While a growing literature on party competition has established that parties react to shifts in public opinion,Footnote 1 to election resultsFootnote 2 and to voters’ issue priorities,Footnote 3 so far much less attention has been dedicated to the question of how parties react to each other’s behavior. Although some studies have started looking into the question of how parties’ left–right shifts affect other parties,Footnote 4 we know comparatively little about parties’ interaction with each other and its effect on policy positions and specific issues.Footnote 5 Taking the idea of competition seriously, it should be of fundamental importance to scholars of party competition if parties ‘only’ react to voters’ preferences and signals thereof, or if the behavior of competitors has an independent effect on what other parties do. Hence, in this article we investigate how parties shift their policy positions in response to successful radical right challengers and use a novel empirical strategy to causally identify this effect.
New radical right parties have now become successful and established actors in a large share of European democracies. This phenomenon has received wide attention from political observers and has spurred a considerable amount of academic research. For a long time the most salient question in this line of research has been why these parties are electorally successful in some countries while they fail to achieve the same amount of success in others.Footnote 6 While the debate about these determinants and their interaction with each other is still ongoing,Footnote 7 several scholars have turned their attention to the previously much less studied question of the effect of radical right success on the behavior of other parties. The main focus in this line of research lies on the potentially ‘contagious’ effect of radical right parties, i.e. the question if their success causes other parties to adopt more anti-immigrant and culturally protectionist positions.Footnote 8 Most quantitative comparative studies seem to confirm such an effect on anti-immigrantFootnote 9 or more broadly second dimension positions,Footnote 10 while other studies shed some doubt on the existence of radical right contagion and argue that their effect has been overstated.Footnote 11 The key argument against the contagion hypothesis states that the impact of the radical right has been overestimated since it coincides with other developments that have led to a change in public opinion and have thus caused mainstream parties to shift their position on immigration issues.Footnote 12 And, indeed, while the comparative assessments by van Spanje, Abou-Chadi and Han find empirical support for the claim that mainstream parties react to radical right success by emphasizing more anti-immigrant positions, these studies cannot fully rule out that their findings are driven by changes in public opinion rather than identifying a causal effect of the success of the radical right. This debate thus reflects the broader and fundamentally important question whether we can causally identify an effect of one party’s success on another party’s behavior. Empirically establishing that one party reacts to another party’s success is challenging. Since parties’ vote shares and public opinion are likely correlated, we do not know if the success of one party causes another one to shift its position or if success and positional shift are simply the result of a shift in public opinion that causes both.
In this article we tackle this issue by using a novel research design to study the contagious effect of radical right parties. Employing a regression discontinuity design based on thresholds of representation allows us to study the effect of radical right success independent of the effect of public opinion. Electoral thresholds which are mainly a function of the electoral system determine which vote share will lead to parliamentary representation of parties in the form of seats.Footnote 13 Exploiting these thresholds allows us to compare cases with similar radical right vote share which differ in the representation of these parties in parliament. Parliamentary representation provides radical right parties with additional resources and media attention which increases the threat that they constitute for established parties. Since we exploit exogenous variation based on electoral thresholds we can assess the causal effect of this threat independent of public opinion, which we argue cannot be reliably controlled for in standard regression approaches.
Analyzing party positions in twenty-three European countries in the period from 1980 to 2014 based on data provided by the CMP/MARPOR group with this design, we find a substantive and statistically significant effect of radical right success on the positions of mainstream parties. This effect is robust against a number of alternative operationalizations and specifications. We can also provide evidence that the mainstream right as well as the mainstream left respond to radical right success by shifting toward a more cultural protectionist position.
These findings have important implications for the literature on the new radical right as they demonstrate that their success causally affects other parties. These parties’ electoral successes are, thus, not only a symptom or by-product of the ‘right turn’ in European democracies,Footnote 14 but a factor contributing to this development. Of course this should not be understood as an argument that public opinion does not matter for the fate of the radical right or party competition more generally. The goal of this article, however, is to empirically demonstrate that radical right parties as agents shape the behavior of mainstream parties and to causally identify this effect.
