Government Formation and the Radical Right: A Swedish Exception?

Abstract

This article tests the claim that government cooperation between mainstream parties and radical right parties can be explained by coalition theory. It does so by analysing three Swedish cases of coalition formation where the radical right Sweden Democrats (SD) have remained excluded despite holding a pivotal position in the parliament. It argues that, with the right analytical tools, this exclusion can be explained by coalition theory: cooperation with the SD has been unattractive in terms of policy, and unnecessary because the mainstream parties have been able to form viable minority governments. This argument requires three things: first, that we consider the two-dimensional nature of Swedish politics; second, that we shift the focus from majority government to viable government; and third, that we acknowledge strategic time horizons that extend well into the future. The findings contribute to our understanding of coalition formation and of how mainstream parties respond strategically to the radical right.

Why are some radical right parties included in government coalitions while others are not? One answer to this question can be found in standard coalition theory: radical right parties that are systematically excluded from government cooperation are not qualitatively different from other parties – they simply lack the characteristics in terms of size and policy positions needed to make them attractive coalition partners (de Lange 2008, 2012). This argument also appears to be supported by the empirical record in Western Europe. In a number of countries, including Austria, Denmark, Finland, Italy, Norway and the Netherlands, radical right parties that have been able to provide the mainstream right with a parliamentary majority have become coalition partners or support parties to governments (Bale 2003; de Lange 2012; Jungar 2021; Twist 2019; van Spanje 2010). Some radical right parties are, at least initially, treated as ‘pariahs’ and shunned as a matter of principle (Akkerman and Rooduijn 2015; Downs 2001; Moffitt 2021; van Spanje 2010). However, whether such isolation is strategic or principled, it tends to be upheld only as long as this is not costly for the mainstream right (Twist 2019). It seems, then, that radical right parties are much like other parties when it comes to coalition formation.

Not all countries appear to fit this pattern, however. In Sweden, the mainstream parties have gone to considerable lengths to exclude the radical right party the Sweden Democrats (Sverigedemokraterna, SD) from government cooperation, despite its pivotal position in the parliament. In 2014, the other parties concluded the so-called ‘December agreement’ with the intention of making the SD irrelevant for coalition formation (Bäck and Hellström 2015; Bjereld et al. 2016). After the 2018 election, parties from both the established blocs of left and right instead chose to engage in cross-bloc cooperation with the explicit intention of isolating the SD (Teorell et al. 2020). While mainstream parties in other countries rely on the radical right, they argued, Sweden ‘chooses a different path’ (Dagens industri 2019).

The apparently deviant nature of the Swedish case calls into question coalition theory's ability to fully account for the exclusion of the radical right from government cooperation. If Sweden really is different, it may be that the standard explanation is bounded by scope conditions. For example, some radical right parties – such as those that, like the SD, have roots in extreme-right movements – may be qualitatively different (cf. Ivarsflaten 2006). Another possibility is that mainstream parties that have previously committed themselves to non-cooperation (e.g. made a radical right challenger out to be a ‘pariah’) fear the electoral costs of appearing unprincipled or opportunistic (Backlund 2020; Strøm et al. 1994). Although such alternative explanations are of great interest, they are not the main focus of this article. Instead, I focus here on the more fundamental question: Is Sweden really different? To answer this question, I analyse three cases of coalition formation in the presence of the SD, following the 2010, 2014 and 2018 elections. The small-N research design answers the call for more case-oriented coalition research to complement the dominance of large-N statistical studies (Andeweg et al. 2011; Bäck and Dumont 2007).

