Black–Blue or Bahamas? Explaining CDU, CSU, FDP and AfD Voter Attitudes Towards a Common Governmental Coalition Before the 2017 German Federal Election

ABSTRACT Will the Alternative for Germany (AfD) soon be actively involved in forming Germany’s governments? The findings of this article illustrate that the established German parties would be well advised not to form a coalition involving the AfD as their own voters strongly oppose it. For the first time, coalition preferences for a so-called Black–Blue coalition (CDU, CSU and AfD) as well as a Bahamas coalition (CDU, CSU, FDP and AfD) are examined. Using regression models to explain the emergence of preferences for such coalitions, the findings indicate the following: (1) while an identification with the AfD leads to a higher desirability for such coalitions, the opposite is the case if an individual identifies with the CDU; (2) a positive impression of the AfD’s candidates leads to a stronger coalition preference, whereas the rejection of the CDU’s then chancellor-candidate Angela Merkel has no significant effect; and (3) opposing migration leads to a stronger preference for such coalitions.


Introduction
Being relatively stable over the last years, the founding and electoral success of the Alternative for Germany (Alternative für Deutschland, AfD) has begun to shape the German party system.The AfD was not only the strongest opposition party in the German Bundestag after the 2017 German federal election, it was also the only opposition party that held seats in all of the 16 German state parliaments.Observing the AfD since it was founded, it is now clear that the party has transformed itself from 'a right-wing populist movement' (Berbuir, Lewandowsky, and Siri 2015, 173) at the very beginning to a clearly recognizable radical right-wing populist (RRP) party after 2015 (Hobolt and Hoerner 2019).
With the AfD 'explicitly positioning itself to the right of the Christian Democratic Union (CDU) and Christian Social Union (CSU)' 1 (Dilling 2018, 84), it can be stated that 'the expansion of choice in Germany (…) has led to an increase in turnout, especially among those voters who hold right-wing views' (Hobolt and Hoerner 2019, 15-16).Indeed, voting for the AfD in the 2017 German federal election was highly motivated by individual attitudes towards migration and refugees but also by anti-establishment attitudes reflecting dissatisfaction with democracy (Hansen and Olsen 2019).Especially the first motivation mentioned was one of the main driving factors for CDU/CSU and SPD voters to defect to the AfD (Wurthmann et al. 2020).The combination of boththe offer of specific policies and the rising mobilization of disaffected parts of the electorateis highly important in understanding the electoral success of RRPs like the AfD (Schulte-Cloos and Leininger 2021, 10).While one would suspect that the AfD's electoral success might cause a higher satisfaction with democracy and the political system, there are no such effects observable, at least not in a clearly noticeable manner (Reinl and Schäfer 2020).Besides lots of other key factors as for example populist attitudes that lead to a higher probability of voting for this party (Loew and Faas 2019), the AfD generally serves a concrete substantive political demand within the population (Schwarzbözl and Fatke 2016).Moreover, one can even identify regional differences, saying that the AfD enjoys particular support in eastern German states (Pesthy, Mader, and Schoen 2020) or even within specific groups of the society, as for example the resettlers of Russian origin (Goerres, Spies, and Kumlin 2018).While it has been illustrated that mainstream parties adjust their policy positions because of the mere existence of RRPs (Abou-Chadi and Krause 2020), specific research on how established parties deal with such parties is still in its infancy.If the aim is to make RRPs disappear, mainstream parties are facing a significant challenge.This is especially the case because there are no concrete instructions to follow and the handling strategies differ a lot.Studies have shown in the past, using Austria as an example, that neither a strategy of exclusion nor a strategy of cooperation or inclusion guarantee that right-wing populists will lose votes in subsequent elections (Fallend and Heinisch 2015).Nevertheless, focussing on the German case of the AfD, especially in eastern German states, politicians of the CDU have more frequently 'started pushing for an openness towards AfD as a coalition partner' (Weisskircher 2020, 620), mainly in the hope or indeed expectation of winning back voters.
At present, the only imaginable governmental cooperation that is most likely to be formed with the AfD is a so-called Black-Blue 2 coalition made up of the CDU, CSU and AfD or a Bahamas 3 coalition consisting of CDU, CSU, FDP and AfD.Although there have recently been certain rapprochements between the mainstream parties and the AfD, little to nothing is known about how their supporters assess these rapprochements and this openness to future coalitions.This is of particular importance, however, as it is known that 'coalition preferences matter for individual vote decisions' (Meffert and Gschwend 2010, 348).These can then have both negative and positive effects on voting behaviour for specific parties included in such coalitions.It therefore seems sensible for parties to take into account corresponding attitudes of their own voters.This leads to this article's research question: How do CDU, CSU, FDP and AfD voters perceive such coalitions and what can be identified as influencing such a coalition preference?
