The last few decades have been marked by a pan-European rise in radical parties. To the right on the left–right heuristic, the radical right family has increased its electoral and political strength significantly in both Western and – more inconsistently – Central and Eastern Europe, and several parties – the Italian Northern League (2001–6, 2008–11), the Bulgarian Attack (2009–14), the Dutch Party for Freedom (2010–12) and the Norwegian Progress Party (2013–) to name but a few – have gained experience of executive power or acted as support parties (Akkerman et al. Reference Akkerman, de Lange, Rooduijn, Akkerman, de Lange and Rooduijn2016a; Minkenberg Reference Minkenberg2017: Chs 5–6). A similar, albeit weaker, trend can also be observed at the other end of the left–right heuristic: the electoral support of the European radical left family has slowly recovered from the rock-bottom levels reached in the (early) 1990s, and several parties – such as, for instance, the communist parties of France (1997–2002) and Portugal (2015–), the Finnish Left Alliance (1995–2003, 2011–14) and the Icelandic Left-Green Movement (2009–13) – have either been included in coalitions or supported minority governments (Chiocchetti Reference Chiocchetti2017: Ch. 3; March Reference March, March and Keith2016).
The growing importance of radical parties in European government formation processes (see e.g. Akkerman et al. Reference Akkerman, de Lange and Rooduijn2016b; Albertazzi and McDonnell Reference Albertazzi and McDonnell2015; Bale and Dunphy Reference Bale and Dunphy2011; de Lange Reference de Lange2008, Reference de Lange2012; Dunphy and Bale Reference Dunphy and Bale2011; Hough and Verge Reference Hough and Verge2009; March and Keith Reference March and Keith2016; Minkenberg Reference Minkenberg2013, Reference Minkenberg2017; Olsen et al. Reference Olsen, Koß and Hough2010b) is puzzling, especially in light of the allegedly (see March Reference March2012: 8–9; Mudde Reference Mudde2007: 31) anti-system nature of these actors. Although there have been signs of deradicalization within both families in recent years (e.g. Akkerman Reference Akkerman, Akkerman, de Lange and Rooduijn2016a; Keith and March Reference Keith, March, March and Keith2016a), radical right and radical left parties still tend to – at least rhetorically – oppose the prevailing political order and undermine the legitimacy of the existing political regime (Sartori Reference Sartori1976: 133). In the words of coalition theorists, they should hence be ‘non-coalitionable’; they should, as a general rule, be ‘excluded [from government] through the reactions of the other [pro-system] parties’ (Budge and Keman Reference Budge and Keman1990: 49, 52).
Against this backdrop, this article sets out to conduct a broad comparative assessment of how – or, put differently, under what circumstances – radical right and radical left parties are able to overcome the obstacles associated with governing and, accordingly, enter coalition governments (as junior coalition partners).Footnote 1 By means of a fuzzy-set qualitative comparative analysis (fsQCA) of 37 radical parties in 207 coalition formation instances in 22 European countries between 1990 and 2017, the article sheds light on the different paths to government inclusion (and exclusion) of radical actors. The empirical evidence indicates that electoral success in combination with ideological similarity with a weak prime minister party is sufficient for government inclusion. The paths to government exclusion, by contrast, underline the importance of ideological distance in combination with size-related factors. Taken together, the findings provide important new insights to scholarly work on (radical) political parties and party competition, to research on government formation and government participation and, by extension, to the analysis of the state and future of democratic governance.
This article is organized into five sections. The next section provides a review of previous theoretical and empirical research on government formation and, in particular, on radical parties’ coalition behaviour. The third section discusses case selection, introduces configurational comparative methods and presents the data, the measurement techniques and the calibration strategies. The fourth section reports and discusses the results from the empirical analysis while the fifth section concludes and suggests avenues for further research.
Radical parties in power: how do they get in?
What do we know about the government participation of parties occupying the fringes of the conventional left–right distinction? Or, more precisely: Which are the factors that may enable radical right and radical left parties to – despite their inherent criticism of current liberal democratic principles – enter government coalitions in liberal democracies? To approach this problem, I apply a theory-testing design. The main focus is on a number of propositions related to the ‘size and ideology’ framework that emphasizes the role of party size (Riker Reference Riker1962: 32–3) and party ideology (de Swaan Reference de Swaan1973: 88) in the coalition formation process (see online Appendix A for a brief review of this literature). Inspired by previous observations indicating that the phenomenon under investigation is both complex and multicausal (e.g. de Lange Reference de Lange2008: 15–17; Olsen et al. Reference Olsen, Hough, Koß, Olsen, Koß and Hough2010a: 173–4), I expect conjunctural patterns where combinations of multiple factors explain a single outcome. Next, I review the existing empirical research on the behaviour of radical right and radical left coalition candidates and identify a number of potentially important (non-idiosyncratic) explanatory factors.
The first three of these factors are related to party size. As emphasized by Paul Warwick (Reference Warwick1996), considerable parliamentary strength is a liability rather than an asset when it comes to prospective coalition partners’ chances of getting into office. Instead of large parties, the formateur party (henceforth the prime minister party)Footnote 2 tends – in order to maximize both its own influence and the working capacity of the coalition – to prefer medium-sized partners (see de Lange Reference de Lange2008 and, especially, Olsen et al. Reference Olsen, Hough, Koß, Olsen, Koß and Hough2010a) such as, for example, the Finnish Left Alliance (in coalition 1995–2003, see Dunphy Reference Dunphy, Olsen, Koß and Hough2010) or the Slovak National Party (in coalition 2006–10, see Gyárfašová and Mesežnikov Reference Gyárfášová, Mesežnikov and Minkenberg2015). Accordingly, my first expectation is: if the parliamentary strength of the radical right (left) party is moderate – that is, neither minuscule nor huge – the party will gain access to government.