The findings suggest that Sweden is not, after all, a case where the mainstream parties have chosen ‘a different path’. Rather, cooperation with the radical right has been unattractive in terms of policy, and it has been unnecessary, because the mainstream parties have been able to form viable minority governments. The explanation provided by coalition theory applies to Sweden as well – as long as we use the right analytical tools. This requires three things: first, that we consider the two-dimensional nature of Swedish politics; second, that we shift the focus from majority government to viable government; and third, that we acknowledge strategic time horizons that can extend well beyond any given bargaining situation. I also find that the transformation of Swedish party competition from unidimensional to multidimensional has made cooperation with the radical right an increasingly attractive option. These findings contribute to our understanding both of coalition formation and of how mainstream parties respond strategically to the presence of the radical right.

The article is structured as follows. In the next section, I outline the coalition theories on which the study is based. I then turn to research design, justifying the choice of Sweden and describing the data and measurements I use. This is followed by the empirical analysis, which consists of two steps. First, I compare the policy cohesiveness of the coalitions that actually formed to hypothetical majority coalitions that include the SD. Second, I analyse how these coalitions came into office and survived. Finally, in the last section I summarize my conclusions and their implications for future research.

Theory

In parliamentary systems, the executive needs support from – or at least toleration by – the parliament. In most such systems, this means being able to survive a vote of confidence (or no confidence), based on a majority criterion. If no single party wins a majority of the parliamentary seats, which is uncommon in proportional electoral systems, the parties need to build coalitions in order to form majority cabinets. Coalition bargaining entails a trade-off between multiple party goals, which are commonly conceptualized as policy, office, and votes (Müller and Strøm 1999; Strøm 1990a): parties are policy-seeking in that they try to influence the government policy output in their preferred direction; they are office-seeking in that they want access to government portfolios; and they are instrumentally vote-seeking in order to increase their bargaining power. Often these goals are conflicting, in the sense that a strategy suitable for pursuing one goal will make it more difficult for the party to obtain another. For example, a party pursuing office by becoming a government coalition partner may alienate some of its voters when it has to compromise on policy with the other parties in the coalition.

The most fundamental coalition theories focus exclusively on office-seeking motivations. The minimal winning theory of political coalitions is based on cooperative game theory (Gamson 1961; von Neumann and Morgenstern 1953) and asserts that if parties are to divide a limited number of ministerial portfolios among themselves, they will include the number of parties necessary to form a parliamentary majority, but no more (Riker 1962). One problem with pure office-seeking theories, however, is that they cannot account for minority cabinets, which empirically constitute around a third of all coalitions in Western Europe (Bassi 2017). If parties are motivated only by the pursuit of office, minority governments should not form because this would imply that the majority opposition is giving away rewards that it could claim for itself.

To solve problems of this kind, later coalition theories introduced policy-seeking motivations. According to Robert Axelrod (1970), parties seeking to minimize conflicts of interest will form coalitions that are ideologically ‘connected’. Such minimal connected winning coalitions consist of parties that are adjacent to each other along the main dimension of political conflict. Michael Leiserson (1966) predicts instead the formation of minimal range coalitions, which are as ideologically compact as possible (although not necessarily connected). Differing policy positions also mean that some parties are strategically advantaged in terms of coalition formation (de Swaan 1973). For example, the party controlling the median legislator on the main dimension of political conflict is expected to be included in the government, since it cannot be defeated by majorities to either left or right (Laver and Schofield 1998: 111).

If parties are also motivated by policy, minority governments may be viable because the majority opposition cannot agree on a replacement cabinet (Budge and Laver 1986; Laver and Schofield 1998: 74–81; Laver and Shepsle 1996: 262–263). This is particularly likely if the opposition is divided along the main dimension of political conflict. Under such circumstances, an opposition party may choose to support a governing coalition of which it is not a part if this government is viewed as the ‘lesser evil’ (Budge and Keman 1993: 50). Stated otherwise, some parties may be ‘captives’ that have little choice but to tolerate a certain government, simply because all alternatives would produce an even worse policy output (Bale and Bergman 2006). Due to the electoral costs of governing, moreover, parties may also have vote-seeking motivations for supporting the government from the outside (Strøm 1984, 1990b). Such support can range from formal agreements to ‘shifting majorities’, where the government negotiates new legislative coalitions to reach a majority on an issue-by-issue basis.

