Electoral strategies in multilevel systems: the effect of national politics on regional elections

ABSTRACT Elections are at the heart of representative democracies. Whereas the study of national elections is a prominent field in comparative politics, we still know little about subnational political actors’ behaviour. We seek to close this gap by applying a dictionary coding approach to analyse parties’ issue-based content of 743 subnational manifestos in Austria and Germany. We show that subnational parties emphasize regional topics less if regional elections happen close to national elections but focus more on both regional and mixed topics if their national party organization is in government. This has important implications for electoral competition in multilevel systems.


INTRODUCTION
Parties strive to gain as many votes as possible in elections to get in a strong bargaining position once the process of forming a government occurs . Besides offering policy positions that are identical or close to voters' policy preferences, parties highlight specific issues in their election campaigns which they believe voters care about. Which specific issues parties (de-)emphasize in their election manifestos has been investigated in various ways: (1) parties emphasize issues they 'own' and deemphasize issues where voters consider rival parties as being more competent to handle the job (Hobolt & De Vries, 2015;Klüver & Sagarzazu, 2016;Petrocik, 1996;Spoon & Klüver, 2014); (2) parties de-emphasize issues when their supporters are divided on these issues ; or (3) particularly challenger parties highlight issues that drive a 'wedge' between dominant parties and coalitions (e.g., De Vries & Hobolt, 2020;Wagner & Meyer, 2014).
Due to the methodological nationalism  that still prevails in the literature on party competition, however, a multilevel perspective on parties' decisions to (de-)emphasize issues has largely been neglected. We know little about how partiesto be specific, subnational partiesfactor in the formal division of power in the institutional setting when drafting their manifestos. From a legal viewpoint, policy jurisdiction sets the framework of subnational party competition and, thus, limits the number of meaningful campaign issues. From a strategic viewpoint, however, parties might have incentives to emphasize issues irrespective the policy jurisdiction. A crucial condition for voters to reward or punish (government) parties is that voters can hold politicians accountable by knowing which actor has been responsible for which political decision. Hobolt et al. (2013), for instance, demonstrate that it is 'government clarity' (i.e., the fact that voters can easily identify the incumbent government) rather than 'institutional clarity' (i.e., the horizontal and vertical power dispersion within a multilevel system) that matters for voters' performance voting. Yet, even in a more or less ideal situation of clear governmental accountability in a federalized system (Canada) during a period where one issue was highly salient (the beginning of the COVID-19 pandemic in 2020), receiving a lot of media attention, and when almost all political actors between and within parties agreed on the ultimate goal, even then were voters not able to correctly assign responsibilities to and hold accountable various levels of government (Kennedy et al., 2022). These findings might be lamentable for democratic theorists; however, they give political actors strategic leeway in drafting election manifestos and in designing electoral campaigns, especially when campaigning in 'marble cake'-like, federal multilevel systems (Kropp & Behnke, 2016) with intertwined horizontal and vertical power dispersion (on power-sharing systems, see Loughlin, 2013;Watts, 2013). If voters are not capable of clearly identifying which political level is legally responsible for which policy issue, then why not highlight particularly those issues in election manifestos that parties think they can benefit from on election day, even if the respective political layer the party is campaigning on is not legally in charge of these issues? Furthermore, and arguing from a party-organizational perspective, if subnational units possess considerable competencies regarding policy-making, taxation and financial matters, then they should have more incentives 'to prioritize their own goals and identity' (Thorlakson, 2020, p. 74) and, thus, emphasize beneficial topics to a greater extent. Consequently, we are interested in answering the following research question: To what extent and under which conditions does national politics affect subnational parties' regional issue emphasis in their manifestos?
We expect that the temporal proximity between subnational and national elections as well as a party's government participation at the national level will affect a subnational party's emphasis of subnational issues. We differentiate between a set of regional topics that fall exclusively under the jurisdiction of the states, a set of regional and mixed topics where states and country share responsibilities, and a set of national topics under national control. 1 Our empirical test is based on the cases of Austria and Germany: two prime examples of joint federalism settings, where financial matters are largely centralized at the federal level and where subnational regions possess low fiscal and legislative autonomy but display strong administrative interdependence with the federal level (e.g., Thorlakson, 2020).
Using a novel data set on subnational parties' relative issue emphasis in 743 regional election manifestos in Austria and Germany between 1990 and 2019, we show that German parties considerably talk more about regional topics in their election manifestos than Austrian parties. This is consistent with the fact that German regions have more political power in the multilevel political system than Austrian regions (Hooghe et al., 2010). The multivariate analyses highlight that subnational parties put less emphasis on solely regional topics if the regional election is held close to a national one. Furthermore, they put more emphasis on regional as well as mixed topics in their election manifestos when their national party organization is in government.