On a more general level these findings contribute to the literature on party competition because they demonstrate that parties’ behavior has an impact on other parties’ strategies that goes beyond a signaling of public opinion. Moreover, they indicate that these reactions do not only take place on the main left–right axis of political competition but also on other issues such as immigration. These findings should thus encourage scholars of party competition to further investigate the complex interactions of parties as they matter independent of public attitudes.
RADICAL RIGHT SUCCESS AND MAINSTREAM PARTY BEHAVIOR
In order to explain the causes of parties’ policy shifts, scholars have largely relied on the spatial conception of party competition first introduced by Downs.Footnote 15 Following from a rational act of voting that is based on ideological proximity (voters vote for the party ideologically closest to them), parties have an incentive to react to changes in public opinion and adjust their policy positions accordingly. However, facing fundamental uncertainty about voters’ preferences, vote-seeking parties have to rely on signals about these preferences. Elections constitute the most important of these signals.Footnote 16 Building on these assumptions a growing literature has investigated how parties react to public opinion, elections and sub-groups of the population.Footnote 17 However, the same framework indicates that in order to understand parties’ policy shifts, it is necessary to take into account other parties’ behavior – a question that so far has seen far less scholarly attention.Footnote 18
As already Downs points out using the example of the States’ Right Party or ‘Dixiecrats’, new challenger parties can play a crucial role for this dynamic.Footnote 19 While the goal of the new party might not even be to gain office or win an election, it has the potential to make other parties shift in the direction of its preferred policy position. What is crucial here is that the challenger party affects other parties’ behavior independent of the voter distribution. Its success causes other parties to shift their position because the competitive space has changed. The established party does not shift its position because of a shift in voter preferences but to affect the fate of another party – ‘to take the wind out of its sails’.Footnote 20
Studying the success of radical right parties, several authors indeed argue that radical right success should cause mainstream parties to shift their positions. Leaving the left–right dimension and focusing on party competition in the two-dimensional post-industrial political space, scholars have argued that the success of radical right parties constitutes a fundamental challenge for established parties and that their reactions will largely concern issues such as immigration and multiculturalism.Footnote 21 Faced with a credible threat from a niche party, mainstream parties have an incentive to move their position toward the position of the niche party in order to keep the niche party from stealing their votes at subsequent elections. In the words of Bonnie Meguid, mainstream parties can apply an accommodative strategy in order to steal a niche party’s issue ownership.Footnote 22 This strategy is especially electorally appealing as niche parties mostly compete on a single issue. If they lose their appeal on this issue they are likely to lose a lot of electoral support. Similar to Downs’ example about the Dixiecrats, the success of a radical right party here causes other parties to shift their position independent of the underlying voter distribution. In this account, the radical right party itself is an actor and not just a signal or by-product of public opinion.
Empirically, several small-N studies have documented and portrayed how the success of radical right parties leads to the adaption of other parties’ behavior.Footnote 23 Because competition with the radical right is not limited to the left-right dimension and concerns cross-cutting issues such as immigration, these reactions are not limited to the mainstream right but can be equally found among the mainstream left.Footnote 24 In contrast to this, some recent analyses cast doubt on the contagious effect of the radical right and argue that their impact has been overstated in academic and public debates.Footnote 25 While there is little doubt that immigration has become a salient political issue in most European democracies, these studies argue that this development cannot be attributed to the success of the radical rightFootnote 26 and that in some countries mainstream parties started competing on this issue even before there was a successful radical right party.Footnote 27 Following this interpretation, especially parties of the mainstream right have contributed to politicization of the immigration issue by increasingly adopting an anti-immigrant stance.Footnote 28
In contrast, three large-N comparative studies have found a positive effect of radical right vote share on mainstream parties’ anti-immigrant or anti-multiculturalism positions.Footnote 29 Using party positions on the immigration issue from expert surveys, van SpanjeFootnote 30 demonstrates that where radical right parties were stronger in the 1990s, mainstream parties show more anti-immigrant positions in the 2000s. Similarly, based on party positions extracted from manifesto data, Abou-Chadi and HanFootnote 31 demonstrate that mainstream parties emphasize more anti-immigrant or anti-multiculturalism positions when radical right parties received higher vote shares at the previous election. While the effects are (slightly) more pronounced for parties of the mainstream right, all three studies show that parties of the mainstream left are also affected by radical right success.
