Introduction
Notwithstanding that ethnic minorities (henceforth: EMs) are still underrepresented, the past two decades have witnessed their increasing presence in legislative chambers across Europe. While the political left is particularly renowned for its parliamentary diversity, EM politicians’ recent emergence has been documented across the whole political spectrum (Bird et al., Reference Bird, Saalfeld and Wüst2010; Sobolewska, Reference Sobolewska2013). Simultaneously however, other studies have shown that during this same time span, entire party spectra have shifted toward more restrictive positions on immigration and multiculturalism (Green-Pedersen and Krogstrup, Reference Green-Pedersen and Krogstrup2008; Bale et al., Reference Bale, Green-Pedersen, Krouwel, Luther and Sitter2010; Alonso and Fonseca, Reference Alonso and Fonseca2012; Van Heerden et al., Reference Van Heerden, de Lange, van der Brug and Fennema2014; Akkerman, Reference Akkerman2015; Han, Reference Han2015; Odmalm and Hepburn, Reference Odmalm and Hepburn2017; Abou-Chadi and Krause, Reference Abou-Chadi and Krause2020; Dancygier and Margalit, Reference Dancygier and Margalit2020): a seeming paradox.
We seek to provide an explanation for this conundrum of increasingly restrictive party platforms coupled with ethnic diversification of parliamentary parties. Based on models of dynamic representation (e.g., Stimson et al., Reference Stimson, MacKuen and Erikson1995; Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004; Ferland, Reference Ferland2020), we theorize that if the public mood on immigration policy swings to the right, some parties become pressured to substantially shift their party platform in the same direction. Yet, due to the emergence of an anti-racism norm across Western societies (Blinder et al., Reference Blinder, Ford and Ivarsflaten2013; Sobolewska and Ford, Reference Sobolewska and Ford2020), these shifts are risky as they can estrange both EM and ethnic majority voters by inviting allegations of intolerance and racism. To explain how parties can protect themselves, we build upon Ivarsflaten’s (Reference Ivarsflaten2006) notion of reputational shield. While she shows how a party’s historic legacy can fend off accusations of racism, we link this concept to present EM representation. That is, our Reputational Shield Hypothesis posits that parties that radically shift to the right in order to converge to the shifting public mood, will simultaneously increase the descriptive representation of EM MPs to minimize backlash. The ethnic diversification of the parliamentary party thus serves as a shield with which the party can deflect accusations of having become intolerant despite shifting their policy platforms in that political direction.
Secondarily, the shield could sway ethnic majority voters who hold pro-diversity norms, and especially those for whom it is an important part of their self-identification that they do not act in a prejudicial way, as the literature shows that these voters do not vote for an anti-immigration platform partly because they cannot justify it to themselves and/or their social environment (Ivarsflaten et al., Reference Ivarsflaten, Blinder and Ford2010; Blinder et al., Reference Blinder, Ford and Ivarsflaten2013; Harteveld and Ivarsflaten, Reference Harteveld and Ivarsflaten2018). We propose that this internalized anti-prejudice norm can be offset if the party simultaneously signals its commitment towards diversity, as such parties will no longer seem to be in breach of anti-racism norms and, instead, might appear consistent with pro-diversity norms as a result of being ethnically diverse.
Altogether, this gives parties a unique strategic incentive to do both, meaning that the parties who shift to the right are the same parties who increase their numbers of EM parliamentarians. This offers a novel explanation for the simultaneous increase of EM representation and shifts to more restrictive party platforms.
While we are not the first proposing that parties may use EM politicians as reputational shields (Bird, Reference Bird2005; Sobolewska, Reference Sobolewska2013), we are the first to provide a systematic empirical test, by looking at policy shifts and descriptive representation simultaneously and across country contexts. We test our Reputational Shield Hypothesis using data from eight countries and 78 parliamentary parties for each legislative period between 1994 and 2015, applying three innovations. First, data on EM representation is from a novel dataset (Morales et al., Reference Morales, Vintila, Geese, Mügge, van der Pas and van de Wardt2017), containing information on every elected member of eight national parliaments in Europe: Belgium, France, Germany, Greece, Italy, the Netherlands, Spain, and the UK. Second, to gauge public opinion shifts on immigration policy, using Stimson’s (Reference Stimson1991, Reference Stimson2018) dyad-ratios algorithm, we created a new dataset with uninterrupted time series on the public mood comparable across countries (English et al., Reference English, Van Hauwaert, van de Wardt, Sobolewska and Morales2017). Third, contrary to the extant platform change research cited above, we distinguish measurement error from significant platform changes (Benoit et al., Reference Benoit, Laver and Mikhaylov2009; Meyer, Reference Meyer2013). In so doing, we provide robust evidence for our Reputational Shield Hypothesis: Parties that radically shift their platform to the right in response to a public mood swing toward more restrictive policies will also ethnically diversify their parliamentary party.
Theory
The descriptive representation literature has traditionally argued that the most decisive factors explaining variation in the number of EM MPs across parliaments are the openness of the political system, electoral rules, and party processes of candidate selection (Norris and Lovenduski, Reference Norris and Lovenduski1993; Bird et al., Reference Bird, Saalfeld and Wüst2010). More recently, factors such as public opinion, ideological contexts, and social networks have emerged as equally important factors (Ruedin, Reference Ruedin2009; Bloemraad, Reference Bloemraad2013; Durose et al., Reference Durose, Richardson, Combs, Eason and Gains2013; English, Reference English2019; Tolley, Reference Tolley2019). Within this larger literature, our study speaks to the research on electoral incentives to promote diversity. To date, studies here have mostly focussed on how EM candidates can be a strategic resource to attract EM voters in particular (Bergh and Bjørklund, Reference Bergh and Bjørklund2011; Martin, Reference Martin2015). Yet, recent research also shows how they can be an asset or liability to attract the wider electorate (Bloemraad and Schönwälder, Reference Bloemraad and Schönwälder2013; Sobolewska, Reference Sobolewska2013; Stegmaier et al., Reference Stegmaier, Lewis-Beck and Smets2013; Street, Reference Street2014; Dancygier et al., Reference Dancygier, Lindgren, Oskarsson and Vernby2015; Fisher et al., Reference Fisher, Heath, Sanders and Sobolewska2015; Thrasher et al., Reference Thrasher, Borisyuk, Rallings and Webber2017; Portmann and Stojanović, Reference Portmann and Stojanović2019; Martin and Blinder, Reference Martin and Blinder2021).