This is an important question on two grounds: first, it expands the body of knowledge on what determines the coalition preferences of individuals and secondly, it can be understood, especially for the parties examined here, as an indicator of how their electorates feel about the formation of such coalitions.The research presented here additionally expands the knowledge about coalitions including the AfD with data that were gathered on the national level before the 2017 German federal election.
The article is structured as follows: Section 2 contextualizes what is known about the strategies towards mainstream parties, what kind of role coalitions play in such consideration and what recent rapprochements there have been between mainstream parties and the AfD.Section 3 then discusses the importance of coalition preferences in general.In this context, what is known about coalition research so far is explained as well as how the Michigan Model as a theory of voter choice may be used as an approach to identify influencing factors on the emergence of a coalition preference.Section 4 then describes the data and method used in this paper.Section 5 presents the results of an empirical analysis; the latter are summarized and concluded together with the research desiderata in Section 6.

How to deal with RRPs and the recent case of Thuringia
In what ways can parties respond to the emergence of new radical right parties?In this context, one can differentiate between disengaging strategies and engaging strategies (with several sub-dimensions) with a view to typologising the strategies of mainstream parties towards RPPs.Meguid (2005, 349-350) provides a basic conceptualization of how mainstream parties can respond to such a development, distinguishing between so-called dismissive, accommodative and adversarial strategies.A dismissive strategy means, for example, that parties can decide not to react to newly emerging parties and their political agenda.In contrast, the accommodative or adversarial strategies mean that mainstream parties move either towards or away from the positions of given niche parties.On this basis, Heinze (2018) then develops a typology in which she not only differentiates between socalled disengage strategies and engage strategies, but at the same time adds various sub-dimensions that serve to typologise the strategies of mainstream parties towards RRPs.While the disengagement strategies consist of six paths, ranging from ignoring to isolating or demonizing RRPs, parties can also focus more on their core electorates and therefore highlight their high competence in specific policy fields; the engagement strategies consist of the adoption of specific topics or the collaboration with an RRP (Downs 2012, 31;Heinze 2020, 44-54;Heinze 2018, 288-290).In particular the collaboration with an RRP, for example through forming an executive coalition, is a great risk for parties as 'there is no resilient proof that demystification by government participation is an effective strategy' (Grabow and Hartleb 2013a, 405).Instead, it can even lead to higher legitimation of these parties.In this context, some authors argue that moderate 'populists can be the stimulus for political debate and coalition-building.Extremist populists, on the other hand, appear dangerous and certainly not capable of forming coalitions with parties of the centre' (von Beyme 2019, 18).Speaking specifically about the AfD as an RRP, this party can no longer be considered as moderate (Hansen and Olsen 2019, 3), which makes any form of cooperation highly risky for mainstream parties.This applies in particular to the period from 2015 onwards, after AfD leader Lucke was expelled from the party and the AfD began to develop increasingly into a radical right-wing party (Jäger 2021, 486).
The preliminary stage to a formal coalition is either legislative cooperation or the formation of a minority government which is supported by an RRP (Heinze 2020, 44).These are usually formed primarily for political or tactical reasons to disenchant the competitors from the right wing (Grabow and Hartleb 2013b, 39-40) even though such an effect is not guaranteed.Weisskircher (2020, 620) argues for good reasons that if there is cooperation between the Christian Democrats and the AfD, it will be more likely to happen in one of the eastern German states.This is especially true against the background of the events in the eastern German state of Thuringia: In early 2020, Thomas Kemmerich, member of the Free Democratic Party (FDP), was the first German politician to be elected as premier of a German state thanks to AfD votes.Kemmerich had decided to run against a candidate from the previously ruling coalition of the Left (Die Linke), Social Democrats (SPD) and Greens (Bündnis 90/Die Grünen) and against a candidate that was nominated by the AfD after neither of the two candidates was able to achieve a parliamentary majority in two ballots.The AfD decided not to vote for their own candidate even though he was still on the ballot; instead, they opted to support Kemmerich unanimously, on which basis he managed to achieve a majority together with the votes from both to CDU and FDP. 4 Only the left coalition voted for their candidate.In spite of the AfD's votes, Kemmerich nevertheless accepted his election and consequently became the first German premier to be elected by the AfD which then led to national and international outrage (Ellyatt 2020;Hill 2020;Oltermann 2020).These incidents are highly relevant for two reasons.On the one hand, the AfD had already developed into a party of the radical right years before this election.On the other hand, the AfD in Thuringia, with its völkisch-authoritarian orientation and its leader Björn Höcke, must be classified as extreme right-wing (Häusler 2019, 26).
