Electoral Behaviour in a European Union under Stress

Is electoral behaviour affected by the current challenges of the EU and, if it is, through which channels and mechanisms? This study offers a cross-national analysis together with a broad understanding of both the crisis phenomenon and electoral behaviour. To investigate this research question appropriately, we first distinguish at the most general level between the two main behavioural alternatives at play when it comes to electoral behaviour, namely abstention and vote choice. Second, and no less important, we differentiate between the mechanisms that mediate the relationship between the ‘EU under stress’ and electoral behaviour, namely egocentric and sociotropic economic voting motivations. Drawing on data from the European Election Study 2014, our article provides important insights into the study of electoral behaviour in an EU under stress. First, we are able to show that the multiple crises that have hit the EU have the potential to determine both turnout and the decision to vote for a Eurosceptic party. Second, different mechanisms are in play for each of the two behavioural alternatives: Turnout is clearly related to egocentric determinants and thus depends on individuals’ personal exposure to the financial crisis. Conversely, the decision to vote for a Eurosceptic party is based on a different mechanism. Voters—without necessarily being personally affected by the crisis—have a higher propensity to vote for a Eurosceptic party if they perceive their country to be threatened by such an EU under stress. These findings add to a better understanding of the EU’s multi-level democracy.


Introduction
Starting with the financial crisis which hit the world in 2007-2008, followed by the economic and the sovereign debt crisis, the EU has faced numerous events that together have brought the European system of multi-level governance into a permanent state of crisis. Although the ensuing recession was certainly a significant event, the Brexit referendum and related discussions around the procedure of the UK's withdrawal from the EU as well as continuing debate among EU member states within the policy field of immigration have also con-tributed to the air of crisis. In short, the EU is "in the midst of its worst crisis since its inauguration in 1993" (Verdun, 2013, p. 45). Scholars have already begun to investigate the conditions and consequences of such an EU under permanent stress (e.g., Cramme & Hobolt, 2015). The bottom line is that the multiple crises had and still have far-reaching implications not just for the political systems of EU member states but also for the European level of governance. Moreover, empirical studies have shown that not only has the political system been affected but also the dimension of wider EU politics. Political parties adapt their policy positions towards the EU in times of crisis (Braun, Popa, & Schmitt, 2019;Conti, Hutter, & Nanou, 2018), while citizens adjust their perceptions of the EU system of multi-level governance (Braun & Tausendpfund, 2014;Hobolt & Wratil, 2015). Not only have citizens' perceptions changed in times of crisis but so too has their political behaviour. Empirical studies show that the crisis has consequences for individual voting behaviour (Hernández & Kriesi, 2016), although in particular economic effects on incumbent support are less crisis-driven than has previously been assumed (Talving, 2018). The bulk of empirical studies focuses on the vote of government and mainstream parties as well as Eurosceptic parties in times of crisis, whereas scholars have traditionally been less interested in voters' electoral participation. Although we agree that the investigation of the Eurosceptic vote or respectively the defection from mainstream or government parties is highly relevant, this is only one possible citizens' reaction. At least theoretically, voters can choose between two main behavioural alternatives. Hirschman (1970) aptly labels these two basic behavioural alternatives "exit" or "voice" as possible options if consumers (in our case voters) are no longer satisfied with the quality of the product (in our case the EU under stress).
Against this background, the aim of our article is to provide a comprehensive and systematic theoretical framework as well as an in-depth empirical study on electoral behaviour in the EU under stress. Since electoral behaviour in the EU can best be studied by examining elections to the European Parliament (EP), we draw on data from the 2014 European Elections Study (EES; Schmitt, Hobolt, Popa, Teperoglou, & European Parliament, 2015). To effectively examine the idea of the EU that has been much challenged in recent years, we look at the economic crisis and study its implications on electoral behaviour, namely the two main behavioural alternatives of abstention and vote choice. To put it simply, our empirical study seeks to investigate whether and how (i.e., through which mechanisms) electoral behaviour is affected by an environment in a state of permanent crisis or under stress. In line with major insights from economic models of voting (e.g., Lewis-Beck & Stegmaier, 2007), we argue that two different mechanisms are in play when studying the implications of the series of EU crises on electoral behaviour. Either individuals decide to abstain or to cast their vote with reference to their personal economic situation ('egocentric voting'), or they take into account the national economic conditions ('sociotropic voting'). In order to present a comprehensive and systematic overview of electoral behaviour in the EU under stress, we integrate the two behavioural alternatives together with the mechanisms into one analytical framework. Our findings show that the existence of an EU under stress has implications for both turnout and the decision to vote for a Eurosceptic party. Moreover, different mechanisms are present for each of the two behavioural alternatives: Turnout is clearly related to egocentric determinants and thus depends on personal ex-posure to the financial crisis. Conversely, the decision to vote for a Eurosceptic party is based on a different mechanism. Voters-without necessarily having been personally affected by the crisis-have a higher propensity to vote for a Eurosceptic party if they perceive their country to be threatened by such an EU under stress. These findings add to a better understanding of EU multilevel democracy.

