It’s no longer the economy, stupid! Issue yield at the 2017 German federal election

Abstract This article demonstrates that the issue-yield concept is able to predict the electoral strategies of mainstream and challenger parties at the 2017 German federal election. While the electorate of mainstream parties favour valence issues, the Greens and the AfD can gain more by concentrating on socio-cultural positional issues. Relying on a unique survey covering 17 positional issues and 10 valence issues as well as an analysis of Twitter accounts, the article shows that contemporary Germany is characterised by a centrifugal competition on the socio-cultural dimension. At the same time, an asymmetric ideological confrontation persists on the socio-economic dimension, because the Left and the SPD still refer to their traditional welfare issues while the bourgeois parties no longer counter this with a contrasting free-market ideology. Thus, the economy is currently not the decisive issue in German politics. Migration, integration, and other socio-cultural issues are rather driving electoral competition.

issues. However, German parties' programmatic positions continue to refer also to the traditional socio-economic dimension. Second, German party competition is characterised by the effects of a so-called Grand Coalition of two large mainstream parties located in the ideological centre. After the early election in 2005, which marked the end of the red-green Schr€ oder government, two out of three governments were Grand Coalitions of the CDU/CSU and SPD. Hence, discontent with the government could only be expressed by voting for parties not located at the centre, which, at least to a substantial degree, led to the formation of yet another Grand Coalition after the election analysed in this contribution.
There are many ways to conceptualise political competition and its consequences for party behaviour and party system configurations. Identical to the other contributions in this special issue, we rely on the concept of issue yield (e.g. De Sio and Weber 2014) as our general framework. In short, the concept means that campaign strategies of political parties are not simply driven by the systemic (i.e. overall) salience of political issues and a given party's competence profile. Instead, parties focus on issues where their position is almost unanimously shared by their supporters and widely shared by the electorate at large, so that they can expect electoral gains. Hence, issue yield should influence parties' electoral strategies and campaigns while informing us about the nature and structure of political competition at the same time. In a modification to the original approach, the electoral potential of valence issues can also be taken into account by incorporating explicit measures of party credibility (De Sio and Weber 2019) whichas we argue belowis very important for an adequate representation of current political competition in Germany.
We focus on three major research questions: first, we examine to what extent the classical ideological conceptualisations of the twentieth century are still valid for German parties. We do so by looking at the issue-yield patterns in terms of more left-wing and more right-wing policy stances for both the economic and the cultural dimensions of political conflict. Second, we analyse the different vote-maximising potentials of position and valence issues for governing and opposition parties. Third, we test one of the core assumptions of the issue-yield theory in that parties are less driven by systemic salience than by their party-specific risk-opportunity calculations. We rely on a unique pre-election mass survey providing information on 17 position and 10 valence issues and on Twitter data of electoral campaign frontrunners for each party (see online appendix for details). The survey data is used to calculate issue-yield scores while the Twitter data refers to party campaign strategies as it allows different levels of issue emphasis to be distinguished.
We demonstrate that contemporary Germany is characterised by centrifugal party competition on the socio-cultural dimension. At the same time, an asymmetric ideological confrontation on the socio-economic dimension persists because the post-communist Left Party (Left) and the SPD still refer to their traditional welfare issues even though the Christian Democrats do not contrast this with a free-market ideology. Further, we show that the German governing parties CDU/CSU and SPD mainly address valence issues, while especially the right-wing populist Alternative for Germany (AfD) and the Greens concentrate on socio-cultural position issues. The AfD and the Greens had unified support within their electorates, particularly regarding immigration, while the support of the CDU/ CSU and SPD was split over this issue. Hence, the mainstream governing parties avoided position issues in order not to alienate parts of their own electorate. The Liberals (FDP) faced a specific strategic situation because they needed to fight for re-entry into the Bundestag after their electoral meltdown in 2013. Yet their core issues belong to the declining economic dimension in which they are set apart on the right side of the dimension from all other partiesexcept the AfD (Franzmann 2019). However, we also show that the FDP benefits from valence issues more than all other smaller parties. The article concludes with an outlook on future problems of government formation in Germany due to the asymmetries in party competition.

The transforming political space of Germany
The central idea of issue-yield theory is that parties adopt their issue stances according to their individual risk-opportunity profile. Hence, according to this model, concentrating on issues with the highest systemic salience is not the most promising strategy. Instead, parties will focus on those issues which unite their core voters and, at the same time, are widely supported by the overall electorate (De Sio 2018: 398). Put differently, policy goals that could attract many votes from the overall electorate are only emphasised if they do not alienate the core electorate. Therefore, issue yield can be defined 'as the degree to which an issue allows a party to overcome the conflict between protection and expansion of electoral support' (De Sio and Weber 2014: 871). The interaction of political parties with the electorate is analysed as acting on an 'issue-market' where policy-goals are 'exchanged' via communication (Franzmann 2011). Due to the process of de-ideologisation, parties are relatively free to combine each kind of issue beyond traditional cleavage politics. Parties can pursue problem-solving or conflict-mobilisation strategies, while the latter is not necessarily related to the traditional cleavages (De Sio and Lachat 2019). Furthermore, an important general assumption within the issue-yield framework is that parties can actively change the character of an issue as being valence or positional. Depending on the current state of the public debate an issue can be perceived as commonly shared. In this case, party competition over this particular issue concentrates on problem-solving credibility, even if this issue could theoretically be framed as a positional .
German party competition provides a very interesting case for applying the issue-yield concept. In contrast to other European democracies discussed in this special issue, Germany has not moved towards de-ideologised politics. Instead, socio-cultural issues drive a process of reideologisation while economic issues are currently of lower salience as a result of a flourishing economy. At the same time, we show how the issue-yield concept can help us to better understand the state of German party competition.
In order to formulate expectations of how German parties react to this changing environment, we will first discuss the transformation of the demand side (the electorate), and then we will show that the supply side of the issue-market is still struggling with adapting to this transformation.