While previous research has focused more on the strategic decision making from the parties' perspective, it remains unknown what supporters of specific parties think about such cooperation, especially future coalition building.Therefore, it is highly relevant to focus as well on the demand side to elicit what supporters of established parties think about cooperation or coalitions with an RRP like the AfDfor example in the form of coalition preferences.Empirical research was able to prove that individuals indeed generally anticipate the formation of a specific coalition and therefore base their vote on these specific considerations (Duch, May, and Armstrong II 2010).The only research published about this specific topic deals with the case of Thuringia right before the state's election, saying that for example 73% of CDU voters and 53% of FDP voters would not be in favour of such a coalition.Furthermore, 60% of the CDU supporters expressed their clear stance that they would no longer vote for the Christian Democrats if they formed a coalition with the AfD (Salheiser, Quent, and Klaßen 2019, 10-11).For this reason, it is imperative to obtain a deeper insight into what shapes an individual's coalition preference for such a coalition.One possible way to explain this is introduced in the following.

Determinants of Individual Coalition Preferences
Classically, coalition research is primarily concerned with the question of the reasons why parties become part of a government and how stable or unstable they are (Kropp 2008), how long governments last in different institutional settings (Schofield 1995) and/or how coalition signals may influence an individual's voting behaviour (Meffert and Gschwend 2011).On the one hand, prior to an election, coalitions must be seen as more of a hypothetical construct from an individual's perspective (Plescia and Aichholzer 2017).On the other hand, both the strategic dimension of coalition building, especially how the sub-national level functions as laboratory for new cooperations (Switek 2010), and the specific role of wings within a party for coalition building processes (Debus and Bräuninger 2008) have attracted interest.
The literature on coalitions has recently been updated through work on coalition agreements and the issues they cover (Klüver and Bäck 2019), what kind of stabilizing functions they have for governments (Krauss 2018) and what kind of a key role they play even on regional or local levels (Gross and Kraus 2019).Nevertheless, all of these important findings undermine the fact that coalition research is too often focused solely on the supply side of politics and less on how the demand side actually thinks about these coalitions.When coalitions must be considered as such a hypothetical construct from the demand side, there are two aspects that are relevant to mention which are highly important for such an estimation.First, 'reasoning about coalitions is thought to be more complex than reasoning about parties, because voters vote for parties not coalitions, at least not directly' (Plescia and Aichholzer 2017, 260).Second, which relies more on how parties communicate, and which coalition signals they possibly send, the 'choice set, as the voters perceive it, does not necessarily correspond to the set of all theoretically possible coalitions' (Debus and Müller 2014, 55).Therefore, individuals must rely on what their preferred parties signal what might be a lucrative option for a future coalition.The greatest challenge is that still only little is known about the emergence of specific coalition preferences, although research about this question has recently been intensified (Bytzek 2013a;Debus and Müller 2012;Debus and Müller 2014;Plescia and Aichholzer 2017;Wurthmann, Marschall, and Billen 2019).
The limited knowledge that so far exists about coalition preferences is only about what effect they have on voting and that they are kind of a political object whose support is expressed by an individual's voting behaviour.Voting becomes so to say to an expression of coalition preferences as individuals vote for both parties and coalitions (Meffert et al. 2011).Voting is for this reason a reflection of individual coalition preferences (Blais et al. 2006;Bytzek 2013a, 55;Huber 2017).For their part, however, coalition preferences also influence voting behaviour, as a study was able to prove for the 2009 German federal election (Debus 2013).Electoral behaviour and coalition preferences are thus in a circular process in which they can continuously influence each other.