More recent research has also shown that institutions matter for government formation, meaning that coalitions are not formed in an unconstrained environment (Martin and Stevenson 2001; Strøm et al. 1994). Constraints can be ‘hard’, such as the formal rules governing investiture votes (Bergman 1993) or parliamentary committees (Strøm 1990b); or they can be ‘soft’, such as self-imposed pre-commitments to form certain coalitions (Golder 2006) or the legitimacy that comes with electoral gains (Mattila and Raunio 2004).

When radical right parties win enough votes to gain parliamentary representation, their presence by definition will affect the coalition formation equation. According to Sarah de Lange (2008, 2012), however, we need no special theory of mainstream party responses to the radical right in terms of government formation. On this argument, radical right parties are not qualitatively different from other parties, and their inclusion or exclusion from governing coalitions can be explained using standard theories of coalition formation. If this is correct, we can expect radical right parties to be included in coalitions when they fulfil two criteria: holding policy positions that make them an attractive partner, and controlling enough seats to contribute to a winning coalition.

Research design

Case selection

While most coalition research takes place in a large-N statistical setting, the benefits of case studies are being increasingly recognized (Andeweg et al. 2011; Bäck and Dumont 2007). Cases that do not appear to be accounted for by the established explanations provided by coalition theory – puzzling cases – are particularly appropriate for the purposes of theoretical refinement (Dumont et al. 2011). Earlier research has identified a number of radical right parties that have faced systematic boycott by the political mainstream in Western Europe (Akkerman and Rooduijn 2015; van Spanje and van der Brug 2007). However, these are mainly cases where the exclusion has not been particularly costly, either because the party has won very few seats or due to disproportionalities introduced by the electoral systems (such as in France and the United Kingdom). Kimberly Twist (2019: 163–164) makes a similar argument about Belgium, where the regional party systems and related institutional features have reduced the likelihood of the radical right's being included in government. The case that is most comparable to the Swedish one may be Germany, where the relative newcomer Alternative for Germany has been excluded from government cooperation despite a strong parliamentary position (Arzheimer 2019).

The present study focuses on three Swedish cases of coalition formation where the exclusion of the radical right appears to have come at a cost: that in 2010 (Reinfeldt II), that in 2014 (Löfven I) and that in 2018 (Löfven II). As shown in Table 1, the SD remained excluded from government cooperation after all three elections. In 2004, the four centre-right parties formed a pre-electoral alliance called ‘Alliance for Sweden’, or just the ‘Alliance’ for short. This marked the start of a period of increased polarization between the right and left blocs – between the Alliance and the ‘Red–Greens’ (Aylott and Bolin 2007, 2015). Having governed in majority from 2006 to 2010, the Alliance was deprived of its majority in 2010 when the SD entered the parliament. The SD won a pivotal position between the established blocs, which it retained in the two subsequent elections, meaning that majorities could only be achieved either by cooperating with it or by cooperating across the bloc divide. Note, however, that under Sweden's ‘negative parliamentarism’ (Bergman 1993) a government need only be tolerated – rather than actively supported – by an absolute majority of 175 votes. In other words, abstention equals passive support in the investiture vote.

Table 1. Summary of the 2010, 2014 and 2018 Elections

Notes: The Swedish parliament has a total of 349 seats, with 175 required for a majority. Seat counts in bold indicate government coalition parties.

* = Formal support party.

