Appearing moderate or radical? Radical left party success and the two-dimensional political space

Abstract Challenger parties’ electoral successes have attracted increasing scholarly attention. Based on the example of West European radical left parties, this article investigates whether and how centripetal and centrifugal positional movements on different conflict dimensions influence the election results of these parties. Depending on parties’ issue-linkages, these strategies will have a different effect for the economic and the non-economic issue dimension. Due to radical left parties’ long-term commitment and a strong party-issue linkage on economic issues, more moderate positions will play to their electoral advantage. In contrast, far-left parties compete with social democratic and green-libertarian parties for party-issue linkages on the non-economic issue dimension. Here, they benefit from promoting centrifugal strategies. Based on time-series cross-section analyses for 25 West European far-left parties between 1990 and 2017, the empirical results show that the success of radical left parties’ positional strategies varies with the conflict dimension in question and that this effect is only partly moderated by the positions of competing mainstream left parties.

programmatic transformation with the consequence that they have seen remarkable electoral comebacks (Hudson 2012;March and Keith 2016a). On average, the electoral support for the West European radical left has increased from 7% in the early 1990s to 12% in most recent elections. Although this trend is far from uniform across countries, 1 most RLPs have seen a remarkable increase in their public support in recent decades, with Syriza in Greece and Podemos in Spain being the most striking examples. 2 As a result, RLPs even defeated their social democratic opponents in some of the last elections. 3 This development raises the question of what explains the electoral success of this party family.
An increasing number of researchers have started to investigate the causes of RLPs' electoral performances. Two main approaches to identifying the determinants of radical left party success are discussed in the current literature. First, studies have dealt with current trends and events in West European politicssuch as European integration or the 2008 financial crisisto explain the electoral fortunes of the far left (e.g. Beaudonnet and Gomez 2017;Dunphy 2004). Second, research has investigated the influence of individual-level factors from the perspective of cleavage theory (Ramiro 2016), theories focusing on economic distress and deprivation (March and Rommerskirchen 2015;Visser et al. 2014), or modernisation theory (Charalambous and Lamprianou 2017;Gomez et al. 2016). Although some of these studies also address factors related to the political supply side, little is known about the success and failure of RLPs' programmatic strategies. More concretely, no study has yet accounted for the question of how and to what extent RLPs' programmatic offers affect their electoral performances. This is surprising since the West European far left has undergone a profound process of programmatic re-orientation (March 2011). While a minority of (post-)communist parties continue to promote a rather orthodox programmatic profile, others have moderated their stances on economic issues and/or incorporated 'New Politics' issues into their programmes.
This study investigates how the supply side of electoral politics has influenced the electoral performance of radical left parties. To this end, the analysis focuses on the positions of radical left parties on the two dominant axes of conflict in Western Europe, i.e. an economic and a non-economic dimension. The core argument of this article is that radical left parties benefit from different strategies on these two dimensions during election campaigns. Relying on the Downsian framework of political competition, as well as on literature on niche and challenger parties, I argue that radical left parties profit electorally from centrifugal movements, i.e. more extreme positions, when it comes to non-economic issues, while they benefit from centripetal shifts, i.e. more moderate positions, on the economic conflict dimension. Moreover, the article demonstrates how the success of these strategies depends on the positions occupied by social democratic competitors. Using data between 1990 and 2017 for 17 West European countries provided by the MARPOR research group, the empirical analysis shows that the success of RLPs' positional strategies does indeed vary with the conflict dimension in question.
The article at hand contributes to two strands of the literature. First, it facilitates a better understanding of the conditions explaining support for RLPs. More specifically, this is the first article to go beyond the analysis of demand-side factors to explain RLP support on a broad comparative basis. Second, this article adds to the ongoing debate on the electoral success of challenger parties. Few studies have considered both dimensions of political conflict (economic and non-economic) simultaneously when analysing the effect of positional shifts on parties' electoral performances. 4 I argue that the success of parties' positional strategies depends on the conflict dimension in questionan aspect that previous literature has not explicitly dealt with. The framework developed in this article combines different perspectives of party competition and is thus not restricted to radical left parties, but can be applied to other challenger parties as well.
The article proceeds in four steps. First, I discuss how centrifugal and centripetal movements influence the electoral fortunes of challenger parties in general. The second section presents a series of hypotheses to explain RLPs' electoral fortunes from a supply-side perspective. Then the data, operationalisation, and methodological issues are discussed. The subsequent section presents the empirical results of the statistical analysis. The last section discusses the implications of the findings.