If a coalition preference is therefore a sufficient but not necessary determinant for voting behaviour, and voting behaviour can, but does not necessarily have to be, an expression of a coalition preference, this represents a conceptual challenge.Voting behaviour and coalition preferences are both objects that need political supportspecific support, to be precise (Easton 1975).The most relevant question is: What is this specific support ultimately based on?Since coalition preferences as a phenomenon are closely related to voting behaviour, the attempt has recently been made to explain the emergence of coalition preferences by adopting an approach from the field of electoral research, particularly the so-called socio-psychological approach or Michigan Model (Bytzek 2013a;Debus and Müller 2012;Wurthmann, Marschall, and Billen 2019).According to the Michigan Model, there are three key factors identified in influencing voting behaviour: a long-term party identification (PID) and two short-term influencing factors, namely issue opinions and candidate images.While they all have a direct impact on voting behaviour, the PID also has an indirect impact through issue opinions and candidate images (Campbell et al. 1960).For this reason, it can be assumed that this model also fits when explaining the emergence of specific coalition preferences.
Indeed, at least the effect of a party identification on the emergence of a coalition preference has already been examined in various studies and proven to exist (Debus and Müller 2012, 176;Plescia and Aichholzer 2017, 263-264).The empirical evidence shows not only that a coalition is better rated when it includes the most preferred party (Bytzek 2013a, 46) but also that coalition preferences are higher for these possible government formations (Huber 2019).Nevertheless, this is not necessarily true for each coalition as there has been evidence as well that supporters of the Christian Democrats evaluate a coalition with the AfD as being worse than a leftleaning coalition consisting of SPD, Greens and The Left (Wurthmann, Marschall, and Billen 2019, 305-307).One possible explanation could be that a coalition with the AfD would violate core values that led to identification with the CDU in the first place.According to this, identification with the Christian Democrats leads to warding off possible threats to both the party and its associated identity.This leads to the first hypothesis to be examined in this paper: H1: A party identification with the AfD is associated with a higher evaluation of a Black-Blue or a Bahamas coalition, whereas such a coalition is viewed less favourably by those who identify with the CDU.
While a positive candidate image is usually associated with a higher propensity to vote for a specific party (Campbell et al. 1960), German Chancellor Angela Merkel (CDU) has become kind of an exceptional case.While her popularity indeed leads to a higher propensity to vote for CDU/CSU (Schoen and Greszki 2014, 263-264), she has furthermore enjoyed an increasing popularity since the beginning of her term in office.Only AfD voters had a much more negative attitude towards her in 2017 than in 2013 (Berz 2019, 550).Bearing in mind that one of the AfD's top candidates for the 2017 German federal election, Alexander Gauland, announced on the election day: 'We will hunt Mrs. Merkel and whomever else.And we will take our country and our people back' (Benner 2017), it is obvious that Merkel has become an enemy image for AfD supporters due to her decision to take in refugees in Germany during the refugee crisis of 2015.Merkel's refugee politics was also one of the main reasons for voters in favour of the governing parties CDU, CSU and SPD to defect to the AfD in the 2017 German federal election (Wurthmann et al. 2020).As candidate images are highly influential for voting behaviour, one would expect the same to apply to a coalition preference, particularly so because a potential governmental coalition would be formed by the parties' most relevant politicians.This leads to the following hypothesis: H2: The higher an individual's opinion of Angela Merkel is, the less desirable a coalition with the AfD is rated.Huber et al. (2009) were able to prove that the expected probability of a coalition to be formed also affects individual coalition preferences.Therefore, similar results should also be expected concerning a Black-Blue or a Bahamas coalition, leading to the third hypothesis: H3: The more likely the formation of a Black-Blue or Bahamas coalition is perceived, the higher the preference also is for such an alliance.
Since a negative stance on migration issues is one of the pull factors to vote for the AfD, it is very likely that there is also an observable effect towards coalitions involving the AfD in a government coalition.Moreover, 'AfD sympathizers are not only pronounced welfare chauvinists, but they are also highly critical of class-based redistribution via welfare and taxation' (Goerres, Spies, and Kumlin 2018, 261).Interestingly, at least on the party level, Bräuninger et al. (2019, 91) show in their analysis that the weighted ideological heterogeneity of a Bahamas coalition, along the economic and the immigration policy dimension, is smaller than the one of the later formed grand coalition between CDU, CSU and SPD.One could therefore suspect that this might translate into the individual level and that, if the emergence of a coalition preference depends as much on issue orientation as voting behaviour, one might be able to measure such an effect.This leads to the following two research hypotheses: H4: The more critical an individual is of migration, the more desirable a Black-Blue and a Bahamas coalition are rated.and H5: The more an individual rejects redistributive policy, the more desirable a Black-Blue or Bahamas coalition is rated.