While all three studies, thus, show an association between the threat posed by a successful radical right contender and mainstream parties’ anti-immigrant positions, as a consequence of the applied designs, important doubts remain about the identification of a causal effect of the radical right on other parties’ behavior. Taking a counterfactual perspective the crucial question is: in absence of a successful radical right challenger would we not have seen a shift in policy position? Hence, a main concern for causally identifying the effect of radical right parties on the behavior of other parties lies in the effect of public opinion as a potential confounder.Footnote 32 If public opinion shifts towards a more anti-immigrant stance then this should cause both an increase in radical right strength as well as stronger anti-immigrant positions by mainstream parties. There is broad empirical evidence that indeed both of these mechanisms seem to be prevalent within multiparty competition. First, several studies demonstrate that mainstream parties are responsive to shifts in public opinion.Footnote 33 While these studies focus on parties’ left–right movements, we should expect this to be equally the case for issues surrounding immigration. Secondly, attitudes toward immigration have been shown to be the number one predictor of voting for a radical right party.Footnote 34 At the aggregate level, we should thus expect anti-immigrant attitudes and radical right vote shares to covary. Hence, finding a correlation between radical right success and anti-immigrant positions of mainstream parties is potentially spurious.Footnote 35 In addition to that, public opinion is a latent construct that is difficult to measure and hardly possible to control for. HanFootnote 36 does include a measure of public opinion based on the aggregation of a survey data item. However, the problem with such a measure remains that it is nearly impossible to account for all variation in public opinion that might be a source of confounding. Hence, what is necessary to identify if mainstream parties, indeed, react to the success of a radical right party, is a research design that permits analyzing the effect of radical right success on mainstream party positions independent of public opinion.
RESEARCH DESIGN
In this article we employ a regression discontinuity design and use radical right parties’ representation in parliament as an indicator of the threat that they pose to established parties. While a radical right party gaining a minimum of one seat might not seem a dramatic event, parliamentary representation provides radical right parties with a significant increase in resources and media attention, and will make their survival as political contenders much more likely.Footnote 37 Hence, when radical right challengers achieve representation in parliament we should expect mainstream parties to react more strongly to their success as they have proven to be a viable contender that now commands an increased amount of resources. Parliamentary representation will also affect future voting decisions by reducing strategic voting against these parties. In short, parliamentary representation provides radical right parties with a multitude of monetary, strategic and symbolic resources which make these parties a more credible challenger in future contests – and indeed Dinas et al.Footnote 38 demonstrate that parliamentary representation profoundly increases the likelihood of new party survival. Having crossed the watershed of parliamentary representation, radical right parties thus demonstrate that under current circumstances they are likely to remain a competitor for the long run. Previous to radical right parties’ representation in parliament mainstream parties may consider the politicization of immigration as too risky and might still hope for, for example, a disappearance of the radical right party due to within-party struggles. However, once radical right parties signal their presence as a stable and credible challenger, mainstream parties have a strong incentive to accommodate their main issue.
Parliamentary representation is of course as much a function of public opinion as vote share per se. Moreover, selection into parliament will depend on factors such as the quality of parties’ leaders or their organizational capacitiesFootnote 39 which, in turn, should affect other parties’ reactions. However, representation in parliament is also a function of a country’s electoral system. Since this threshold of representation is exogenous to party behavior, we can exploit the variation that is caused by it to identify the effect of radical right success on other parties’ anti-immigrant positions.