Challenger parties' success strategies: centripetal and centrifugal movements
This article relies on the insight that challenger parties' 5 positional strategies can be distinguished into two groups. The first group of strategies rests on the Downsian idea of political competition and therefore focuses on parties' proximity to the median voter to explain electoral success and failure. From this perspective, centripetal incentives are present for all parties within a party system, as voters will vote for the ideologically closest party (Downs 1957). The second group of strategies centres on centrifugal positional strategies. Especially political parties that introduce new issues into the political debate have been found to benefit from policy stances that deviate from those proposed by competing parties (Adams et al. 2006;Ezrow 2008). Here, centrifugal incentives drive parties to more extreme policy stances, as these outsider positions are likely to increase parties' public visibility. Based on this, parties underline their ideological distinctiveness with the long-term perspective of establishing issue ownership (Wagner 2012).
In general terms, the Downsian perspective maintains that parties compete with each other by putting forward different stances on a policy dimension. Parties' vote shares are the result of the underlying voter distribution on this dimension, as peopleconceptualised as rational actors seek to maximise their utility function by voting for the party that is ideologically most proximate to them. Political parties are thus best advised to propose centrist policy positions close to the median voter to attract as many votes as possible. These centripetal incentives equally prevail for challenger parties as they 'are most likely to survive when there is an opportunity for them to cut off a large part of the support of an older party by sprouting up between it and its former voters' (Downs 1957: 128). Challenger parties thus increase their public support by shifting toward the position of their closest mainstream competitor. Nevertheless, putting forward more moderate policy stances does not imply that these parties abandon their challenger status. Instead, centripetal movements are a means for such parties to present themselves as more viable voting options to those parts of the electorate affiliated with mainstream parties.
Challenger parties do not only gain electoral successes by competing on established policy dimensions but also by introducing new issues into the political arena. They can thus act as issue entrepreneurs by underlining the relevance of issues that have been neglected by other competitors (Meguid 2008). In contrast to the Downsian approach to party competition, this perspective assumes that challenger parties do not improve their electoral performance by proclaiming more moderate political programs, but that extreme positions lead to vote gains (Adams et al. 2006(Adams et al. , 2012Ezrow et al. 2010). Non-centrist, radical stances are thus a way of attracting public attention by increasing a party's ideological distinctiveness as perceived by the voters (Hobolt and de Vries 2015). Based on this 'product differentiation' (Kitschelt 1994: 118), extreme stances might also help to establish a long-term association with a policy field in the voters' minds that may then translate into issue ownership in the long run (Petrocik 1996). Both product differentiation and (prospective) issue ownership provide strong centrifugal incentives for small parties and are considered crucial in convincing citizens to vote for challenger parties (Wagner 2012).
The question remains under which circumstances centrifugal or centripetal movements play to the electoral advantage of challenger parties. Although research widely acknowledges the importance of economic and non-economic issues in modern societies, analyses that focus on the simultaneous relevance of centrifugal and centripetal strategies in multidimensional political spaces are rare. Most of the recent work that is concerned with challenger and niche parties either focuses on the strategic behaviour of political parties on single non-economic issues such as immigration, environmental protection, or ethno-regionalism (Meguid 2008;Meyer and Miller 2015;Wagner 2011) or is concerned with positions on the general left-right dimension (Adams et al. 2006;Ezrow 2008). Few contributions deal with the impact of parties' positions on more than one dimension. However, these studies are mostly concerned with questions related to the predominance of single issues or dimensions rather than parties' strategic positioning within this space (Elias et al. 2015;Rovny 2013;Rovny and Edwards 2012). These perspectives do not take the dynamics of issue competition into account, which leads political parties to pursue more extreme or more centrist positions on different conflict dimensions.