These results add to the increasing literature on the nationalization, regionalization and second-orderness of subnational elections (e.g., Schakel, 2021;, as well as to the literature on subnational parties' electoral strategies (Cabeza, 2018;Cabeza et al., 2017;Müller, 2013;Stecker, 2015). Furthermore, our theoretical arguments and empirical findings speak to the literature on integrated and independent politics in multilevel systems (most prominently, see Thorlakson, 2020) by highlighting how national-level factors and the institutional setting in countries with joint federalism are impacting not only subnational parties' position-taking but also their salience strategies when drafting their election manifestos.

A SUPPLY-SIDE PERSPECTIVE: HYPOTHESES ON SUBNATIONAL PARTIES' ISSUE EMPHASIS
Parties behave strategically when formulating their election manifestos (e.g., Adams & Somer-Topcu, 2009;Somer-Topcu & Zar, 2014;van de Voorde et al., 2014). They adjust their policy positions to successful policy positions of rival parties, they emphasize issues they 'own', they highlight issues that are currently the concern of voters, and they de-emphasize those issues which they think would only benefit their opponents.
One might argue that subnational issue emphasis is the best strategic choice in subnational campaigns due to insights from the pledge fulfilment literature. In a large comparative study, Matthieß (2020) has convincingly shown that low levels of pledge fulfilment lead to electoral losses and, clearly, pledges that fall outside one's own jurisdiction are much harder to fulfil (Praprotnik, 2017). Yet, issue emphasis and issue clarity are two distinct concepts and parties may easily choose to highlight a certain issue without making specific pledges and hence without risking being punished for a lack of pledge fulfilment. This is in line with the results from the literature on position blurring showing that parties strategically blur issues, especially those of less priority (Bräuninger & Giger, 2018;Han, 2020;Somer-Topcu, 2015).
Looking at the scarce evidence of the literature on party competition at the subnational level, we find evidence that backs these claims. Cabeza et al. (2017) demonstrate for regional election manifestos in Spain that parties running for elections at both the national and regional levels strongly emphasize national-level issues in regional campaigns to gain regional voters' electoral support. Additionally, Gross and Jankowski (2020) show for local election manifestos in Germany that some parties even emphasize issues related to national and international politics, respectively. Even for coalition agreements, Gross and Krauss (2021) have shown that parties at the local, state and federal levels in Germany do include those topics that they are legally responsible for, but political actors also emphasize issues legally belonging to other political layers. Consequently, these studiesas well as the literature dealing with spillover effects between elections at different political levels in multilevel systems (e.g., Bechtel, 2012;Golder et al., 2017;Schakel, 2018Schakel, , 2021Schakel & Romanova, 2021)indicate that 'electoral arenas cannot be studied in isolation' (Schakel, 2021, p. 644).
Our theoretical expectation, therefore, is that parties competing at the subnational level emphasize issues in their election manifestos that lie exclusively in the realm of subnational political decision-making. They also emphasize issues where they share legislative powers with national political actors or have no power at all. We Electoral strategies in multilevel systems: the effect of national politics on regional elections 845 base our theoretical line of argumentation on the following assumptions. Subnational parties are rational actors that pursue policy-, office-and vote-seeking goals. The priorities attached to these goals vary between parties, but at the time of writing an election manifesto and during the election campaign, vote-maximizing efforts most certainly will prevail . The reason for this is that a vote-seeking strategy in the run-up to an election will facilitate the fulfilment of both policy and office goals in the subsequent legislative period. It is thus instrumental for achieving parties' intrinsic aims in the future. Furthermore, these rational subnational parties are autonomous actors that are (more or less) free in designing such a vote-seeking campaign, including their electoral manifestos. 2 In so doing, they will consider several subnational factors such as their position in the subnational parliament, their top candidate or their perceived issue ownership. However, even though subnational parties are self-determined, they do not operate in a vacuum, but next to the political landscape at the national level. Since studies have shown that a substantial share of voters takes national considerations into account when casting a subnational vote (Cabeza, 2018;Schakel, 2015), subnational parties must come up with an appropriate way of dealing with the upper political level. We identify the national electoral calendar as well as a party's national government or opposition position as the two central factors that drive this decision in an electoral campaign.