The regression discontinuity (RD) design allows us to estimate the causal effect of RRP presence in parliament on mainstream parties’ positional shifts on immigration in a quasi-experimental setting. The basic idea of the RD design is to identify variation in a cause by means of exploiting an exogenous, institutional cut-off point that assigns treatment. In our case, this rule of treatment assignment is the electoral threshold as an idiosyncratic feature of electoral systems: mainstream parties that compete in party systems in which RRPs have closely failed to surpass the electoral threshold are not exposed to parliamentary RRP pressure, while those where radical right parties have just managed to surpass the electoral threshold are forced to deal with RRPs in parliament. Since we are comparing situations with similar RRP vote shares, we can rule out public opinion as a confounder. Hence, we use the RD design as a concept that allows us to employ a counterfactual comparison of mainstream parties’ position shifts with and without RRP presence in parliament. The causal effect is defined as the difference between the potential outcome under treatment and the potential outcome under control:
$$\tau {\ \equals \ }E[\Delta Y(D{\ \equals \ }1){\minus}\Delta Y(D{\ \equals \ }0)],$$
where ΔY(D) is a potential mainstream party-shift while D=1 if a RRP has managed to enter the parliament and D=0 otherwise. The validity of the RD design rests on a continuity assumption,Footnote 40 which implies that mainstream parties, which are just exposed to RRP presence in parliament and those that are just ‘free’ of parliamentary RRP pressure, show no significant differences in their characteristics with the consequence that both groups do not vary from each other except for their treatment status. For instance, the threshold of parliamentary representation is set to 3 per cent in the Greek electoral system. Here, it should be of little relevance for Greek mainstream parties whether a RRP gained a vote share right above or below 3 per cent apart from the fact that the corresponding radical right challenger has managed in the first case to gain parliamentary representation, but not in the latter. Given that this assumption holds, we can exploit the discontinuity induced by the institutional cut-off point in order to identify the causal effect of RRP presence in parliament without having to rely on the assumption – as it is the case in commonly used regression designs – that all relevant confounding factors have been considered. The crucial idea is, thus, that in a narrow window around the threshold confounders do not affect treatment assignment. As a consequence of this, changes in an outcome variable can be attributed to the change in treatment status induced by an assignment variable that determines whether the threshold is surpassed or not. In our case, the treatment status (RRP presence in parliament) is a function of the electoral support for radical right parties. Hence, what the RD design helps us to achieve is that we can exploit variation in the threat of radical right success that is independent of public opinion.
As outlined by Lee,Footnote 41 the continuity assumption also implies that political parties can neither perfectly predict nor control the electoral outcome of RRPs. Otherwise local random assignment of treatment and control groups would not be possible. While the continuity can never be completely tested, it seems plausible in the context of RRP success. In this setting, perfect control over RRPs’ electoral results would only be possible by means of electoral fraud or a strategic manipulation of the electoral threshold.Footnote 42 In these situations, political parties could rig election results in order to push competitors above or below the threshold. In line with previous research,Footnote 43 we assume that electoral fraud is not present in our sample of European democracies. Nevertheless, the history of European party systems knows (rare) cases in which mainstream parties indeed changed electoral laws in order to either facilitate the entry of small parties or to prevent them from gaining access to the national parliament. One famous example for this is the 1985 decision of the then French president François Mitterrand to abolish the two-ballot majoritarian system in favor of party-list proportional representation which guaranteed that the Front National (FN) would enter the Assemblée Nationale.Footnote 44 However, in this case the (effective) electoral threshold was set at a much lower level than the FN was expected to gain. In other cases, mainstream parties increased the threshold in order to prevent small party representation in parliament. However, the newly institutionalized electoral thresholds clearly exceeded the anticipated electoral support of the targeted political parties.Footnote 45 All in all, we assume that political parties cannot perfectly control the entrance of RRPs into parliament in a close window around the threshold and that the basic assumptions of the RD design hold for the following analysis.Footnote 46
The RD design necessitates that assignment to treatment changes discontinuously at a cut-off point. In our case, this assignment is complicated by the rules of European election systems. On the one hand, nationwide electoral thresholds seldom represent exclusive rules and are frequently complemented by additional provisions of the electoral law allowing small parties to enter parliament without surpassing the former. For instance, in German general elections the nationwide threshold is set to 5 per cent of the vote share. However, parties winning at least one single-member district will, independently of their nationwide electoral support, gain one seat in the German Bundestag. On the other hand, several European countries have not instituted a legal threshold at all. In these cases we rely on the effective nationwide threshold proposed by Taagepera.Footnote 47 However, this estimation technique is defined as an approximation to the vote share necessary to gain at least one seat in parliament. Hence, it is possible that parties with a vote share below this estimated threshold gain a seat and vice versa.