One exception is Margit Tavits' work on 'principled' and 'pragmatic' issues. Tavits (2007) has shown that parties' policy shifts are punished and rewarded by the public in different ways depending on the policy domain in question. In her setting, voters welcome policy shifts on pragmatic, i.e. economic, issues, while disapproving shifts on principled, i.e. social, issues. Based on this, the electoral prospects of challenger parties' centrifugal and centripetal movements will vary with the conflict dimension in question. More concretely, I argue that they will depend on parties' reputation on a given issue dimension. The intensity of party-issue linkages will influence whether one of these two strategies plays to their electoral advantage. Party-issue linkages are best understood as co-occurrences of parties and issues in the public debate (Walgrave et al. 2012) and are often considered as the main way to gain issue ownership in the long run (Tresch et al. 2015;Walgrave and De Swert 2007). Thus, the more frequently political parties are mentioned in association with a specific issue, the more will voters accept them as committed actors in the policy realm in question. This linkage manifests itself as a close identification between parties and specific issues in the voters' minds and is thus closely related to the concept of associative issue ownership. 6 The varying intensity of these party-issue linkages will change the electoral prospects of centrifugal and centripetal movements. Zons (2016) shows in this context that challenger parties only benefit from high nicheness and programmatic concentration at the beginning of their lifecyclei.e. when their linkage with their core issue is still lowand that this effect fades away over time and even becomes negative the more established they arei.e. when the corresponding linkage intensity increases. Challenger parties thus profit electorally if they 'mainstream' their programmatic profiles after they have entered the parliamentary arena. In the same vein, Wagner (2012) suggests that putting forward extreme policy stances constitutes a promising strategy when less established political parties need to attract public attention. Due to their lower vote shares and a lack of media coverage, this is especially true for those challenger parties trying to gain visibility on new issues or competing for party-issue linkages with other parties (see also de Vries and Hobolt 2012: 247).
In contrast, the more a challenger party is a publicly accepted player on a particular issue dimension, the stronger its party-issue linkage will be. In some cases, the party might also own some of the issues that are part of the conflict dimension in question. In this situation, the positive effect of promoting extreme positions should diminish. Instead, challengers are incentivised to maximise their vote share to ensure their political survival (Spoon 2011). In line with the Downsian framework of party competition, abandoning policy purism in favour of more moderate positions is likely to increase their public support. Otherwise they run the risk of being perceived as political pariahs without any substantial influence on policies (March and Keith 2016b;Krause and Wagner 2019).
Combining the incentives for centrifugal and centripetal positional strategies in this way conceptualises party competition as a multi-dimensional endeavour. The outlined theoretical framework is thus sensitive to the entire 'positional toolbox' that challenger parties have at hand to gain electoral advantages. In fact, the intensity of party-issue linkages centrally alters the prospects for success of centrifugal and centripetal positional movements. Building on this argument, the next section will elaborate on how these strategies impact RLPs' electoral performances. Moreover, I will argue that these effects will not be uniform for all far left challengers, but will be conditioned by the positions put forward by mainstream left parties.

The impact of differing positional strategies on radical left party success
Radical left parties are well established actors concerning the economic conflict dimension. It is especially the long-term programmatic history of the radical left party family that centres on economic issues such as socio-economic equality, welfare, or labour rights that dominates the public perception of RLPs (March 2011). Rooted in a common communist tradition, they share a distinctly critical view of the capitalist system and are united by the aspiration to transform society while rejecting neo-liberal and market-oriented policies (Charalambous and Lamprianou 2017;Chiocchetti 2017;March 2011). Although their policy goals vary from programmatic purism as put forward by the Greek KKE or the Portuguese PCP to more pragmatic, 'social-democratised' policy stances, all members of the radical left party family have been and continue to be fierce opponents of public spending cuts and welfare-state retrenchment (Hudson 2012). Moreover, far-left parties are frequently seen as the associative issue owners when it comes to state control of the economy (Williams and Ishiyama 2018). Based on this, a strong party-issue linkage on the economic dimension can be expected for RLPs as they are consequently closely affiliated with economy-related issues in the public debate. In line with the Downsian framework of party competition, favourable partisan realignment is thus most probable if far-left parties take a more moderate stance on this conflict dimension. Moreover, centripetal movements may also signal policy pragmatism and prospective governing aspirations that might further boost far-left parties' electoral support (March and Keith 2016b). Moderating their position and pursuing a centripetal strategy is thus likely to increase their support at the ballot boxes.
Hypothesis 1: Radical left parties' vote shares increase the more they moderate their position on the economic conflict dimension.