The first factor that should have an influence on how much emphasis subnational parties put on regional topics in their manifestos is the national election cycle. Election manifestos are not written isolated from each other but rather with an eye on content, positions and issue emphasis of manifestos at other political levels, especially if elections are sufficiently close in time. The theoretical arguments can be derived from two different perspectives: one that focuses on the interplay between voters and parties, and one that looks at party resources when drafting manifestos. Looking at the interplay between voters and parties, Reif and Schmitt (1980) have famously argued that voters and the media consider national elections as important because they determine the national government. European as well as subnational elections, however, are both considered as being less important and therefore suffer from, for instance, lower voter turnout than firstorder elections (however, see Cabeza et al., 2017;and Cabeza, 2018, for the Spanish case). Furthermore, recent research suggests that when subnational elections are held simultaneously or close in time to national elections, then subnational electoral results are more similar to national electoral results compared with cases where subnational elections take place further away from national elections (Bolgherini et al., 2021; see also Schakel, 2013). We expect that due to the increased media attention for the national election, voters will be more aware of issues that are relevant at the national level. Since national elections are better at attracting media and voter attention, subnational parties might be motivated to include more national topics in their manifestos if they think they can benefit from favourable circumstances in national politics (see also Cabeza et al., 2017). The potential to benefit should be larger, the closer the national election is to the subnational election. 3 From a more practical perspective we note that parties only have limited resources in preparing manifestos and the closer in time two elections at different political levels take place, the larger should be the trade-off that parties are facing when deciding where to allocate the personnel and content-related resources. Parties should face the constraints of allocating resources for the preparation of manifestos when national and subnational elections are close in time. For example, this has already been shown for Sinn Féin in the Irish national elections and the Northern Ireland Assembly elections in 2007 (Däubler, 2012a, pp. 55-56).
Hence our first hypothesis reads: Hypothesis 1: The closer a national election is to a subnational election, the less emphasis will subnational parties put on subnational issues.
The national calendar is expected to have an influence on subnational parties' campaign strategies, since it most likely dominates the political discourse even in times of subnational elections. However, it would be naïve to expect an equal effect for all subnational parties irrespective of the specific political situation at the national level. Therefore, we further specify that a party's national government or opposition status will influence its issue emphasis at the subnational level and moderate the effect of the electoral calendar. For example, the research on the preparation of manifestos shows that government parties write longer manifestos and include more issues than opposition parties, not least because '[g]overnment parties simply have more resources available' (Dolezal et al., 2012, p. 872).
Furthermore, previous research has convincingly shown that there is an advantage for incumbent governments in national elections (for an overview, see Carson et al., 2020). Yet, we argue that this incumbency bonus at the national level should not affect subnational parties' share of emphasis of subnational topics, respectively. Even though the advantage from being in national office usually comes with a heightened presence in the media as well as with having a higher probability to deliver on the promises made during the electoral campaign (see also Shikano & Nyhuis, 2019), we argue that subnational party branches of parties governing at the national level are more likely to be punished by voters for the performance of their national party organization in subnational elections. These mid-term losses have been empirically proven for the German case (Dinkel, 1977;Jeffery & Hough, 2001;Kern & Hainmueller, 2006), and for both Austria and Germanythe two empirical cases under study here -'the identity and interest of the federal and subnational branches of parties are often perceived to be so similar that results of Land elections are routinely interpreted as barometers of support for or protest against the federal 846 Martin Gross et al. governing party' (Thorlakson, 2020, p. 3, passim). Hence, we expect subnational party branches of national government parties to anticipate such midterm losses (e.g., Campbell, 1986;Erikson, 1988;Müller & Louwerse, 2020) and, thus, to put more emphasis on subnational topics to counter these expected losses (see also Cabeza et al., 2017). Therefore, our second hypothesis is as follows: Hypothesis 2: If a party is in government at the national level, the subnational party branch will put more emphasis on subnational issues.

RESEARCH DESIGN AND DATA
To put our hypotheses to a test, we rely on subnational parties' issue emphasis in 743 regional election manifestos from Austrian and German parties between 1990 and 2019. 4 We chose Austria and Germany because both countries are characterized by a rather complex multilevel political system, where both the national and regional levels have exclusive but also shared competencies within a joint federalism setting. In both countries, political competition and legislative decision-making are closely interrelated, and the influence goes both from the state to the federal level and from the federal level to the state level (Bussjäger & Johler, 2021;Däubler & Debus, 2009;Karlhofer & Pallaver, 2013;Stecker, 2016;Thorlakson, 2020). This is also reflected in the close ties between subnational party organizations and the federal party organization (Hepburn & Detterbeck, 2013, p. 88). Particularly since the 1990s, Austrian and German parties are similar in their type of multilevel party organizations: they are characterized by a 'federalist strategy', where subnational party branches gained more independence by positioning themselves differently from the federal party organization and emphasizing different issues than their federal counterparts during election campaigns, and started embracing larger autonomy in state policy-making than in the 1970s and 1980s, while still having close ties to the national party organizations (Detterbeck & Hepburn, 2018, pp. 122-126). Yet, even though the countries share similar multilevel structures in general and the state arena only possesses limited exclusive jurisdiction in both countries (Thorlakson, 2020, pp. 61-66), there are differences in the specific distribution of competencies and the countries vary regarding regions' self-and shared-rule powers (Hooghe et al., 2010). While self-rule power reflects a state government's authority within their own region, shared-rule power denotes a state government's authority in the whole nation. Austria has a self-rule power of 14.0 and a shared-rule power of 9.0, whereas for Germany these are 25.7 and 12.0, respectively (Hooghe et al., 2016;Shair-Rosenfield et al., 2021). Thus, Germany holds greater state powers especially with respect to self-rule. Furthermore, Austria and Germany display some interesting variation regarding party competition and government formation at the substate level. In the period under study , both party competition and government formation in Germany has been much more regionalized (with the Left Party and recently the Alternative for Germany (AfD) gaining electoral strongholds in the eastern part of Germany) and displaying a larger heterogeneity in the formation of different government formation patterns at the state level than in Austria (see also Detterbeck & Hepburn, 2010, pp. 110-111).