The demand side: the socio-cultural turn of the German electorate The transformation of the German political space came in two instalments. The most obvious effects of these changes were, first, the emergence of the Greens and, second, the rather 'belated' establishment of the right-wing populist AfD, accompanied by the electoral decline of the SPD and the overall increase of party-system fragmentation. These are all symptoms of a phenomenon several scholars already described during the past decades. Some labelled it as the change from class-based to 'new politics' (e.g. Dalton 1985;Kitschelt and Hellemans 1990;Poguntke 1987), following the idea that a 'silent revolution' established a conflict over post-materialist attitudes (Inglehart 1984). Ignazi (1992) diagnosed a 'silent counter-revolution' that would lead to the establishment of rightwing populist parties across Europe. The importance of economic issues has been declining for decades. Simultaneously, socio-cultural issues have grown increasingly salient. Politics struggles less about economic concerns. Kriesi et al. (2008) argued that this development is boosted by the process of de-nationalisation and globalisation. While at the beginning issues of self-actualisation and environmental protection led to an increased salience of non-economic issues (Inglehart 1984), now issues related to new demarcation/integration conflict (Kriesi et al. 2008) dominate the socio-cultural dimension of political contestation. In the consequence, a 'new cultural divide' (Bornschier 2010) has started structuring party competition. Kitschelt (1994Kitschelt ( , 2003 illustrated this change by an elliptic configuration of voter demands for policies that changed from a horizontal, economic 'left-right' to a diagonal placement within the political space. The poles of the vertical axis of the political space are defined by libertarian or authoritarian attitudes. In the political science literature, we find a wide range of terms for these attitudes such as GAL (green, alternative, libertarian) vs. TAN (traditional, authoritarian, national) by Marks et al. (2006), libertarian-universalistic vs. traditionalist-communitarian values by Bornschier (2010, or cosmopolitism vs. communitarianism by Merkel (2017). Kitschelt (2003: 130) argued that the diagonal is in the process of turning into a vertical column, representing an ideological conflict constellation for which the traditional left-right class distinction would only be of limited importance. Flanagan and Lee (2003) similarly predicted that economic issues will become valence issues, while socio-cultural issues will figure as the decisive positional issues. Germany was long seen as an exceptional case where right-wing populists were not successful (Arzheimer 2015), arguably because the experience of the Nazi past had truncated the ideological spectrum (Smith 1982). Furthermore, after reunification, economic issues continued to dominate the political agenda despite the success of the post-materialist Greens.
This changed drastically during the legislative period from 2013 to 2017. The establishment of the AfD followed the pattern Kitschelt and Bornschier diagnosed for other countries. However, in the German case, there were also important catalytic events. The so-called 'refugee crisis' in 2015 constitutes a critical juncture which led to changing party-voter alignments and a strong politicisation of immigration issues. People opposed to immigration from now on tended to support the AfD instead of the CDU/CSU (Mader and Schoen 2019). The AfD benefited from becoming the issue-owner of the migration issue (Grande et al. 2019). Despite the AfD's early focus on the European integration process and especially on European currency and financial crisis policy, voters were mainly attracted by latent anti-immigration attitudes (Schwarzb€ ozl and Fatke 2016). In this vein, Germany is similar to other European countries (Bechtel et al. 2014). From the beginning, AfD politicians addressed right-wing populist attitudes and hence the demarcation/integration divide, even if they belonged to the rather moderate and bourgeois economist wing (Lewandowsky et al. 2016;Franzmann 2016). With Chancellor Merkel's decision to open the borders to large numbers of refugees in summer 2015, migration became the dominant topic within the German electorate and the public sphere. In combination with the steady decline of unemployment and a rather positive overall economic outlook, this socio-cultural issue dominated the German policy agenda (Forschungsgruppe Wahlen: Politbarometer 2018). Given this shake-up of the demand-side agenda, political competition changed drastically for the political parties compared to earlier elections.
The supply side: towards polarised pluralism The era of Angela Merkel's chancellorship has been characterised by Grand Coalition government (except for 2009-2013). In other words, the defining feature of the supply side of German party competition was a centre coalition and a lack of government alternation whereas German politics had been dominated by a bipolar pattern of party competition before. 1 The electoral weakness of the SPD in conjunction with its refusal to admit the post-communist Left to a coalition at the federal level left a Grand Coalition as the only politically viable option. To be sure, there would have been other majorities in the Bundestag, but these options did not seem politically feasible. Connected in a Grand Coalition, Christian and Social Democrats had to defend the same government policy, moving closer on economic and social policy. Even in the aftermath of the financial crisis, German party competition was dominated by valence issues (Clarke and Whitten 2013), and this did not change when a bourgeois coalition of CDU/CSU and FDP assumed office in 2009. Even more, Merkel fostered a course of de-polarisation, modernising the CDU's position on gay marriage, child care, and nuclear energy, where she switched to a nuclear phase-out. Similar to what had happened to the SPD earlier in the wake of its socio-economic reform agenda, this left the CDU (and to a lesser degree also the CSU) with a substantially weakened core ideology (Poguntke 2014: 960). Nevertheless, Merkel's decision to switch back to the nuclear phase-out was also mainly driven by cost-benefits calculations as substantive segments of CDU sympathisers changed their minds after the Fukushima disaster. Many CDU voters now held inconsistent preferences in that they preferred a nuclear phase-out but at the same preferred the pro-nuclear CDU as a party (Thurner 2017: 175). In order to avoid vote loss, the CDU/CSU changed its government's course of action since there was no competitor promoting nuclear energy in 2011 (Meyer and Schoen 2017). This situation changed with the establishment of the AfD two years later.