Electoral Behaviour in the EU under Stress: Abstaining or Voting for Eurosceptic Parties
What do we know about electoral behaviour in the EU under stress so far? From a predominantly theoretical perspective, the standard model of political participation, the civic voluntarism model (Verba, Schlozman, & Brady, 1995) postulates that political participation depends on motivation, mobilisation, and resources. In times of crisis, when material resources become scarcer, political participation is thus supposed to decrease. According to Verba and Nie (1972), people with access to the necessary resources are more prone to participate actively in politics. Conversely, individuals with fewer resources tend to participate less in political life than others. To put it differently, "actors require resources to be able to participate politically" (Kern, Marien, & Hooghe, 2015, p. 466), primarily because these individuals realise they have fewer possibilities to exert an influence on politics. They thus decide not to take part and become in a sense apathetic with respect to political life. Linking this argument to the current crisis environment of the EU, di Mauro (2016) has been able to show that the economic crisis-one important trigger for the EU under stress-indeed has an effect, albeit limited, on electoral participation.
Not only can electoral participation be affected but also the choice to vote for a specific party. One likely scenario in this regard is that citizens in an EU under permanent stress vote to a higher degree for Eurosceptic parties to show their discontent with the current situation. Empirical studies point to an intensification of a broader trend towards destabilisation within the European party systems during times of EU crisis-e.g., with higher vote shares for populist and Eurosceptic parties (Hernández & Kriesi, 2016). The latter insight dovetails with the finding that citizens behave differently in normal times than in times of crisis (Kern et al., 2015). Individuals decide to vote for their preferred mainstream parties in a crisisfree environment, whereas they tend to avoid choosing established parties in times of economic hardship because they have lost their faith in the main actors of the political system and vote instead for more extreme parties (Hernández & Kriesi, 2016). It has also been shown that motivations to vote for incumbent parties varied at different stages of the economic crisis (Okolikj & Quinlan, 2016), but that voters have certainly defected from government parties in times of crisis as a protest against the EU's fiscal policies (Magalhães, 2016). In addition, voters seem to have been more reactive to government policy decisions in the post-crisis period than before (Talving, 2017). In contrast to these findings, Talving (2018) has also shown that economic effects on incumbent support are surprisingly stable over time, which suggests that economic voting is as pronounced during times of crisis as it is in normal times. Moreover, citizens harmed more by the crisis have a higher propensity to vote for Eurosceptic parties (Hobolt, 2015;Hobolt & de Vries, 2016), a tendency which is driven by two main factors: a general dissatisfaction with mainstream parties on the one hand and, on the other, fundamental concerns of voters about the domestic effects of EU membership or their discontent with the EU's handling of the various crises (Hobolt, 2015;Hobolt & de Vries, 2016;Treib, 2014).