Another factor of potential importance is the recent electoral trend of the radical right or radical left party in question (see Mattila and Raunio Reference Mattila and Raunio2004). The assumption here is that a party that increased its share of seats in the previous election is – regardless of its actual strength – a more serious candidate for office than a party with less fortunate results (see especially de Lange Reference de Lange2008, but see also Olsen et al. Reference Olsen, Hough, Koß, Olsen, Koß and Hough2010a). The Swedish Left Party, for example, managed to sign a support contract with the social democratic minority government after nearly doubling its size in the 1998 general election (Koß Reference Koß, Olsen, Koß and Hough2010), and the Italian Northern League got in to office (in 2008, together with its long-time centre-right ally) after a highly successful result (Albertazzi and McDonnell Reference Albertazzi and McDonnell2015: Ch. 5). Hence, the second expectation is: if the radical right (left) party has increased its share of seats in the most recent national election it will gain access to government.
In addition to factors related to their own electoral fortunes, radical right and radical left parties’ abilities to gain office or to become support parties may also be influenced by the size of the prime minister party (de Lange Reference de Lange2008; Olsen et al. Reference Olsen, Hough, Koß, Olsen, Koß and Hough2010a; see also Strøm et al. Reference Strøm, Budge and Laver1994; Warwick Reference Warwick1996). The (by Scandinavian standards) rather weak Danish Social Democrats, for example, had to rely on support from the Socialist People's Party (and from the Red–Green Alliance) in the 1990s (Christensen Reference Christensen, Olsen, Koß and Hough2010) while the minority coalition led by the Dutch (centre-right) People's Party for Freedom and Democracy counted on support from the Party for Freedom between 2010 and 2012 (Akkerman Reference Akkerman, Akkerman, de Lange and Rooduijn2016b). The expectation is, thus: if the prime minister party is comparatively weak, the radical right (left) party will gain access to government.
The next three factors are related to ideology. The first of these focuses, rather straightforwardly, on the general left–right location of the prime minister party. As a leading actor in the government, the prime minister party plays an important role in the selection of coalition partners. Since radical parties tend to be shunned by parties located at or close to the opposite end of the general left–right heuristic, their chance of getting into office often hinges upon the prime minister party being located on the same side of this super-dimension (e.g. Bale and Dunphy Reference Bale and Dunphy2011; Minkenberg Reference Minkenberg2013; see also Warwick Reference Warwick1996). Hence, radical right parties govern with (or support) mainstream right parties, as in, for example, Latvia (2011–14, see Auers and Kasekamp Reference Auers, Kasekamp and Minkenberg2015) and Norway (2013–, see Jupskås Reference Jupskås, Akkerman, de Lange and Rooduijn2016), and radical left parties govern with – or support – mainstream left parties, as in Denmark (2011–14, see Christensen Reference Christensen, Olsen, Koß and Hough2010) or Italy (1996–98, see Newell Reference Newell, Olsen, Koß and Hough2010). Accordingly, I expect: if a centre-right (centre-left) party leads the government, the radical right (left) party will gain access to it.
A second ideological factor worth considering is party radicalness on the most preferred policy dimension (see e.g. Döring and Hellström Reference Döring and Hellström2013; Savage Reference Savage2014). Previous observations indicate that less rigid sociocultural rightism (radical right) and socioeconomic leftism (radical left) can open up possibilities for government participation. The Austrian Freedom Party, for example, became coalitionable (in 2000) after moderating its sociocultural positions (Heinisch and Hauser Reference Heinisch, Hauser, Akkerman, de Lange and Rooduijn2016). Similarly, the Spanish United Left became acceptable as a support party (in 2004) after a gradual turn towards more ‘realistic’ socioeconomic policies (Verge Reference Verge, Olsen, Koß and Hough2010). Hence, I expect: if the radical right (left) party takes a moderate position on the sociocultural (socioeconomic) dimension it will gain access to government.
The third ideological factor of interest is the distance between the prime minister party and the radical party, again on the most preferred policy dimension (see e.g. Warwick Reference Warwick1996). Regarding the radical right, previous research shows that political affinity between the radical right party and the prime minister party, especially on sociocultural issues, can open up possibilities for radical right government participation. In Poland (2005) and Slovakia (2006), for example, the prime minister parties (the conservative populist Law and Justice and the social democratic Direction, respectively) shared similar restrictive views on nationalism and xenophobia to their radical right coalition partners League of Polish Families and Slovak National Party (Gyárfášová and Mesežnikov Reference Gyárfášová, Mesežnikov and Minkenberg2015; Kasprowicz Reference Kasprowicz and Minkenberg2015). A similar pattern can be observed on the left. Here, an increasing distance between a right-turning centre-right and a left-turning centre-left – as in, for example, Portugal (Freire and Lisi Reference Freire, Lisi, March and Keith2016) – have increased the chances for radical left parties to become serious candidates for government. My final expectation is hence: if the radical right (left) party locates itself close to, or more centrally than, the prime minister party on a sociocultural (socioeconomic) dimension it will gain access to government.