Consequently, a gap opened to the right of the Christian Democrats, and this is where the AfD entered the competition. Figure 1 illustrates the German political space at the 2017 federal election. The two-dimensional position scores are based on the manifesto data (Volkens et al. 2017), applying the context-sensitive approach of Franzmann and Kaiser (2006). It reflects peculiarities of the German party competition, e.g. that support for the 'social market economy' is historically located in the bourgeois party camp (Franzmann and Kaiser 2006). 2 Figure 1 reveals a situation that Marks et al. (2006) identified as a pattern for the whole of Western Europe: all relevant parties are close to a diagonal from the left-libertarian upper left corner to the right-authoritarian lower right corner. In other words, there is one dimension of party competition which accommodates a two-dimensional issue space, and the polar opposites of this axis are defined by the Left and the AfD. However, this figure does not reveal which dimension is more salient for each party. Actually, issue emphasis of the Greens and the AfD is mainly devoted to socio-cultural issues, while the FDP and the Left are more concerned with socio-economic issues. 3 The CDU/CSU is placed at the very centre, a consequence of Merkel's moderate, less conservative course of action over the past years. In principle, this provides an advantage for the CDU/CSU as this shift has shortened the distance to the Greens, which now constitute a potential coalition partner (Poguntke 2014: 960). At the same time, the CDU/CSU allows space for a more conservative political party: the AfD.
After a neo-liberal phase under Chancellor Gerhard Schr€ oder until 2005, the SPD has moved further to the left. However, in the meantime, the SPD and CDU/CSU formed a centre coalition twice, a constellation for which Sartori (1976) predicts a centrifugal competition because  Franzmann and Kaiser (2006). discontent with the government can only be expressed by voting for parties far from the centre. Even though Sartori's argument was originally developed for a one-dimensional space, we can easily transfer it to a twodimensional space of competition: Unsatisfied voters with salient preferences on socio-cultural issues will vote for the Greens, the Left, or the AfD, while those with salient preferences on socio-economic issues will vote for the Left, the AfD, or the FDP. In Figure 1, the box in the middle represents the centrist space occupied by the Grand Coalition. The arrows symbolise the two-dimensional centrifugal dynamics. Translating Sartori's framework into an agent-based model, it can be shown that only coalitions of the left camp (e.g. the SPD, the Greens, and the Left) or the right camp (the AfD, the FDP, and the CDU/CSU) could change the direction of competition from a centrifugal to a centripetal one (Schmitt and Franzmann 2019). The traditionally more conservative Christian Democratic sister party CSU is suffering from this centrifugal dynamics because it is less credible in taking a conservative position on socio-cultural issues than the CDU. Hence, the centrifugal dynamic even works within the Christian Democratic camp. Since both parties build a joint electoral bloc, their core electorate is potentially torn on this issue. Thus, according to issue-yield theory, CDU/CSU will avoid stressing socio-cultural issues too much during the electoral campaigns.
Remarkably, compared to the demand side discussed above, the supply side seems to be delayed in its adaptation. While, as Kitschelt (2003) suggests, the demand side concentrates more and more on the vertical sociocultural dimension, the supply side is still a diagonal, thereby aligning both dimensions of competition. The FDP keeps aloof from this main supply-side competition axis, occupying the upper right corner representing a combination of economic right-wing and libertarian issues. Contrary to other European countries (see D'Alimonte et al. 2019, and De Sio and Lachat 2019), this reflects not a consequence of a de-ideologised political space. The FDP's core issues, a combination of culturally progressive and economically right concerns, leave the party's location in the competition space more or less unchanged in comparison to earlier elections. Based on these observations we will now formulate our expectations.

Party strategies: expectations
What are the rational strategies for the German parties, given the tension between supply and demand sides caused by the transforming political space and the simultaneous potentially centrifugal dynamic? Arguably, the mainstream centre coalition partners avoided putting immigration and related issues on the agenda since this would have divided their electorates. In addition, emphasising positional issues carries the risk of strengthening the centrifugal dynamics of party competitionparticularly if positions on both sides of the spectrum are strongly associated with a certain political party or political camp. According to issue-yield theory, the mainstream coalition parties should have focused on valence issues instead of positional issues. Moreover, as the two biggest mainstream parties had been coalition partners for 8 out of the last 12 years, competing on positional issues had become much more difficult for themnot just because of credibility problems. The good government record on growth and unemployment meant that economic valence issues should provide the best risk-opportunity structure, especially for the major governing party CDU. The SPD, being the junior coalition partner, was unlikely to benefit to the same degree. Furthermore, such a strategy prevented the SPD emphasising crucial differences to the CDU. The situation was somewhat different for the CSU, which plays a specific role as dominant regional party in Bavaria and represents the conservative wing of the sister parties at the federal level. The CSU thus had some reason to mobilise on the immigration issue. However, at the federal level, the CDU/CSU core electorate was split over the immigration issue, which meant that downplaying immigration should have been a rational overall strategy.
We expect the reverse pattern for the two opposition parties representing the poles on the socio-cultural dimension. The Greens and the AfD had great incentives to focus on positional socio-cultural issues in order to benefit from both the centrifugal dynamic of party competition and the transforming demand side. The post-communist Left had the chance to present itself as a challenger party not only opposed to the established 'grand coalition state', but also to the vanishing of political conflict over economic issues. The FDP faced a more complicated situation than the other opposition parties. The declining saliency of economic issues threatened the attractiveness of their positional core issues regarding the free market economy. It would have been rational for the Liberals to concentrate on economic valence issues instead of their free market position in order to attract votes for their economic competence. However, as the FDP had lost all seats in the previous Bundestag elections it was much less visible and credible than before. In addition, it was still struggling to reinvent itselfin terms of both policies and personnel (Giebler 2013).
To summarise, we formulate the following general expectations regarding the three research questions that will guide the empirical analysis of the 2017 election. In a first step, we investigate the degree to which the ideologies of the twentieth century's class struggle might have lost their relevance for political competition. As we argued above, we assume that due to the transformation of the political space, positional issues now mainly belong to the socio-cultural dimension. At the same time, we expect that economic issues were less important. This should have resulted in patterns of attitudes among party supporters which no longer relate clearly to an underlying ideology and mix issues originating from the two different dimensions in terms of high issue-yield scores.
Expectation 1: The German electorate is less aligned along the economic left-right dimension and parties tend to mix left and right with GAL and TAN positions to maximise vote shares.