Additionally, we are aware of within-country variation regarding the major parties' support for decentralization, with the Christian Democratic Union of Germany (CDU) and Austrian People's Party (ÖVP) being more in favour of higher decentralization than Social Democratic Party of Germany (SPD) and Social Democratic Party of Austria (SPÖ) (Hepburn & Detterbeck, 2013, pp. 77-81), which should lead to an increased attention on regional issues by the two conservative parties in their regional election manifestos (e.g., Libbrecht et al., 2009;Maddens & Libbrecht, 2009). Lastly, there is also considerable variation both within and between larger and smaller parties in Austria and Germany regarding the influence and autonomy of state parties (Thorlakson, 2009).
We choose subnational parties' election manifestos as the decisive documents to measure issue salience in regional election campaigns for at least three reasons (cf. Däubler, 2012a; also see Eder et al., 2017;Harmel, 2018;Harmel et al., 2018;Kavanagh, 1981): first, by summarizing their policy positions and making election promises on important issues, parties provide citizens with a document that might serve as a base for their voting decisions. Second, election manifestos serve intra-party purposes by signalling parties' main campaign positions and issues not only to voters but also to party members and supporters. Third, these documents are guidelines for parties' post-electoral bargaining on government formations. Hence, manifestos can be considered as 'authoritative' and 'representative statements for the whole party' (Klingemann et al., 2006, p. 164). 5 Neither Austrian nor German subnational party organizations face any constraints or scrutiny by national party organizations when drafting their election manifestos. Just as subnational political actors hardly have a say in the preparation of national manifestos, no national party organization is massively intervening in the manifesto preparation process at the subnational level (cf. Däubler, 2012b;Dolezal et al., 2012).
Our unit of observation is one manifesto per party per regional election. We include all relevant parties in our dataset which are defined as those parties that entered the subnational parliament at least once in the period under investigation. We obtained the regional election manifestos from the Political Documents Archive (Benoit et al., 2009;Gross & Debus, 2018) and added more recent regional election manifestos to this source. Since many Austrian regional manifestos were missing, we contacted all regional parties as well as regional and national libraries to complement the data. Overall, our dataset includes 743 manifestos from 28 different parties between 1990 and Electoral strategies in multilevel systems: the effect of national politics on regional elections 2019. A total of 161 of these manifestos are from Austrian parties; 582 were written by parties from Germany (see Tables A1 and A2 in the supplemental data online). 6

Operationalization of the dependent variables
We are interested in determining what influences the share of regional topics mentioned in regional parties' manifestos. To do so, we code these manifestos so that we can determine how much emphasis subnational parties put on those regional topics that they are legally responsible for. While there are several techniques to analyse large corpora of texts (for an overview, see, e.g., Grimmer & Stewart, 2013;Laver, 2014;Wilkerson & Casas, 2017), we have decided to rely on a dictionary coding approach (e.g., Grimmer & Stewart, 2013, pp. 274-275;Laver & Garry, 2000). In contrast to hand-coding manifestos, as done by the Manifesto Project (Budge et al., 2001;Klingemann et al., 2006), dictionary coding is a purely mechanical, quantitative approach where key terms belonging to a specific topic are allocated to the respective categories. In a subsequent step, this dictionary is then applied to the documents and the frequency of occurrence of these key words is recorded (Laver & Garry, 2000). Because we are interested in the share of regional topics mentioned in the party manifestos, we have created a dictionary that distinguishes not only between topics but also between the regional and national levels.
While the actual coding procedure is quite quick, especially compared with hand-coding, setting up the dictionary can be difficult and very time-consuming. One general challenge of creating a dictionary is to find words that fit into the category and grasp the underlying theme, but which also exclusively apply to only one specific category. We have decided to use the dictionary provided by Gross and Krauss (2021) as a starting point since they created a dictionary for political actors' issue emphasis of several prominent policy areas in local, regional and national coalition agreements in Germany.