Taken together, the six factors introduced in the discussion above are all assumed to be relevant when seeking to understand the fortunes of radical right and radical left parties in government formation processes. As the structure of the hypotheses (‘if-then’) indicates, their relevance is assumed to be implicational rather than covariational (Thiem et al. Reference Thiem, Baumgartner and Bol2016; see also Schneider and Wagemann Reference Schneider and Wagemann2012). I expect, hence, that the endogenous factor – the governmental status of radical parties – is best understood by focusing on necessary, sufficient and INUSFootnote 3 conditions and by allowing for conjunctural, equifinal and asymmetric causation. To reiterate: the general assumption – and, indeed, the core hypothesis of this study – is that the factors related to party size and party ideology introduced above work together to produce different, mutually non-exclusive, paths to government (non)-participation for radical parties.
Cases, methods and data
Case selection
The study includes 37 radical parties from 22 European countries, observed at, in total, 207 government formation instances between 1990 and 2017. The cases of interest are relevant radical right and radical left parties. Following Cas Mudde (Reference Mudde, Backes and Jesse2005), radicalism is defined as opposition to key features of liberal democracy and support of a root (Latin radix) and branch transformation of society.Footnote 4 Radical right parties are hence ‘radical’ and ‘right’ in the sense that they oppose core liberal democratic values such as pluralism and multiculturalism and, instead, pursue nativist policies associated with the sociocultural right pole of the multidimensional left–right heuristic (Mudde Reference Mudde1996, Reference Mudde2007: Ch. 1). Radical left parties, by contrast, are ‘radical’ and ‘left’ because they oppose the underlying (capitalist) socioeconomic structure of liberal democracies and, consequently, seek to advance socialist policies associated with the socioeconomic left pole of the multidimensional left–right heuristic (Keith and March Reference Keith, March, March and Keith2016b; March Reference March2012: Ch. 1; March and Mudde Reference March and Mudde2005).
Regarding relevance, I follow the ‘possibility principle’ (Goertz Reference Goertz2006: 186) and include, in the main analysis, only radical parties that have either been included in or supported at least one government during the period of interest (i.e. parties where a positive outcome is demonstrably possible).Footnote 5 This strategy excludes ostracized (e.g. the French National Front) and ideologically rigid (e.g. the Communist Party of Greece) parties that have, at least heretofore, been ‘impossible’ coalition partners (even under very favourable conditions). It also excludes minor parties in systems where one of two major parties generally form the government (e.g. the British UKIP). The inclusion of ostracized and rigid parties in the analysis does not significantly alter the results, as shown in online Appendix D3.
To identify radical right and radical left parties, I rely on family classifications in the Manifesto Project Data (Volkens et al. Reference Volkens2017) and in the Chapel Hill Expert Surveys (Bakker et al. Reference Bakker2015). Other sources guiding the identification of radical parties (e.g. in case of discrepancies between or apparent misclassifications in the above-mentioned sources) are previous authoritative works on the radical right (see Mudde Reference Mudde2007) and the radical left (see March Reference March2012). To identify relevant radical parties, I utilize the ParlGov database (Döring and Manow Reference Döring and Manow2016) and the information provided by Tjitske Akkerman et al. (Reference Akkerman, de Lange, Rooduijn, Akkerman, de Lange and Rooduijn2016a: 3), Luke March (Reference March, March and Keith2016: 42) and Michael Minkenberg (Reference Minkenberg2017: 124). The main data set includes observations of 20 radical right and 17 radical left parties.
Regarding single observations, I consider a new government to have been formed if any of the following three events occurs: a national election, a change of the party holding the prime ministership and a major change in the composition of the governing parties. In contrast to much general research on government formation (e.g. Budge and Keman Reference Budge and Keman1990: 14–15; Müller and Strøm Reference Müller, Strøm, Müller and Strøm2000: 12), I hence do not consider simple prime minister swaps where the new prime minister represents the same party as his/her predecessor. Moreover, I also find it feasible to ignore minor changes in government composition, such as mergers of two (or more) of the government parties and withdrawals of minor parties. Finally, I exclude caretaker governments, governments where the prime minister represents a radical party and – due to missing data on party positions – governments formed after a number of recently held elections.Footnote 6 This leads to 108 observations of radical right parties and 99 observations of radical left parties (for details, see online Appendix B).
CCMs: what, why and how?
Configurational comparative methods (CCMs) consist of a number of set-theoretic techniques for the causal analysis of configurational data, introduced to the social sciences by, primarily, Charles C. Ragin (Reference Ragin1987, Reference Ragin2000, Reference Ragin2008). The most well-known and used set-theoretic techniques are crisp- and fuzzy-set Qualitative Comparative Analysis (csQCA and fsQCA, respectively). While csQCA requires the objects to be either members (1) or non-members (0) of a set, fsQCA allows for partial set membership; objects are, hence, also allowed to take values in the interval between the endpoints of full membership (1) and full non-membership (0).
In this study, I use fsQCA for a number of reasons. Following Michael Smithson and Jay Verkuilen (Reference Smithson and Verkuilen2006: 1–2), I note that (techniques based on) fuzzy sets are able to handle vagueness in a systematic way, to rigorously combine set-wise thinking with continuous variables and to take account of both the categorical (qualitative) and the dimensional (quantitative) character of objects. The main reasons for relying on fsQCA are, however, related to the nature of the research problem. As noted in the previous section, recent scholarship on radical parties’ government participation (e.g. de Lange Reference de Lange2008: 15–17; Olsen et al. Reference Olsen, Hough, Koß, Olsen, Koß and Hough2010a: 173–4) as well as studies on government formation in general (e.g. Andeweg et al. Reference Andeweg, De Winter and Dumont2011) suggest that office entry follows a complex pattern where several factors work together in different ways to produce the outcome. Because of this complexity, configurational methods such as fsQCA are good alternatives to standard conditional and mixed logit models and qualitative case studies. To put it differently, fsQCA is an appropriate choice when the relationships of interest are assumed to be set-theoretic (logical) and multicausal (configurational and equifinal) in nature (Schneider and Wagemann Reference Schneider and Wagemann2012: 8–13).