When the Alliance lost its majority in 2010, it chose to govern in minority rather than to cooperate with the SD, even though this meant it became considerably less productive (Demker and Odmalm 2021: 10). Likewise, having won fewer seats than the Red–Greens in the 2014 election, the Alliance tolerated a minority government by the latter so as not to become reliant on the radical right (Reinfeldt et al. 2014). The SD effectively blocked the Löfven I government from passing its budget, however, and to avoid a snap election the other parties concluded the so-called December agreement. According to this agreement, the prime-ministerial candidate supported by a coalition of parties larger than any other conceivable coalition would be tolerated, and thus enabled to form a government (Bäck and Hellström 2015; Bjereld et al. 2016). In essence, the agreement would allow the larger of the two blocs to govern as a minority, regardless of how many seats the SD held; the radical right was simply removed from the coalition equation. The agreement was intended to remain in place until the 2022 elections, but already in October 2015 the Christian Democrats defected (following intra-party conflict), and the agreement was subsequently abandoned by the other parties as well. With the December agreement dissolved, the same problem threatened in the 2018 election. Rather than resulting in cooperation with the radical right, however, the 2018 election led to a split in the right bloc, as parties from both blocs turned instead to cross-bloc cooperation (Teorell et al. 2020).

As described above, the Swedish mainstream parties have gone to great lengths to exclude the SD from government cooperation, despite its pivotal position. The choice of Sweden is justified, therefore, by its puzzling appearance in relation to both theoretical expectations and the empirical pattern in most of Western Europe.

Data and operationalization

To analyse the exclusion of the Swedish radical right, I draw on two key concepts in coalition theory: first, the idea that parties prefer coalitions that are more cohesive in terms of policy; and second, the idea that coalitions can be viable, even if they are not majorities, as long as they cannot be defeated by any alternative coalition that is preferred by a parliamentary majority. Note that this study does not formally test these theories; rather, I use them as heuristic tools in order to explain the three cases of coalition formation. To do so, I proceed in two steps. First, I show that the coalitions that formed are more policy cohesive than are alternative majority coalitions that include the SD. Second, I show how these coalitions constituted viable governments. In the following, I describe how the concepts of policy cohesiveness and viability are operationalized.

Starting with policy, I follow most coalition research in turning to expert survey data. Specifically, I use data from the Chapel Hill Expert Surveys (CHES; Bakker et al. 2015; Polk et al. 2017), where country experts are asked to estimate party positions along predefined policy dimensions scaled from 0 to 10. Although this approach is not without critics (see e.g. Budge 2000), different expert surveys have been shown to correlate well with each other, and additionally with alternative estimates of party positions derived from voter perceptions, election manifestos and elite surveys (e.g. Hooghe et al. 2010; Marks et al. 2007; Ray 2007; Steenbergen and Marks 2007; Whitefield et al. 2007). At the same time, such alternative estimates are problematic for the purposes of this study.¹

To compare the policy cohesiveness of different coalitions, I rely on measurements of policy range. When doing so, the choice of which policy dimension(s) to use is crucial. In Sweden, party competition has historically been structured mainly by economic conflict (Oscarsson and Holmberg 2016). In recent years, however, the sociocultural GAL–TAN dimension, reflecting party positions on issues such as immigration and law and order, has become increasingly important (Demker and Odmalm 2021; Rydgren and van der Meiden 2019).² At the same time, this dimension is only moderately correlated with the economic dimension (0.43 as estimated in CHES 2017).

For this reason, I analyse the three coalitions in terms of party positions on both the economic and the GAL–TAN dimensions. To do so, I place the parties in a two-dimensional policy space where, the closer the parties in a coalition are to each other, the smaller the range of the coalition. More specifically, I measure the range of any given coalition as the maximum Euclidean distance between any two parties included in the coalition.³ In a two-dimensional policy space, however, both dimensions may not be of equal importance to the parties (Benoit and Laver 2006). For example, a party that attaches little importance to the GAL–TAN dimension may perceive a party that is positioned far away in absolute terms to be a viable coalition partner, simply because it is willing to compromise on these issues; a lower salience diminishes the distance between the parties. To account for this, I include a robustness check where the Euclidean distances are weighted by party-specific dimension salience (cf. Debus 2009).⁴

Because formal models of coalition formation struggle beyond a single dimension unless strong assumptions are introduced (Laver and Shepsle 1996), I also make use of unidimensional measurements as a robustness check. This includes results based on the CHES ‘general’ left–right dimension, where experts are asked to condense political conflict into a single dominant dimension. However, because this general dimension has been shown to have questionable validity in the Swedish case (Backlund 2020: 86–91), I also construct a ‘weighted’ left–right dimension where a party's position is equal to the average of its position on the economic and GAL–TAN scales, weighted by the party-specific salience attached to each of them.⁵ The point of the weighted left–right dimension is that its content is more transparent and that it focuses explicitly on the two subdimensions analysed here.