Building on these contributions, we propose that parties’ efforts to increase descriptive representation of EMs can form part of a broader electoral strategy. That is, parties will simultaneously adjust their policy platform to move towards the median voter on immigration as and when they move to the right on that issue, but will also offset this policy adjustment with an increase in visible party diversity. We propose that parties do so to seek to avoid blame from existing voters or party members, and may use EM politicians to legitimize the shift and signal commitment to anti-prejudice norms to attract (or maintain support from) ethnic majority voters. This is not to say that parties only recruit EM politicians for this purpose; we also discuss alternative explanations, like ethnic voting, throughout.
Blame avoidance
Large policy shifts are not without costs in general (Adams et al., Reference Adams, Clark, Ezrow and Glasgow2006). They can estrange the party’s current base, and they may be perceived as unreliable by the wider public (Downs, Reference Downs1957). Given the intrinsically divisive nature of immigration issues, voter blame and punishment might be an even a bigger worry for parties attempting to radically move to the right on this issue. They run the risk of accusations of intolerance or perhaps even racism from political commentators and/or political opponents. These accusations would be highly problematic since people are sensitive to such accusations (Gest, Reference Gest2016; Sobolewska and Ford, Reference Sobolewska and Ford2020). Social acceptance of an anti-racism norm is clearly rising (Blinder et al., Reference Blinder, Ford and Ivarsflaten2013; Sobolewska and Ford, Reference Sobolewska and Ford2020). Hence, voters with ‘progressive’ perspectives on the issue – either among the party’s realized or potential support base – might withdraw their support, or become even less likely to support them, based on such criticisms. Societal blame is not only a serious risk in case of right-wing shifts by mainstream right or radical right parties (e.g., Bale, Reference Bale2008). Arguably, left-wing parties like social democrats would be even more susceptible to criticism, because they are historically associated with multiculturalism and permissive immigration policies (Bale et al., Reference Bale, Green-Pedersen, Krouwel, Luther and Sitter2010). However, for all parties, a significant right turn implies electoral risks despite any median voter calculations.
We know that under such circumstances parties will resort to blame avoidance strategies (e.g., Weaver, Reference Weaver1986). We propose that one such blame avoidance strategy is that by bringing more EM MPs to parliament, parties can avoid blame – or at least to some extent, shield themselves – from criticisms from other parties, their own members, political commentators, and the public, of intolerance on issues of race and diversity. This idea is inspired by the reputational shield concept, originally introduced by Ivarsflaten (Reference Ivarsflaten2006) of how a party’s ideological imprint can protect it against allegations of racism and intolerance when it has a restrictive platform on immigration issues. For example, a party with agrarian origins that adopts an anti-immigrant platform would be better protected against accusations of extremism than one with a fascist past. Similarly, bringing EM politicians to parliament could also be an effective blame-avoidance strategy. And since it is easier for the party to control this, rather than their ideological past, it is a viable strategy for many more parties than historical legacies would be.
Legitimization effect
Apart from mere blame avoidance, EM politicians could play an active role in legitimizing the policy shift. For instance, the involvement of women representatives in the adoption of anti-feminist policies increased the perceived legitimacy of these decisions both among women and men – but especially among men (Clayton et al., Reference Clayton, O’Brien and Piscopo2019). Likewise, an active role of EMs should enhance the substantive legitimacy of a more restrictive party platform both among EM and majority voters, as their presence signals that EMs were involved in deliberations. There are plenty of examples of this. Ayaan Hirsi-Ali, a former MP of Somalian origin from the Dutch liberal-conservative People’s Party for Freedom and Democracy (VVD), linked Islam with violence against women. This rendered greater legitimacy to the VVD’s policy turn to the right in the early 2000s (Van Kersbergen and Krouwel, Reference Van Kersbergen and Krouwel2008). More recently, in the UK, a prominent Labour politician of Pakistani background, Sadiq Kahn, pleaded for tougher integration policy to help defend the party’s tougher platform under Miliband. These examples illustrate Ivarsflaten’s (Reference Ivarsflaten2006: 6) point that it is not only the message that matters - the actor delivering it can make a crucial difference.
Offsetting the anti-racism norm
Finally, the reputational shield strategy can also be seen as a potent electoral strategy where a party wins new voter groups. That descriptive representatives might be a vote-winning strategy has been shown to hold for women (Weeks et al., Reference Weeks, Meguid, Kittilson and Coffé2023) and has been the dominant explanation of EM representation (Farrer and Zingher, Reference Farrer and Zingher2018). However, given the relatively small proportions of EM voters (as opposed to residents) among most European electorates, this explanation is unlikely to be the only electoral gain the parties will aim for.Footnote 1 Therefore, in addition to their importance to minority voters, we propose that EM MPs can also be used to offset the anti-racism norm that has to a large extent been internalized by Western European ethnic majority voters (Blinder et al., Reference Blinder, Ford and Ivarsflaten2013; Sobolewska and Ford, Reference Sobolewska and Ford2020). People- especially women- self-police any negative racial bias that they might hold to the extent that they may not actually exhibit or act upon these views (Ivarsflaten et al., Reference Ivarsflaten, Blinder and Ford2010; Blinder, Ford and Ivarsflaten, Reference Blinder, Ford and Ivarsflaten2013; Harteveld and Ivarsflaten, Reference Harteveld and Ivarsflaten2018). Yet, if political parties adopt a more restrictive platform on immigration issues while simultaneously signaling their commitment towards diversity, these voters could justify supporting such parties to themselves and their social environments.