An fsQCA consists of three analytical phases (Thiem Reference Thiem2017). The first phase calibrates the raw data into configurational data and arranges it in a truth table, the second phase minimizes the truth table into a prime implicant (PI) chart and the third phase decomposes this chart in order to obtain the final solutions. The calibration process (i.e. the first part of the first analytical phase) is discussed in the following subsection. In the subsequent (fourth) section, I then turn to the key analytical phases: the construction (second part of first phase) and analysis (second and third phase) of the truth table.Footnote 7
Data, measurement and calibration
Irrespective of method, investigators occupied with empirical research need to concern themselves with issues related to the definition and measurement of key concepts. In fsQCA, an additional step is also required: the base variables must be calibrated into sets. The calibration of base variable X into fuzzy set A is done using a membership function m A(x) that indexes the degree, usually a number within the closed unit interval [0, 1], to which every x ∈ X belongs to A.Footnote 8 The membership assignment process can follow three different strategies (Verkuilen Reference Verkuilen2005). The direct assignment method uses only substantial human expertise, the indirect assignment method applies a statistical model to human judgements and the transformational assignment method relies on a theoretically motivated mapping. In this article, I use both the direct and the transformational assignment method. To calibrate the (categorical) base variable underlying the endogenous factor, I use the direct assignment method. Of the base variables underlying the six exogenous factors, five are numerical and one categorical. The numerical base variables are calibrated using (linear) transformational assignment, and the remaining categorical base variable using direct assignment.
Endogenous factor
The endogenous factor in this study is the government status (G) of each relevant radical party at each coalition formation instance. To (directly) calibrate this factor, I proceed in three steps. First, I define a government party as a party that has at least one minister with full voting rights (e.g. Budge and Keman Reference Budge and Keman1990: 73; Müller and Strøm Reference Müller, Strøm, Müller and Strøm2000: 16). To identify such parties, I make use of the ParlGov database (Döring and Manow Reference Döring and Manow2016) and, more precisely, its Boolean indicator ‘cabinet party’. A score of 1 on this indicator indicates that the party observation has at least one minister and, hence, that it is fully in the government (m G[in] = 1). In the second step, I focus on party observations scoring 0 on the above-mentioned indicator. While most of these observations can be classified as fully out of the government (m G[out] = 0), some are more ambiguous. These ambiguous observations are the instances where a party does not formally enter government, but, instead, acts as support (or pseudo-opposition) party – that is, agrees to lend a (minority) government explicit, comprehensive and permanent support in exchange for some other concession than access to government portfolios (Strøm Reference Strøm1990: 61–2; see also Andeweg Reference Andeweg, Müller and Narud2013: 108). To identify instances where radical parties have acted as support parties, I rely on information given in previous work by Akkerman et al. (Reference Akkerman, de Lange, Rooduijn, Akkerman, de Lange and Rooduijn2016a: 3), March (Reference March, March and Keith2016: 42) and Minkenberg (Reference Minkenberg2017: 124). The identified instances are examined in more detail in the third step of the calibration process. In this final step, I focus on whether the support party occupies a key position and whether it is the only support party. The most influential support parties are those that: (1) act as the only support party and (2) are able to tip the balance of power in favour of the opposition by retracting its support. These key support parties can – without taking other support parties’ desiderata into account – demand a great influence upon government measures and also determine the future of the supported government on its own. As a consequence, they are considered as clearly more in the government than out of it (m G[strongsupport] = 0.70). If a support party is only one of several and, in addition, not in a key position, its influence on the government is weaker. It is, however, considered to be slightly more in the set of government parties than out of it (m G[weaksupport] = 0.51). Support parties that are either in a key position or act as the only support party (but not both) are considered as somewhat more in the government than out of it (m G[support] = 0.60). A diagram illustrating the reasoning behind the calibration of this five-value fuzzy set is given in Figure C1 in online Appendix C.
Exogenous factors
Next, I focus on the calibration of the six exogenous factors. First, I have argued that medium-sized radical parties (M) tend to be included in coalition governments. In order to measure parliamentary strength, I use data from the ParlGov database and divide the number of seats that the radical party holds in (the lower house of) the national parliament (the variable ‘seats’ in ParlGov) by the total number of seats (the variable ‘election seats total’ in ParlGov). The resulting base variable is numerical, and the membership assignment process is best performed using the transformational assignment method. Since the focus is on medium-sized parties, the set membership scores should first increase, and then decrease, as the values of the base variable increase. The underlying concept is, hence, a positive mid-point concept. Mid-point concepts require six thresholds: lower (τin1) and upper (τin2) thresholds for full inclusion, lower (τex1) and upper (τex2) thresholds for full exclusion and lower (τcr1) and upper (τcr2) crossover points. Despite the somewhat varying definitions used in previous research, a party holding more than 10% but less than 15% of the parliamentary seats can undoubtedly be regarded as fully in the set of medium-sized parties. Hence, τin1 = 10.0 and τin2 = 15.0. Similarly, it should be fairly uncontroversial to classify parties with less than 2.5% of the seats as small and, hence, non-medium-sized (τex1 = 2.5), and, likewise, parties with more than 22.5% of the seats as large non-medium-sized parties (τex2 = 22.5). The crossover points τcr1 and τcr2, finally, are set at 5 and 20, respectively.Footnote 9
The second argument holds that radical parties that succeeded in the previous election (S) tend to be included in coalition governments. Again, the data stem from the ParlGov database, and the base variable is constructed by subtracting the share of seats won by the radical party of interest at election e − 1 from the share of seats won by the same party at the most recent election e (for calculation of seat shares, see above). The resulting base variable is again numerical, with positive (negative) values indicating that the party's share of seats increased (decreased), and with 0 indicating that the seat share remained unchanged (or, alternatively, that the party was not represented in parliament during the previous election term). Transformational calibration of (positive) end-point concepts assumes the establishment of three thresholds – one for full inclusion (τin), one for full exclusion (τex) and one crossover point (τcr). The crossover point τcr is, rather straightforwardly, set at −0.01. Negative values hence indicate that the party is (more) out of (than in) the set of winning parties while neutral and positive values imply that the party is (more) in (than out of) the set of winning parties. τex is set at −1.0 (i.e. losses of one percentage point or more) since even a comparatively small loss should be enough for a party to be fully out of S. τin is set at 4.0 (i.e. wins of four percentage points or more).