Contrary to the approach of Gross and Krauss (2021), however, we decided to allocate single words and not categories to the regional and national levels, respectively. This means that our dictionary is not only horizontally divided into 11 categories but also vertically into different levels of competencies. As mentioned above, the multilevel systems in Germany and Austria are interlinked and there are several issue areas where the regional and national levels have shared competencies. Our dictionary, therefore, has three vertical categories: regional, mixed and national. 7 Table A3 in the supplemental data online presents a snapshot of the three categories 'regional', 'mixed' and 'national' in the data set and provides some examples for key terms associated with these categories. 8 Another difficulty while creating the dictionary came with including two countries, Austria and Germany. While they share a similar multilevel set-up regarding their state structure, the single competencies sometimes vary. To account for these differences, we created two versions of the dictionary: one for Germany and one for Austria. Additionally, this was also necessary because some words are solely used in one country but not in the other (e.g., Krabbelstube versus Hort for describing a toddler group). These two dictionaries account for language differences but also for different distribution of competencies. Therefore, this does not distort our results since we do not systematically add or leave out words in one of the dictionaries. 9 Lastly, creating the dictionaries was not only an a priori approach but also an iterative process. We started off with an initial dictionary, coded the texts, and then looked for potential problems and/or errors. We continued this process until we were confident that we have two dictionaries that can pick up on the differences between the regional and national levels as well as on the differences between Austria and Germany. To ensure as much intercoder reliability as possible, all three authors assigned words to policy areas separately and discussed each case of different coding until a unanimous solution had been reached. Although there is no 'gold standard' of validating a dictionary, our approach circumvents some of the inherent problems of dictionaries regarding intercoder reliability (Grimmer & Stewart, 2013, p. 275): since we are not interested in the tone or the emotion of a statement or a word used within an election manifesto, we simplify the classification problem by relying on a binary codinga word can only exclusively belong to the national, regional or mixed category, or it will not be part of the dictionary at all.
Finally, we ended up with two dependent variables: one only includes the share of regional topics in a party manifesto (Share regional topics), the other includes both regional and mixed topics (Share regional and mixed topics). Theoretically, the dependent variable could range from 0 (no mention of regional or regional and mixed topics) to 1 (the manifesto only includes regional or regional and mixed topics). In our sample, however, Share regional topics ranges from 0 to 0.26 with a mean of 0.11 and a standard deviation (SD) of 0.04 (see Table A4 in the supplemental data online). Share regional and mixed topics ranges from 0.06 to 0.70 (mean ¼ 0.44; SD ¼ 0.07). These descriptive results demonstrate that Austrian and German parties do not primarily focus on subnational issues when drafting regional election manifestos; rather, they heavily focus on topics that exclusively belong to national politics.

Operationalization of the independent and control variables
To test our hypotheses on relative issue emphasis in regional manifestos, we coded the following independent variables. First, examining the effect of the electoral calendar (Hypothesis 1) requires the dates of the national and regional elections. Election dates can be drawn easily from online sources such as official government websites or Wikipedia. Based on this information, we calculated the number of days between the regional and national elections. Our argument is valid for both the previous and the next national election. Accordingly, we have chosen the election that is closest to the regional election. Yet, since we do not expect a linear increase in party actors' considerations regarding the time span between the state and federal elections, we calculated the temporal proximity of state to federal elections (Three months to national election) as a dummy variable, taking on the value 1 if the time between the state election to the previous or next federal election is three months or less. We chose this operationalization because the three months before a national election are the heyday of electoral campaigning and media coverage of national issues. Furthermore, this cutoff point also includes the scarce information we have on when election manifestos are officially adopted by party congresses, which varies between three and six months before the elections take place (cf. Däubler, 2012aDäubler, , 2012bDolezal et al., 2012). 10 Second, we coded parties' government status at the national level (Hypothesis 2). This binary variable Party in government (nat.) differentiates between national parties in office and in opposition. Again, these data can be found easily online. Since we are interested in the effect of national government status during the time regional parties draft their manifestos, we registered the partisan composition of the government at the day of the regional election.
Additionally, we also include five control variables: whether the party is in government at the regional level, 11 if it is the state premier's party; the lagged regional election turnout; if the national election was a snap election; and a dummy variable indicating to which country the party belongs. We expect that parties that are in government at the regional level are more likely to talk about regional topics since they were able to influence public policy at the regional level during the previous legislative term. Similarly, we also expect a higher share of regional topics in manifestos of parties that held the state premier's position before the election. Regarding power, influence and prestige, there is no position that is stronger than the one of the state premiers. As such, the party of the state premier should be highly interested in emphasizing regional topics. We control for the level of turnout in the previous regional election to account for the importance of the election. Due to the second-order election effect, turnout for regional elections is usually lower, and hence, parties might be motivated to include more national topics to encourage citizens to take part in the elections. We include a dummy variable that denotes snap elections at the national level to account for the fact that an unforeseen national election might have less effect on regional manifestos than a regularly scheduled one. Regarding the country of origin, we expect that German parties are more likely to talk about regional topics since German regions have considerably more authority over regional policies than Austrian regions (Hooghe et al., 2010).