Lacking Resources, Egocentric and Sociotropic Economic Considerations in View of the 'EU under Stress'
To sum up, findings from previous studies suggest that an environment of crises or, in the case of our study more precisely, the EU under stress, has important implications for electoral behaviour, namely abstention and a Eurosceptic vote. As summarised in Figure 1, we seek to investigate how (i.e., through which mechanisms) the EU multi-level system under stress influences the two main alternatives of electoral behaviour. Beside the fact that previous studies mainly focused on these behavioural alternatives, we opted for turnout and vote choice as principal variables, building on Hirschman's (1970) general framework which has been applied successfully for electoral research and in the case of EP elections (Weber, 2011). It suggests that voters can generally choose between "exit" or "voice." Applying these two behavioural alternatives to our research question implies that voters can either decide to abstain from an election ("exit") or they can show their disapproval with the current situation by voting for Eurosceptic parties ("voice"). Our study seeks a better understanding of the precise motivation for each of these decisions in times of permanent crisis. We study in particular the underlying mechanisms in play: egocentric or sociotropic economic motivations. The subsequent sections serve to explain in more detail these mechanisms and their role in the case of electoral behaviour in the EU under stress. To abstain from elections as well as to vote for Eurosceptic parties is related to some degree to the fact that the EU is under stress. But, to study this link appropriately, we need to take into account insights from additional scholarly work. To capture each of the mechanisms appropriately, we draw both on economic models of voting (Duch & Stevenson, 2008;Lewis-Beck & Stegmaier, 2007)  Together with the general civic voluntarism model and the empirical findings reported above, insights from two additional strands of literature can tell us a lot about the particular mechanisms in play when economic considerations unfold in the case of electoral behaviour. First, empirical studies on EU support have highlighted for a long time the predominant role of utilitarian considerations when determining public attitudes towards European integration (Eichenberg & Dalton, 2007;Gabel, 1998). Elkink et al. (2019, p. 335) mention in this regard that "sociotropic motivations-an altruistic driveare a significantly more potent driver of the voter visà-vis utilitarian motivations." The latter differentiation has been inspired by the second strand of literatureeconomic voting (Duch & Stevenson, 2008;Lewis-Beck & Lobo, 2017;Lewis-Beck & Stegmaier, 2007)-which is also highly relevant to our task of disentangling the mechanisms behind electoral behaviour. The economic voting model posits that voters will reward the incumbent government with their vote if the economy is good, and conversely punish it by casting their vote for another party if they perceive their economic environment as less successful. Additionally, it is a common claim that individuals either cast their vote with reference to their personal economic situation, i.e., based on egocentric considerations ('pocketbook voting'), or they take into account the perceived or objective national economic conditions ('sociotropic voting'). In the latter case, voters behave not in an egocentric way, but display a more altruistic and therefore sociotropic behaviour (Lewis-Beck & Stegmaier, 2007). Important to note, however, research shows that 'sociotropic voting' has a greater impact than egocentric pocketbook voting (Lewis-Beck & Stegmaier, 2013).
Overall, these thoughts from a) the civic voluntarism model, b) the economic voting literature, as well as c) insights from utilitarian accounts in the EU support literature, lead to the following theoretical assumptions: Based on the theoretical account of the civic voluntarism model, we assume for the most fundamental type of political participation-namely, electoral turnout-that individuals who have suffered personally in times of crisis and economic downturn or who have more generally experienced the EU under stress are assumed to participate less often in elections because they feel unable to exert any influence on politics. This is a reasonable assumption since unemployment rates have increased in many, though not all, EU member states in the years following the financial crisis. Moreover, at least in the member states that have been hit hardest by the subsequent waves of financial, economic, and Euro currency crises, material resources have become scarcer. In addition, the literature on EU attitudes and economic voting theory suggests that economic-egocentric as well as sociotropic-considerations apply when voters decide to vote for a particular party, in our case affecting the vote for Eurosceptic parties. As such, we have created the following hypothesis: H1a: The greater the individual exposure to the financial crisis and its consequences (e.g., experienced through reduced income, less money, lost job), the less inclined are respondents to cast a ballot in EP elections (lack of resources egocentric nonvoting); H1b: The greater the individual exposure to the financial crisis and its consequences (e.g., experienced through reduced income, less money, lost job), the more inclined are respondents to vote for Eurosceptic parties (lack of resources egocentric Eurosceptic voting); H2: The more citizens perceive the financial crisis and its consequences as a threat for their country (i.e., are dissatisfied with the current and future state of their national economy), the more inclined they are to vote for a Eurosceptic party (sociotropic Eurosceptic voting, individual level); H3: The more strongly a country has been hit objectively by the financial crisis and its consequences (e.g., measured via levels of unemployment, gross domestic product (GDP), national debt), the less inclined are respondents to cast a ballot in EP elections and the more inclined are respondents to vote for a Eurosceptic party (direct non-voting and Eurosceptic voting).
