The EU Multi-Level System and the Europeanization of Domestic Blame Games

Blame games between governing and opposition parties are a characteristic feature of domestic politics. In the EU, policymaking authority is shared amongmultiple actors across different levels of governance. How does EU integration affect the dynamics of domestic blame games? Drawing on the literatures on EU politicisation and blame attribution in multi-level governance systems, we derive expectations about the direction and frequency of blame attributions in a Europeanized setting. We argue, first, that differences in the direction and frequency of blame attributions by governing and opposition parties are shaped by their diverging baseline preferences as blame avoiders and blame generators; secondly, we posit that differences in blame attributions across Europeanized policies are shaped by variation in political authority structures, which incentivize certain attributions while constraining others. We hypothesize, inter alia, that blame games are Europeanized primarily by governing parties and when policy-implementing authority rests with EU-level actors. We test our theoretical expectations by analysing parliamentary debates on EU asylum system policy and EU border control policy in Austria and Germany.


The Politics of Blame in the European Union
A central feature of politics is that vote-seeking politicians engage in blame games, attributing blame for contested policies to one another. The oppositiongovernment divide has traditionally structured blame games in domestic politics (Hansson, 2017;Weaver, 2018). EU integration and the concomitant political authority wielded at the EU-level has introduced additional political actors to the policymaking arena, but also created ample opportunities for blame attribution to external EU actors, such as EU institutions or foreign EU member state governments (Gerhards, Roose, & Offerhaus, 2013, pp. 110-112;Hood, 2011, p. 83;Rittberger, Schwarzenbeck, & Zangl, 2017). As EU policies are increasingly entering the arena of mass politics, they are becoming more salient and contested in the wider public (Hooghe & Marks, 2009;Hutter, Grande, & Kriesi, 2016;Kriesi et al., 2006). As a consequence, EU policies become focal points for the domestic politics of blame. How, then, does EU integration affect the dynamics of blame games and hence blame attributions among domestic actors?
Two literatures have, so far, touched upon this question. First, the literature on blame attribution in multi-level governance systems suggests that blame games between government and opposition are, at least partially, Europeanized as politicians from governing parties can take advantage of the complex EU multi-level governance system and shift blame for neg-ative outcomes onto external EU actors (Gerhards et al., 2013;Hobolt & Tilley, 2014;Kumlin, 2011;Roose, Scholl, & Sommer, 2018;Schlipphak & Treib, 2017;Sommer, 2019;Vasilopoulou, Halikiopoulou, & Exadaktylos, 2014). Conversely, the Europeanization of policymaking offers the opposition more points of attack vis-à-vis governing parties, since the latter are directly involved in EU policymaking as members of the Council, but only partially control EU policy choices and are thus likely to fall short of their proclaimed goals (Novak, 2013).
Second, the literature on EU politicisation is, inter alia, interested in the salience of EU issues and suggests that the frequency with which national politicians address EU policies generally depends on the conflict potential these policies carry for their parties' constituencies. Opposition politicians, and particularly those from challenger parties, have a heightened interest in politicizing EU issues in order to drive a wedge through mainstream government parties. As a consequence, politicized EU policies are likely to be the focus of heightened blame activities by the opposition. Conversely, politicians from governing parties will seek to avoid and depoliticize EU issues in order to prevent intra-party divisions, voter alienation, and avoid being the target of frequent blame attributions (de Vries & Hobolt, 2012;Hobolt & de Vries, 2015;Hutter & Grande, 2014;Hutter et al., 2016;van de Wardt, 2015;van de Wardt, de Vries, & Hobolt, 2014).
Both literatures have their merits: The blame attribution literature develops expectations about the direction of politicians' blame attributions, while the politicisation literature derives expectations about the frequency with which politicians engage in blame attributions. While the issue saliency literature has difficulties to explain why governments do, at times, address contested EU policies rather frequently (Braun, Hutter, & Kerscher, 2016;Rauh & de Wilde, 2018), the blame attribution literature has difficulty to account for the observation that governments sometimes refrain from shifting blame to external actors (Hobolt & Tilley, 2014), and that opposition parties, at times, blame EU-level actors (Vasilopoulou et al., 2014, p. 396).
Drawing on and addressing the limitations of existing explanations, the main ambition of this article is analytical. To better understand the Europeanization of domestic blame games, we develop expectations about political parties' blame attribution behaviour for EU policies, taking into consideration both the direction as well as the frequency of blame attributions (Section 2). We argue that differences in the direction and frequency of blame attributions can be accounted for by two factors. First, governing and opposition parties hold different baseline preferences, the former being blame avoiders and the latter blame generators. Secondly, blame attributions are shaped by the policy-specific distribution of policymaking authority, which incentivizes certain attributions while constraining others: Whether implementation of a contested EU policy is conducted by national-level or EUlevel actors affects the direction and frequency of politicians' blame attribution. We test our theoretical expectations by conducting a content analysis of blame attributions by members of parliament (MPs) from governing and opposition parties in the Austrian Nationalrat and the German Bundestag for two publicly contested EU migration policies: the EU's asylum system policy and EU border control policy (Section 3). We conclude by summarizing our findings and contributions to the literatures on blame attribution, politicisation, and EU accountability (Section 4).