The rise of non-economic issues since the late 1970s not only challenged traditional patterns of party competition as such (Inglehart 1977;Kriesi et al. 2008;M€ uller-Rommel 1984) but also altered the competitive environment of radical left parties. As already explained in the groundbreaking work by Herbert Kitschelt (1994), these new issues opened up new opportunities (and risks) to attract (and lose) voters in an increasingly fractionalized electorate. Although economic problems constitute the bread and butter of West European RLPs, they frequently started to pay attention to new cultural conflicts within society. With the increasing importance of socio-culturally liberal voters, RLPs underwent a profound process of programmatic re-orientation and renewal that focused on including cultural issues, such as environmentalism, gender equality, or minority rights (Hudson 2000). The most common rationale behind this process has been to attract new left-leaning voters who were no longer predominantly concerned with the question of social inequality from the perspective of class analysis, but who prioritised an idea of equality that focuses on non-economic categories (Fagerholm 2017). Hence, several RLPs have started to mobilise those segments of the middle class that are composed of younger, better educated and more libertarian voters (Chiocchetti 2017;H€ ausermann et al. 2013;Oesch 2013). 7 Although far-left parties remain divided on their willingness to respond to the rise of left-libertarian values, several studies confirm the relevance of non-economic issues for their electoral performance (see e.g. Hudson 2012; Spierings and Zaslove 2017). Some parties became issue entrepreneurs or created long-lasting electoral alliances with green parties and movements, e.g. in Iceland, Portugal, Spain, Norway, Denmark, or Greece (Gomez et al. 2016). 8 Other far-left parties are directly competing with other libertarian and green parties, such as in Germany, France, or Sweden. In both situations, RLPs need to compete for issue linkage on these less established 'New Politics' issues (see e.g. Charalambous and Lamprianou 2017). Pursuing a centrifugal strategy in this situation implies the chance to a) appeal to new voter groups and b) to establish a credible linkage with these new issues, which can brand them as prospective issue owners. Otherwise, they possibly lose culturally concerned voters to other leftist parties or even provide breeding ground for the emergence of further libertarian challengers. 9 Hypothesis 2: Radical left parties' vote shares decrease the more they moderate their position on the non-economic conflict dimension.
While the primary focus of the above explanations is about RLPs' positioning in the political space, these supply-side factors do not operate in an isolated way. As maintained by Spoon (2011), challenger parties' optimal position can only be understood in relational terms; hence together with the position of their ideologically closest mainstream competitor. While, e.g. centrist shifts can have a beneficial impact on challenger parties, this does not imply that these parties can modify their positions without limitations. Rather, they are best advised to avoid both positional convergence with mainstream parties as well as too radical a stance that discourages more voters than it attracts. Whether and how the depicted strategies affect RLPs' electoral support thus depends on the positioning of mainstream parties. Downs (1957: 131-2) argues that mainstream parties confronted with parties on their flanks are well advised to adopt some of the challenger's position to regain lost votes. Hence, mainstream party movements toward the extreme end of a conflict dimension are assumed to diminish electoral support for challenger parties. Although previous work has identified accommodative behaviour as a frequently used strategic tool of mainstream parties to deal with radical challenger parties (see e.g. Abou-Chadi and Krause 2018; van Spanje 2010), we have little empirical evidence indicating that policy-adaption indeed reduces challenger parties' vote shares. Instead, Hino (2012) has shown that challenger partiesonce they are more established actorsare unaffected or even benefit electorally if mainstream parties pay increased attention to the issues they raised. Although Hino's study focuses on issue salience rather than positions, it indicates that accommodative strategies might not affect challenger parties as expected by the literature.
Two aspects are likely to prevent the success of accommodative strategies. First, the agenda-setting power of mainstream parties in the context of multi-party competition might play an important role. As also acknowledged by Meguid (2008), mainstream parties have the capability to either put specific issues on the public agenda or to keep them out of the political debate. They therefore substantially influence which issues are at stake and which positions are viable options during election campaigns. Mainstream parties exert this influence not only on the basis of greater legislative experience and governmental efficacy but also by means of higher media coverage (Merz 2017). As a consequence, smaller political parties will suffer electorally if their ideologically closest mainstream competitor does not pay attention to issues and positions raised by them. Second, a closer stance between challenger and mainstream parties increases the chances of future cooperation in government. Hence, if mainstream parties promote non-centrist positions, voters possibly tend to vote for a challenger party because of future government options.
Based on these considerations, I assume that the success of far-left parties' centripetal and centrifugal strategies is conditioned by the policy position of their closest mainstream competitors, i.e. social democratic parties: Hypothesis 3: The effect of RLPs' centripetal shifts on the economic conflict dimension will increase the more mainstream left parties are positioned toward the leftist end of this scale.
Hypothesis 4: The effect of RLPs' centrifugal shifts on the non-economic conflict dimension will increase the more mainstream left parties are positioned toward the libertarian end of the scale.