This logic has been offered as a possible explanation as to why right-wing parties increase EM MPs in their parliamentary parties (Bird, Reference Bird2005; Sobolewska, Reference Sobolewska2013). We argue that this applies across the whole political spectrum: any party deciding to radically shift its position to the right in response to a median voter shift, and particularly left-wing parties that lack ownership on restrictive party platforms, can only electorally benefit from this shift if they simultaneously manage to offset the anti-racism norm prevalent among their (potential) electorate.
There could also be some electoral risks to the reputational shield strategy. It requires recruiting politicians who may be unpopular with voters, as some studies suggest (Stegmaier et al., Reference Stegmaier, Lewis-Beck and Smets2013; Martin, Reference Martin2015; Portmann and Stojanović, Reference Portmann and Stojanović2019; Martin and Blinder, Reference Martin and Blinder2021). However, other studies only find such ‘ethnic penalties’ against particularly disliked groups (Fisher et al., Reference Fisher, Heath, Sanders and Sobolewska2015; Martin and Blinder, Reference Martin and Blinder2021), or refute such effects altogether (Brouard and Tiberj, Reference Brouard and Tiberj2010; Street, Reference Street2014; Dancygier et al., Reference Dancygier, Lindgren, Oskarsson and Vernby2015). Moreover, there are also electoral strategies available to circumvent potential electoral penalties. For example, English (Reference English2022) shows that British parties exclude minority candidates in specific areas where public opinion is less liberal. Further, Sobolewska (Reference Sobolewska2013) shows that parties seeking to strategically diversify place their minority candidates in their ‘safe’ seats where the risks from any electoral penalty are negligible.
Role of the policy mood
It is a crucial part of our explanatory model that political parties will use the representational shield strategy to keep pace with changes in public opinion. Parties performing an arbitrary and unprovoked move to the right, and not one forced by pressure from the electorate, are unlikely to worry about either of the three mechanisms: blame avoidance, legitimization or persuading voters with an internalized anti-racism norm. Such parties are more likely to pursue other aims than vote maximization by following the median voter, and thus, the dual reputational shield strategy will not be their obvious choice. Parties that substantially shift to the right regardless of public opinion movements, may be already extremely punitive on issues of immigration and diversity, and as such, the shielding mechanisms would not suffice: these parties are unlikely to be able to recruit voters with internalized anti-racism norms whatever the ethnic diversity of their politicians. Finally, parties that aim to win EM voters as their primary source of a new electorate, are likely to be recruiting EM politicians regardless of their policy platform responses to public opinion.
Taken together, we expect that when aggregate public opinion on immigration shifts to the right, some political parties become incentivized to shift their policy platforms in the same direction, that is, to the right. As elaborated below, this does not apply to all parties: parties can also decide to hold their position or even move to the left. However, and this is our hypothesis, parties that do respond with a rightward policy shift, will also increase their representation of EM MPs.
HYPOTHESIS 1: Reputational Shield Hypothesis. Parties shifting to the right in response to a public immigration mood shift to the right are likely to simultaneously increase their number of EM MPs.
Dynamic representation mechanism
Before testing this hypothesis, we further discuss the role of public mood shifts within our theory and examine which parties become pressured to shift their platform to the right.
In line with the dynamic representation model (e.g., Stimson et al., Reference Stimson, MacKuen and Erikson1995; McDonald and Budge, Reference McDonald and Budge2005) and spatial models of elections (e.g., Downs, Reference Downs1957), public opinion shifts cause party policy change. Spatial theory posits that voter preferences and the policies offered by parties can be mapped in a dimensional space. Assuming that people vote for the party that stands ideologically closest to them, parties have an incentive to react to aggregate changes in public opinion. This has been substantiated in many different policy areas (for a review see Adams, Reference Adams2012). In 21st-century politics, with the huge amount of voter information available to parties through polling and the 24-hour news cycle, parties noticing and responding to public opinion change becomes even more plausible (Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004: 608).
That said, not every party has an incentive to shift its platform in response to a public opinion shift. This depends on where it stands vis-à-vis public opinion (Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004; Ferland, Reference Ferland2020). Only when the shift is such that public opinion is moving away from a given party, it can be considered disadvantageous for them, and necessitate a response (Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004). Hence, nationalist parties that already offer the most restrictive platform on immigration should be more likely to stay put if the public mood swings to the right and in their direction – unless, of course, they are less incentivized by votes and have a preferred policy position (e.g., due to a more radical position of their core supporters) even further to the right (Ferland, Reference Ferland2020: 5). In that case, the mood shift ensures that they can make this move without electoral losses (if the shift is such that the distance to the median voter stays constant).
Hence, parties renowned for pro-immigrant policies have the clearest incentive to shift to the right if the mood goes further right, as they clearly stand on the wrong side of the mood. Only a purely policy-seeking disadvantaged party will hold its position. And a significant shift in the opposite direction to the mood shift seems out of question for a disadvantaged party.
Which party families would be ‘disadvantaged’ by a public opinion shift to the right, and have an electoral incentive to shift their platform to the right? In Fig. 1 we depict the parties’ platforms by party family as recorded by the CMP. It may appear surprising that ethnic-regionalist parties have by far the most progressive stance. As discussed below, however, we are confined to party positions on multiculturalism. Hence, these policies are not only in the interest of immigrant-origin minorities but also of ethno-regional minorities without a recent immigration history. For the remaining party families, there are no surprises: green, social democrat and communist parties are more multiculturalist than liberal, conservative and Christian democrat parties. Yet, except for ethnic-regionalist parties and nationalist parties, differences are not that large.