The following sections will show that none of these specific cases are present in our dataset so that we are dealing with a so-called sharp RD design.Footnote 48 The treatment status D is, thus, a deterministic function of the assignment variable x with the property that D=0 if x<c and D=1 if x≥c, where c denotes the electoral threshold. This allows us to identify the treatment effect (τ) by the following regression formula:Footnote 49
$$\eqalignno{ \Delta Y_{m} {\ \equals \ } & \alpha {\plus}\tau D_{m} {\plus}\beta _{1} (x_{i} {\minus}c_{i} ) {\plus} \beta _{2} (x_{i} {\minus}c_{i} ){\asterisk}D_{m} {\plus}{\epsilon}_{m} , $$
where m refers to mainstream parties and i to election specific characteristics. Importantly, the causal effect identified by means of the RD design only refers to elections where RRPs just managed or failed to enter the parliament. Hence, the causal impact of the forcing variable is not generalized to the entire population. The analysis identifies local average treatment effects (LATE).
Two ways of modeling the conditional expectations of the RD design have been discussed in the literature: non-parametric, local linear regressions and parametric linear models.Footnote 50 In the first case, the sample is restricted to a certain bandwidth h away from the cut-off point so that c i−h≤x i≤c i+h. Based on this, a local linear function on the resulting set of observations is estimated. We employ a triangular kernel in order to give more weight to election results closer to the cut-off.Footnote 51 In the second case, the entire sample along with a global high-order polynomial function of (x i−c i) is used. As pointed out by Lee and Lemieux,Footnote 52 both approaches can lead to valid results given the specific characteristics of the underlying data. Since our dataset is characterized by a small sample size around the cut-off point, we use a variety of bandwidths and polynomial specifications to verify the stability of our results. All our models additionally include country fixed effects and standard errors are clustered within parties and elections.
DATA AND OPERATIONALIZATION
Three indicators constitute the central variables of the RD design: the outcome variable, the assignment variable and the cut-off-point. Our outcome is a mainstream party’s policy shift on the immigration issue between election t and election t−1. To assess parties’ policy positions we make use of the data provided by the CMP/MARPOR project.Footnote 53 The CMP/MARPOR dataset has been the object of considerable debate. Its advantages and disadvantages have been discussed elsewhere,Footnote 54 but it remains the only dataset available that spans a time period long enough to evaluate our hypothesis.Footnote 55 Unfortunately the CMP data do not include items directly referring to immigration policy positions. We thus construct our main dependent variable based on the items per607 (‘Multiculturalism: Positive’) and per608 (‘Multiculturalism: Negative’), representing appeals to cultural protectionism, a core issue of the radical right agenda.Footnote 56 In addition, these items also constitute the main dependent variables in the studies by Han and Abou-Chadi, which assures comparability between our results and theirs. In the Appendix, we additionally demonstrate that our findings equally hold for broader measures of policy positions, which have been suggested to represent anti-immigrant positionsFootnote 57 and the core of the radical right agenda.Footnote 58 Following the suggestions by Lowe et al.Footnote 59 that concern the scaling of positions extracted from manifesto data, we logit transform our dependent variable.Footnote 60 Positive values on this scale then indicate more culturally protectionist positions.
We conduct our analysis for a sample of twenty-three European democracies. Since we want to focus on mainstream parties’ reactions, and because several scholars have demonstrated that small and niche parties follow different incentives from their mainstream competitors,Footnote 61 we include only those parties with an average vote share of at least 10 per cent in general elections since 1960. Radical right parties are excluded from this sample independently of their electoral performance.
The electoral support of radical right parties in election t−1 constitutes the assignment variable. In order to obtain vote shares that are as exact as possible, it is necessary to include radical right parties that have gained vote shares well below the level usually reported by datasets on elections in Europe. For that reason, the election results of the corresponding strongest radical right parties have been collected for all general elections since 1980.Footnote 62 The categorization of radical right parties has been discussed intensely in the literature.Footnote 63 For the present analysis, we build on the categorizations put forward by MuddeFootnote 64 for Western Europe and BustikovaFootnote 65 with regard to the Eastern European countries of our sample.
The Appendix gives a detailed outline on which parties and countries were included or had to be excluded from the analysis.