Electoral performance and radical left parties' policy positionsempirical analysis
Operationalisation and model specification The research interest of this study on the determinants of RLPs' electoral performances requires information on vote shares and policy positions of West European RLPs between 1990 and 2017. 10 The dependent variable measures the percentage of electoral support for RLPs. 11 Following the literature on radical left parties, opposition to neoliberalism and exceeding capitalism constitutes the core ideological feature of these parties (Beaudonnet and Gomez 2017;Gomez et al. 2016;March 2011). In line with the academic consensus regarding the members of the West European radical left party family, 25 parties have been identified. 12 Radical left parties' positions on the economic and non-economic conflict dimension are the first set of independent variables of interest in this study. The MARPOR project (Volkens et al. 2018) provides information to measure parties' positions on both problem dimensions. MARPOR uses manual coding of party manifestos, which are divided into quasi-sentences, whereby each sentence is assigned to a pre-designed list of policy positions or areas. As a result, the percentage of quasi-sentences referring to each code is computed. This dataset covers the vast majority of elections and parties of interest in this study. To estimate spatial positions, salience scores for left (libertarian) and right (authoritarian) categories are combined. 13 The use of logit transformed scales as proposed by Lowe et al. (2011) solves two problems related to the construction of position scales with MARPOR data. First, former calculations have also taken the total number of quasi-sentences in a party manifesto into account. As a consequence, centrist tendencies have been indicated even though the numbers of quasi-sentences referring to leftist and rightist categories have remained constant (Fernandez-Vazquez 2014). Second, using log-odds ratios acknowledges that the marginal effect of additional quasi-sentences is decreasing in the amount that a party has already attributed to an issue.
Higher scores indicate more rightist (authoritarian) stances on the respective dimension, while lower values implicate more leftist (libertarian) positions. 14 Figure 1 shows the distribution of RLPs' economic and non-economic positions during the time period that is under investigation. Although the majority of RLPs' is positioned in the left-libertarian  quadrant of the plot, the distribution reveals considerable variation on both conflict dimensions. The question is which of the positions are electorally advantageous.
Hypotheses 3 and 4 assume that mainstream left party policy stances condition the effect of radical left parties' positional movements on both conflict dimensions. For this purpose, the positions of the electorally strongest mainstream left parties in their respective party systems are taken into account. 15 Finally, control variables that potentially influence the electoral performances of RLPs are added to the model. While macro-economic trends tend to explain only a small portion of the variation of RLPs' vote shares (see e.g. March and Rommerskirchen 2015), it is necessary to control for these factors. Given that the ideological focal point of RLPs relates to issues such as social justice, job security, and welfare extension, these parties are expected to do better in times of economic downturns. Increasing unemployment rates (International Labour Organization 2019) and decreasing GDP per capita (World Bank 2019) should influence RLPs' electoral performances positively. In the latter case, the log of the variable is used to account for its skewed distribution. Next, the composition of ruling governments is likely to affect RLPs' electoral fortunes. Participation in government is likely to result in electoral losses due to lacking capacities to realise promised policy proposals. In addition, I control for RLPs that lend support to minority governments and left-wing governments without RLP participation. A measure differentiating between these three types of government coalitions prior to the election in question will be added to the model. The reference category subsumes all governments with at least one non-left-wing party. Moreover, in some countries multiple relevant radical left parties compete for votes. Vote gains and losses of rival RLPs are likely to affect a party's electoral support. I control for this factor by adding a continuous measure indicating the vote share gained by all rival RLPs in the election in question. Lastly, I control for voter turnout since lower participation rates in national elections are likely to benefit small parties. Table  1 gives the summary statistics of all variables.
The regression model includes a lagged dependent variable to deal with problems of heteroskedasticity and autocorrelation, as suggested by Beck and Katz (1995). 16 Moreover, the model is estimated using OLS with standard errors clustered by election date and party since the error terms are possibly correlated within units and over time. Party fixed effects are added to the right-hand side of the equation to eliminate country-and party-specific heterogeneity. Eliminating in this way the between-unit variation has the essential advantage that the regression models provide an estimate of a variable's average effect within units (parties) over time. Table 2 presents the results for the regression models. Model 1 includes all variables apart from the independent variables of interest. Model 2 tests hypotheses 1 and 2 by controlling for the positions of RLPs and mainstream left parties on both conflict dimensions. Model 3 adds two interaction effects to test hypotheses 3 and 4.