Figure 1. Party policy positions on multiculturalism by party family. Triangles denote the mean position of the party family, while squares and diamonds reflect the family’s 25th and 75th percentile. The CMP estimates are rescaled following recommendations from Lowe et al. (Reference Lowe, Benoit, Mikhaylov and Laver2011). The dashed line depicts the median platform of all parties in our sample.
The average positions of the other party families are not that far removed from the median party position in our database (dashed line). This supports studies documenting that the party spectrum has converged and moved in the direction of the radical right (Odmalm and Hepburn, 2017).
Despite the relatively small differences, ethnic-regional, green, communist and social democrat parties are still more multiculturalist than the other party families in Fig. 1. Hence, based on the disadvantaged parties model (Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004; Ferland, Reference Ferland2020), these families should be more likely to shift towards monoculturalism if the mood shifts to the right, and thus, most likely away from them. That social democratic parties would consider such a move has already been established by the literature cited earlier. Perhaps it is more controversial that we also expect green, ethnic-regionalist and communist parties to potentially move to the right. Yet, we see no reason to a-priori assume that green parties and other niche parties would be immune to median voter shifts. Based on Ferland (Reference Ferland2020), only a disadvantaged party that is purely policy-seeking would hold its position; hence, a green party that also values votes could certainly consider moderating its support for multiculturalism if the median voter moves to the right.
Despite these predictions, it has not yet been empirically proved that parties systematically shift their position in response to public opinion shifts on the specific issue of immigration. Extant studies mainly consider other antecedents of immigration platform change (Bale et al., Reference Bale, Green-Pedersen, Krouwel, Luther and Sitter2010; Van Spanje, Reference Van Spanje2010; Alonso and Fonseca, Reference Alonso and Fonseca2012; Han, Reference Han2015; Abou-Chadi and Krause, Reference Abou-Chadi and Krause2020; Dancygier and Margalit, Reference Dancygier and Margalit2020). And the only two studies focussing on public opinion shifts report non-findings (Han, Reference Han2015; O’Grady and Abou-Chadi, Reference O’Grady and Abou-Chadi2019).
Still, based on the disadvantaged parties model, we expect party families usually associated with permissive policies (e.g., social democrats and green parties) to be more likely to respond with a significant right-wing platform shift if the public mood shifts toward more restrictive positions.
HYPOTHESIS 2: Disadvantaged Parties Hypothesis: Ethnic-regional, communist, green and social democrat parties are more likely to shift towards monoculturalism when the public immigration mood shifts right.
This is not to say that the reputational shield strategy is by definition the exclusive domain of these disadvantaged party families. Any party that drastically shifts to the right to adapt to a median voter’s shift to the right is incentivized to combine this with strategic ethnic diversification of the party; even if it belongs to a party family that does not systematically respond in this way (this is what we test with the Reputational Shield Hypothesis). We are simply proposing that disadvantaged parties are the ones most likely in need of this electoral strategy, as a public mood swing to the right will put them in a more disadvantaged position than other party families.
Data, methods and operationalizations
We test our hypotheses on a cross-national dataset of eight European countries and 78 parliamentary parties, for each legislative period between 1994 and 2015. Table A2 of the Supporting Information (SI) depicts the case selection.
Variables
The dependent variable testing our core hypothesis (HYPOTHESIS 1) captures whether a party has changed its number of EM MPs (0 = increase, 1 = hold, 2 = decrease) between legislative period t and t − 1. Here we rely on the new database (Morales et al., Reference Morales, Vintila, Geese, Mügge, van der Pas and van de Wardt2017). Information on MPs comes from parliamentary websites and speeches, MPs’ personal websites, party websites, Wikipedia entries, newspaper articles and interviews, Facebook pages and LinkedIn profiles. EM MPs are operationalized as those who could be perceived by voters as such based on ‘visible’ traits. This can include perceptions of ‘non-whiteness’ and perceptions of ‘foreignness’ due to their names and physical appearance. When our coders answered this question with a “yes”, or “ambiguous”, we classified the MP as EM.
We converted the EM representational shifts into three categories (see above), subsuming all increases or decreases in the category ‘decrease’ or ‘increase’. We do not focus on the magnitude of change because here we have hardly any variation to explain. As shown in Figure A6 of the SI, in 48 % of cases parties make no change; in 12 % of cases they decrease their EM MPs by one; and in 13% of cases we witness an increase by one. 73% of cases fall within a range of −1, 0 or +1 EM MP. Besides having little variation, the magnitude of the shift should be largely a function of party size: a party with more parliamentary seats to assign will likely diversify with bigger upward shifts (and may have a bigger pool of candidates) than a party with very few seats. Therefore, we rely on a categorical dependent variable.
Looking closer at parties that increase their EM MPs, 49% of those cases have an increase of one MP and in 21% of cases, an increase of two. Thus, 70% of increases are of no more than two MPs. However, the average total number of EM MPs within parliamentary parties that made an increase of one or two MPs is 2.7 (range between one and 15) and 3.9 (range between two and 10), respectively. In 44% of cases where parties increase EM MPs by one, this was also their only EM MP. All of this shows that these apparently small increases are symbolically impactful within most parties. The examples given in the theory section also illustrate how individual EM MPs can have a large impact on the image of the party, suggesting that a potent reputational shield does not necessarily require high numbers of EM politicians. Critical acts by one individual could suffice.
Our first key independent variable captures a party’s platform shift between election t and t − 1. For data we rely on the CMP (Volkens et al., Reference Volkens, Krause, Lehmann, Matthieß, Merz, Regel and Weßels2019), which limits us to only consider a party’s platform on multiculturalism defined as “favourable mentions of cultural diversity and cultural plurality within domestic societies. [It] may include the preservation of autonomy of religious, linguistic heritages within the country including special educational provisions” (Volkens et al., Reference Volkens, Krause, Lehmann, Matthieß, Merz, Regel and Weßels2019: 20). Ideally, we would also have included the party’s position on immigration; namely, whether the party wants to allow more or less new immigrants. While we know that the CMP’s subcategories per601_2 and per601_2 tap into this, these categories are only occasionally available and offer no solution.