The second research question addresses the patterns of valence and position issues between parties. Due to the increased salience of sociocultural issues and the potentially centrifugal dynamic induced by the Grand Coalition, both mainstream parties CDU/CSU and SPD should reveal the highest issue yield on valence issues. Therefore, the most promising strategy for both parties should be to concentrate on these issues. Opposition parties, especially those with core issues belonging to the socio-cultural dimension (AfD and Greens), should benefit mainly from emphasising position issues. The Left, on the other hand, is a party with core issues belonging to the economic dimension. Hence, the most successful strategy for the Left should be to stick to its traditional core economic position issues, opposing the declining relevance of economic issues on the political agenda. Furthermore, in contrast to their party elite, the Left's voter base in East Germany has traditionally been split on socio-cultural issues. The FDP as an established, long-lasting government party also had the possibility to emphasise its competence on (economic) valence issues.
Expectation 2a: As long-standing governing parties, CDU/CSU, SPD and FDP strategically adapt best to the voters' demands by a concentration on valence issues. Expectation 2b: As opposition parties, the AfD and the Greens strategically adapt best to the voters' demands by a concentration on socio-cultural positional issues, in which their electorate is not split (compared to economic issues).
Expectation 2c: As opposition party with its core electorate split on sociocultural issues, the Left strategically adapts best to the voters' demands by a concentration on socio-economic positional issues.
Finally, we test the issue-yield assumption, which posits that parties' behaviour is less driven by systemic salience than by issue yield, that is, an individual risk-opportunity calculation for each single issue (see D'Alimonte et al. 2019). In other words, we expect that parties focus on those issues which promise to generate most support for them (high issue yield) rather than on those which are highest on the overall public agenda (high systemic salience). The concept of issue yield is not only strongly relying on a rationalist (or electoralist) perspective on parties. In looking at citizens' preferences in positional as well as valence terms, it also assumes that voters pick parties or electoral candidates with the goal of utility maximisation. Hence, the concept should be ideally suited for and work better in contexts lacking any kind of traditional party alignments or, at least, in contexts showing strong dealignment. However, Germany is a case where to a certain extent the 'old' linkages are still alive. Despite the declining electoral support of the two major parties, the CDU/CSU and the SPD, these remain the strongest parties at the federal level. If the issue-yield concept also works in Germany, this would have to be understood as strong support for the more dynamic issue-yield view on party competition.

Research design, measurement, and methodological approach
The subsequent analysis uses the issue-yield model, which calculates the specific gain that a party can expect from emphasising a certain issue in an electoral campaign. As all country studies in this special issue are based on the issue-yield concept as theoretical framework, this section only highlights peculiarities of the German study. More general information, especially regarding the set-up of the population surveys, the approach to the Twitter data, and the calculation of the core indicators like the issue-yield score can be found in D'Alimonte et al. (2019) and De Sio and Weber (2019).
This study covers all major German parties that won seats in the 2017 election. In order of party size, these are CDU/CSU, 4 SPD, AfD, FDP, the Left, and the Greens, representing roughly 95% of the votes cast. In order to address our three research questions, we need a comprehensive issue battery covering all relevant aspects of political competition in Germany. Due to the requirement of comparability for cross-national issue-yield analysis, we included European Union membership, job market reform, immigration, and referenda. The challenge was to identify all additional issues that could become relevant during the electoral campaign in Germany. First, we analysed the regular surveys of the Research Group Elections (Forschungsgruppe Wahlen: Politbarometer 2018), which provide information on the most relevant issues in the German public debate. As a second step, we investigated the electoral programmes of all major parties in order to assess the particular importance of these issues for the 2017 electoral campaign. We decided to remove only education since it is a domain of German L€ ander (states) and therefore not of major interest at the federal level. While defence spending and the state of NATO are inherently relevant, all German parties avoided discussing these issues during the electoral campaign. It was also not an issue of major concern to the German public. Altogether, we identified 17 salient positional issues that can be assigned to either of the two dimensions of political competition (economic vs. cultural); for all of them a progressive and a conservative stance can be formulated. 5 Furthermore, we selected 10 valence issues: affordable homes, fighting poverty of elderly people, infrastructure, fighting against terrorism, unemployment, environmental protection, supporting families and children, fighting against crime, social justice, and economic growth. Theoretically, these issues could also be framed as positional issues. However, as an inspection of the electoral manifestos shows, they represent common goals for all partiesa situation which is reflected in public opinion data. In the online appendix we give an overview of all 27 issues and their allocation to the socio-economic and socio-cultural conflict dimension if applicable (see Tables A3 and A4).
Clearly, the validity of our findings on current German party competition depends on the comprehensiveness of our list of positional and valence issues. We were able to rely on the most comprehensive issue battery regarding German elections ever administered. It was composed on the basis of theoretical reasoning and the results from previous opinion polls. While it may be surprising that important issues such as digitalisation and items regarding foreign policy are missing, it needs to be emphasised that these policy areas played no role during the electoral campaign. 6 Hence, we are confident that our 27 issues cover all relevant issue aspects and allow meaningful conclusions.