Altogether, we assume that the financial crisis and its consequences that marked the starting point of the EU's current permanent state of tension has implications for electoral behaviour (H1 to H3). More specifically, we believe that different motivations are in play in this regard: A (perceived) lack of resources which is almost identical to 'egocentric' considerations in the perspective of economic voting theory can be linked theoretically to both abstention and a Eurosceptic vote (H1). 'Sociotropic' considerations, in contrast, might in particular occur in the case of Eurosceptic voting (H2) which is related to the idea of winners and losers of globalisation (Kriesi et al., 2008). In short, the assumption is that the so-called losers of globalisation (and to a similar degree of European integration) do not benefit from these new trends but rather fear them. Although transnational cooperation in general is thus perceived as a threat to their national environment, these feelings intensify in times of permanent crisis. Finally, one could of course assume that the crisis itself has direct implications for electoral behaviour (H3). Figure 2 summarises the overall model of our analysis for the two dependent variables. While H1 and H2 are dedicated to testing the individual-level effects of the financial crisis and its consequences on the two dependent variables, H3 is formulated to examine the aggregatelevel link. Moreover, to single out the genuine effects of what we have labelled 'EU under stress,' it is necessary not only to take into account economic indicators that shine a light on the crisis itself. We also need to control for relevant determinants of electoral behaviour as well as for the particular character of EP elections. According to classical theories of voting behaviour (Arzheimer, Evans, & Lewis-Beck, 2017) and vote abstention (Cancela & Geys, 2016;Smets & van Ham, 2013) traditional determinants of electoral behaviour are political interest, internal political efficacy, party identification, and left-right placement, as well as demographic features such as education, social class, gender, and age. With regard to the particular character of EP elections, we take into account satisfaction with the EU and the national government. This is important insofar as EP elections represent a particular set of elections since these contests are considered to be second-order elections (Reif & Schmitt, 1980;Schmitt & Teperoglou, 2017)mainly because there is less at stake compared to national first-order elections. Equally important in terms of EU-level particularities is the individual perception of responsibility on the part of the EU for the series of crises (Hobolt & Tilley, 2014). Moreover, in EP elections, voter turnout is also to a large degree determined by factors related to the electoral systems of the member states and factors connected to the EU such as compulsory voting (Mattila, 2003), EU membership duration (Flickinger & Studlar, 2007) and satisfaction with EU membership (Steinbrecher & Rattinger, 2012;Stockemer, 2012). Accordingly, we control at the individual level for socio-demographic background, general political attitudes and EU support. At the aggregate level, we take into account the duration of EU membership, compulsory voting, post-communist membership, and the electoral cycle.

Design of the Study
In line with the model illustrated in Figure 2 we include in our analysis individual-level and context-level variables. Hence, we need to apply a multi-level approach (Hox, Moerbeek, & Schoot, 2018). This method estimates both the influence of individual-level and con-textual factors in a simultaneous and statistically accurate manner. Both dependent variables are binary variables. Therefore, we use multi-level logistic regression. Following the logic of hierarchical modelling, we present a series of models, with each model building upon the preceding one. Taking as our starting point the 'empty' model, which excludes independent variables, we can establish the variance to be explained for the micro and macro levels (intra-class correlation). In the next step, we include the individual variables (H1, H2), before considering the contextual indicators (H3).
For the empirical analyses, we draw on data from the EES 2014 . From 30 May to 27 June 2014, a national post-election survey was conducted in each of the 28 member countries of the EU. The sample size is roughly 1,100 people in each EU member state, with the exceptions of Cyprus, Luxembourg, and Malta, where the sample size is 500, as well as the UK and Germany, where the sample is approximately 1,400 and 1,600 citizens respectively. Overall, 30,064 citizens across all member states of the EU were interviewed (by means of a computer assisted personal interview).