The third argument is that radical parties tend to be included in coalition governments when the prime minister party is comparatively weak (W). As in the calculation of M above, I use data from ParlGov and construct a numerical base variable by dividing the number of seats held by the prime minister party (or the largest party in government, if the prime minister is not affiliated with any political party) by the total number of seats. The thresholds used in the transformational calibration of this negative end-point concept are τex = 35.0, τcr = 30.0 and τin = 25.0: prime minister parties with more than 35% of the seats are, hence, undoubtedly out of the set of comparatively weak prime minister parties while parties with less than 25% of the seats are undoubtedly in the same set.
Moving to the factors related to ideology, the fourth argument holds that radical right (left) parties tend to be included in coalition governments when the government is led by a centre-right (centre-left) party (C). The base variable here is binary, and the ‘calibration’ is performed using direct assignment. The set membership score in the crisp set C is, hence, 1 if the government is led by a party locating itself on the same side of the left–right heuristic as the radical party of interest (m C[same] = 1) and 0 if it is not (m C[different] = 0). Here, agrarian, Christian democratic, conservative and liberal parties are classified as centre-right parties, and social democratic/socialist (and green) parties as centre-left parties.
According to the fifth argument, radical right (left) parties tend to be included in coalition governments when they take moderate (R) positions on the sociocultural (socioeconomic) dimension. The data source here is the Manifesto Project Data, which provides estimates of how much space parties’ election manifestos assign to different policy categories (Volkens et al. Reference Volkens2017). To measure parties’ sociocultural positions, I focus on four categories associated with sociocultural rightism and four categories associated with sociocultural leftism. To capture parties’ socioeconomic positions, I focus on five socioeconomic right and five socioeconomic left categories (see Tables C3 and C4 in online Appendix C for details). The resulting two numerical base variables illustrate party positions on the sociocultural and socioeconomic dimensions, with negative values indicating leftist positions and positive values rightist positions. The sociocultural scale (R sc) is hence a negative end-point concept. Using data on radical right parties only, the thresholds are set at the 80th percentile (τex = 3.65), the mean (τcr = 2.56) and the 20th percentile (τin = 1.63). The socioeconomic scale (R se), in contrast, is a positive end-point concept. To calibrate, I use data on radical left parties and set the thresholds at the 20th percentile (τex = −2.31), the mean (τcr = −1.21) and the 80th percentile (τin = −0.05).
The sixth and final argument holds that radical right (left) parties tend to be included in coalition governments when they locate themselves relatively proximate to the prime minister party on a sociocultural (socioeconomic) dimension (P). The data source is the Manifesto Project Data, and the scales used are the same as for R above. To compare party positions on the preferred dimension, I, for radical right parties, subtract the sociocultural position of the radical right party from the sociocultural position of the prime minister party and, for radical left parties, subtract the socioeconomic position of the prime minister party from the socioeconomic position of the radical left party. This operation results in two numerical base variables (positive end-point concepts) where negative (positive) values indicate that the radical party locates itself closer to its ‘own’ endpoint (to the opposite endpoint) of the scale than the mainstream party. After transforming positive values to 0, I use data on radical right (for P sc) and radical left (for P se) parties to set the thresholds at the 20th percentile (τex[P sc] = −2.25; τex[P se] = −3.80), the mean (τcr[P sc] = −1.08; τcr[P se] = −2.29) and the 80th percentile (τin[P sc] = 0.0; τin[P se] = −0.47).
Results
After the calibration of the base variables, the resulting configurational data can be summarized in a truth table – a matrix listing all the 2k unique conjunctions derivable from the k exogenous factors included in the study. With six exogenous factors, the truth table consists of 26 = 64 unique conjunctions (i.e. minterms, or matrix rows). Each of these must be associated with a value on the endogenous factor. To do this, it is customary to rely on sufficiency inclusion scores and frequencies.Footnote 10 Following Carsten Schneider and Claudius Wagemann (Reference Schneider and Wagemann2012: 128, 279), the inclusion cut-off should be specific to every research project and vary with, among other things, the number of observations, the quality of the data and the precision of the theoretical expectations. Considering the general theoretical propositions and the large number of (especially negative) observations, I use a relatively liberal sufficiency inclusion cut-off value of 0.750 and a low frequency cut-off value of 1. The resulting truth table is partially revealed in Table 1.