Answering our three research questions requires two different data sources in order to calculate all relevant measures for the analysis. First, a online survey (N ¼ 1001) was administered by Demetra opinioni.net Italy between 30 August and 13 September 2017, using predetermined quotas for age/sex combinations, level of education, and geographical region. 7 This data was used to calculate issue yield values for all 27 issues. In contrast to earlier approaches to issue yield (De Sio and Weber 2014), the modified approach allows for the comparison of positional as well as valence issues by taking into account parties' credibility. The revised measure of issue yield is presented in detail by De Sio and Weber (2019). The most important consequence for our analysis here is that we can rely on scores measuring how promising campaigning on certain issues actually is for different parties regarding positional and valence issues. 8 For the positional issues, we have used a scale ranging from 1 to 6 throughout. 9 As is shown in the online appendix, this information is then combined with information on whether a respondent thinks a respective party is credible dealing with the issue. Issue-yield scores for valence issues are also based on twofold information. First, we asked which parties are competent to deal with the issue and, secondly, we asked how important the issue is to the respondent. In addition, the data was used to calculate the systemic salience for all issues. 10 As a second data source, we looked at party communication in order to identify their campaign strategies. This allows us to answer our third research question: namely, whether a party's campaign strategy is mainly driven by systemic salience or by issue yield. To this end we conducted extensive analysis of political competitors' Twitter accounts. This allowed us to generate measures of parties' issue emphasis. Also, this measure takes account of the changing nature of campaigning. We selected both the official party accounts and the frontrunners' Twitter accounts or a relevant substitute to measure these strategies for all six parties. 11 The coding scheme mirrors our 27 pre-defined policy issues. Altogether, we collected 3942 tweets published during the month before the election of 24 September 2017. It is important to note that the coding scheme is based on issues and not on specific issue positions. 12 While this is irrelevant for valence issues, it means that we assume that parties are rational actors and have information on their ideal position regarding the 17 positional issues. As a result, their communication should be in line with more left/GAL positions or more right/TAN positions depending on their specific issue-yield scores. 13 Regarding Twitter use by parties, the issue-yield approach follows the empirically corroborated 'press release assumption' , namely that parties use Twitter to communicate their desired messages to the media. Thus, the absolute number of followers is not decisive, but parties can use it to address the media.
To answer research question 1 (do parties' profiles still follow certain ideological patterns and traditional political conflicts in terms of issue stances?) we used issue-yield values for all parties and all issues. The same is true for investigating which parties should, according to their issue-yield profiles, focus more on valence or on positional issues in their campaigns (research question 2). However, in addition to the original formula, we applied a centring approach to the scores to answer research question 1. Similar to the approach of Franzmann and Kaiser (2006) for Comparative Manifesto Project (CMP) data (Volkens et al. 2017), this gives us an insight into the relative issue yield. For example, if the vast majority of the electorate support staying in the EU and the majority of all voters of each party do so as well, all parties will have a large issue yield for this issue position. However, campaigning on this issue position would not promise vote gains as the parties will not be able to distinguish themselves from their competitors. Centring is based on subtracting the average issue yield score of all left/GAL (or right/ TAN) positions regarding a specific issue from a given party's score. 14 As a consequence, positive values represent an issue-yield score higher than the average score of all parties competing with the same issue position while negative values indicate the opposite.
Finally, our third research question looks at the factors driving parties' campaign behaviour. We use the proportion of Twitter posts regarding a specific issue as the dependent variable measuring party behaviour. We collected the Tweets within the two months before the election date. It is safe to assume that the party elite had at least rough information based on surveys about the electorate's preferences. The two explanatory variables are issue yield and systemic salience. Depending on which independent variable is better able to explain party behaviour we can conclude that such behaviour is primarily driven by a party-specific strategy (issue yield) or systemic salience. To control for the clustered data structure, we also include party dummies and, as the dependent variable's value range is bound between 0 and 1, we estimate the model in a Tobit set-up. Our findings based on these models are robust for various model specifications. 15

Ideology underlying parties' positional issue profiles
We start our analysis by inspecting the issue-yield profiles of German parties regarding positional issues. How far do parties' issue profiles still reflect the traditional political conflicts of the twentieth century? To this end, we calculate issue-yield scores which indicate the vote gain relative to the magnitude of potential vote losses a party can expect from emphasising a certain issue. Regarding economic issues, the ideologically left camp still benefits from its left-wing stances (see Table A5 in the online appendix). Both the SPD and the Left reveal high issue yield for more or less all left-wing positions on economic issues. Interestingly, the CDU/ CSU also reveals substantive yields for left-wing economic positions but still slightly less than the Greens. Only the FDP and AfD benefit from right-wing stances on economic issues; however, for the AfD, this applies only to the issues of 'using budgetary surplus to reduce tax levels', 'decreasing job market regulations', and 'not paying to save the euro'. The FDP's positional profile clearly points to focusing on right-wing stances but, at the same time, the respective issue-yield scores only range between 0.06 and 0.13. In contrast, the AfD has a much higher yield (0.26) for opposing any additional payment to save the euro. Nevertheless, positional stances on economic issues no longer provide a promising strategy for very polarised electoral competition. With the exception of the three issues mentioned, it only sets the FDP against all other relevant parties. This shows that the traditional economic left-right conflict does not discriminate very much between parties.
Inspecting the centred issue-yield scores provides important additional information (Table 1). As described above, these scores inform us about a party's issue yield relative to the issue yield of all parties putting forward the same stance on the issue. This follows from the fact that a party's electoral result is always relative to the results of its competitors. If two parties present the same position during an electoral campaign but one party has a much lower issue yield regarding this issue than the other party, the latter party will win many more votes relative to the risk of losing votes. Knowing that parties operate with limited resources, it would be much better for the party with the lower issue yield to focus on a different issue in which it actually has a higher issue yield than the other party. Table 1 shows that four parties have a left-wing stance regarding budgetary surplus. In contrast, the FDP and the AfD see budgetary surplus as a possibility to reduce taxesa right-wing positionindicated by the grey boxes. For both groups of parties positive values refer to a higher than average issue yield in comparison to other parties with the same position. If we look to the budgetary-surplus issue, we see that the SPD has a score of 0.11, which is the highest score of all parties with a left-wing stance. In other words, the SPD would benefit the most from campaigning on this issue. Clearly, there is a sharp distinction between the left-wing party camp of the SPD, the Left, and the Greens, and the CDU/CSU and AfD. While the latter two parties have more or less no incentive to campaign on economic issuesindicated by the majority of negative valuesthe former group of parties should have benefited from emphasising left-wing economic positions. However, in the left-wing camp, the Greens show not only positive values. This is particularly pronounced for the reduction of income differences and for keeping job-market regulations. Given that the Green electorate tends to be characterised by a very high socio-economic status, this is not surprising. In relative terms, the Green's high issue-yield scores all belong to the socio-cultural dimension, especially same-sex marriage and all issues related to environmental politics (see Table 2). 16 Interestingly, the Greens are not the only party benefiting from a GAL profile. The SPD and the Left also benefit more from GAL positions than from TAN positions. However, while the SPD has at least some large and positive scores, for example, keeping gay marriage legal or staying a member of the EU, the Left would not benefit much in relative terms from concentrating an electoral campaign on these issues. In contrast, the profile of the AfD is strongly dominated by cultural issues favouring TAN stances. Everything related to conservative positions on migration and refugees reveals a relatively high yield. This supports our argument that the AfD and the Greens represent the poles of the new and increasingly dominant sociocultural dimension. It also shows that the SPD and the Left have more or less no overlap when looking at this dimension. The CDU/CSU and FDP show a mixed profile with slight majorities for GAL positions. None of the two parties would benefit much from emphasising TAN positions. While the FDP has no positive score on cultural issues, the CDU/CSU shows very high scores for GAL positions related to the EU. In general, Table 2 supports our assumption that the socio-cultural dimension is now more important for electoral competition than the economic dimension. Here, we see larger differencesin terms of public opinion, party positions, and direction (GAL vs. TAN). Interestingly, with the exception of the AfD, we find that all parties on the right benefit from emphasising some GAL and some TAN positions. This might indicate an ongoing process of reshuffling in the conservative camp.