Before we start to test our theoretical assumption, we should shed some light on the dependent variables of our study: abstention and the choice to vote for a Eurosceptic party. To map abstention, we use the following question: "EP elections were held on the [date]. For one reason or another, some people in [our country] did not vote in these elections. Did you yourself vote in the recent EP elections?" 17,217 respondents participated in the European election, while 12,778 persons stayed away from the ballot box and 69 respondents answered with "don't know." The latter group was coded as non-voters. Turnout, measured via survey data, varies considerably between the EU member states (see Figure 3). Moreover, it becomes obvious that turnout in EP elections is overestimated when we draw on data from the EES 2014 (see also Mattila, 2003, p. 453  turnout is on average 12 percentage points lower than that based upon the post-election survey. Although this phenomenon can be clearly classified as 'over-reporting,' the correlation between official turnout and our survey data is very high (r = .89; N = 28). In the end, at least for the investigation of our research question, the issue of over-reporting is negligible. The second dependent variable in our analysis is the choice to vote for a Eurosceptic party. To classify the national political parties as Eurosceptic, we use the information provided by Hobolt (2015, p. 13; see Table A1 in the Supplementary File). We coded the vote of a Eurosceptic party as 1 (N = 3,286), whereas the vote of non-Eurosceptic parties was coded as 0 (N = 13,931); non-voters were excluded from this particular analytical step. As in the case of abstention, the Eurosceptic vote varies strongly between EU member states (see Figure 4). While more than 60 percent of voters in Hungary voted for a Eurosceptic party, the share of Eurosceptic votes in Malta was 0 percent (as a result, no party in Malta has been classified as a Eurosceptic party). Again, the pattern differs between the official results and the findings provided by the EES 2014. What we can learn from this rather descriptive lesson is that drawing on survey data alone can lead to an underestimation of the share of Eurosceptic parties (this is true for all countries except Latvia). Nonetheless, the correlation between the aggregated individual data and the official data is r = .97 (N = 28), whereby the strong correlation indicates that possible biases are also negligible.
As previously discussed, when it comes to our main independent variables, we need to distinguish carefully between factors operating at the individual and the con-textual level. At the individual level, we use five variables to capture the individual economic situation in the aftermath of the financial crisis. As indicators for the idea of egocentric considerations due to lack of resources (H1), we use three different questions: First, whether the respondent or someone in the household of the respondent has lost his or her job during the last two years; second, if the household has experienced a decrease in income during the last two years; third, how often the respondent faced difficulties in paying bills at the end of the month during the last twelve months. These three indicators ideally map different nuances of economic deprivation in the aftermath of the crisis. As indicators for 'sociotropic voting' (H2), we use the individual perception of the current national economy and the evaluation of the general economic situation over the following twelve months. Higher values indicate a negative perception of the national economy. To map the implications of the crisis at the contextual level (H3), we use GDP per capita, the unemployment rate, and the national debt as objective indicators for the economic situation. In contrast to previous studies, we have opted not only to use static values (e.g., the economic situation in 2013) but also to allow for a dynamic perspective taking into account changing values over time (2008 to 2013). Furthermore, we include a dummy variable to indicate whether a country is a member of the Eurozone (yes = 1) or a bailout country (yes = 1).
In addition to the economic variables, empirical research on turnout and vote choice suggests a number of other potentially influential factors, especially at the individual level. Therefore, we include several variables designed to control for other potentially confounding fac-  tors: the individual perception of responsibility of the EU for the current economic situation; satisfaction with the EU and the national government; political interest, internal political efficacy, party identification, and the leftright placement as classical determinants of turnout and voting. Finally, we consider demographic features such as education, social class, gender, and age. At the macro level, we control for compulsory voting, election cycle, duration of EU membership and communist history. All indicators are coded identically with low values indicating low levels and high values indicating high levels of the respective characteristics. All variables are rescaled-the lowest value is 0 and the highest is 1. Whereas in the case of turnout, we consider all member states of the EU (N = 28), in the case of the Eurosceptic vote, Malta will be excluded due to the lack of a Eurosceptic party in this country (N = 27). Descriptive information concerning the independent variables can be gathered from Table A2 in the Supplementary File and the question wording is also documented in the Supplementary File's Table A3.