Turning to the question of viability, the standard threshold in coalition research is the majority criterion. Coalition theory assumes that, other things being equal, parties will have a preference for majority status, since this allows them to win votes predictably. Indeed, as noted above, earlier studies point to the ability of the radical right to contribute to a majority on the right as crucial for its inclusion. Because all parties need not be included in the executive, formal support parties can be considered de facto coalition members. There are, however, both theoretical and empirical reasons to go beyond a majority criterion. First, other things may not be equal: the addition of more partners to a coalition comes at the cost of spreading office benefits more thinly and increasing the need for policy compromise. Second, Swedish cabinets have an overwhelming tendency to be minority ones, including all three of the cases studied here.

For this reason, I analyse these coalitions in terms of policy viability. A policy-viable coalition is one that represents a policy package that cannot be defeated by any other coalition preferred by a parliamentary majority (Budge and Keman 1993; Budge and Laver 1986; Laver and Schofield 1998). Assuming unidimensionality, policy-viable coalitions can be defined formally as those that include the median party (Strøm and Leipart 1993). Because my analysis is two-dimensional (and of the small-N kind), I approach the concept informally, drawing on media reports to infer policy viability from party behaviour. I focus on two observable implications. First, I assess whether a minority coalition takes office with support from a party that can be considered ‘captive’ in the sense that it tolerates the government without receiving any policy concessions in return (i.e. because this constitutes the ‘lesser evil’). In the Swedish institutional setting, abstention equals passive support, but it is less costly in terms of party unity than is active support (Aylott and Bergman 2011). Second, I assess the extent to which the survival of a minority coalition can be attributed to ideological division among the opposition parties, making them unable to agree on a replacement cabinet.

Analysis

An unattractive coalition partner?

Figure 1 shows Swedish party positions along the economic and GAL–TAN axes as estimated in 2010 and 2017. The SD, as we see, takes an economically centrist position between the left and right blocs. On the vertical GAL–TAN dimension, by contrast, the party has by far the most culturally conservative position. Figure 1 also shows how the Alliance parties have become polarized along the GAL–TAN axis, resulting in increased dealignment between the two dimensions. Marie Demker and Pontus Odmalm (2021) trace this transformation to the parliamentary entry of the SD, which increased the salience of the immigration issue, and the subsequent breakdown of the Alliance, which allowed the parties to differentiate in this regard. The largest absolute change can be observed for the Centre Party, which has shifted from a neutral position to one roughly matching the Greens and the Left Party.

Figure 1. Swedish Party Positions in 2010 and 2017

Source: CHES 2010 and 2017 (Bakker et al. 2015; Polk et al. 2017).

Turning to the coalitions, Table 2 lists the coalitions that actually formed following each election, together with the number of seats they controlled. Note that no distinction is made here between a party's being in government and its being a formal support party for a government. Table 2 also shows the policy range of each coalition, measured in the two-dimensional space illustrated in Figure 1. For each election, Table 2 also lists a hypothetical right-wing majority that includes the SD, together with its seat share and policy range. These are suitable as objects of comparison, since they correspond to the type of coalition that is typically formed between mainstream parties and radical right parties.

Table 2. Actual and Hypothetical Coalitions, 2010–18

Source: CHES 2010, 2014 and 2017 (Bakker et al. 2015; Polk et al. 2017).