Research Design
The data used for this paper were gathered during a research project for the 2017 German federal election.An online panel survey 5 was designed for this purpose and carried out nationwide by the market research company Respondi.The data used in this article were collected during the first survey wave which took place between August 21st and August 29th 2017 one month before the German federal election. 6The respondents, aged between 18 and 69, were recruited using a quota sample based on the Respondi Access Panel and its 90,000 registered members.When comparing socio-demographic characteristics between representative census data and the respondents, only minor deviations can be determined (see Online Appendix, Table 1).2,380 people were recorded, of which 68 respondents were excluded from the analysis because they stated that they either were not entitled to vote in the 2017 election or they were not adequately informed about their current status as to whether or not they were allowed to vote.
The purpose of this paper is to examine not only what German society as a whole thinks about a Black-Blue and a Bahamas coalition but also what the attitudes of the CDU, CSU, FDP and AfD electorates are towards such a coalition.This is of specific interest as their preferred party would be part of such a government.Moreover, through a subgroup analysis where only the electorates of these four parties are analysed together, there should be indicators identified that have an impact on such coalition preferences.In order to do so, there are two linear regression models employed: one for a Black-Blue coalition comprised of CDU, CSU and AfD, and one for a Bahamas coalition comprised of CDU, CSU, FDP and AfD.How these regression models are composed is explained in the following (see Online Appendix, Table 2 for a detailed operationalization list).
Although the data was collected for the 2017 federal election, we are able to infer at least two things from the above analysis.Firstly, after the German federal election in 2017, the AfD became more radicalized, which has also triggered surveillance measures by the Office for the Protection of the Constitution.Accordingly, if the supporters of the CDU, CSU and FDP already had a low preference for a Black-Blue coalition or a Bahamas alliance in 2017, it may be assumed that this effect only intensified in the further course of time.Secondly, the analysis also clarifies whether and to what extent the events in Thuringia were negligent; this is because these appear to have run counter to the interests of the CDU and FDP constituencies there.Nevertheless, the timing of the data collection should be borne in mind.

Dependent Variables
There are two dependent variables used to examine individual coalition preferences.Respondents were presented with a list of different coalition options and asked the following question: 'Regardless of the outcome of the federal election, how desirable do you personally consider the following coalition governments to be?' Part of this list was not only a Black-Blue coalition but also a Bahamas coalition.Respondents could then express their position on a scale ranging from (1) −5 not at all desirable to (11) +5 extremely desirable.

Independent Variables
The individual party identification was measured through an open question asking for the leaning towards a specific party 'Do you lean towards a certain partyin general?And if so, which one?' The variable was then used to create four dummy variables that represent a respective party identification with the CDU, the CSU, the FDP or the AfD; (0) stands for no identification with this party and (1) stands for 'I identify with this party'.
Candidate images where measured through asking the respondents 'What do you think of the top candidates of the parties for the upcoming German federal election?'They could then express their opinion in a scale that ranged from (1) −5 I am not in favour of this person at all to (11) +5 I am very much in favour of this person.A total of four candidates are included in the analysis.These are made up of the joint candidate of the CDU and CSU, Angela Merkel, the FDP party leader Christian Lindner and the AfD candidates Alexander Gauland and Alice Weidel.
The respondent's assumed probability of a specific coalition formation after the German federal election was measured through asking them how likely they think that a governing coalition will be formed out of a specific combination of parties.The Black-Blue and the Bahamas coalition were two out more than ten combinations.The scale used to help the respondents express their stance ranged from (1) −5 very unlikely to (11) +5 very likely.
The policy positions of the respondents were measured through two questions.One of them specifically deals with the individual's position towards migration, asking 'And what is your position on the topic of immigration opportunities for foreigners?'The scale used for the measured ranged from (1) The possibility of immigration for foreigners should be facilitated to (11) The possibility of immigration for foreigners should be restricted.The second policy item measures the socio-economic attitudes of the respondent towards their position on taxes and welfare state benefits.The corresponding scale ranks from (1) Less taxes and duties, even if that means fewer welfare state benefits to (11) More welfare state benefits, even if that means more taxes and duties.

Control Variables
The regression analyses used here require the inclusion of various control variables.The items used are sex, age and educational level of the respondents.As there were differences observed between eastern and western Germany concerning the support for the AfD, it only seems logical to control for similar effects on coalition preferences that involve the AfD as a coalition partner.In addition to that, the individual level of political interest is included to control for the effects on coalition preferences.