Table 1. Partial Truth Table (outcome G)
Notes: The endogenous factor (G) is coded as 1 if INCL ≥ 0.750 and if N ≥ 1. Observed minterms with INCL ≤ 0.500 (N = 43; range of INCL = [0.050, 0.476]) and logical reminders (N = 14) are omitted from the table. A complete truth table is accessible from the author upon request. ‘·’ = and so on.
a Proportional reduction in inconsistency, calculated using the SetMethods package for R (Medzihorsky et al. Reference Medzihorsky2017). Higher scores indicate that the inclusion scores for G and ¬G are very different from each other. Ideally, (the product of INCL and) PRI should be fairly high (Schneider and Wagemann Reference Schneider and Wagemann2012: 241–4).
The next subsection presents the minimization process and discusses the paths to government inclusion of radical parties while the subsequent one focuses on the negation of government inclusion – that is, government exclusion.
Paths to government inclusion (G)
The next stage is to identify prime implicants (PI) by minimizing the truth table. The minimization is performed using the enhanced Quine–McCluskey algorithm (Dușa and Thiem Reference Dușa and Thiem2015). I present and interpret the parsimonious solution because, as demonstrated by Michael Baumgartner and Alrik Thiem (Reference Baumgartner and Thiem2017b), it has the lowest risk of committing causal fallacies among QCA's three solution types. The PI chart (see online Appendix D2) shows six prime implicants of G: ¬MWCR, SWCP, SWCR, SW¬RP, WCR¬P and MWC¬RP.Footnote 11 Of these, SW¬RP and WCR¬P are essential PIs; they must be included in the final solution model because they are the only PIs that cover minterms 58 and 47, respectively. MWC¬RP, by contrast, is redundant since it only covers a minterm that is already covered by the essential PI SW¬RP (i.e. minterm 62). ¬MWCR, finally, is also redundant: the only minterm it covers (32) is also covered by SWCP and SWCR, both of which also cover the remaining minterm 64. This results in the following two possible solution models:
$${sm_{G.1} :SW\neg RP\,\vee\, WCR\neg P\,\vee\, SWCR\Leftrightarrow G \;{\rm and}} $$
$$ \hskip-2pt {sm_{G.2} :SW\neg RP\,\vee\, WCR\neg P\,\vee\, SWCP\Leftrightarrow G.\quad\ } $$As is evident, the two models are closely related. The only difference is to be found in the last conjunction of each model, where sm G.1 offers SWCR while sm G.2 suggests SWCP. Due to somewhat better summary statistics, sm G.2 is preferred instead of sm G.1.Footnote 12 Hence, expression (2) will be subject to a more detailed discussion. To guide interpretation, this expression can be factorized as follows:
$$ {sm_{{G.2}^{\prime}}\!\! :SWP(C\,\vee\, \neg R)\, \vee\, WCR\neg P\Leftrightarrow G.} $$Expressions (2) and (2′) as well as Table 2 hence indicate that there are three paths to government for radical parties, all of which include the factor W. The first two paths both suggest that radical parties get into government when their electoral fortunes are good (S) and when the prime minister party is comparatively weak (W) and takes fairly similar policy positions to the radical party (P) on the policy dimension preferred by the radical party. In addition, the path SWPC also emphasizes the importance of a location on the same side of the general left–right heuristic (C). This path, henceforth labelled the ‘feasible allies’ path, thus combines favourable size conditions – a victorious radical party and a weak prime minister party – with close ideological affinity. The inclusion score of 0.815 indicates that this path is a consistent sufficient condition; SWPC is an almost (albeit not fully) perfect subset of G. The coverage scores of 0.234 (raw coverage) and 0.135 (unique coverage) are, however, rather low. The path is, in other words, able to explain only a small part of the observations.
Table 2. Government Inclusion of Radical Parties (parsimonious solution)
Notes: For illustrations of cases, see Figures D2a–c in online Appendix D2. Conservative solution: MSW¬RP ∨ M¬SWCR¬P ∨ SWCRP.
a True logical contradictions, i.e. observations with high (> 0.5) membership in the outcome and low (< 0.5) membership in the condition. The true logical contradictory observations here are FrP in 1997(a) and LVP in 1995(b).
The second path – SWP¬R – adds the absence of radical party moderation on the preferred policy dimension (¬R) to SWP. This path – I label it the ‘radical partners’ path – hence combines favourable size conditions with radical policy convergence; the radical party and the prime minister party are fairly close to each other on the policy dimension preferred by the radical party and, moreover, take positions that are comparatively radical. With an inclusion score of 0.775, SWP¬R is a relatively consistent subset of G. The low raw (0.123) and unique (0.041) coverage scores indicate, however, that the path is able to explain only a limited part of the outcome of interest.
The final path, WCR¬P, conjuncts a moderate position on the preferred policy dimension (R) with the presence of a comparatively weak (W) and distant (¬P) prime minister party that, nevertheless, is located on the same side of the left–right heuristic as the radical party (C). This ‘moderate followers’ path seems, hence, to emphasize the role of ideological moderation on the preferred ideological dimension; despite a large distance between the prime minister party and the radical party, the radical party may get access to executive power if it takes moderate positions and if the prime minister party is weak and ideologically related to the radical party. The inclusion score for this path is a low 0.595, and the low raw (0.059) and unique (0.018) coverage scores indicate that its overall importance may be trivial (see online Appendix D3 for a discussion on robustness).