We also know that Lehmann and Zobel (Reference Lehmann and Zobel2018) and Dancygier and Margalit (Reference Dancygier and Margalit2020) have recently introduced two new datasets on party positions on immigration. However, the first of these studies only includes three of our eight countries and focuses on a smaller time span, while the second includes six of our eight countries and focuses only on a smaller sample of parties within each country. Hence, these datasets offer no viable alternatives. This is somewhat unfortunate because, as explained in more detail below, our public mood measure does contain items on immigration. That said, the data of Lehmann and Zobel (Reference Lehmann and Zobel2018) and Dancygier and Margalit (Reference Dancygier and Margalit2020) reveal relatively high correlations between the parties’ positions on immigration and multiculturalism of 0.6 and 0.65, respectively. Hence, we can safely assume that it is valid to relate our public mood measure that also includes questions on immigration to a party platform measure exclusively tapping into a party’s stance on multiculturalist policies.
Our platform change measure aims at distinguishing significant from non-significant shifts and is calculated in three steps. First, based on per607 (=favorable mentions of cultural diversity and multiculturalist policies) and per608 (=negative mentions) of the CMP, we first calculate the parties’ positions. Rather than simply subtracting the percentages of positive and negative quasi-sentences, we follow Lowe et al. (Reference Lowe, Benoit, Mikhaylov and Laver2011) and convert these quantities into a logit scale. Second, utilizing the method proposed by Benoit et al. (Reference Benoit, Laver and Mikhaylov2009), we calculated standard errors around these estimates utilizing bootstrapping. This method assumes that text authorship as well as the parsing and coding of these texts by human coders is a stochastic process. Third, again following Benoit et al. (Reference Benoit, Laver and Mikhaylov2009: 504), we compute a measure of change by comparing a given party’s platform between two consecutive elections. We consider a party as moving to the left (right) only if the 95% confidence interval of its multiculturalism platform at t is further to the left (right) than its confidence interval at t − 1. Our approach’s important strength is that we thus distinguish real platform change from noise due to measurement error. Another, theoretical reason underlying our approach is that parties should not require a reputational shield if their platform does not significantly differ from their platform in the previous election.
While this approach’s validity has been documented elsewhere (Benoit et al., Reference Benoit, Laver and Mikhaylov2009), we can cast some light on its face validity here. For instance, the seven occurrences of significant party platform change towards monoculturalism that we observe in the Dutch 2003 elections correspond well with descriptions of how the Dutch spectrum has shifted toward monoculturalism in the aftermath of the electoral rise of the radical right ‘Pim Fortuyn List’ (Van Kersbergen and Krouwel, Reference Van Kersbergen and Krouwel2008). In Section 1.5 of the SI, we provide a qualitative illustration of what significant platform change looks like by comparing the VVD’s statements in their 2002 and 2003 programs, respectively. For the full distribution of the platform change variable by party family, we refer to Figure A1 of the SI. As shown, we witness significant right-wing moves within each party family.
Our second key independent variable gauges public mood changes on immigration policy between election t − 2 and t − 1. This is a new measure (English et al., Reference English, Van Hauwaert, van de Wardt, Sobolewska and Morales2017). First, we gathered time-series data on immigration attitudes from high-quality cross-national survey sources. All questions and sources are listed in Table A8 of the SI. Second, we harmonized these questions by recording the proportion of respondents providing a conservative (or more restrictive) response to that question (of all non-neutral, valid answers). Third, we used Stimson’s (Reference Stimson1991) dyad-ratios algorithm, a process similar to a factor analysis, to produce a single, smoothed estimate of public mood over time. Hence, this allows combining time series relating to different aspects of the broader immigration issue that may also cover different time spans. This is a well-established practice within existing literature (e.g., Ford et al., Reference Ford, Jennings and Somerville2015). Lastly, we note that most topics included in our mood measure overlap quite well with the CMP’s coding scheme from which we derive party platform shifts. As shown in Table A8 of the SI, they cover perceptions vis-a-vis the impact of cultural diversity on the country’s society. Yet, as we already acknowledged above, there are also items on the desired numbers of new immigrants (see “NUMBERS”), a policy domain not covered by the CMP.
We used lagged shifts in the public opinion mood for two reasons. First, writing and changing an election manifesto already takes place over a time of two or three years (Adams and Somer-Topcu, Reference Adams and Somer-Topcu2009: 832). Similarly, it may take time before a party wishing to diversify can actually do so. Not every party will have a readily available pool of EM candidates at its disposal (Norris and Lovenduski, Reference Norris and Lovenduski1993; Sobolewska, Reference Sobolewska2013). Second, we lag the mood shifts to mitigate endogeneity concerns (Stimson et al., Reference Stimson, MacKuen and Erikson1995; also see Han, Reference Han2015).
Controls
We control for various party characteristics, foremost, party family based on the CMP (Volkens et al., Reference Volkens, Krause, Lehmann, Matthieß, Merz, Regel and Weßels2019). Yet, in some instances we disagreed with the classification and made revisions (see Table A2 of the SI). Additionally, we consider whether a party is an incumbent or opposition party. Opposition parties may be likelier to adapt (to diversify their parliamentary party), as they find themselves in the domain of losses (Somer-Topcu, Reference Somer-Topcu2009; but see Schumacher et al., Reference Schumacher, Van de Wardt, Vis and Klitgaard2015). Additionally, we include the number of seats won or lost after election t. Parties that gained seats have more resources to diversify.