Results
At the time the data for this study was gathered, governmental cooperation with the AfD, either as a Black-Blue or as a Bahamas coalition, was quite unpopular.Especially the Christian Democratic Union's supporters are highly critical of a Black-Blue coalition (m [mean] = 2.13 [95% CI = 1.89, 2.37]), followed shortly after by the Christian Social Union voters (m = 2.18 [95% CI = 1.75, 2.60]).The Free Democratic Party's supporters are also very critical of such a coalition (m = 2.59 [95% CI = 2.09, 3.09]) but less than those of CDU und CSU which is an interesting finding.One would usually expect electorates of a party involved in a government formation to view such a coalition more favourably than supporters of parties which are not involved.In contrast, the findings regarding the Alternative for Germany's supporters confirm what would be expected: they are highly in favour of such a coalition (m = 7.09 [95% CI = 6.64, 7.55]).The findings become more interesting focusing on the electorate's attitudes towards a Bahamas coalition.While the CDU (m = 2.11 [95% CI = 1.88, 2.34]) and CSU (m = 2.11 [95% CI = 1.76, 2.57]) voters are even more critical towards such a government formation, the supporters of the FDP are still critical towards such a coalition but much less than they are towards a coalition to be formed without them (m = 3.10 [95% CI = 2.54, 3.67]).In contrast, the AfD's voter attitudes towards a Bahamas coalition is somewhere between neutral and positive (m = 6.18 [95% CI = 5.72, 6.65]) although they are much less in favour of such a coalition compared to a Black-Blue government formation (see Figure 1).
Even a so-called Red-Red-Green coalition, which would be formed by Social Democrats, the Greens and the Left, seems at least for the electorates of CDU (m = 2.46 [95% CI = 2.25, 2.67]) and CSU (m = 2.48 [95% CI = 2.09, 2.87]) to be more attractive than a coalition with the AfD.This is of great importance because the CDU and CSU usually describe horror scenarios right before every major election depicting what a Red-Red-Green coalition would mean for Germany and why one should vote for CDU/CSU to prevent such a government formation.The image of Red-Red-Green coalitions serves these parties as an ideal enemy to mobilize their own supporters and makes it more meaningful when they nevertheless evaluate it as being better than a possible coalition which might include their own party.In contrast, FDP voters prefer (m = 2.71 [95% CI = 2.32, 3.10]) a Bahamas coalition to a coalition consisting of the SPD, the Greens and the Left.Furthermore, there are no significant differences observable between the CDU, CSU and FDP electorates concerning their attitudes towards a Black-Blue and a Bahamas coalition while they all differ very significantly from the AfD's electorate (p < 0.001).Overall, it does not seem like forming a coalition with the AfD either with the FDP or without the Free Democrats would be a good idea from a Christian Democratic or Christian Social position.Their electorates are definitely not in favour of such a coalition.Similar results may be found for the FDP voters.Their supporters are nevertheless more in favour of such cooperation than those who support CDU and CSU.Supporters of the AfD, in contrast, are very supportive of governing with the CDU or CSU and they are a little less in favour of an additional FDP in future government coalitions.
But what determinants can be taken into account to explain such coalition preferences with the AfD as an RRP?The findings show that various factors of the Michigan Model can indeed prove to be explanatory for the emergence of coalition preferencesin a positive and also a negative way.While a party identification with the CDU leads to a decrease of desirability of a Black-Blue or a Bahamas coalition, it leads to the opposite when one identifies with the AfD.The desirability of such coalitions increases when an individual identifies itself with this party.Further effects of a PID cannot be observed, identifying with the CSU or the FDP therefore has no effect on attitudes towards such coalitions.For this reason, as there was nevertheless no such effect expected, H1 can be confirmed.In contrast, H2 must be rejected since a positive evaluation of Angela Merkel (CDU) has no effect on an attitude towards either a Black-Blue or a Bahamas coalition.Only a positive evaluation of the AfD's candidates Alexander Gauland and Alice Weidel has a positive and significant effect on the emergence of such coalition preferences.Moreover, there is a positive effect associated between a positive evaluation of FDP's party leader Christian Lindner and a more positive evaluation of a Bahamas coalition.It is indeed no big surprise that a positive evaluation of AfD politicians and the FDP party leader affect a coalition preference involving the parties these politicians belong to.Due to the fact that Angela Merkel is particularly unpopular among AfD supporters, but at the same time very popular among CDU, CSU and FDP supporters, an effect would have been expected here.However, the effect of this would have been contrary to previous assumptions that appreciation for a candidate would lead to support for a coalition involving that person and their party.The lack of support or the non-emergence of such a coalition preference could be an expression of the low esteem in which a Black-Blue or Bahamas coalition is held within the CDU and CSU electorates (see Table 1).