EMPIRICAL ANALYSIS
In this section we present the analysis of our two hypotheses. We will start with a description of the distribution of our dependent variables. We then discuss our modelling choices before finally testing our hypotheses by relying on ordinary least square (OLS) regressions.
4.1. Descriptive results of subnational parties' emphasis of regional (and mixed) topics Figure 1 presents individual parties' relative emphasis of regional topics in their regional election manifestos. The upper part shows the box plots for the regional topics whereas the lower part shows the share of regional and mixed topics. In the upper part, we see that German parties put more relative emphasis on regional topics than Austrian parties. We find further evidence for this pattern in Figure A1 in the supplemental data online, where we show the distribution of the dependent variables by country. Here, the means for Germany are considerably higher than for Austria. Although the differences are rather marginal between the German parties, the data show that the CDU and the Free Democratic Party (FDP) emphasize regional topics to a greater extent than the SPD, which is in line with the historical party stances towards a more federalized (CDU, FDP) versus a more centralized (SPD) political system (see also Detterbeck & Hepburn, 2018;Hepburn & Detterbeck, 2013). We no longer see this clear distinction between the two countries in the second graph. Here, there is no clear order between the German and Austrian parties. Furthermore, Figure A2 in the supplemental data online displays the distribution of our dependent variables by state. Here we again see that the German states put more emphasis on regional (and mixed) topics compared with Austria, but with one notable exception: Tyrol.

4.2.
Multivariate results explaining subnational parties' relative emphasis of regional (and mixed) topics The main aim of this paper is to analyse whether the national government status and the temporal proximity to the national election have an influence on subnational parties' relative emphasis of regional (and mixed) topics in their regional election manifestos. To do so, we run several OLS regressions. Our data are hierarchically clustered into countries, regions and parties. We account for potential country differences by including a dummy variable in the analysis. Regarding regions, we argue that there should not be substantial differences between them aside from country differences. Unlike in Spain, all Austrian and German regions have the same competencies, rights and duties. What is more interesting are the differences between parties. There are parties that are traditionally more focused on regional differences and should therefore put more emphasis on these topics. These considerations lead us to the conclusion that there are more meaningful differences between parties rather than between regions. Ideally, we would have liked to run multilevel analysis with parties as the second level. Since our dataset also encompasses parties for which we only have very few manifestos, however, this is not an option. Ultimately, Electoral strategies in multilevel systems: the effect of national politics on regional elections 849 we therefore decided to run OLS regressions with clustered robust standard errors. The results of our analysis are presented in Table 1. In model 1, the share of regional topics is the dependent variable, whereas the share of regional and mixed topics is the dependent variable in model 2.
In line with our theoretical expectations in Hypothesis 1, the coefficient for our explanatory variable Three months to national elections is negative and statistically significant in model 1. This means that subnational parties put less emphasis on regional topics if the regional election happens three months before or after a national election. We do not find similar effects for regional and shared topics in model 2. 12 Hypothesis 2 states that subnational parties should put more emphasis on regional topics if they are in government at the national level. The coefficients are positive and statistically significant in both models. The results thus indicate that parties put more emphasis on regional topics if they are in government at the national level. This finding corroborates our theoretical expectations formulated in our second hypothesis for both the share of regional topics and the share of regional and mixed topics. Note: The information in the box plots only includes parties with at least 20 regional election manifestos in the data set. Grüne ¼ Green party in Austria; B90/Grüne ¼ Green party in Germany.
Regarding the control variables, we find that only the country of origin of the party has a substantial influence on the share of regional topics in the manifesto in both models. The result of the analysis confirms what Figure  1 has shown graphically: parties in Germany talk substantially more about regional topics compared with Austrian parties. In model 2 we also find a positive, significant influence of voting turnout as well as a negative effect for being the state premier's party at the regional level.
Finally, we run several robustness checks (see Tables A5-A10 in the supplemental data online). The first four additional models assess if there are differences between the countries regarding our two hypotheses. Figure 2 shows the average marginal effects from the different models presented in Table A5 online. It demonstrates that there are not only differences regarding the relative emphasis on regional/mixed topics in the manifestos in Austria and Germany, but that the influence of our main explanatory variables varies in the different countries. The upper left graph shows that in Germany the variable indicating the temporal proximity to the national elections has a negative influence on the share of regional topics, while the variable is not significant in Austria. The same pattern holds for the graph in the lower right corner: being in government at the national level has a positive and significant influence on the share of regional and mixed topics in Germany but not in Austria.