As our cut-off-point we use the nationwide thresholds of representation of the European election systems. We have already mentioned that only a part of the country sample under study makes use of a legally defined nationwide threshold. In the remaining countries, thresholds cannot be directly inferred from electoral laws. Based on previous research,Footnote 66 we calculated in these cases the so-called effective nationwide threshold. Building on the ‘threshold of inclusion’ first suggested by RokkanFootnote 67 and the ‘threshold of exclusion’ proposed by Rae et al.,Footnote 68 Taagepera defines the effective threshold as ‘an average threshold of representation . . . where parties have a 50–50 chance of winning their first seat’.Footnote 69 For that purpose, Taagepera identifies the total number of seats in the assembly (S) and the number of electoral districts (E) to calculate the effective threshold:
$$T{\ \equals \,}{{75\,\%\,} \over {\left( {\left( {{S \over E}{\plus}1} \right){\asterisk}\sqrt E } \right)}}$$
The validity of the effective threshold has been questioned regarding its use in the context of plurality-majoritarian systems.Footnote 70 Following this criticism, plurality-majoritarian as well as mixed-member majoritarian systemsFootnote 71 are excluded from the analysis. Table 1 lists our resulting sample as well as the corresponding legally defined or estimated effective thresholds.Footnote 72 Since the effective threshold is not purely determined by electoral institutions but also depends on the behavior of parties, we additionally run our models limiting our sample to cases with a legal threshold. Lastly, it is important to mention that the vote shares of radical right parties have been centered around the corresponding national thresholds in order to ensure that the effect of entering the parliament can be measured across countries.
Election Thresholds in Europe, 1975–2014
Notes: * Levelling seats and seats reserved for minorities have been subtracted from the total number of seats in parliament.
**In the case of Luxembourg, the empirical threshold (Taagepera Reference Taagepera1989) has been applied in order to account for misspecification resulting from electors’ option to split their vote between different parties (see e.g. Dumont and Poirier Reference Dumont and Poirier2005).
RESULTS
A first way of evaluating effects within a regression discontinuity design is a graphical representation of the outcome depending on the values of the treatment variable. In Figure 1 we show the changes in parties’ anti-immigration positions depending on the vote share of radical right parties at the previous election. The dashed line marks our cut-off point depending on the electoral threshold and the solid lines represent cubic fits with 95 per cent confidence intervals. First, we can see that the application of a sharp RD design is justified. All radical right parties left of the cut-off point did not gain any seats in their corresponding national assemblies while all RRPs to the right did. Second, a ‘jump’ at the cut-off point is clearly visible. In cases where radical right parties surpass the threshold of representation we witness an on average much larger shift in mainstream parties’ anti-immigrant positions. Hence, Figure 1 provides a first strong indication for a causal effect of radical right success on other parties’ policy positions that is independent of public opinion.
Mainstream party position change on cultural protectionism Note: The solid lines plot the cubic fit and the shaded areas denote the 95 per cent confidence bands.
We should emphasize again that the vote shares on the x-axis are calibrated with the effective threshold. This means that negative values indicate how much the vote share was below and positive values how much it was above the threshold. This also means that we cannot simply interpret the scatter plot as a linear relationship between radical right vote share and mainstream party position shifts. Hence, while a first look at Figure 1 may seem to indicate that increasing vote shares above the threshold do not lead to stronger shifts, this interpretation should not be made from the graph as values on the x-axis are calibrated to the threshold and are thus not one to one transferable to vote shares.Footnote 73
Table 2 presents our findings for the local average treatment effect using parametric as well as non-parametric estimation techniques. For both techniques and different order polynomials we find a highly statistically significant effect of a radical right party’s representation in parliament on other parties’ anti-immigrant positions. This effect is not only statistically significant but also substantively meaningful. The local average treatment effect of 3 (resulting from a non-parametric specification with a first order polynomial) is about twice as high as the standard deviation of positional shifts in our sample. This is in size similar to the positional difference between the Dutch center-right VVD and the center-left PvDA in the year 2010.
Mainstream Party Position Change on Cultural Protectionism
Note: Country-fixed effects and two-way clustered standard errors used. Bandwidth estimation according to Imbens and Kalyanaraman (Reference Imbens and Kalyanaraman2009). *p<.1, **p<.05, ***p<.01.