Notes: Cabinet composition including support parties. Range computed as Euclidean distances in a two-dimensional policy space (economy/GAL–TAN).

As the hypothetical majorities show, the mainstream right could have formed majority coalitions with the SD following each of the elections. In 2010 and 2014, this would have required all four Alliance parties; in 2018, either the Liberals or the Christian Democrats could have been dropped while retaining a majority (the choice of which of the two to include does not affect the policy range). In terms of range, the hypothetical majorities become less cohesive over time, which is primarily due to the increased polarization on the GAL–TAN axis. Comparing these coalitions with the ones that actually formed, we can see that the latter are more policy cohesive in all three cases. In 2010, for example, the Alliance coalition has a much smaller range when it excludes the SD. Likewise, the 2018 cross-bloc coalition is more policy cohesive than is a centre-right majority including the SD. The same conclusions apply if the distances are weighted by party-specific dimension salience, as shown in Table A.1 in the Online Appendix.

The unidimensional results, presented in Table A.2 in the Online Appendix, show conflicting patterns. If we rely on the CHES general left–right dimension, the hypothetical majorities with the SD appear very policy cohesive. This measurement leaves us unable to explain, for example, why the Centre Party and the Liberals would not have preferred to form a majority coalition with the SD in 2018, which would have had a smaller range than the cross-bloc minority that actually formed. If party positions on the economic and GAL–TAN scales are instead weighted into a single dimension according to salience, the actual coalitions remain more policy cohesive than the majorities; however, the differences are much smaller because a great deal of variation is discarded. These results indicate that both the economic and the GAL–TAN dimensions are required to accurately reflect coalition formation in Sweden. In sum, the results above show that, if both dimensions are taken into account, the SD has not been a particularly attractive coalition partner in terms of policy.

We turn now to the question of viability, for even if the inclusion of fewer parties can reduce the need for policy compromise, government coalitions still need to take office and survive. For example, by excluding the SD in 2010, the Alliance parties could form a very cohesive coalition, but they also shut the door for a stable parliamentary majority on the right. The institutional rules at the time, however, did not require an investiture vote unless the prime minister resigned or was brought down in a vote of no confidence. Because the Red–Greens and the SD were unwilling to join forces in such a vote, the Alliance could simply remain in office (Dagens Nyheter 2010; Svenska Dagbladet 2010).

A constitutional amendment dictating that an election be followed by an investiture vote came into effect after the 2014 election. In this vote, as noted earlier, all four Alliance parties abstained in favour of a Red–Green minority government. This choice is more difficult to explain, because even if policy differences ruled out a coalition between the mainstream right and the SD, the former could still have opted for cooperation across the established blocs. Indeed, the cross-bloc coalition that formed after the 2018 election would have been even more policy cohesive in 2014 (not shown here) – and it would in fact also have been a majority government. By choosing instead to go into opposition, the Centre Party and the Liberals thus rejected policy and office rewards that were arguably within their grasp. In the next section, I address this puzzle in more detail.

From the December agreement to a cross-bloc coalition

Why did the Alliance parties tolerate a Red–Green government in 2014? First of all, an Alliance minority cabinet would have been potentially costly in terms of policy. Unlike in 2010, the right bloc held fewer seats than the left, and the SD demanded significant concessions in return for support (Åkesson 2014). At the same time, the costs of forming a cross-bloc coalition were also high. The Alliance cooperation had been highly effective as an office-seeking strategy, and the parties were not keen on ending it. Indeed, Stefan Löfven's appeals for a cross-bloc coalition were seen as a strategy for breaking up the Alliance permanently in order to secure control of the executive for the foreseeable future (Expressen 2014). In 2014, then, the outcome can be explained by the fact that the costs of both a cross-bloc majority and a within-bloc minority were prohibitive. Instead the Alliance parties opted for a third option: the December agreement.