Explaining the Direction and Frequency of Political Parties' Blame Attribution Behaviour
Who do national politicians blame for negative EU policy outcomes and how intensely do they engage in attributing blame? We start from the assumption that voteseeking politicians are boundedly rational actors who are concerned with the costs and benefits associated with their actions. This includes the decision on whether or not to blame another political actor in public for contested policies. Politicians have to decide whether to "speak up" or keep a low profile (frequency of blame), and whether to blame domestic or external actors and thus Europeanize the domestic blame game (direction of blame). We first theorize how being member of a governing or opposition party shapes blame strategiesblame avoidance or blame generation-and thereby differences in blame frequency and direction. Second, we explore how the policy-specific authority structure shapes blame frequency and direction across EU policies.

Government and Opposition: Blame Avoiders and Blame Generators
The institutional position of a political party in the political system-whether it is in government or oppositionis a central determinant for its "blame preferences." Drawing on Weaver's (2018) differentiation between blame avoidance and blame generating strategies, we conceptualize the distinct blame motivations of government and opposition parties. While government parties tend to be blame avoiders, opposition parties tend to be blame generators. This difference has distinct implications for both the frequency and direction of political parties' blame attributions.

Governing Parties as Blame Avoiders
Since governing parties exercise policymaking authority, they are prime targets for blame attributions. The motivation of governing parties is therefore to avoid or at least minimize blame for contested policies. Governing parties thus typically engage in a strategy of blame avoidance and thus of 'minimizing their responsibility for unpopular actions taken' (Weaver, 2018, p. 260; see also, Hood, 2011;Weaver, 1986). Once policies are adopted, the most important tool of blame avoidance are presentational strategies, 'attempts to deflect, avoid or mitigate blame through public rhetoric, argument or news management' (Hood, Jennings, & Copeland, 2016, p. 543). Most prominently, policymakers seek to downplay their own responsibility for contested policies while emphasizing the responsibility of other actors through blame-shifting, i.e., 'deflect[ing] blame by blaming others' (Weaver, 1986, p. 385; see also, Gerhards et al., 2013;Hood, 2011, pp. 50-53;Sommer, 2019). As regards the direction of blame attributions, the direct involvement of governing parties in EU policymaking means that they have strong incentives for avoiding blame for contested EU policies by downplaying their own responsibility and emphasizing the responsibility of others. The EU's multilevel system provides them with ample blame-shifting opportunities as they share policymaking responsibility with EU institutions (the European Commission, the European Parliament, the European Council, the European Court of Justice, the European Central Bank, or EU agencies) and foreign EU member state governments (Gerhards et al., 2013, p. 114;Roose et al., 2018, pp. 49-51;Sommer, 2019). Governing parties will thus direct blame for contested EU policies to external EU actors.
Regarding the frequency of blame attributions, governing parties face a trade-off between actively engaging in the blame game or maintaining a low profile. Engagement in public blame attributions comes with risks and might even have 'reverse effects' (Hood, 2011, p. 65). For instance, frequent blame attributions can come with a 'retribution risk' (Weaver, 2018, p. 270) as they might attract the opposition's attention and thereby escalate into a 'blame showdown' (Boin, Hart, & McConnell, 2009, p. 89). Similarly, the politicization scholarship on issue saliency points out that emphasizing EU issues is a risky strategy for politicians from governing parties in general (van de Wardt, 2015). As decisions in the EU are characterised by compromise, EU policymaking outcomes stray from the governing parties' ideal points (Rauh & de Wilde, 2018, p. 199). What is more, when domestic constituents are Eurosceptic or divided on EU issues, governing parties are expected to downplay the agreed upon policy outcomes to avoid electoral costs (Heinkelmann-Wild, Kriegmair, Rittberger, & Zangl, 2019;van de Wardt et al., 2014, p. 988). The benefits of blame-shifting might thus be outweighed by its (potential) costs and prompt governing parties to maintain a low profile (Hood, 2011, pp. 58-62;Hood, Jennings, & Dixon, 2009, p. 715). This strategy of non-engagement 'deal[s] with blame by saying as little as possible' and 'sit[ting] out a blame firestorm until it passes over and public attention comes to be focused on something else' (Hood, 2011, p. 59). Hence, governing parties tend to be blame avoiders since the putative costs of frequent blame attributions normally outweigh the benefits. They prefer, all else equal, to keep a low profile and engage in blame attribution rather infrequently and only when responding to blame attributions from other political actors. When they do engage in blame attributions, they prefer blaming external EU actors, such as EU institutions, and foreign EU member state governments over national authorities.

Opposition Parties as Blame Generators
Since opposition parties seek to challenge the government of the day, one prime motivation of opposition parties is to stick blame to the government. They therefore typically engage in a strategy of blame generation, 'generat[ing] negative messages against other politicians' in order to inflict political costs on the target, for instance, by inducing 'defections among members of the audience who would otherwise support the target' (Weaver, 2018, pp. 267-268). By generating blame, opposition parties also signal to their constituents that they fulfil their main task of holding public officials accountable and that they possess superior problem-solving competence. Regarding the direction of blame attribution, opposition parties are likely 'seeking to frame policy failures as the responsibility of current officeholder…targets' (Weaver, 2018, p. 283). They prefer to attribute blame to "their" national government as well as subordinated public actors (Hansson, 2017, p. 1;Roose et al., 2018, p. 51).
Regarding the frequency of blame attributions, opposition parties generally benefit from "speaking up" since they have much to win and little to lose. While opposition parties might also face costs from frequent blame attributions, such as a backlash from their own supporters, they should be less risk-averse than governments (van de Wardt, 2015, p. 94). In addition, by engaging in blame attributions for negative EU policy outcomes, opposition parties may benefit from merely emphasizing an issue. Even if not targeting the government directly, increasing the salience of an issue can drive a wedge through government parties' constituencies when it is divided on EU integration in general or on the issue at hand in particular (van de Wardt et al., 2014, p. 988). Hence, opposition parties are likely to attribute blame more frequently than governing parties, and they prefer blaming their national government over external EU actors.
In sum, all else equal, governing parties are more likely than the opposition to direct blame to external EU actors. Moreover, opposition parties are generally more inclined to engage in blame attribution behavior than their counterparts in government. While government parties tend to be blame avoiders and opposition parties blame generators, we argue, in the next section, that these blame attribution motivations are affected by the policy-specific political authority structure in the EU's multi-level system: H1: The share of blame attributions directed at external EU actors is higher for governing parties than for opposition parties; H2: Governing parties engage in blame attribution behaviour less frequently than opposition parties.

Political Authority Structure: Blame Plausibility and Blame Pressure
In this section we argue that the policy-specific authority structure, in particular the level of government where policies are implemented, affects political parties' ability to pursue their preferred blame strategy in two ways.