The findings underline those of Huber et al. (2009): the assumed probability of a specific coalition formation indeed has an influence on the emergence of coalition preferences, although the effect is not the same for both a Black-Blue and a Bahamas coalition.In fact, the more likely the formation of a Black-Blue coalition is perceived, the less desirable the formation of a Bahamas coalition appears to be from an individual perspective.Interestingly, no such effect through an assumed probability of such a coalition to be built can be observed towards a Black-Blue coalition preferenceneither positive nor negative.In contrast to this, the assumed probability of forming a Bahamas coalition not only has a significantly positive effect on the preference for a Black-Blue coalition, it also positively affects the preference for a Bahamas coalition to a much greater extent.One possible reason for this difference may be found in the generally perceived probability for these governmental coalitions to be formed.While supporters of the AfD expect a Bahamas coalition less likely to be formed after the 2017 German federal election than a Black-Blue coalition, all other electorates perceive this differently.Although the differences are marginal and these voters also do not estimate a Bahamas coalition as likely to be built, these results are still noteworthy.For this reason, H3 can be confirmed for a Bahamas coalition but must be rejected for a Black-Blue coalition.
Finally, the findings about the influence of attitudes towards specific issues on coalition preferences, such as immigration and social welfare distribution, are interesting as well.Individual attitudes on immigration do indeed affect the emergence of a coalition preference that would include the AfD in a future government coalition.The more likely an individual is to think that immigration should be more restricted, the more likely s/he is to believe that either a Black-Blue or a Bahamas coalition is desirable.The effect on the former is greater than on the latter.One explanation for this may be the fact that there are obviously voters for the parties mentioned who would like to reverse the refugee policy of Angela Merkel's government from 2015.H4, which stated such a relationship, can therefore be confirmed, whereas H5 can only be confirmed for a Black-Blue coalition but not for a Bahamas coalition.The rejection of social and welfare state redistribution measures thus only has an effect on the emergence of a Black-Blue coalition preference, but not on a Bahamas coalition.In light of the fact that the AfD and the FDP are parties that both share a more market-liberal orientation (Franzmann 2019), a dismantling of redistributive measures involving the FDP would be more likely.From this point of view, a Bahamas coalition should seem to be more desirable than a Black-Blue coalition, but the existing variance cannot be explained completely.
In addition to that, women and older individuals are less likely to be in favour of such coalitions involving the AfD.Political interest and whether one comes from the western or eastern part of Germany do not matter for specific coalition preferences.Especially the fact that the regional aspect does not matter is interesting as this factor definitely plays a role in voting behaviour differences between western and eastern Germany.
All in all, the predictive power of the models as an explanation for the emergence of coalition preferences is very high.While both coalition preferences can be explained quite well, the model fits even better for a Black-Blue coalition (Adj.R² = 0.548) compared to a Bahamas coalition (Adj.R² = 0.503).Especially for social sciences, such high values for the adjusted R² seem to be quite unusual, but they are fairly normal for the few studies that have been published on coalition preferences within the German party system (Bytzek 2013b, 48;Wurthmann, Marschall, and Billen 2019), especially when the Michigan Model is used as the theoretic foundation for the measurements.

Discussion
While knowledge in social science research about voting behaviour in favour of the AfD is growing, the way mainstream parties deal with radical right parties is largely unexplored.As various studies have shown, dealing with RRPs like the AfD differs significantly across countries (Grabow and Hartleb 2013a;Heinze 2018).As can be demonstrated by the example of Thuringia, with the election of the FDP Prime Minister Kemmerich through support of the AfD, intentionally or unintentionally, German politics has reached a preliminary stage for the formation of a formal coalition with the AfD.In fact, by accepting to be elected though the votes from the AfD, until Kemmerich's resignation, the party was politically legitimized to an extent that previously seemed unthinkable.Nevertheless, this also raises the question of how the formation of formal government coalitions with the AfD is assessed, which was the starting point of this article.For this reason, two specific coalition preferences, one for a Black-Blue coalition consisting of CDU, CSU and AfD and one for a Bahamas coalition consisting of CDU, CSU, FDP and AfD were the main interest of investigation.The descriptive findings indicate that of the electorates examined, those of CDU and CSU are the most critical of a Black-Blue or a Bahamas coalition.The FDP voters are also critical of such a coalition but much less than the CDU and CSU voters.This is specifically of interest as the usual assumption is that voters evaluate a coalition better when their most preferred party is part of such a coalition.Indeed, FDP voters conform with the assumptions described.Conversely, the significance of CDU and CSU voters preferring a coalition consisting of SPD, Greens and the Left over a coalition of which they would form part, be it a Black-Blue or a Bahamas coalition, should not be underestimated.In addition to the findings for the FDP voters, those for the AfD voters are also in line with the expectations.AfD voters evaluate coalitions that would include their party much better than alternative coalition models.
