0.09)** Bündnis 90/Die Grünen 0.070 (0.08) Die Linke -0.068 (0.11) Intercept -0.782 (0.03)*** -0.739 (0.07)*** -0.783 (0.09)*** N 832 832 832 R 2 0.007 0.142 0.184 Log likelihood -369-360 -357 Notes: Cell entries represent unstandardized GLM regression coefficients, with robust standard errors in parentheses. Dependent variable is an index between 0 and 1. References: no immigrant origin, female, mean age, low education, mean duration of party membership, federal level, SPD; * p<0.1, ** p<0.05, *** p<0.01. Source: German candidate study and state-level candidate surveys. 4 These are: local and state party office, national party office, local legislative office, state legislative office, national legislative office, MP s employee, mayor and state executive office. 16