Empirical Analysis
Following the logic of hierarchical modelling, we present a series of models, where each builds on the preceding one. Starting with the empty model, the share of cross-national variance on the total variance amounts to 14.6 percent (turnout) and 21.8 percent (vote for a Eurosceptic party). The estimates also indicate that turnout and the vote for a Eurosceptic party depend on both individual-and context-level indicators. In a further step, we consider the effects of our core individual-level determinants of vote abstention and vote for a Eurosceptic party. The results of the individual models are presented in Table 1 via stepwise models: M1 displays the effects for egocentric considerations only; M2 displays the effects for sociotropic considerations only; and M3 combines the two models into one final model. The coefficients in M1 confirm the impact of egocentric considerations due to lack of resources (H1) and can be interpreted as follows: Egocentric considerations decrease electoral participation (H1a) and raise the likelihood of voting for a Eurosceptic party (H1a). Moreover, viewed through the lens of the notion of sociotropic voting (M2), our results are interesting in two ways: First, a respondent with negative current perception and/or future evaluation of the economic situation is less likely to cast a ballot in the EP elections; second, a negative current perception and/or future evaluation of the economic situation also increases the likelihood of voting for a Eurosceptic party (H2). In summary, our results thus support the idea of both sociotropic and egocentric voting.
For ease of interpretation, we now include our individual-level control variables in the subsequent model. The key message of the results in Model 3 in Table 1 is that under the influence of relevant individual characteristics, the effects of the economic indicators are weaker than in Models 1 and 2 and sometimes no longer significant, albeit without changing substantively. Put simply, the economic indicators continue to exert an independent effect, whereby the loss of a job and problems in paying bills reduces the likelihood of electoral participation. This pattern clearly confirms our first hypothesis (H1), showing that egocentric voting has a sig-  (27) Notes: Multi-level logistic regression. Coefficients are not standardised; all variables are rescaled (lowest value is 0; highest value is 1). Coefficients indicate the change associated with moving from the lowest to the highest value. Significance level: * p < 0.05; ** p < 0.01; *** p < 0.001. Source: .
nificant impact only on electoral participation, but not on the vote for Eurosceptic parties. In contrast to this observation, a negative future evaluation of the economic situation increases the likelihood of voting for a Eurosceptic party which in turn confirms our second hypothesis (H2), that the mechanism of sociotropic voting is relevant in the case of voting for Eurosceptic parties. Finally, the effects of the control variables on electoral participation and the vote for a Eurosceptic party are in line with previous research. Perception of responsibility on the part of the EU regarding the economic situation has a positive effect on electoral participation, but the effect on party vote is statistically not significant. Satisfaction with the EU also promotes turnout, whereas being dissatis-fied with the EU and the national government fosters voting for a Eurosceptic party. The findings confirm the idea that national politics has an impact on voting behaviour in EP elections, whereby higher political interest, education, and party identification have a positive effect on turnout.
In the next step, we include the context-level variables in our model in order to detect whether the crisis also has a direct impact (H3), in addition to the effects via the individual level already shown. At this point, we distinguish between the current economic situation and a dynamic perspective. Initially, we consider the indicators separately in the models, i.e., we estimated 24 different multi-level models, one model for each indi-cator. Individual variables are always included but not shown for reasons of clarity (compared to Model 3, the results of the individual-level indicators remain stable). The results of the multi-level regression are presented in Table 2. Two crisis indicators show statistically significant effects. We can see that the greater the national debt, the more citizens cast their ballot in EP elections. A fast economic recovery after the economic crisis-operationalised by the change in GDPreduces the likelihood of a Eurosceptic voting decision. With an eye to the control variables at the contextual level, our findings suggest that compulsory voting and a longer membership in the EU increases turnout, whereas a communist heritage reduces electoral participation. Moreover, in Eurozone countries, the probability of voting for a Eurosceptic party is somewhat lower than in non-Eurozone members.
In a final step, we merge the significant context-level indicators into one multi-level model (M5; Table 3). The results of the multi-level regression show that compulsory voting has a positive effect and a communist heritage has a negative effect on electoral participation. National debt no longer has a statistically significant effect. In terms of the vote for a Eurosceptic party, we find two statistically significant effects: First, a fast economic recovery after the economic crisis reduces the likelihood of a Eurosceptic voting decision; and second, the likelihood of voting for a Eurosceptic party is lower in countries of the Eurozone compared with non-Eurozone countries. To sum up, our final results corroborate that there is indeed a direct relationship between the financial crisis and its consequences and electoral behaviour (H3). Nevertheless, this holds true only in the case of the vote for a Eurosceptic party, but not for electoral participation. What does this mean substantively? First, in cases where a country is a member of the Eurozone, the likelihood of voting for a Eurosceptic party is lower. Second, the faster the economic recovery of a country, the less inclined are respondents to vote for a Eurosceptic party.