One of the central concerns of this article has also been to be able to explain the coalition preferences for a Black-Blue and a Bahamas coalition.It should be noted that while a party identification with the AfD leads to the expected higher rating of a Black-Blue and a Bahamas coalition, the opposite is the case with the CDU voters.The desirability for such a governmental coalition decreases in a meaningful way when an individual identifies himself with the Christian Democrats.For the CSU and the FDP voters, no significant effects can be determined.Furthermore, the most notable effects refer to the effect of candidate images and attitudes towards specific policies.While one expectation was that a positive image of Christian Democrat politician Angela Merkel would make it less likely to evaluate coalitions with the AfD as being desirable, there are no significant effects observable, whereas a positive evaluation of AfD's candidates Alexander Gauland and Alice Weidel indeed increases the desirability of a Black-Blue or a Bahamas coalition significantly.In addition to that, the more one thinks that migration to Germany should be more restricted, the more desirable a governmental coalition with the AfD seems.This is in line with expectations that such cooperation in the government might cause a shift within German politics.Finally, while assuming the formation of a Bahamas coalition to be very likely after the German federal election of 2017 leads to an increasing desirability of such a coalition, no observable effect through a Black-Blue coalition can be found.The only noticeable effect is that the more likely the formation of a Black-Blue coalition appears, the less desirable the formation of a Bahamas coalition is perceived.
All these findings and the overall explanatory power indicate that the Michigan Model as a theoretic framework is highly relevant for the emergence of coalition preferences.Nevertheless, more research should be done in this field.It seems obvious to investigate to what extent long-term oriented social value orientations are relevant as a predictor for coalition preferences.Since recent research shows that for example religious individuals are less likely to support the AfD (Dilling 2018, 97;Marcinkiewicz and Dassonneville 2021, 7), it would be interesting to gather more knowledge whether the same effects can be measured as determinants for the emergence of coalition preferences.Moreover, as parties can differ greatly at the state level, it would be interesting whether regional differences can by observed towards coalition preferences in general but also towards the AfD in specific.It would also be important for surveys capturing such coalition preferences to be conducted in the form of large-scale computer-assisted personal interviews in the future.Nevertheless, the German mainstream parties should, all in all, take the findings of this article seriously.A coalition with the AfD was already far from desirable in 2017, and even more so as the party has further radicalized itself thereafter.The parties should listen to their supporters.

Notes
1.While the CSU is a party that only competes for votes in Bavaria, the CDU competes for votes in the whole country except for Bavaria.Since 1949, both parties have always formed a common party group in the Bundestag, Germany's national parliament.2. In German politics, parties are usually associated with specific colours.For the CDU and CSU, black became the externally attributed colour as priests used to wear black cassocks and both parties claim to represent Christian ideals, whereas the colour blue was chosen by the AfD on their own.3.In addition to the CDU's and CSU's black and also AfD's blue, the FDP is mainly identified through its yellow party colour.All three of these colours are part of the national flag of the Bahamas.The description as 'Bahamas' coalition was first used in the beginning of September 2013 (Welt Online 2013).
4. The possibility of such a feint was widely discussed in the media in the run-up to the election.5.The author wishes to acknowledge the funding of the panel study by the Fritz Thyssen Foundation, Cologne.6.Only in Wave 2 were coalition preferences still surveyed.However, since this wave was surveyed directly before the Bundestag election, it could be strongly influenced by short-term coalition signals and/or current polls.For this reason, the decision was made to use Wave 1 as the data basis.

Figure 1 .
Figure 1.Coalition Preferences for a Black-Blue Coalition and a Bahamas Coalition.Source: Author's own calculation and presentation.

Table 1 .
Coalition preferences for a Black-Blue and a Bahamas Coalition.