Next, we consider electoral and societal changes. First, we control for the radical right’s electoral success, since parties could take this success as a proxy that they had best exercise caution with diversification policies (e.g., Geddes, Reference Geddes and Saggar1998). Second, a higher number of asylum applications at t − 1 could politicize the immigration issue, which, in turn, may have consequences for EM representation. Asylum application data is derived from the OECD Stats database (2020). Third, we also include a dummy variable which switches to one after the 9/11 terrorist attacks have taken place, a major driver behind Europe’s assimilationist turn (e.g., Bauböck, Reference Bauböck2002). Fourth, we tap into ethnic voting by including the number of people that newly acquired citizenship at t − 1. To gauge their electoral potential, we divide this number by the total population eligible to vote.
Finally, we consider the number of days elapsed between two elections. When elections follow each other closely, parties have limited time to diversify. The descriptive statistics, operationalizations and model specifications are shown in Table A1 and A3 of the SI.
Statistical approach
Given our categorical dependent variable, HYPOTHESIS 1 is tested with a multinomial regression where we explain whether parties will increase or decrease their EM MPs as compared to staying put. To test HYPOTHESIS 2, party platform change (see above) becomes the multinomial dependent variable and the mood shifts are interacted with the party family.
To control for observed and unobserved heterogeneity between countries (e.g., the electoral system or particularities of the immigration regime), we add country fixed effects. And because we have repeated observations within parties, we specify clustered standard errors at the party-level.Footnote 2
Results
Figure 2 depicts how the public immigration mood has evolved in the countries studied. Higher values indicate a more negative public mood. Notwithstanding the different trends, each country has experienced periods where the public mood clearly became more negative (see the upward movements). This suggests that parties have been incentivized to adopt more restrictive positions, triggering the need for reputational shield tactics.
Figure 2. The public mood towards immigration.
Source: own calculations.
Do parties increase the representation of ethnic minorities as a reputational shield?
The multinomial logistic regression coefficients are in Table A4 of the SI. To test our hypotheses we need to calculate the predicted probabilities. Figure 3 depicts the difference in the predicted probability of parties to increase, decrease or hold their number of EM MPs for parties that either moved their platform to the left, right or that held their position in case the lagged public mood shift increases along its interquartile range (IQR): from its 25th to it 75th percentile. Hence, we compare a situation where the mood has shifted 3.15 units to the left (the 25th percentile) between t − 2 and t − 1 to a situation where it has shifted + 6.49 units to the right (the 75th percentile). In absolute terms: a shift of 9.64 units in the rightward direction.
Figure 3. Predicted changes in the probability that a party will increase, decrease or hold its number of EM MPs (x-axis) if we move from a situation where the mood shifted 3.15 units to the left (25th percentile) to one where it shifted 6.49 units to the right (75th percentile) (x-axis) conditional upon the platform shift made by the party (y-axis). Pseudo R 2 = 0.35, N = 238. 90% (95%) confidence interval inside (outside) brackets. The regression coefficients for the full models are in Table A4 (Model 2) of the SI.
As shown, the Reputational Shield Hypothesis (HYPOTHESIS 1) is clearly confirmed. Parties that shifted to monoculturalism in the situation of a substantial public turn (an IQR increase could be seen as such) toward more conservative policies are significantly more likely to bring more EM politicians into parliament. Specifically, the probability that they will increase their number of EM MPs increases with a probability of 0.24 (on a scale from zero to one). This is a sizeable effect. In Section 2.2 of the SI, we show that in absolute numbers, these increases on average correspond with one additional EM MP. This is consistent with the discussion above that the reputational shield strategy usually does not require large numbers of EM MPs.
We move on to our second, auxiliary ‘Disadvantaged Parties Hypothesis’. Table A5 of the SI again depicts the regression coefficients. Based on these coefficients, Figure 4 depicts the difference in the predicted probability of each type of platform shift if we increase the lagged mood shift variable along its IQR. As shown, in line with HYPOTHESIS 2, disadvantaged parties – the party families most in favor of multiculturalism – are the most likely to respond by significantly shifting towards monoculturalism if the public makes this IQR move to the right. For ethnic-regional, green and social democrat parties, the effect significantly differs from zero at the 95 confidence level, while the effect for communist parties is significant at the .90 level. Notably, the effect for green parties is the largest. To gain an impression of the effect size: if we increase the lagged public mood shift variable along its IQR, the probability (between zero and one) that a green party moves toward monoculturalism increases by 0.27. Hence, also HYPOTHESIS 2 is confirmed by the data.
Figure 4. Predicted changes in the probability of a party engaging in a particular platform change type (x-axis) if we move from a situation where the mood shifted 3.15 units to the left (25th percentile) to one where it shifted 6.49 units to the right (75th percentile). Pseudo R 2 = 0.26, N = 241. 90% (95%) confidence interval inside (outside) brackets. The regression coefficients for the full models are in Table A5 of the SI.
Together these analyzes suggest that parties that significantly shift their platform to the right in response to a rightward public mood swing, are more likely to concurrently increase EM representation. While this is regardless of party family (we control for this), our evidence for HYPOTHESIS 2 suggests that disadvantaged parties may be overrepresented among those in need of this strategy. In fact, Table A7 of the SI presents the full list of parties that engaged in reputational shield tactics in response to rightward public mood shifts: 17 cases in total. The social democrats are clearly overrepresented (6 cases). Together with the other disadvantaged party families, ethnic-regionalist ecologist, and communist parties, they constitute 12 (out of 17) cases. Hence, reputational shield tactics is mostly a strategy of the disadvantaged, more left-wing parties. Still, we also observe several liberal and one Christian democrat using it. That there are no cases of nationalist parties on the list is likely due to the fact that it makes little sense for them to make a significant shift to the right in response to a right-wing mood shift. Figure 4 indeed shows that they are significantly more likely to stay put if this happens. Hence, reputational shield tactics are mainly a resort for parties clearly disadvantaged by such shifts.