Distribution of Political Authority Limits Plausible Blame Targets
First, political authority structures lend plausibility to specific blame targets and thus shape the direction of blame attributions. The institutional structure governing a particular issue area incentivizes or constrains blame attributions to certain actors and not to others, even if this contradicts political parties' baseline preferences for blame attribution. Claims about the responsibility for contested policies need to remain plausible: Parties can attribute blame according to their baseline preferences only in so far as their blame attributions are able to maintain the 'illusion of objectivity' (Kunda, 1990, pp. 482-483). Specifically, the direction of blame attributions is circumscribed by the institutionalised responsibilities in the policymaking process (Heinkelmann-Wild & Zangl, 2019; Schwarzenbeck, 2015, p. 37).
EU policymaking generally increases the number of potential blame targets and its overall complexity obfuscates the clarity of individual policymakers' responsibility (Hobolt & Tilley, 2014, p. 45;León, Jurado, & Garmendia Madariaga, 2018, p. 661). Recent literature suggests that the complexity of EU policymaking puts one set of actors in the spotlight: implementing authorities (Heinkelmann-Wild, Rittberger, & Zangl, 2018;Rittberger et al., 2017). When a political actor is clearly responsible for "on the ground" implementation, she is likely to be identified with a policy outcome and becomes focal in the public domain. The EU's political authority structure thus incentivizes blame attributions to be directed at implementing actors, since blame attributions to non-implementing actors tend to be less plausible (Heinkelmann-Wild & Zangl, 2019). If policy-implementation authority is squarely located at the EU-level, governing and opposition parties have a heightened incentive to attribute blame for negative policy outcomes to a specific group of external EU actors: implementing EU institutions. Conversely, if EU policies are implemented by national-level authorities, governing parties can shift blame onto other EU member states and their respective domestic implementing authorities, while opposition parties can stick blame to "their" national government in line with their baseline blame attribution preferences. While opposition parties are thus likely to Europeanize blame solely in the case of EU-level implementation, the governing party can plausibly shift blame to external EU actors. Hence, all else equal, if a policy is implemented by EU-level actors the overall share of blame attributions directed at external EU actors by governing and opposition parties is higher than when EU policies are implemented by nationallevel actors.

Distribution of Political Authority Impacts Blame Incentives
Second, the policy-specific political authority structure not only affects the direction, but also the frequency of blame attributions. When the responsibility for implementing an EU policy rests with national-level authorities, opposition parties can plausibly blame their preferred target: "their" national government. The frequency of opposition parties' blame attributions should thus be higher compared to situations characterized by EU-level implementation.
National-level implementation therefore puts blame pressure on governments. When blame is predominantly targeted at a particular actor, she cannot remain inactive but seeks to deflect blame by blaming others: 'Higher levels of blame will be likely to lead to active attempts to reduce or remove it than "do nothing" or "no comment" responses' (Hood et al., 2016, p. 544). When under blame pressure, governing parties thus have an incentive to mount their blame attribution activities to avoid electoral costs (Traber, Schoonvelde, & Schumacher, 2019, pp. 3-4). When the responsibility for implementing an EU policy rests with national-level actors, the blame pressure on the government is thus likely to prompt governing parties to enter the blame game more forcefully-either in anticipation of their focality as implementing actors, or in evasion of responsibility once they become focal in the opposition's blame attributions. Conversely, if an EU-level actor carries responsibility for policy-implementation, blame pressure on governing parties will be more moderate and they will be less inclined to engage in blame-shifting, hiding out behind the complexities of EU policymaking. Hence, if a policy is implemented by national-level actors, the overall frequency of blame attributions by all parties is higher than when EU policies are implemented by EU-level actors.
In sum, all else equal, we expect that EU-level implementation increases the overall share of blame attributions to external EU actors. At the same time, we expect the overall frequency of blame attributions to be lower when a contested policy is implemented by EU-level actors compared to national-level actors: H3: If EU-level actors implement an EU policy, the share of blame directed at external EU actors is higher than if it is implemented by national-level actors. H4: If EU-level actors implement an EU policy, the frequency of blame is lower than if it is implemented by national-level actors. Table 1 summarizes all expectations about the direction and frequency of blame attributions in the EU's multilevel system.