Hence, our findings provide strong empirical evidence that radical right success has a causal and substantive impact on the positions of other parties. Our design allows us to rule out public opinion as a potential confounder and thus underlines the claim that radical right success is not simply a by-product of a right-turn in Western Europe but in itself a driving force of this process. On a more general level, those findings, thus, also demonstrate that parties react to other parties’ success independent of public opinion. We should emphasize that this does not mean that parties generally do not react to public opinion. It solely indicates that there is an effect of the competitive space that parties interact in that cannot be reduced to public attitudes and signals thereof. We should also emphasize that this does not mean that vote shares capture all variation in public opinion. What our design allows us to do is to rule out that the variation in radical right vote share that determines reactions of mainstream parties is confounded by public opinion or other unobserved third variables.
An additional interesting question in this regard is if mainstream left as well as mainstream right parties react to the success of radical right parties. While based on the traditional spatial model of party competition one should expect that parties of the mainstream right react more strongly to the success of the radical right, research in this field has repeatedly pointed out that radical right and mainstream left parties also compete for a similar segment of voters.Footnote 74 In the two-dimensional space of post-industrial societies, both mainstream left and mainstream right parties have incentives to shift toward a more anti-immigrant and culturally protectionist position when faced with a successful radical right contender.Footnote 75
Tables 3 and 4 show the results of the RD analysis for samples divided between mainstream left and mainstream right parties.Footnote 76 We find a significant positive effect for both mainstream left as well as mainstream right parties, which is robust and substantively meaningful. For mainstream left parties the substantive effect varies between 3 and 4 and thus is quite similar to the overall sample. While the values for mainstream right parties are generally higher, we can see relatively strong fluctuation depending on the exact specification. One reason for this is of course the substantially reduced sample size. Nonetheless, we can clearly see that there is a substantive and significant effect for mainstream left as well as mainstream right parties. Hence, if radical right parties surpass the threshold of representation then mainstream left and mainstream right parties alike shift their position. Again, our design allows us to causally identify this effect and rule out public opinion as a potential confounder.
Mainstream Left Party Position Change on Cultural Protectionism
Note: Country-fixed effects and two-way clustered standard errors used. Bandwidth estimation according to Imbens and Kalyanaraman (Reference Imbens and Kalyanaraman2009). *p<.1, **p<.05, ***p<.01.
Mainstream Right Party Position Change on Cultural Protectionism
Note: Country-fixed effects and two-way clustered standard errors used. Bandwidth estimation according to Imbens and Kalyanaraman (Reference Imbens and Kalyanaraman2009). *p<.1, **p<.05, ***p<.01.
ROBUSTNESS AND PLACEBO TESTS
In order to demonstrate the robustness of our findings in the following section, we present several additional analyses. Table 2 has already demonstrated that our results are not sensitive to varying specifications of the polynomial control function. One further issue that surrounds the validity of RD designs is the choice of the optimal bandwidth. If the estimated local average treatment effect were sensitive to varying sizes of this window, this would cast serious doubts on the validity of the presented results. Figure 2 shows how our local average treatment effect varies with different bandwidths ranging from 1.5 per cent to 10 per cent. The optimal bandwidth used for the previous calculations has been 3.3. We can see that for any other specification we similarly find a statistically significant and substantively meaningful effect of radical right success on anti-immigrant positions. The predicted effect consistently lies between 2 and 5. None of the confidence intervals includes the value 0.
LATE under varying bandwidths
Tables A6, A7 and A8 in the Appendix present three placebo tests which further underline that our analysis, indeed, constitutes a causal identification of the effect of radical right success on mainstream parties’ policy shifts. In table A6 we can see our analysis using arbitrary cut-off points that differ from the nationwide threshold. Following Imbens and LemieuxFootnote 77 we test for discontinuities at the median of the two samples on both sides of the cut-off point. We should not expect a jump at any of these points. And, indeed, for none of these fake cut-off values do we find a significant effect.
Table A7 shows our analysis for a different dependent variable, namely changes in positions on environmental protection.Footnote 78 Since radical right parties are barely associated with these issues, their success should not have an impact on other parties’ positions on environmental issues. Looking at the results in table A7, we can see that there is, indeed, no statistically significant effect.