Regarding the controls (see Table A4 of the SI), parties gaining seats are more likely to increase EM representation. This makes sense, as their increased representation less likely comes at the expense of its extant politicians. Also we find that (regardless of mood shifts) social democrat parties are significantly more likely to increase EM representation compared to Christian democrats.
Sensitivity analyses
Because we only focus on statistically significant party platform change, we have relatively few right-wing platform shift events. As shown in Table A6 of the SI, based on our total N of 238, 50 cases make a significant shift to the right. We note, however, that our findings are robust when we rely on Meyer’s (Reference Meyer2013) more permissive approach to calculate standard errors around the parties’ positions. In that case, we have 82 parties that made a significant shift towards monoculturalism. And as shown in Figure A5, we find continued support for our Reputational Shield Hypothesis.
In the SI (Sections 1.6 and 1.7) we also evaluate the main (unconditional) effects of policy mood shifts and party platform change on increases in EM representation. Importantly, party platform shifts are statistically unrelated to changes in EM representation in a model specification that also includes policy mood shifts. This means that platform change has no effect on EM representation if mood shifts are held constant, strengthening the conditional theory that we advocate in this paper: a negative public mood swing is required for making some parties move significantly to the right and increase EM representation as a reputational shield.
Discussion and Conclusion
We endeavored to explain the seeming paradox of an increased presence of EM politicians in national parliaments while many political parties have adopted more restrictive platforms on immigration issues over the same time. We proposed and tested a novel idea that the parties’ efforts to increase descriptive representation of EMs forms part of a broader electoral strategy aimed at the median voter. Based on data from eight countries between 1995 and 2015, we confirmed our Reputational Shield Hypothesis, positing that parties that radically shift their platform to the right in response to conservative public mood swings, will simultaneously increase their EM MPs. By contrast, parties displaying no such shifts, or shifts in the opposite (multiculturalist) direction did not show an increased probability of adding to their numbers of EM MPs. We also found that reputational shield politics are mainly the domain of disadvantaged party families historically associated with more permissive policies.
First, our findings speak to the EM descriptive representation literature, and more specifically to research on parties’ electoral incentives to ethnically diversify. Studies here have either focussed on how EM candidates serve as a strategic resource to attract EM voters (Bergh and Bjørklund, Reference Bergh and Bjørklund2011; Martin, Reference Martin2015) and if they are a lability to attract the wider electorate (Bloemraad and Schönwälder, Reference Bloemraad and Schönwälder2013; Sobolewska, Reference Sobolewska2013; Stegmaier et al., Reference Stegmaier, Lewis-Beck and Smets2013; Street, Reference Street2014; Dancygier et al., Reference Dancygier, Lindgren, Oskarsson and Vernby2015; Fisher et al., Reference Fisher, Heath, Sanders and Sobolewska2015; Thrasher et al., Reference Thrasher, Borisyuk, Rallings and Webber2017; Portmann and Stojanović, Reference Portmann and Stojanović2019; Martin and Blinder, Reference Martin and Blinder2021). We propose a radically new idea, namely that increases in EM representation can also be part of a strategy where parties seek to reap the electoral rewards from their more restrictive immigration platform, by offsetting some of the costs of such platform change. Contrary to the logic of ethnic voting, this strategy is aimed at the wider electorate. Additionally, in contrast to studies on electoral penalties, we propose that parties can view EM politicians as a winning formula to attract ethnic majority voters skeptical of immigration, at least if the party simultaneously adopts more restrictive positions.
Hence, our study changes the theoretical understanding of representation. To be truly meaningful, descriptive representation must lead to improvements in EMs’ substantive representation (Mansbridge, Reference Mansbridge1999). Empirical research has shown this relationship to hold in most cases (e.g., Saalfeld and Bischof, Reference Saalfeld and Bischof2013; Lowande et al., Reference Lowande, Ritchie and Lauterbach2019). The assumed link between substantive and descriptive representation also remains one of the main motivations for descriptive representation research. All representation devoid of this substantive content is deemed tokenistic, although Pitkin (Reference Pitkin1967) allowed for even this kind of representation to be meaningful: EM audiences may create meaning from their mere presence in parliaments. Yet, how this symbolic meaning of representation can affect audiences outside of the represented group remains under-explored. Our paper indicates that political parties take advantage from the meaning that can be created by EM politicians among majority voters: They can shield themselves from the negative consequences of programmatic shifts to the right, as ethnic minority politicians signal their tolerance toward diversity.
Second, we have wedded spatial models of elections (e.g., Downs, Reference Downs1957) with models of strategic descriptive representation (e.g., Weeks et al., Reference Weeks, Meguid, Kittilson and Coffé2023). Typically, spatial models only focus on how parties maximize their votes by adopting a particular policy position. Extensions also bring in the role of issue salience (Meguid, Reference Meguid2008). Yet, despite a few notable exceptions (e.g., Van Spanje and De Graaf, Reference Van Spanje and De Graaf2018), studies based on the spatial model almost exclusively focus on parties’ programmatic responses to their environment. And no study on party platform change has so far considered strategic descriptive representation as part of the toolkit. While several innovative studies exist focussing on how parties strategically use descriptive representation of women (Weeks et al., Reference Weeks, Meguid, Kittilson and Coffé2023) and ethnic minorities (e.g., Ferland, Reference Ferland2020) to increase their vote, they fail to link this with party platform change. Our study suggests that parties employ integrated strategies combining programmatic change with strategic descriptive representation.
This paper also has several limitations posing significant questions for future research. As discussed earlier, there are new datasets going beyond party positions on multiculturalism (Lehmann and Zobel, Reference Lehmann and Zobel2018; see Dancygier and Margalit, Reference Dancygier and Margalit2020). Yet, unfortunately, these excellent initiatives did not contain sufficient data for the countries and parties in our dataset. This implies that our public mood measure contains different elements of immigration policy (including multiculturalism), while we could only study the parties’ responses to it on the narrower policy area of multiculturalism.