Regarding the covered cases, I note that no fewer than 11 of the 18 positive observations are about Latvian parties. The ‘feasible allies’ path and, to a lesser extent, the ‘radical partners’ path seem, hence, to be rather well suited for explaining government inclusion of radical parties in Latvia – in the 1990s (TB in 1995[a–b], LVP in 1995[a] and TB/LNNK in 1998[a–c]) as well as in the 2000s (TB/LNNK in 2006[c–d]) and in the 2010s (NA in 2011[a–b] and 2014). Of all the positive Latvian observations, these two paths are able to explain some 79% (11 of 14). The solution model is, hence, suitable for explaining radical (right) government inclusion in Latvia. It is, however, less suitable for explaining government inclusion of radical right parties in other countries, although it – and especially the ‘feasible allies’ path – is relevant also for a number of observations from Austria (FPÖ in 1999), Denmark (DF in 2005 and 2007), Estonia (ERSP in 1992), Italy (LN in 1994) and the Netherlands (LPF in 2002). Finally, the solution model is rather ill suited for explaining government inclusion of radical left actors; the only radical left parties included are the (already mentioned) Latvian LVP (in 1995[a]) and the Finnish VAS (in 1999), the latter of which is uniquely covered by the ‘moderate followers’ path.
Taken together, the model presented in expressions (2) and (2′) is doing pretty well in predicting government inclusion of radical parties – the model inclusion score of 0.762 indicates that most observations that are members of the configuration are indeed also members of the outcome G. As the – even for comparatively large-N configurational settings – low model coverage score (0.293) demonstrates, the solution model is, however, not able to deliver a comprehensive explanation of the radical government participation; only 26% of the positive observations are covered (see online Appendix D1 for further details). There must, hence, be other paths for radical parties to enter government – paths that cannot, at least not fully, be discovered with the help of a QCA focusing solely on the conventional ‘size and ideology’ framework. To improve the model and, hence, further broaden our understanding of radical government participation, a companion article (see Fagerholm Reference Fagerholm2019) focuses on ‘typical’ and ‘deviant’ (unexplained) cases, as suggested by Schneider and Rohlfing (Reference Schneider and Rohlfing2013, Reference Schneider and Rohlfing2016; see also Goertz Reference Goertz2017; Lieberman Reference Lieberman2005 and, regarding nested analysis in coalition research, Bäck and Dumont Reference Bäck and Dumont2007).
Paths to government exclusion (¬G)
Next, I focus on the paths that lead to government exclusion of radical parties. To examine government exclusion, I construct a truth table with the same exogenous factors as in Table 1, but with a negated endogenous factor (¬G). In contrast to outcome G, its negation ¬G holds, naturally, a large number of positive observations. Accordingly, the cut-off values for sufficiency inclusion (and frequency) can be raised. With the sufficiency inclusion cut-off value set at 0.850 and the frequency cut-off value at 3, 10 positive minterms are identified (see Table 3).
Table 3. Partial Truth Table (outcome ¬G)
Notes: The endogenous factor (¬G) is coded as 1 if INCL ≥ 0.850 and if N ≥ 3. Observed minterms with INCL < 0.850 and N ≥ 3 (N = 20; range of INCL = [0.272, 0.846]) as well as minterms with N < 3 (N = 20; range of INCL = [0.303, 0.976]) are omitted, as are the ‘true’ logical reminders (N = 14). A complete truth table is accessible from the author upon request. ‘·’ = and so on.
a See note a in Table 1.
The (parsimonious) minimization of this truth table reveals 12 prime implicants (see online Appendix D2). ¬M¬C is essential as the only PI that covers minterm 20. In addition, it also covers minterms 17, 19, 3, 1, 9 and 25. The remaining three minterms are orphan columns covered by multiple (8) inessential PIs. This leads to 14 solution models, with inclusion and coverage scores ranging between 0.872 and 0.913 and 0.442 and 0.504, respectively (for a complete list, see Table D5 in online Appendix D2). Of these models, two consist of three disjuncts while the remaining 12 include four disjuncts. The two models consisting of three disjuncts show the highest inclusion scores, display satisfactory coverage scores and are easier to interpret than most of the remaining models. I also note that all 14 solution models also include ¬M ¬C and either ¬W ¬C¬P or S¬C¬P. Hence, I focus on the first two models:
$$ {sm_{\neg G.1}\! : \;\neg M\neg C\,\vee\, \neg S\neg CR\,\vee\, \neg W\neg C\neg P\Leftrightarrow \neg G\;{\rm and}} $$
$${sm_{\neg G.2}\! :\;\neg M\neg C\,\vee\, \neg S\neg CR\,\vee\, S\neg C\neg P\Leftrightarrow \neg G.} \quad\quad\quad $$Of these, sm ¬G.1 shows slightly better model statistics than sm ¬G.2. Hence, sm ¬G.1 and, more specifically, its factorized version
$${sm_{{\neg G.1}^{\prime}}\! :\; \neg C(\neg M\,\vee\, \neg SR\, \vee \, \neg W\neg P)\Leftrightarrow \neg G} $$will be subject to a more detailed discussion.
Expressions (3) and (3′) suggest three paths (see also Table 4), all of which include the factor ¬C. Path ¬C¬M, firstly, proposes that radical parties are excluded from government when the government is formed by an ideological rival located on the other side of the left–right heuristic (¬C) and when the radical party is not medium-sized – that is, when its parliamentary size is either too small or too large (¬M). In other words: ideologically distant radical parties that, due to their size, are not able to make a significant contribution to the government tend to be excluded from coalition governments. With an inclusion score of 0.907 and coverage scores of 0.325 and 0.108, this ‘useless rivals’ path is a highly consistent and non-trivial sufficient condition for ¬G.