The December agreement can be understood as a kind of ‘institutional design’ by which actors ‘enlarge their strategy space and choose a previously unavailable option’ (Tsebelis 1990: 10). The purpose of the agreement was essentially to remove the SD from the coalition-formation equation, allowing the larger of the two traditional blocs to govern as a minority. It ensured predictability for the formation and survival of minority cabinets, with the one parameter determining which bloc would govern being their relative size. For the Social Democrats and the Greens, this meant they could remain in power at the cost of allowing the Alliance to govern in 2018 (if electoral fortunes were to change). For their part, the Alliance parties renounced the rewards of office in the short term, in the expectation that these could be gained at a lower cost in the longer term – with fewer policy compromises and the successful Alliance cooperation intact. However, the December agreement was negotiated under severe time constraints (the threat of a snap election), and the leaders of the Alliance parties arguably failed to anticipate its true costs in terms of intra-party conflict (Bjereld et al. 2016). When such conflict led to the premature dissolution of the agreement in October 2015, this ‘third way’ was closed. Although this meant the Alliance was now free to remove the Löfven I cabinet with the help of the SD, disagreement about what would replace it prevented any action from being taken (Dagens Nyheter 2016).

By the time of the 2018 election, the Alliance parties had diverged along the GAL–TAN axis and the SD expressed an interest in supporting, or being part of, a conservative Moderate–Christian Democrat government. For the Centre and Liberal parties, meanwhile, a cross-bloc coalition had become a more attractive option than it had been in 2014. After prolonged negotiations, the Social Democrats and the Greens offered the Centre Party and the Liberals sufficient concessions (in a 73-point policy agreement) to secure their formal support. In addition to being a victory in terms of policy influence for the latter two parties, the agreement reduced the vote-seeking cost of backing a prime minister for whose removal from office they had campaigned. To this end, they emphasized that this ‘January agreement’ was nothing like the unpopular December agreement. Whereas the latter had focused almost exclusively on form, the former was all about content, allowing the enactment of some of the parties' most salient policy proposals. With the cost of providing cross-bloc support much reduced as compared to 2014, then, such a coalition now formed.

Still, it could not take office without being tolerated by the Left Party. The January coalition correctly assumed that the Left was ‘captive’ in the sense that it would not risk the formation of a Moderate–Christian Democrat government supported by the SD; somewhat begrudgingly, the party abstained in favour of the cross-bloc coalition (Fokus 2019). Not only was the Left Party not offered any policy concessions in return, the January agreement also stated explicitly, in what came to be known as the ‘humiliation clause’, that it was aimed at depriving the Left of influence. Pushing back, the Left announced that a few key proposals in the agreement were simply unacceptable, and that it would support a vote of no confidence if the government attempted to implement them. Although many commentators doubted the credibility of this threat, the party did indeed come to act on it – under new leadership – when the government announced a proposal for unregulated rents on newly produced public housing. Seizing any opportunity to remove the incumbent, the other opposition parties also voted against Löfven, who in the summer of 2021 became the first ever Swedish prime minister to be removed from office in a vote of no confidence. The Left's actions were aimed at policy influence and not the removal of the government as such, but because it was shut out from policy bargaining, it risked the latter in order to exert policy influence while in opposition.

Following this episode, the Liberals – also under new leadership – withdrew from the January agreement. Struggling at the polls, the party aligned instead with the Moderates and the Christian Democrats in anticipation of the 2022 elections. Still, because Löfven retained the larger of his two support parties, and was again tolerated by the Left once the public housing proposal had been dropped, he could promptly reassume office.

Conclusions

In this article, I have tested the validity of the claim that the inclusion or exclusion of radical right parties from government cooperation can be explained by coalition theory (de Lange 2008, 2012). To do so, I have analysed three Swedish cases of coalition formation where the radical right party the SD has held a pivotal position between the established blocs of left and right, yet has remained excluded from government cooperation (Reinfeldt II, Löfven I, and Löfven II). My findings show that, given the right analytical tools, the exclusion of the SD can, indeed, be explained by coalition theory. Cooperation with the SD has not only been unattractive due to policy differences, it has also been unnecessary, because the mainstream parties have made use of strategies allowing them to govern in minority. This argument relies on three main points.