Conclusion
The financial crisis which turned quickly into a global economic crisis was followed by severe political consequences. This led to a certain transformation of the EU multi-level system, but also the dimension of EU politics. Our study adds to a growing literature on EU politics illustrating that not only citizens' perceptions towards the   (27) EU but also different types of electoral behaviour in EP elections have been and are still being affected by these turbulent times, to be more specific the EU which is currently under stress. While most of the conventional determinants still hold true, our findings most importantly show that the crisis and its implications indeed have an independent effect on electoral behaviour as the crisis has provoked EU citizens to participate to a lesser degree in the 2014 EP elections and increased the vote for Eurosceptic parties. Moreover, our empirical study enables us to get a better idea of the precise mechanisms in play. As hypothesised, lacking resources and therefore the fact that the individual life of a person is affected by the crisis-a phenomenon which we labelled 'lack of resources egocentric non-voting' (H1)-indeed has an impact on electoral participation, whereas these egocentric considerations do not affect the decision to vote for a Eurosceptic party. In contrast, the mere perception of worsening economic conditions, i.e., the idea of 'sociotropic' voting (H2) as well as some objective indicators of the crisis (H3) affect the decision to vote for a Eurosceptic party, but not the non-participation in the 2014 EP elections. What does this pattern tell us? A citizen may be individually affected by the crisis and its eco-nomic and political consequences, but does not (automatically) vote for a Eurosceptic party. Instead of voting for a Eurosceptic party, in that case he or she decides not to vote at all. Moreover, the 'mere existence' of the crisis in a country as well as its perception-but not necessarily whether an individual feels personally affected by the crisis-heightens the propensity to vote for Eurosceptic parties. This could indeed be interpreted as an intensification of a broader trend towards destabilisation within the European party systems (Hernández & Kriesi, 2016). Instead of voting for established parties in times of economic hardship, voters make their choice for parties outside the mainstream since they have lost their faith in the main actors of the political system. The findings of our study are not only of interest for scholars of voting behaviour, but will also further the understanding of EU multi-level politics. To gain a different perspective beyond the findings reported above, Figure 5 maps the entire process from a multi-level point of view. A European-wide event, the financial crisis followed by the Euro currency crisis, hit most of the member states, but not all of them equally hard. A few years later, the 2014 EP elections took place and, as our findings indicate, this crisis and its implications affected  individual-level electoral behaviour (see in more detail the introduction to this thematic issue, Braun, Gross, & Rittberger, 2020). To be more specific, the national economy, affected by a European crisis, is related to the outcome of the EP elections. Although these elections take place at the European level, they are widely viewed as being second-order national elections (Reif & Schmitt, 1980)-and this second-order character still holds true for the 2014 EP elections . This means that EP elections might be different from national elections since they produce different results and always need to be interpreted with an eye to national elections. Nevertheless, in the end the results are meaningful for the two electoral levels: the EU level since the composition of the EP represents the result of the EP elections; and the national level as the EP elections are always to a certain degree the test balloons for national first-order elections. The lesson we can thus draw from our study beyond the undoubtedly interesting mechanisms in operation is that electoral behaviour in the EU multi-level system is far from independent from the different levels in play. Accordingly, each of these levels needs to be taken into account when investigating electoral behaviour (but see also Golder, Lago, Blais, Gidengil, & Gschwend, 2017;Reif & Schmitt, 1980;Schmitt, Sanz, Braun, & Teperoglou, 2020) or party politics in the EU (Braun & Schmitt, 2018). The 2019 EP elections should provide scholars of EU politics with a fertile resource for appropriately studying these and related mechanisms regarding electoral and party politics. Nonetheless, we do not have any reasons to expect a completely different pattern for each of the described mechanisms in the 2019 EP elections. The slightly higher levels in turnout could be interpreted by a decrease in 'egocentric' considerations whereas the unabated high levels of Eurosceptic vote should still be related to 'sociotropic' considerations.