Results
First, the government indicators show a clear pattern across all models. Participating in coalition governments plays to the electoral disadvantage of RLPsthe corresponding b-coefficients are negative and significant in all models. It is also substantial in size, indicating that RLPs' electoral support decreases by about six percentage points after government participation. Interestingly, supporting minority governments does not have a similar impactthe effects are close to zero and indicate that this strategy comes with less electoral costs for RLPs. The coefficients of the leftwing government indicator (without RLP participation) have a negative impact. Hence, RLPs suffer electorally if social democrats are in a ruling position (compared to governments with right-wing participation). This is in line with previous findings showing that RLPs especially profit electorally if right-wing governments have been in power (March and Rommerskirchen 2015).
Next, unemployment shows a positive impact on RLPs' vote shares. This is in line with previous expectations claiming that far-left parties whose core issues relate to questions of social justice and equality do better in times of socio-economic downturns. In contrast, GDP per capita shows a statistically significant positive impact on RLP vote share indicating that RLPs' public support rises if the economy is doing better. Also moderate in size, this effect contradicts previous expectations. This seeming paradox possibly reflects the unequal development of economic growth and income inequality in Western democracies. Rising economic growth is often linked to increasing income inequality, causing increasing levels of dissatisfaction among voters (e.g. Oishi and Kesebir 2015). Under these circumstances, RLPs' might attract those voter groups that are critical of these trends. Considering the further control variables, the b-coefficients of rival RLPs' vote shares indicate, as expected, a negative impact on the dependent variable. In contrast, the effects of voter turnout are close to zero. Let us now turn to the impact of this article's variables of interest: RLPs' policy positions as shown in models 2 and 3. Two hypotheses were formulated regarding the impact of RLP positions. While more moderate positions on the economic dimension should influence RLPs' electoral fortunes positively, the contrary is expected for the non-economic dimension. These two hypotheses find support in model 2. The impact of promoting more moderate policy positions on the economic and non-economic dimension does, indeed, differ. More moderate positions on the economic dimension have a positive impact on RLPs' vote shares. The statistically positive effect indicates that RLPs benefit from putting forward a more centrist stance when it comes to issues such as the welfare state, redistribution, or state interventionism into the economy. On a scale ranging empirically from À5 to 1, if a RLP shifts by one unit to the centre of the economic conflict dimension, its electoral support increases by roughly 1:5 percentage points. In contrast, moderating their position on non-economic issues influences RLPs' electoral support negatively. Here, the scale ranges empirically from À5 to 4:5 and a one unit change comes with a :5 percentage point decrease. Hence, as predicted by the theoretical framework, radical left parties indeed benefit from a more moderate stance on economic issues while they suffer if they pursue a similar strategy on the non-economic conflict dimension.
To investigate the effects proposed in the remaining hypotheses, two interaction terms have been added to the model. As stated in hypotheses 3 and 4, the identified effects should be moderated by the positions of mainstream left parties. The effect sizes of the base terms remain stable and statistically significant in the third model. Considering the interaction terms, mainstream left parties' positions only matter with regard to the economic dimension. Here, the b-coefficient isas expectednegative and statistically significant on the 1% level. In contrast, the effect of RLPs' policy positions on non-economic issues is not moderated by the policy stances of mainstream left parties. The coefficient of the interaction term is not statistically significant and close to zero. Hence, while RLPs lose support at the ballot boxes if they promote more moderate stances on non-economic issues, this electoral punishment is not conditional on the strategic behaviour of competing social democratic parties.
Concerning hypothesis 3, Figure 2 shows the corresponding marginal effect plot to ease interpretation. The marginal effect is positive and statistically significant if mainstream left parties promote more leftist stances on economic issues. Moreover, this effect becomes negative if social democrats promote more rightist policy positions, thus undermining the legitimacy of leftist ideas concerning the economy. Increased levels of positional competition on the left side of the party spectrum therefore play to the electoral advantage of far-left parties. The corresponding dot plot shows the distribution of mainstream left parties' positions along the economic conflict dimension. As can be seen, the majority of social democratic parties areas expectedpositioned around the centre of the scale. However, one case possibly drives the identified effect, as it is positioned toward the rightist end of the scale. Re-running the analysis while excluding this case does not affect the empirical results. 17 The presented results illustrate that RLPs' electoral fortunes indeed depend on different electoral strategies. The success of centrifugal and centripetal movements varies with the conflict axis in question. To substantiate this finding, a variety of additional models have been estimated.