First, we need to acknowledge the two-dimensional nature of Swedish politics. Although the SD holds a centrist position on economic issues, its outlier position on social and cultural (GAL–TAN) issues has made it an unattractive coalition partner. As measured in a two-dimensional policy space, the three coalitions that actually formed are more cohesive than are hypothetical coalitions where the mainstream right cooperates with the radical right. Although multidimensionality poses a problem for formal coalition theory, I have shown that a similar – if weaker – result can be obtained by weighting together positions on the economic and GAL–TAN scales into a single dimension according to party-specific dimension salience. In fact, this weighted left–right succeeds where a general left–right does not in accounting for the exclusion of the SD (as measured in the CHES); whether this construct is useful beyond the Swedish case, however, remains an open question.

Second, we need to shift focus from majority government to viable government. In this study, I have shown how coalitions that divide the opposition constitute a viable (but not necessarily effective) alternative to cabinets that rely on the radical right. Minority coalitions can come into office because they do not need to pass an investiture vote, such as the Reinfeldt II cabinet. Or they can do so with support from ‘captive’ opposition parties – such as the Left Party in 2018 – that are prepared to back them even though they receive no policy concessions in return, simply because they consider them to be the ‘lesser evil’ of government alternatives. Once in office, the three coalitions studied here have survived because, in the absence of an alternative coalition, the ideologically divided opposition has been unable to oust the incumbent despite controlling a parliamentary majority. Although the concept of policy viability has limited use for the purpose of prediction (in particular in a multidimensional setting), these findings show how it can be useful for the purpose of explanation. The fact that Prime Minister Löfven was, for a brief time, removed from office following a vote of no confidence in 2021 also illustrates the risks involved for a minority coalition relying on policy viability.

Third, we need to acknowledge that the parties' strategic time horizons can extend well beyond any given bargaining situation. Otherwise, it is difficult to explain why all four parties in the centre-right Alliance chose to tolerate a centre-left minority in the investiture vote following the 2014 election. This outcome can, however, be accounted for by the logic of the subsequent December agreement, which granted executive power to the largest of the two established blocs. Finding the costs both of governing as a minority and of forming a cross-bloc coalition to be prohibitive, the Alliance parties opted instead to restore predictability to the coalition formation process, in the hope of reaping office rewards in the future at a lower cost. Still, the choice of this non-standard strategy can only be understood by extending the strategic time horizons. In other words, we need to consider how the parties' actions in one bargaining situation affect the set of strategies available to them in future ones.

In the end, this study supports the argument that radical right parties are just like other parties when it comes to coalition formation (de Lange 2012; Twist 2019). Given that the SD is encumbered by a legacy of right-wing extremism – that it lacks a ‘reputational shield’ (Ivarsflaten 2006) – this is a significant conclusion. Although the party has been labelled a ‘pariah’ by the political mainstream, it has also been an unattractive partner from the perspective of coalition theory. As the mainstream parties have become polarized along the GAL–TAN dimension, however, this is changing: since the 2018 election, the Moderates, the Christian Democrats and the Liberals have all opened the door for governing with support from the SD.

Seeing that a party that has been isolated for alleged racism and extremism can become accepted as a support party when the costs of exclusion become too high, we should be sceptical of arguments claiming that cooperation with certain radical right parties is ‘impossible’, even in highly unfavourable contexts. These findings are also consistent with the argument that the comparatively late transformation of the Swedish political space from unidimensional to multidimensional delayed radical right success in Sweden, which has in recent years been catching up with the rest of Europe (Rydgren and van der Meiden 2019). Sweden, it seems, is not an exception; rather, it is a case where the strategic incentives required for inducing government cooperation with the radical right were not yet in place.