Perhaps more importantly, we could only observe the outcomes of party behaviors. While we judged them consistent with the logic of our Reputational Shield Hypothesis, further research is needed to confirm some of the causal mechanisms, both at the level of parties and voters. To start with the party-level, the first question is whether the reputational shield strategy is indeed a conscious strategy targeted at both ethnic minority and ethnic majority voters with an internalized anti-racism norm (Ivarsflaten et al., Reference Ivarsflaten, Blinder and Ford2010; Blinder et al., Reference Blinder, Ford and Ivarsflaten2013; Sobolewska and Ford, Reference Sobolewska and Ford2020). An alternative interpretation for the observed empirical pattern could be that parties cater to the white, majority voter by switching towards monoculturalism, while simultaneously trying to reap the ethnic vote by bringing EM politicians to parliaments. In this case the party wants to have it both ways; yet, the two acts of moving to the right and increasing EM representation are not intrinsically linked. We do not believe this is a likely explanation for our findings. In most countries EMs form a very small proportion of the electorate, likely in a region of only 10%, making any strategy aimed solely at minority voters electorally marginal nationwide (although in some major cities EMs might be a more powerful electorate). Moreover, it remains puzzling how a party would sway ethnic voters if it simultaneously makes a radical shift towards monoculturalism, as these voters seem unlikely to be bought off with such a tokenistic gesture.
Nonetheless, we invite future research to further uncover how party strategists think, which audiences reputational shield tactics are meant to target, and whether they are actually aware of the internalized anti-racism norm. This research may also want to consider if reputational shield tactics are dictated by the party leadership. While we have theorized that it is the party that increases EM representation, to mitigate the risk/reap the electoral reward from a monoculturalist policy shift, EM politicians opposing multiculturalist policies may have considerable agency in pulling their parties to the right. Hence, it would be interesting to abandon the unitary actor assumption and to further zoom in on the respective roles of the party leadership and EM politicians (see also Greene and O’Brien, Reference Greene and O’Brien2016). That being said, since we find that the whole process of platform change combined with ethnic diversification is set in motion by public mood shifts, it appears rather unlikely that observed instances of platform change would mostly be driven by EM politicians. The latter should be energized by different incentives than median voter shifts. And even if EM politicians would turn out as the most important instigators behind monoculturalist policy shifts, our main message that they act as a reputational shield to their parties still stands.
A final recommendation for party-level research would be collecting data at the individual candidate-level. Although some such efforts exist, they usually center on a single country (e.g., Tolley, Reference Tolley2019; Dancygier et al., Reference Dancygier, Lindgren, Nyman and Vernby2021), or even a single sub-national region or locality (e.g., Van Trappen et al., Reference Van Trappen, Vandeleene and Wauters2021) or on single parties like the UK Labour party (e.g., Akhtar and Peace, Reference Akhtar and Peace2019; Ashe, Reference Ashe2019). Unfortunately, we only have data on the candidates ultimately elected. While we can certainly expect elected MPs to a large extent to reflect the strategic intentions of parties before the elections, candidate-level data would still be the gold standard. This would help observing even more closely whether parties intentionally put EM candidates on the ballot to engage in reputational shield tactics. Here it is important to consider whether these candidates are selected to contest their more winnable seats or list positions, as that would provide firm evidence of parties engaging in the reputational shield strategy. Winnable positions are heavily competitive, and so, the assignment of these positions to EM candidates would undoubtedly reflect a strategic choice.
At the voter-level, future research is needed to confirm that ethnic majority voters are influenced by parties’ ethnic diversity in their judgements. We believe there is enough indication making this plausible: Using conjoint experiments, Martin and Blinder (Reference Martin and Blinder2021) recently showed that an EM candidate will be more heavily penalized for having a pro-minority platform than a white, British candidate. Similarly, we should explore whether anti-immigration platforms are seen as more acceptable if endorsed by an EM candidate.
Despite these questions, we hope that our evidence that political parties at least behave in line with the Reputational Shield Hypothesis has helped to gain a better understanding of how and when parties strategically improve their ethnic diversity.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S175577392400002X.
Acknowledgements
Results presented in this article have been obtained within the project, Pathways to Power: The Political Representation of Citizens of Immigrant Origin in Seven European Democracies (PATHWAYS)‘. This project was funded by the ANR (France), DFG (Germany), ESRC (UK) and NWO (Netherlands) under the Open Research Area (ORA+) framework. It has been joined in 2015 by a Belgian sister project funded by UCLouvain, ULB and VUB universities and the Fund of Scientific Research. The PATHWAYS consortium is formed by the University of Amsterdam (Prof. Jean Tillie), the University of Bamberg (Prof. Thomas Saalfeld), the University of Leicester (Prof. Laura Morales), and the CEVIPOF Sciences Po Paris (Prof. Manlio Cinalli). The author(s) want to explicitly acknowledge the work of the following researchers in the production of the original datasets: Pierre Baudewyns, Karen Celis, Didier Chabanet, Manlio Cinalli, Jérémy Dodeigne, Silvia Erzeel, Jorge Fernandes, Edward Fieldhouse, Lucas Geese, Simona Guerra, Anne-Sophie van Gulpen, Steven van Hauwaert, Chloé Janssen, Angeliki Konstantinidou, Cornelia Kristen, Siresa Lopez, Laura Morales, Liza Mügge, Daphne van der Pas, Constanza Sanhueza Petrarca, Jean-Benoit Pilet, Luis Ramiro, Huub Ramler, Thomas Saalfeld, Diana Schacht, Maria-Jimena Sanhueza, Carsten Schwemmer, Jean Tillie, Daniela Vintila, and Andreas M. Wüst. The authors also thank the Manifesto Core Group for their help with our queries on the ManifestoR program and to Heiko Giebler and the anonymous reviewers for their valuable feedback on the manuscript.
Open access