First, although Sinn F ein is a member of the GUE/NGL group in the European Parliament since 2004, several studies on the far left exclude this party from their analyses due to its primary commitment to questions of national unity and independence (e.g. Beaudonnet and Gomez 2017;Gomez et al. 2016). Dropping this party from the analysis does not change the results in substantial ways. The same holds if Syrizathe farleft party with the most drastic increase in its vote share during the observation periodis excluded from the analysis. 18 Hence, the empirical results are not driven by possibly influential cases.
Second, including unit-fixed effects and a lagged dependent variable at the same time in the model possibly produces inconsistent estimates (Nickell 1981). Excluding the lagged dependent variable from the main models does not affect the substantial interpretation of the empirical results. Next, using clustered standard errors for clusters with small sample sizes might unexpectedly deflate the size of the standard errors. For that reason, the models have been re-estimated while using only partyclustered standard errors. Lastly, I also re-run the analysis using country- fixed effects instead of party-fixed effects. In all these cases, the results do not change substantially. 19 Third, Table A6 in the Online appendix tests for alternative specifications of RLPs' and MLPs' policy positions. Entrepreneurial strategies possibly do not only depend on promoting extreme positions but also on the salience political parties attribute to these issues (Hobolt and de Vries 2015). The robustness of the results has thus been tested by using an alternative measure for issue entrepreneurship that combines RLPs' noneconomic policy stances with the corresponding salience scores. 20 Again, the substantial interpretation of the results remains unaffected if this alternative specification is applied. Lastly, it is possible that social democratic parties are not always RLPs' most relevant mainstream competitors. I applied a more dynamic approach by replacing social democrats' positions with those of the most leftist/libertarian non-RLP party in an election. As a consequence, the positions of green parties are used in some cases. Nevertheless, the results again remain stable and do not change in fundamental ways.
To sum up, radical left parties' electoral fortunes depend on different strategies regarding the economic and the non-economic conflict dimension. While they benefit from centripetal movements on economic-related issues, they suffer electorally if they pursue the same strategy on non-economic issues. These different effects can be explained if the perspective of issue competition is taken into account. Since RLPs have built a stable party-issue linkage in the economic realm, they benefit from moderating their stances on these issues with the goal of attracting social democratic voters. Concerning non-economic issues, they compete with other leftist and libertarian parties for a credible party-issue linkage in the voters' minds. For that reason, they are punished at the polls if they promote moderate stances that fail to present clearly distinguishable programmatic alternatives. The effect of these strategic considerations varies partly with the positioning of mainstream left parties. In contrast to the widely assumed effect of accommodative mainstream party behaviour (Meguid 2008), more moderate positions of social democratic parties decrease the marginal effect of RLPs' positional strategies on the economic dimension. If the mainstream left takes a noncentrist stance, they seem to legitimise leftist ideas on the economy and thus foster the positive effect of RLPs' positional shifts. Hence, far-left challengers profit electorally if mainstream parties politicise their core issues and signal their relevance to the electorate. Considering the non-economic dimension, this effect could not be observed. Here, RLPs are punished electorally if they moderate their policy stance. This is true independent of the positional strategy put forward by mainstream parties. RLPs are thus best advised to promote an outsider profile on this issue dimension.

Conclusion
This article has investigated the impact of RLPs' policy positions on their electoral support in the context of multi-dimensional party competition. Moreover, it has shown how these effects vary with the positions of mainstream left parties. These findings have important implications for the analysis of multi-party competition and challenger party success. This is the first study that sheds light on the electoral supply side as a decisive factor explaining the electoral success and failure of West European radical left parties. While previous work has predominantly focused on demand-side factors such as voters' attitudes and socio-demographic factors (e.g. Charalambous and Lamprianou 2017;Gomez et al. 2016;March and Rommerskirchen 2015;Ramiro 2016;Visser et al. 2014), the presented work adds the electoral supply side to these approaches. Future analyses of parties' electoral support will thus have to focus on both voters' preferences and parties' offerings.
Second, the results of this study highlight the importance of considering multiple positional strategies simultaneously when explaining parties' electoral performances. Previous studies have found that niche and challenger parties do better electorally when they promote radical policy stances (e.g. Adams et al. 2006;Ezrow 2008). However, these studies have been constrained to inspecting parties' policy shifts on a single left-right dimension. Theory on party competition and voting behaviour gives little reason to assume that parties' policy shifts have a similar effect on all subdimensions of political conflict. Rather, strategies that work on one conflict axis may backfire if applied to a different dimension (e.g. Tavits 2007). The study at hand provides a more fine-grained picture of this relationship and shows that parties' reputation on specific conflict dimensions (party-issue linkages), along with the policy positions of competing parties, play a crucial role when explaining the success and failure of positional strategies. The results of this article provide empirical evidence that simultaneously considering the prospects of centrifugal and centripetal strategies is crucial to understanding the electoral support of West European radical left (and likely other challenger) parties.