Blame Attributions in EU Multi-Level Politics
To test our theoretical arguments empirically, we analyse the blame attributions of German and Austrian opposition and governing parties in two cases of contested EU migration policy. We first introduce our case selection and describe the procedure of data collection and coding. We then present our empirical findings on the direction and frequency of blame attributions.

Research Design
We evaluate our hypotheses by comparing political parties' blame attributions in the Austrian Nationalrat and the German Bundestag in two instances of EU migration policies. The focus on parliamentary debates permits an analysis of the direction and frequency of blame attributions by government and opposition parties. Parliamentary debates are a likely venue for blame attributions since they pit governmental and opposition parties against each other. They are also relevant for the broader public due to their communicative function with regard to EU policies (Auel, 2007;Rauh & de Wilde, 2018).
We selected two EU policies that are prominent and highly contested in the public to ensure sufficient coverage in the parliamentary debates (Hood, 2011, p. 8;Weaver, 2018, pp. 282-283). First, the EU asylum system policy regulates which EU member state is responsible for administering asylum claims by people entering the EU and defines minimum standards for reception conditions and procedures. It was heavily criticised for a long row of dysfunctions (Rittberger et al., 2017). Second, EU border control policy has been designed to save lives at sea, to strengthen the EU's external borders and to disrupt the business of traffickers and human smugglers. It was criticised, inter alia, to have resulted in the deaths of thousands of migrants crossing the Mediterranean on their way to the EU (Rittberger et al., 2017). We collected parliamentary debates from the German Bundestag and the Austrian Nationalrat starting with the official decision to adopt the respective policy. Since neither of the two policies under analysis has been discontinued or replaced, the data collection ends with the initiation of the coding process on 13 August 2018.
The structure of policy-implementation varies across the two cases: The EU asylum system is implemented by EU member states, while EU border control policies are primarily implemented by an EU actor (i.e., Frontex). Moreover, the cases display similarities which allow for the control of possible confounding factors across policies and countries. First, the two policies belong to the same issue area, i.e., the EU's migration regime. Second, their policymaking structures are similarly complex: Following a proposal by the Commission, the Council decides on the respective policy and the EP is involved either through co-decision or consultation procedures. Third, Austria and Germany are both considered destination states of migration movements and were thus similarly affected by the two policies (Biermann, Guérin, Jagdhuber, Rittberger, & Weiss, 2019).
To gauge the blame attributions of governing and opposition parties, we coded the blame attributions voiced by individual MPs in parliamentary debates, which we subsequently aggregated to the level of governing and opposition party. To analyse MPs' blame attributions, we combined automated data collection with a qualitative content analysis. Previous studies have mostly engaged in qualitative content analysis to identify politicians' blame attribution (e.g., Gerhards et al., 2013;Hobolt & Tilley, 2014, pp. 100-119;Mortensen, 2012Mortensen, , 2013. A common pitfall of these studies constituted the reliable identification of rare responsibility statements in large text corpora (e.g., Gerhards, Offerhaus, & Roose, 2007, pp. 117-118;Schwarzenbeck, 2015, pp. 78-82). Inter-coder reliability tests often show low levels of agreement with regard to the identified statements. Contrariwise, once a blame statement is identified, agreement between coders is usually high regarding the components of a specific blame attribution such as its sender or target.
Our approach combines automated and manual coding procedures and thereby helps to overcome the twofold challenge of identifying relatively rare statements in large text corpora and, at the same time, accounting for their context-sensitivity. Using a blame-related dictionary, we first automatically coded potential blame attributions within text segments that referred to the Table 1. Expectations about the direction and frequency of blame attributions.