Table 4. Government Exclusion of Radical Parties (parsimonious solution)
Notes: Conservative solution: ¬M¬C¬R¬P ∨ ¬M¬W¬C¬P ∨ ¬MS¬W¬CR ∨ S¬W¬C¬R¬P ∨ M¬S¬W¬CRP ∨ M¬SW¬CR¬P.
a True logical contradictions, i.e. observations with high (> 0.5) membership in the outcome and low (< 0.5) membership in the condition. The true logical contradictory observations here are PRM in 1997(a–c), Attack in 2013 and ANEL in 2015.
Path ¬C¬SR, secondly, suggests that radical parties are excluded from government when the government is formed by a rival party (¬C) and when the radical party is moderate (R) and lost seats in the previous national election (¬S). This ‘unattractive losers’ path hence suggests that a poor result in the recent election keeps radical parties out of governments formed by parties located on the other side of the left–right heuristic, even if the policy position of the radical party is quite moderate. This path is highly consistent with the outcome (0.902). The raw (0.184) and unique (0.070) coverage scores are, however, rather low.
The third and final path, ¬C¬W¬P, proposes that radical parties are excluded from government when the government is formed by a rival party (¬C) that, in addition, has a strong position in the parliament (¬W) and is distant also on the preferred ideological dimension (¬P). These radical parties are, hence, ‘redundant antagonists’: the prime minister party does not need them because of its considerable size and does not want them because of the radical party's ideological remoteness. The path is a consistent subset of ¬G (0.937), and the raw (0.228) and unique (0.049) coverage scores are moderate (see online Appendix D3 for a discussion on robustness).
The inclusion score of the solution model in expressions (3) and (3′) is 0.912, thus indicating that most observations that are members of the configuration are also members of the outcome ¬G. Its coverage is 0.452; of the 139 negative observations in the data, the solution model covers 37.4% (52). The solution model is more adequate for negative radical left observations (covering 47.9%, or 35/73) than negative radical right observations (25.8%; 17/66), and also more adequate for negative Western European observations (45.1%; 46/102) than for negative Central and Eastern European observations (16.2%; 6/37).
Discussion, conclusions and the way ahead
The substantive findings from the empirical analysis can be summarized in five main points. First, the empirical analysis shows that radical right parties in both Western and Central and Eastern Europe (and particularly in Latvia) may get access to coalition government following two paths where factors related to size and ideology are combined in somewhat different ways. Both paths include: (1) success in the most recent national election and (2) policy affinity, with (3) a weak prime minister party as INUS conditions. The first path (‘feasible allies’) also emphasizes the importance of (4a) a (roughly) similar general left–right location while the second path (‘radical partners’) underlines the importance of (4b) shared radical views on issues related to the policy dimension emphasized by the radical party. The second main finding concerns radical left parties. Here, the ‘size and ideology’ framework is clearly not very relevant in explaining government participation, although (1) a weak prime minister party may seek help from (2) a distant radical left party if the radical left party (3) holds moderate views and (4) is located on the same side of the left–right heuristic as the prime minister party (‘moderate followers’). These two observations lead to a third, more general, conclusion. The low coverage scores, in combination with the high occurrence of observations from a single (small) country, means that the overall conclusion from the analysis of government inclusion must remain cautious: although the models provide intriguing explanations of single cases, it is unlikely that they are able to tell us much about general European trends.
Regarding government exclusion, two additional substantive conclusions appear. The first of these holds that (1) ideological remoteness is a condition included in all paths to government exclusion: radical parties – to the left and to the right, in the west as well as in the east – are hence generally eliminated from government if the government is formed by a party located on the other side of the left–right heuristic. This condition needs, however, to be combined with other factors. The first path hence combines remoteness with (2a) the overtly radical party being either too weak or too strong (‘useless rivals’), the second path with the (2b) overtly radical party being (3) too unsuccessful (‘unattractive losers’) and the third path with the radical party holding (2c) too distant policy positions to be taken (3) seriously by a strong prime minister party (‘redundant antagonists’). The second, more general, conclusion from the analysis of government exclusion holds that the solution is – in contrast to the solution model for government inclusion – relevant also from a broader perspective: by explaining about 40% of the instances where radical parties are excluded, a model based on the ‘size and ideology’ framework is, indeed, able to tell us something important about the general trends in radical government exclusion.
Taken together, this article constitutes an essential first step in a systematic and comparative search for factors explaining government inclusion and exclusion of radical parties. As the – to my knowledge – first study focusing on both left and right and on both Western and Central and Eastern Europe, it has provided intriguing insights: the study emphasizes the complex and conjunctive nature of government inclusion and exclusion, it provides a first step in underlining different patterns in radical left and radical right government participation, it highlights the importance of ideological distance for radical government exclusion and it points to the need for future research to consider factors other than those related to party size and party ideology. By this, it provides new insights to the research on party competition and government participation and, more indirectly, to the analysis of the state and future of democratic governance.
As indicated, the study also highlights the need to look beyond size and ideology. A number of other factors have been stressed, but rarely systematically scrutinized, in previous (comparative) research: parties’ historical roots and trajectories, strategic and tactical considerations, informal relationships between prominent party representatives, previous successful cooperation on the regional or local level, issue convergence and institutional constraints. A common denominator for these factors is their highly elusive, and in some cases even idiosyncratic, nature: they cannot easily be included in large cross-case studies. As a consequence, it is of imperative concern that the important initial steps taken in this study are followed by additional theory-revising research where the relevance of factors other than those directly related to size and ideology is examined.
Supplementary information
To see the supplementary information for this article, https://doi.org/10.1017/gov.2019.24
Acknowledgements
The author would like to thank the anonymous referees for their suggestions and useful comments.