Third, recent scholarly work suggests that voters do not adjust their perceptions according to parties' position shifts (Adams et al. 2011(Adams et al. , 2014Fernandez-Vazquez 2014). The findings of this article paint a more optimistic picture of the determinants of voters' choices at the ballot box: parties' election results do vary substantially with the positions these parties promote. Similar to Spoon (2011) and Ezrow (2008), this article demonstrates this effect for the group of non-mainstream challenger parties. The question remains as to whether the voters of challenger parties tend to observe parties' policy shifts more accurately than voters of mainstream parties. Analysing far-left voters' perceptions of RLPs' positional movements would help to understand the individual-level mechanisms that are at the basis of the effects identified in this study. Moreover, it seems a promising endeavour to investigate whether and under which circumstances centripetal movements pay off electorally for other challenger party families, such as green parties and the radical right. Notes 1. In a few cases, such as the United Kingdom, Belgium or Austria, RLPs have failed to gain electoral ground. 2. Figure A1 in the Online appendix gives a more detailed overview of the development of RLPs' electoral support in Western Europe. 3. Such as in Greece since 2012; in Spain in 2015; in Ireland in 2016; in the Netherlands in 2017; and in France in 2017. 4. Up to now, scholarly work has first and foremost focused on party families whose public support rests predominantly on politicising issues within the non-economic dimension of conflictsuch as green, radical right and ethno-territorial parties. See the growing literature on niche parties: e.g. Abou-Chadi (2016), Elias et al. (2015), or Meguid (2008). 5. The term challenger party refers to those parties that usually do not participate in government and compete by promoting extreme or 'niche' issue positions. Both terms, challenger and niche party, are used interchangeably in this article. See Hobolt and Tilley (2016: 973-6) for a detailed discussion of challenger parties and their characteristics. 6. Here, one common survey question to measure the extent of associative issue ownership is as follows: 'Can you indicate for the following issue which party you spontaneously think about when you think about the issue? This does not have to be the party whose position on that issue you find most compelling' (Walgrave et al. 2012). 7. In support of this, recent studies inspecting the effect of socio-demographic characteristics on RLP support find a U-shaped curvilinear relationship between voters' levels of education and support for radical left parties and ideas (see e.g. Ramiro 2016;Visser et al. 2014). 8. See the country studies in Hudson (2012). 9. Some of the most successful European green parties were able to flourish while radical left challengers have been absent or electorally insignificant, such as in Austria, Belgium, Germany, or Luxembourg. 10. See the Online appendix for a discussion of the spatial and temporal limitations of the analysis. 11. Unless otherwise stated, all variables are provided by the database "Parties, Elections and Governments" of the WZB Berlin Social Science Center (WZB 2019). 12. See Table A1 in the Online appendix for the corresponding list of radical left parties. 13. A vivid debate surrounds the calculation of positional scores from the manifesto scores. It is beyond the scope of this article to discuss the various proposals and their criticisms in detail. See e.g. Gemenis (2013) and Meyer (2013) for critical evaluations of the MARPOR data. 14. See the Online appendix (Tables A2 and A3) for a discussion of the MARPOR items selected for constructing the economic and noneconomic scale. 15. Table A1 in the Online appendix lists all mainstream left parties that are included in the analysis. 16. The corresponding Durbin-Watson-Test statistic indicates that serial correlation is present if no lagged dependent variable is included. Importantly, this is no longer the case once the lagged dependent variable is added to the model. 17. See models 1 and 2 in Table A4 in the Online appendix 18. See models 3 to 6 in Table A4. 19. See Table A5 in the Online appendix for the corresponding regression results. 20. Hobolt and de Vries (2015) propose to estimate this entrepreneurship measure by multiplying parties' salience scores on an issue with the distance between its party position and the average position of all parties in the system on an issue. As the presented analysis already controls for the positions of mainstream left parties, I calculate this score by multiplying RLPs' salience and position scores on non-economic issues. To ease interpretation of the resulting regression coefficient, both values are divided by 10.