Institutional position of blame sender
Policy-specific authority structure (government vs. opposition) (national-level vs. EU-level implementation) Direction of blame (share of blame to external Gov senders > Opp senders (H1) National-level < EU-level (H2) EU actors)

Frequency of blame
Gov senders < Opp senders (H3) National-level > EU-level (H4) (absolute blame attributions) two policies. Based on the pre-selected sample, we then manually coded blame attributions that comprised three criteria: • Blame object, i.e., a contested policy for which blame is attributed. For the purposes of this article, the policies under consideration were the EU border control policy and the EU asylum system policy; • Blame sender, i.e., an actor that attributes blame for a policy failure. For the purposes of this article, we focus on MPs. Blame senders are assigned to one of the two categories: (i) MPs from a governing party; or (ii) MPs from an opposition party; • Blame target, i.e., the actors to whom blame is attributed. For the purposes of this article, blame targets are assigned to one of two categories: (i) external EU actors, such as EU institutions (i.e., the Commission, the Council and EU agencies like Frontex) and foreign EU member state governments and their representatives; or (ii) domestic public actors (i.e., representatives of the national government or national MPs).
Overall, we identified 558 blame attributions in 390 debates (for an overview, see Table A.1 in the Supplementary Material).

Assessing the Direction of Blame Attributions
With regard to the direction of blame attributions, we expected that the share of blame attributions from governing parties targeting external EU actors is higher than for opposition parties (H1); and that the overall share of blame attributions directed at external EU actors is higher in the case policies implemented by EU-level actors than for policies implemented by national-level ac-tors (H3). To evaluate these expectations, we display the share of blame attributions that target domestic actors and external EU actors respectively for MPs from governing and opposition parties in the two cases. The overall pattern of MPs' blame attribution lends supports to our expectations (see Figure 1). First, the share of blame targeted at external EU actors is higher for governing parties than for opposition parties in both cases. In the EU asylum system case (see Figure 1a), the predominant share of governing parties' blame attributions (60%) targeted external EU actors, while a minority of their attributions assigned blame to domestic actors (40%). Conversely, only a minor share of opposition party MPs' blame attributions targeted external EU actors (33%). They predominantly directed their blame to domestic actors (67%). In the EU border control case (see Figure 1b), the predominant share of blame attributions by governing parties was again directed at external EU actors (86%) while only one statement targeted a domestic actor. While opposition parties also predominantly assigned blame to external EU actors (70%), their share of blame to external actors is lower than that of governing parties. A minor share of their blame attributions targeted domestic actors (30%).
Second, irrespective of a parties' institutional position in the political system, the overall share of blame attributions directed at external EU actors is higher in the EU border control case, where policy-implementation is carried out by an EU-level actor, compared to the EU asylum system case, where policy-implementation is in the hands of national-level actors. The share of blame attributions targeting external EU actors from governing parties (86%) and opposition parties (70%) is higher in the EU border control case than the respective shares from governing parties (60%) and the opposition (33%) in the asylum system case.

Assessing the Frequency of Blame Attributions
With regard to the frequency of blame attributions, we expect that, overall, blame attributions from governing parties are less frequent than those from opposition parties (H2). We also expect that blame attributions are overall less frequent when an EU policy is implemented by EU-level actors than when it is implemented by national-level actors (H4). To evaluate these expectations, we display the average number blame attributions from governing and opposition parties per 100 debates. The overall pattern of blame attributions again lends supports to our expectations (see Figure 2). First, governing parties attribute blame less frequently than opposition parties in both cases. In the EU asylum system case, opposition parties attributed blame 129 times per 100 debates while the governing parties only attributed blame 41 times per 100 debates (see Figure 2a). Similarly, in the EU border control case, opposition parties attributed blame 104 times per 100 debates while governing parties only attributed blame four times per 100 debates (see Figure 2b).
Second, blame attributions are less frequent in the EU border control case, where policy-implementation rests with an EU-level actor, than in the EU asylum system case, where policy-implementation is carried out by national-level actors. The overall frequency of blame is higher in the EU asylum system case than the EU border control case for both the opposition (129 > 104) and the government (41 > 4). Moreover, we find that the difference in blame frequency between the two cases is larger for the government (700%) than for the opposition (24%). This finding tentatively suggests that increased blame pressure on the government has a stronger influence on the frequency of blame attributions than the heightened incentive for opposition parties to plausibly blame their preferred target. However, further data is necessary to substantiate this conclusion.