Overall, German politics are rather dominated by progressive GAL and economically left-wing issue positions. However, the FDP in economic and the AfD in socio-cultural terms clearly deviate from this general trend. Moreover, there is a clear divide in terms of political camps: right/ TAN positions are completely irrelevant for centre-left parties (the SPD, the Left, and the Greens), while they are of some importance for parties on the right. Figure 2 summarises these results within the two-dimensional space of competition. 17 We see slightly more polarisation regarding cultural issues as well as no extreme position on the right-hand side of the economic conflict dimensions. The CDU/CSU appears on both dimensions as slightly progressive, reflecting that Merkel's course of action finds support within the electorate and is, or, at least, was, electorally promising. We need to keep in mind, however, that the baseline of our estimates of potential electoral gains is the current core electorate. Some would argue that this might have been larger with a more conservative course over the past years. Compared to the two-dimensional supply-side space that we have illustrated in Figure 1 (based on manifesto data), the Greens' electoral potential is economically more to the right than their official party manifesto. Figure 2 suggests that a potential 'Jamaica coalition' of CDU/ CSU, Greens, and the FDP should have been able to find an agreement on economic issues because they did not constitute an electorally promising battlegroundunlike socio-cultural issues. Contrary to its overall public presentation, the FDP reveals electoral potential with only slightly right-wing economic stances while it is largely indifferent to cultural issues. Due to this unique profile, the 'Jamaica coalition' would probably have been costly for either the FDP or the CDU/CSU and the Greensdepending on character of the coalition agreement on economic policies. The SPD and the Left still represent one pole of the traditional economic cleavage. Both parties rely on similar, but not the same economic issues. While the Left is holding its strongest yield in classical labour-market issues, the SPD has its stronghold in welfare state issues. Nevertheless, a missing opponent makes it difficult to actually campaign on these issues.
There is a final and very important difference between the supply and the demand sides regarding the relationship of the two dimensions. While there is a diagonal relationship present in terms of the supply side (Figure 1), party positions based on issue yield lead to a spatial representation of German parties which leaves much of the economically right-wing space empty. This is in line with Kitschelt's (2003: 130) expectation of a horizontal dimension shifting into a vertical one. Remarkably, regarding the electorate, the Left is the most progressive party on the economic dimension. The Greens and the SPD reveal more progressive positions in terms of cultural concerns. Finally, the AfD clearly is a unique actor in its positioning in the two-dimensional space even more when looking at issue-yield-based positions.
Valence or positional issues: which promise more benefits for which of parties?
In our second analysis, we ask whether parties show higher issue yield for positional or valence issues and whether there are any patterns confirming our expectations 2a to 2c. In general, parties' individual profiles change somewhat when we take valence issues into account. Valence issues dominate across the entire electorate. This is reflected by the issue-yield pattern of the governing parties. With the exception of the EU membership issue, valence issues reveal the highest yield (Figure 3). This is in particular true for the CDU/CSU and the FDP. The SPD presents a similar picture but positional issues are important as well; expectation 2a is only partly confirmed. We also see substantive differences between the CDU/CSU and SPD in terms of which valence issues they could benefit from. The CDU/CSU shows higher yield in fighting crime and terrorism and supporting economic growth while the SPD potentially does well regarding welfare state issues. All three parties share EU membership as the issue with the highest yield, signalling that an overwhelming majority of the German electorate still supports EU membership.
The Left and especially the AfD have more to gain from positional issues as they are not (yet) considered as potential coalition partners on the federal level by their competitors and have, therefore, never been in power. Likewise, with the exception of environmental policies, the electorate of the Greens also primarily concentrates on positional issues. In accordance with our expectations formulated above, important positional issues belong to the socio-cultural dimension for both the AfD and the Greens. Also in accordance with our expectation, valence issues are indeed important for the FDP. However, while we see that economic positional issues would be beneficial for the Left, we do not see a similar pattern for the FDP.

Issue yield vs. systemic issue salience: what determines party strategies?
Finally, we analyse party strategy based on several Tobit regressions as described above. These models allow us to investigate whether the parties' campaign communicationunderstood as a measure of the individual party's campaign strategyis influenced (more) by party-specific issue yield or general systemic salience of an issue. As a reminder, our dependent variable is the share of tweets a party sends out on a particular issue. Hence, we are left with 27 issues for six parties resulting in 162 observations. Models 1 and 2 presented in Table 3 Figure 3. Issue-yield scores by issue type and party.