Since parties take considerable time drafting their manifestos, if mainstream parties truly react to radical right parties, there should not be a discontinuous effect of the vote share of a radical right party at the same election. Table A8 thus presents our analysis using radical right vote share at election t instead of t−1. Again, we do not find a statistically significant effect. All three placebo tests thus support the idea that the findings of our main analysis constitute a causal effect of radical right success on mainstream parties’ positions.
In order to assure that our findings are not driven by a single country, we conduct jackknife analyses.Footnote 79 Table A9 in the Appendix shows how effect size and statistical significance of our treatment effect vary with the successive exclusion of single countries from the analysis. The resulting differences in the size of our τ-coefficients are marginal and for the first order polynomial, for example, only vary between 2.8 and 4.1. We also find statistically significant effects for all cases.
As discussed before, using the effective threshold as a cut-off point for a regression discontinuity design involves the problem that these thresholds are not only determined by electoral institutions and are thus not completely exogenous. Hence, in the following section we repeat our analysis with a sample limited to countries with a legal threshold.
LEGAL THRESHOLDS
Figure 3 shows the relationship between radical right vote share at t−1 and changes in mainstream parties’ positions now only for cases with a legal threshold. Again we can see a clear jump around the cut-off point indicating a causal effect of radical right success.
Mainstream party position change on cultural protectionism, countries with legally fixed threshold. Note: The solid lines plot the cubic fit and the shaded areas denote the 95 per cent confidence bands.
The results in Table 5 confirm a statistically significant local average treatment effect. Although the number of cases is substantially reduced, we find this significant effect for all four different specifications. The substantive size of the effect is very similar to the one we found using the effective threshold.
Mainstream Party Position Change on Cultural Protectionism, Countries with Legally Defined Electoral Thresholds
Note: Country-fixed effects and two-way clustered standard errors used. Bandwidth estimation according to Imbens and Kalyanaraman (Reference Imbens and Kalyanaraman2009). *p<.1, **p<.05, ***p<.01.
In sum, several robustness tests as well as an analysis limited to legal thresholds support our initial findings. The success of radical right parties causes other parties to shift towards more cultural protectionism. We can rule out that these shifts are caused by public opinion as a confounder.
CONCLUSION
In this article we demonstrate that established parties react to the success of radical right challengers by emphasizing more anti-immigrant and culturally protectionist positions. While there is general agreement in the literature that immigration has become one of the most salient issues in the European political space, the question of how the success of the radical right has contributed to this phenomenon has remained far more debated. Although several comparative studies have demonstrated an association of radical right success and other parties’ anti-immigrant positions, they could not fully exclude the possibility that radical right success is a by-product instead of a causal factor of these developments.
Using a regression discontinuity design, in this study we are able to causally attribute shifts in mainstream party positions to the success of the radical right. If these parties gain representation in parliament we observe a strong move towards an anti-immigrant stance by other political parties. These findings have important implications for the literature on the radical right as well as party competition more generally. First, they show that the radical right as an actor plays a fundamental role in the politicization of the immigration issue – they do not simply constitute a symptom of a larger development. Second, they demonstrate that the transformation of the political space in Western Europe that we are currently witnessing is not simply a reaction to shifting preferences of the European electorate, but is a result of the strategic interaction of political parties. Third, in relation to a more general literature on party competition they underline that parties do not only follow shifts on the demand side of the electoral market, but react to other parties’ behavior.
Now it has been established that radical right success causally affects established parties’ anti-immigrant positions, further research should turn to questions of contextual factors and especially sequencing effects. Are we largely witnessing breakthrough effects or are these dynamics playing out similarly when radical right parties have established themselves as parliamentary actors? What are the strategic responses of radical right parties against these accommodative strategies of mainstream parties?
The findings of this article not only have important implications for the political science literature but for the broader debate surrounding the success of radical right parties. They show that these parties’ ascent has a substantial effect on the behavior of established parties, and that institutional representation seems to constitute a watershed within this process. Hence, the transformation of political discourse and, more generally, of the political space that we are currently witnessing is likely one that is fundamental and lasting, rather than short-lived and superficial.