These additional analyses highlight that although both Austria and Germany are federalized countries where the regions have a comparatively high degree of self-and shared-rule (Hooghe et al., 2010), the differences between these two countries regarding self-and shared-rule seem to impact subnational party competition in different ways, at least with regard to subnational parties' relative issue emphasis strategies in regional election manifestos. One potential explanation for these differences could be the variation in power of the second chamber in Austria and Germany. While the German Bundesrat has considerable powers and influence on policy-making, this is not the case for the Austrian second chamber (e.g., Brunner & Debus, 2008;Karlhofer & Pallaver, 2013).
In the further robustness checks, we exclude those cases where regional and national elections took place at the same day (see Table A8 in the supplemental data online), we cluster by states instead of parties (see Table  A9 online), and we include the number of days to measure temporal proximity between state and federal elections (see Table A10 online). Lastly, we also control for the existence of a regional competitor party since one could assume that the presence of a strong, regional party might motivate the other parties to also devote more attention to regional topics. In a first test, we include a dummy variable for all the parties that compete in an election with a competitor party (see Table A11 online). In a second test, we exclude all competitor parties from our analysis (see Table A12 online). The main results remain substantially the same in all these robustness checks.

CONCLUSIONS
To what extent and under which conditions do subnational parties emphasize subnational topics in their manifestos? In this study, we have tried to answer this question by arguing that the national level plays an important role when analysing the content of subnational party manifestos because of the complex multilevel structure of current European representative democracies. We hypothesized that the temporal proximity between regional and national elections as well as the government status of the party at the national level have an influence on the share of regional (and mixed) topics in party manifestos. More specifically, we hypothesized that party manifestos should include lower shares of regional topics the closer the time between regional and national elections. Additionally, we established the hypothesis that regional parties represented in the national government will put more emphasis on subnational issues than regional parties that are in opposition at the national level.
The empirical analysis has been based on the comparison of two prime examples of 'marble-cake' federalist countries, Table 1. Explaining subnational parties' emphasis of regional (and mixed) topics.
Model 1 (regional topics only) Model 2 (regional and mixed topics)

Hypothesis 1
Three months to nat. election Note: Ordinary least squares (OLS) regression models; the dependent variable in model 1 is Share regional topics; the dependent variable in model 2 is Share regional and mixed topics; clustered robust standard errors are shown in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01.
Electoral strategies in multilevel systems: the effect of national politics on regional elections 851 Austria and Germany, where subnational and national political actors are constantly coordinating both among and between their levels of government (Kropp & Behnke, 2016, p. 669). The empirical findings show that German parties considerably talk more about regional topics in their election manifestos than Austrian parties. This is in line with the theoretical expectation because German regions have more political power in the multilevel political system than Austrian regions, especially with respect to selfrule powers (Hooghe et al., 2010). Furthermore, they put less emphasis on regional topics if the regional elections happen within three months before or after a national election. At the same time, we find that subnational parties put more emphasis on regional topics (and the combination of regional and mixed topics) if they are in government at the national level. Corroborating the descriptive results, we also find country differences in the multivariate analyses. In contrast to our main analysis, however, we find a negative effect of the timing of the national elections for the share of regional topics only in the German case.
These results indicate that scholars need not only to take the federal institutional design into account when analysing its effects on party organization, party systems and voter behaviour (e.g., Thorlakson, 2020) but also need to focus more on such effects on subnational parties' relative issue emphasis in their election manifestos. The distinction in self-rule and shared-rule powers (Hooghe et al., 2010) needs to be considered more carefully. One interesting example for further analyses might be the case of Spain where regional entities vary considerably in their legal powers. Spain has even more clear-cut divisions of competencies between the national and the regional level than Austria and Germany, but additionally displays varying degrees of self-and shared-rule between regions (e.g., Libbrecht et al., 2009), thus providing scholars with the possibility to additionally analyse within-country variation regarding 'marble cake'-like and 'layer cake'-like regions (e.g., León, 2011;León & Orriols, 2016), and party competition is significantly shaped by non-nationwide parties leading to regional variations that could be leveraged (Detterbeck & Hepburn, 2010, p. 114). Yet, as it is the case with manually coding a large number of election manifestos, the development of a dictionary regarding the Spanish case might also be very time-consuming (Gross & Jankowski, 2020, pp. 76-78) due to the several different languages and dialects in which subnational election manifestos are written in Spanish regions.