Note: The figure presents the 15 issues with the highest issue yield for each party.
effect of issue yieldrepresenting a party's issue-specific strategic decisionand systemic salience, which refers to the proportion of respondents indicating that the specific issue is of high importance to them. Both variables have been standardised following Gelman's (2008) approach, which means that we can compare the coefficients directly. Clearly, issue yield outperforms systemic salience. In general, parties use Twitter communication strategically as more issue yield is associated with more tweets regarding that issue. The explained variance is more than three times higher for issue yield than for systemic salience. 18 However, Model 3 shows that estimating a joint model provides the highest proportion of explained variance. Issue yield and systemic issue salience independently contribute to parties' communication strategiesalbeit with issue yield as the dominant factor. We expand Model 3 in two different ways to shed even more light on the role of issue yield and systemic salience for party strategies. In a fourth model, we interact party dummies with both explanatory variables to test for party-specific differences. In a final model, we add an indicator distinguishing between positional and valence issues and we interact this new variable with issue yield and systemic salience. Result tables are presented in the online appendix (Table A7); here we present marginal effect plots to visualise the results for the interactions (Figure 4). When we integrate interaction effects of individual parties with their issue yield and the systemic salience, the latter's main effect becomes insignificant. Inspecting the marginal effects for each party, we find that especially the AfD but also the Left and the Greens rely primarily on an issue-yield strategy. For the mainstream parties CDU/CSU and SPD we find no significant difference. However, and in contrast to all other parties, the FDP is clearly driven by systemic salience. This might be the result of only low importance of the economic dimension, particularly the right-wing economic positions, and the FDP's status as a former and long-term governing party (Franzmann 2019).
Finally, we analyse the marginal effects of the two explanatory variables depending on the issue type. As the lower panel in Figure 4 shows, issue yield is much more important than systemic salience for valence issues. We assume that this pattern is driven by issue ownership. Regarding valence issues, it might not be sufficient to focus on issues popular in and  Table A4. relevant for the electorate. A party needs some credibility in dealing with the respective valence issue. As credibility is represented in the conceptualisation of the modified issue-yield approach (De Sio and Weber 2019), we should indeed expect a stronger effect on valence issue and a different pattern for positional issues. Looking at positional issues, we find significant effects for both explanatory variables. Overall, issue yield is very important for parties' campaign behaviourand more important than systemic salience. However, our analysis shows that this is not universally the case for all parties. At the same time, distinguishing issue types supports the high relevance of issue yield, especially for valence issues.

Discussion and conclusion
Of course, our analysis is not without limitations. First, we need to rely on a single survey for a single election. Second, the issue battery had to be developed before the peak of the electoral campaign and hence may not have fitted perfectly the state of public debate close to election day. Both problems could be overcome by collecting more data, ideally panel data, including either a more extended issue battery or allowing for a dynamic adaption of the questionnaire during the fieldwork period. Third, since we focus on the last weeks of the electoral campaign, we are not able to analyse party strategies and change between elections. Linked to that: which role do events play for the issue-yield approach and party competition in general? The literature suggests that a corresponding cost-benefit logic might also be at work between elections (see Meyer and Schoen 2017). A pathway for future research might be to collect and analyse data at different points during a legislative term in order provide a full picture of how parties are able to adapt strategically to changing voter environments.
These caveats notwithstanding, our analysis of the current German party competition has indicated that primarily socio-cultural issuesfirst and foremost linked to migrationcontributed to the polarisation within the German electorate. These issues also shape the central conflicts on the supply side of politics, much more than other socio-cultural and economic issues. Even the issue of European Union membership, despite opposition by the AfD and to a certain extent by the Left, was not really contested since the German electorate overwhelmingly supports membership. Clearly, ideology has lost much of its structuring power for the economic dimension. At the same time, the analysis of the cultural dimension indicates either a transition period or a new polarisation caused by the antagonism between the GAL Greens and the TAN AfD. However, elections are not only won and lost due to positional issues. A clear pattern emerges regarding issue types: while the main governing parties CDU/CSU and SPD focus on valence issues, opposition parties can gain more from mobilising on positional issues. AfD, Greens, and the Left act strategically in the sense that these parties tend to exploit their issue yield by concentrating on these issues. The FDP is an exception because it concentrates on valence issues and, at the same time, reacts primarily to systemic salience.
Furthermore, our analysis using issue yield has revealed interesting tensions between the supply and the demand sides of party politics. Importantly, German party competition is asymmetric in three ways. First, there is an asymmetry between the demand and supply sides of competition, since parties' programmatic offers are less concentrated on socio-cultural issues than the demands of the electorate. The centrifugal direction of party competition, as described by Sartori (1976) for a centrist 'Grand Coalition', only manifests itself on the socio-cultural dimension. Second, we found an asymmetry between the governing and the smaller opposition parties as only the latter tend to concentrate on positional issues. The CDU/CSU and SPD show higher issue-yield scores for valence issues, which makes it more reasonable for them to campaign and compete on them. Third, there is an asymmetry within the electoral competition space or, more precisely, regarding the economic dimension. The electorates of the Left and SPD still emphasise economic issues without having a counterpart that effectively opposes left-wing positions with stances linked to free-market ideology. While the FDP takes such positions in its manifesto, this does not lead to a correspondingly high issue yield on the economic dimension as the majority of the German electorate favours more left-wing policies.
Finally, one could argue that German politics has always been twodimensional, with socio-cultural issue being even more important than socio-economic issues (Pappi 1984). However, this 'old' two-dimensionality was mainly driven by issues of religiosity and church-state relations. The transformation of the political space causes a new situation in two regards. First, salient issues on the cultural dimension have changed. Second, political competition does not simply reinforce the traditional pattern, but especially cuts through the electorates of the mainstream parties. Hence, while all parties have to adapt to this new situationwhich allowed the AfD to win a large number of votes only four years after its foundationthe CDU/CSU and SPD find it very difficult to campaign on the socio-cultural dimension as their own electorate is quite divided when it comes to issues like migration.