Discussion
The observed blame attribution patterns for the contested EU policies corroborate our expectations about the direction and frequency of national parties' blame behaviour. Blame to external EU actors is most prominent in the attributions by governing parties, and when policies are implemented by EU-level actors. Yet, the frequency of blame attributions in these instances is comparatively low. Overall, opposition parties attribute blame more frequently than governing parties. Moreover, the frequency of blame is higher in cases of national-level implementation than in cases of EU-level implementation.
The statistical tests included in the Supplementary Material substantiate our argument that there is a significant relationship between our two independent variables (institutional position of the blame sender and policy-specific implementing authority) and the direction and the frequency of blame attributions. Specifically, we calculated chi-square tests for the direction of blame (see Supplementary Material, Tables A.4-7) and Wilcoxon rank-sum tests for the frequency of blame (see Supplementary Material, Table A.12). In addition, the findings still hold when we analyse the blame attributions in the Austrian Nationalrat and in the German Bundestag separately. Despite minor differences, the blame attribution patterns are not only similar and in line with our expectations, but we can also reject the null hypothesis about a random match on this level of analysis both for the direction of blame (see Supplementary Material, Tables A.8-11) and the frequency of blame (see Supplementary Material, Table A.13). We are thus confident that the results are not driven by our selection of countries. In sum, these results bolster our confidence that the position of a party in the political system, as well as the policyspecific authority structure shape the direction and frequency of national parties' blame attributions.

Conclusion
In this article we argued that variation in the frequency and direction of blame attribution for contested EU policies can be explained by a political party's government or opposition status and by the policy-specific authority structure. Our article contributes to a better understanding of policy-specific political conflict and elite communication in the EU's multi-level system. In particular, the findings allow us to inform claims from three different literatures. First, the EU politicisation literature holds that governing parties tend to restrain themselves from "speaking up" on EU issues in order to avoid putative electoral costs that come from their politicization. Yet, we find that frequent blame attributions are by no means limited to opposition parties. Government parties frequently attribute blame (to external EU actors), especially when the government holds policy-implementation authority and when the blame pressure on the governing parties is thus high.
Second, the literature on blame attribution in multilevel governance systems posits that policymakers are rather unconstrained in attributing blame to EU-level actors. Yet, we find that governing parties only engage in frequent blame-shifting when policies are implemented by national-level actors and they are consequently exposed to high blame pressure. Otherwise, they prefer to hide out in the complexities of EU multi-level policymaking. Moreover, opposition parties even refrain from directing blame to "their" government when EU-level actors are policy implementers.
Finally, the EU accountability literature diagnoses a national "opposition deficit" (Rauh & de Wilde, 2018, p. 210) with regard to EU policies (Auel, 2007;Kiiver, 2006;O'Brennan & Raunio, 2007;Raunio, 2011). Indeed, we find that governing parties can successfully "duck and cover," avoid the opposition's blame and thus evade accountability when policy-implementation authority rests with EU-level actors. By contrast, when domestic executive agencies hold policy-implementing authority, governing parties are unable to avoid the opposition's blame, but-rather than owning their mistakes-they shift the blame directed at them to external EU actors. In this sense, blame attribution patterns re-produce the EU's much-lamented democratic accountability deficit.
Our findings come with two caveats. First, since our focus is on two cases of highly contested and salient EU policies, we cannot conclude that less contested policies come with similar blame attributions patterns. Future research should thus look at cases that vary in salience and across issue-areas. A second and related caveat pertains to the affectedness of the public by a particular policy issue. As destination states with high migratory pressure, both Austria and Germany were directly affected by the EU's asylum policy (Biermann et al., 2019), whereas the effects of the EU border control policy were more distant. The selected cases are thus not suitable to rule out a link between the degree of affectedness of the analysed countries and the frequency of blame attributions. Future research should look into cases that vary in their degree of affectedness while controlling for the political authority structure. The insights presented above are, therefore, only a starting point for the analysis of blame attribution behaviour in the EU multi-level system.