Closely related to the Spanish case, another route worthwhile to take would be to apply our theoretical argument to other subnational party organizations in countries where nationwide parties compete with independenceseeking parties, such as the New Flemish Alliance (N-VA) in Belgium, the various Catalonian parties in Spain, or the Scottish National Party (SNP) in the UK (e.g., Elias et al., 2022;Fabre, 2011;Massetti & Schakel, 2016). Whereas parties from our two cases are solely Note: AMEs of parties' relative emphasis of regional topics (left) and regional and mixed topics (right), dependent on regional parties' government status at the national level (upper) and the temporal proximity between regional and national elections (lower). Estimated results are based on the models in Table A5 in the supplemental data online.

852
Martin Gross et al. competing with more regionalist-oriented parties such as the Christian Social Union (CSU) in Bavaria or the Citizens' Forum Austria (FRITZ) in Tyrol, parties in Belgium, Spain and the UK are confronted with parties striving for the independence of the region in which they are competing (Detterbeck & Hepburn, 2018, p. 134). We would expect that in such regions parties will emphasize regional issues more compared with parties competing in other regions. This might also be the case in some of the Italian regions 'where there is stronger support for sub-state nationalism' (Hepburn & Detterbeck, 2013, p. 88). Lastly, there is still a lot of research to be done on subnational parties' issue emphasis, both in other countries with joint federalism (e.g., Switzerland, Spain), where joint federalism is absent (e.g., Australia, Canada, the United States), but also in non-federal countries where subnational party competition nevertheless exists (e.g., Czech Republic, France, Poland). For example, future research could take a closer look at the causal mechanism between subnational parties' increased emphasis of regional topics when their national party organization is in government. Is this a party-based strategy, that is, do subnational parties not only have leeway in positioning themselves differently to a certain degree than their national party organization (Müller, 2013;Stecker, 2015), but do they also have the leeway in focusing on specific issues in their regional election manifestos? Or is this a strategy for subnational branches of parties that are governing at the national level to reap electoral benefits in situations when the national government is popular at the time of the regional election (Cabeza et al., 2017)?
1. It is not the purpose of the article to explain why subnational parties emphasize specific topics more than others. We use the term 'country' to refer to the national level and 'state' to refer to the subnational (regional) level. 2. Subnational party branches have some leverage in positioning themselves in the political space; however, research consistently shows that they do not position themselves very far from their national party organizations (Giger et al., 2011;Gross & Jankowski, 2020;Mueller & Bernauer, 2018;Müller, 2009Müller, , 2013Nyhuis & König, 2018;Stecker, 2015). Additionally, in some countries such as Belgium or Spain, subnational party branches must use some parts of so-called 'framework programmes' written by upper-level party organizations (Cabeza et al., 2017;van de Wardt, 2018). 3. The argument is about the fact of closeness in time between elections at the subnational and national levels, not about the sequence of these elections. 4. Our theoretical arguments and empirical strategy are not limited to the regional level but could also be applied to the local level. However, because Austrian parties only occasionally draft local election manifestos, we refrain from including local election manifestos in the following analyses. 5. We remain agnostic about the ultimate purpose of an election manifesto. Manifestos can serve the purpose of 'contract/mandate manifestos', 'advertisement manifestos' or 'identity/principle manifestos' (Bennie & Clark, 2020, p. 413). 6. We treat the CDU and CSU as one party in the following empirical analysis. Even though the CSU belongs to the set of 'federal-oriented regionalist parties' (Detterbeck & Hepburn, 2018, p. 134) and runs as a separate party organization as a state-based party (Thorlakson, 2009), both parties do not compete with each other in either federal or state elections, and the CSU 'is part of the statewide political camp of Christian Democracy' (Detterbeck & Hepburn, 2010, p. 110). Furthermore, both parties form a joint parliamentary group in the federal parliament (Bundestag) and are described as 'sister parties' (Schwesterparteien). 7. In some cases it was not possible to decide if the word belonged to the regional or national level (e.g., the German word for economy, Wirtschaft). For these instances we created a fourth vertical category in which we allocated those words that are either relevant for both the national and the regional levels, or where we could not find a clear legal competency. 8. This means that we are dealing with the 'relative issue emphasis' subnational parties devote to regional topics in comparison with the 'mixed', 'national' and fourth vertical category. 9. For a more detailed description of the dictionary, see Table A3 in the supplemental data online. 10. However, we use an alternative operationalization in one of the robustness checks (see Table A10 in the Electoral strategies in multilevel systems: the effect of national politics on regional elections supplemental data online) where we calculate the number of days between the state and national elections as a measurement for the temporal proximity between these elections. 11. Some Austrian states use/used proportional representation as a principle of government formation (i.e., reaching a certain share of votes, every party is represented in government). It is, however, common that some of these government parties additionally agree on a working agreement. We only categorize signatory parties of these agreements as regional government parties. 12. This non-finding could be a hint that parties use shared topics as an electoral strategy making use of voter's difficulties to clearly assign responsibilities to various political levels (e.g., León, 2011).