Despite the serious attempt to form a so-called 'Jamaica coalition' consisting of the CDU/CSU, FDP, and Greens after the 2017 election, our findings suggest that government formation processes in Germany will become increasingly more complicated. The Green Party and its voters are not far away from CDU/CSU and FDP on economic issues. However, this is no longer the decisive dimension of competition in German politics. It is now the socio-cultural dimension and especially migration issues. The Greens, with very liberal stances, contribute to the polarisation over these topics, making a potential government compromise with CDU/CSU and FDP difficult. At the same time, with its economically bourgeois constituency, the Greens are increasingly less a natural coalition partner for Social Democrats or even the post-communist Left. Irrespective of its current status as a fundamental opposition party, the TAN-populist AfD might become a future member of government coalitions together with the FDP and CDU/CSU. We have already seen similar arrangements of bourgeois mainstream parties coalescing with right-wing populists in the Netherlands and Scandinavia. The only alternative is a Grand Coalition. That is what we have observed in Germany since 2005 on a regular basis. Most likely, the two main moderate parties in government will continue to neglect the rising relevance of socio-cultural issues for strategic reasons while opposition parties will continue to occupy polar opposites on exactly these issues. As a result, we can expect further party system fragmentation. The rational behaviour of the two Grand Coalition parties leads to a situation in which nothing else is possible but forming another Grand Coalition. Given that the post-2017 Grand Coalition controlled only 53.3% of the popular vote, this may turn out to be a dead end as a situation in which even the former 'grand coalition formula' will fail to gain a majority becomes ever more likely.

Notes
1. The only other Grand Coalition in (West) German post-war history was in office from 1966 to 1969. 2. In addition to all manifesto coding categories belonging to the economic dimension, the following categories are also included for calculating the socio-economic party positions: per303 (efficiency), per504 and per505 (pro and contra welfare state expanses), per701, per702, per703, per704 (pro and contra trade unions, agriculture, middle class and professional groups). All other categories are regarded as contributing to the sociocultural dimension. 3. The scores based on manifesto data (Volkens et al. 2017)  5. Figure A2 in the online appendix presents the distribution of positions regarding all 17 positional issues. Albeit to differing degrees, there is polarisation in public opinion. 6. Less than 3% of German citizens consider them as one of the decisive problems or issues in Germany (Forschungsgruppe Wahlen: Politbarometer 2018) and neither parties nor media nor public opinion put these topics on the agenda during the election campaign. 7. Albeit there have been tremendous improvements, quota samples based on online access panels still face the danger of being less representative than data collected with other modes and methods. Fortunately, this is not the case with our data. In the online appendix, we present some comparison based on the face-to-face pre-election survey conducted by the German Longitudinal Elections Study GLES (Roßteutscher et al. 2018) in terms of party vote recall ( Table A1 in the online appendix). In fact, the data used in this analysis is only slightly less representative than the GLES data. Taking into account that the GLES survey is based on more than double the number of respondents and constitutes the most high-quality data available, this makes us very confident regarding the validity of our findings. Finally, all calculations of aggregated values using our survey data make use of a post-stratification weight combining socio-demographic characteristics (sex, age, geographical area, education) and past vote recall. The survey and Twitter data were collected in the framework of the ICCP project (https:// cise.luiss.it/iccp/) and are available as GESIS study ZA7499, see . 8. Following the logic of issue yield, for positional issues, all analyses are restricted to the party-specific stance with a higher issue yield. For example, the CDU/CSU has very high issue yield advertising for Germany to stay in the EU. Consequently, the opposite positionleaving the EUhas a very low issue yield and is no longer considered in the analyses. However, this selection is party-specific and, hence, a left-wing position regarding a socioeconomic issue might be included due to its high yield for one party while another party's right-wing stance on the same issue also shows a high yield and is therefore included as well. 9. In a few cases, we have decided to formulate rather extreme positions even though this did not reflect the main thrust of the public debate (e.g. an entire ban on diesel engine cars or leaving the EU). This was done in order to capture all potential positions. Moreover, as Figure A2 in the online appendix shows, we find substantive support for even extreme positions in the German public for more or less all issues. 10. Systemic salience is defined as the percentage of respondents stating in the mass survey that a certain issue is of 'high' importance, with respondents reporting a 'medium' importance counted as half. 11. In general, we rely on the official party accounts plus those of the particular frontrunner. CDU and CSU frontrunners had no individual Twitter accounts. Therefore, the personal speaker of Angela Merkel, Steffen Seibert, and the CSU general secretary's account of Andreas Scheuer were chosen as they are fulfilling the role of communication for their respective frontrunner on Twitter. Merkel's spokesperson Seibert is not a party official. However, since he is the one who puts Merkel's issues on the agenda, we decided to include his official speaker account. The missing account of Alexander Gauland was replaced with the AfD's party account since he was the decisive person in the party headquarters. 12. On the Twitter data, we calculated a very high score of intercoder reliability (Cohen's kappa ¼ 0.91). 13. Consequently, testing the effect of issue yield on Twitter issue salience is conservative as the correlation isif at allbiased towards 0 because we cannot distinguish between left/GAL positions and right/TAN positions for each issue. 14. If there is only one party with a left/GAL or right/TAN stance, we kept the original score. 15. We present several robustness checks in the online appendix. For example, the interpretation of our findings does not change if we estimate the model with cluster-corrected standard errors (clustered by issue; see Table A2). There is also no substantive difference when running 27 additional modelseach leaving out one of the issues. The latter test would detect strong effects of a single issue which is, as Figure A1 shows, not a problem for our study. 16. The original values can be found in Table A6 in the online appendix. 17. We standardised all issue-yield scores between 0 and 1 for each party and dimension (economic vs. cultural) based on their empirical minimum and maximum. In a second step, the resulting values were multiplied with -1 in case of economic left-wing and socio-cultural conservative positions. We then calculated the average value for each party for both dimensions. 18. Calculating values of variance explained for Tobit regressions is far from straightforward. Like the other contributions to this special issue, we rely on a computational approach developed by De Sio and Weber (2019).