When and where did the great recession erode the support of democracy?

It is likely that ten years of economic crisis have eroded the support of democracy in Europe. But how much? The existing research is divided on this issue. Some claim that the degree of satisfaction with democracy has declined across the whole of Europe during the Great Recession. Other researchers have found no empirical evidence that the support of democracy as a core value has declined across Europe. They claim that merely the specific support has decreased in some countries. This article will use the data from the European Social Survey to verify both claims. It shows that the Great Recession did not lead to a legitimacy crisis of European democracies and that the diffuse support of democracy remains high in most regions. The degree to which the specific support of democracy has been weakened is moderated by the type of welfare regime. In countries where the economic crisis did strike hard and the welfare state is weakly developed, the support of democracy has dropped dramatically. This outcome takes a middle position between two extremes in the ongoing academic debate on the support of democracy. Both positions regarding the increase or decrease of support of and satisfaction with democracy are in need of more nuance by taking into account the impact of welfare regimes. Existing research often assumes a uniform European context that shows either increasing or decreasing levels of satisfaction with democracy. Our research has shown that the response of citizens to the Great Recession has been influenced by the welfare regime.


Introduction
During the last decennium many articles have been published that examine the linkage between the prolonged economic crisis within the Eurozone after 2007 ("Great Recession") and the support of and satisfaction with democracy. In many Western countries the crisis involved two or more consecutive quarters of GDP contraction resulting into ongoing hardship as a result of austerity programs and growing unemployment. Often, the tone was quite cumbersome: it was signalled that satisfaction with government and democracy was declining and that even the support of democracy as a political system was weakening. After ten years of economic crisis we may ask to which extent this pessimistic conclusion is (still) justified given the most recent available data.
The Great Recession has had not only financial and economic consequences, but also political ones. It has weakened the problem-solving capacity of democratic systems which increased dissatisfaction with politics and performance and the trust in politicians in many established democracies. Examples are the Southern European democracies which have been hit hard by this crisis (according to the Economic Crisis Index which is calculated as: GDP Growth -Unemployment + Deficit-Surplus, see Kriesi 2013), but also a number of Eastern European countries like Hungary in which voters are increasingly dissatisfied with the functioning of party democracy. In other countries, the support for democracy has not changed, like in Sweden and Germany. These variations within Europe underline the importance of taking into account the national and regional context that impacts on the type and degree of dissatisfaction. Particularly important are the different types welfare regimes of these regions that may strengthen of weaken the effects of the economic crisis on the support of democracy. By focusing on the moderating effect of welfare states on the support of democracy, this research distinguishes itself from most other work on this topic that neglects the role of welfare regimes. The Great Recession urges for insights into the recent developments in the satisfaction with democracy and the trust in politics and politicians. Does the Great Recession gradually undermine the support of democracy among certain groups in the electorate? In the literature we find different answers to this question. Since these findings are conflicting they are in need of a reconsideration now that the Great Recession has lasted for ten years: to which degree did it erode satisfaction with democracy in different welfare state regimes in Europe? The research seeks to answer this question by analysing the degree of satisfaction with democracy and government on the basis of data of the European Social Survey (ESS). The results show a mixed situation in which there is an increase in some countries and a decrease in other countries. There is a systematic correspondence between the level of support of democracy and the type of welfare regime. The variation in welfare regimes is relevant for the degree to which the Great Recession erodes the support of democracy.
One of the main reasons that the literature arrives at different conclusions regarding the support of democracy is that different research designs have been used. The choices which are made in the research design (such as the case selection, type of data, data sources, time period, operationalisation) and the theoretical and normative propositions all have an impact on the findings. In order to limit the chance of biased results, our research design includes ample variation in the dependent and independent variables, selects all relevant cases and uses data on several aspects of democratic support that enable both a cross-sectional and a longitudinal analysis of the patterned variations.

Theories and assumptions
What is support of democracy? In this research we focus on the effects of the economic crisis on the democratic legitimacy of European democracies. There is a formal and informal aspect of the support of democracy. The distinction between "formal support" and "informal support" is needed in order to explain why there is variation in support for democracy in countries that are equally democratic (from a formal point of view). The formal definitions refer to 'rightfulness' of the legal exercise of public authority. Such legitimate exercise of political power is bound to several criteria (Beetham 1994). First, power should be exercised within a set of established responsive and accountable institutions that help to arrive at an agreement in case of conflicting ideas or interests. Second, institutionalized power must be open for control and correction by means of the Rule of Law and forms of checks and balances. These rules and mechanisms secure the political and civil liberties of the individual and of minority groups (Stone Sweet 2008).
The informal definitions emphasize the importance of consent of the people for legitimacy, also called 'willing obedience ' (Levi et al. 2009, p. 355). Consent does not necessarily mean agreement, but more importantly the willingness to accept decisions and to use the channels offered by the democratic system to express (dis)content, for example by means of voting or protesting. Since support of democracy is based on consent of the people, it is not static but may change over time as a response to the varying economic and social performance of governments. A decline of support of democracy may lead to instability of a political system due to decreasing trust and support. This may have far reaching consequences for the ways in which elites and masses interact. If authority is accepted by citizens as just there is no need for much coercion. Instead there is compliance with public regulation and enacted governmental policies in a non-coercive authoritative manner. This enables the government to steer economy and society without much force (Beetham 2013).
Since support of democracy is multi-facetted, most definitions put a different emphasis on the formal and informal aspects. Some emphasize the formalized procedures to safeguard liberties and constraining public contestation, others tend to stress the need of popular participation and (fair) representation of interests and regions. Beetham (2013), for example, argues that an authority is considered legitimate if its exercise of power is established and exercised in accordance with legal rules, justified by shared beliefs of the population, and acquired through the consent of citizens.
Legitimacy is mainly found in (liberal) democracies which are characterized by the rule of law and civil and political rights of the individual (Dahl 1989). This is the formal situation. In practice there may be ample variation in the degree of legitimacy between these countries due to different degrees of popular acceptance and consent with the political regime. This variation implies that perceptions and trust are important for the evaluation of the legitimacy of democratic systems. If citizens are disappointed or disagree with policy outcomes, but at the same time feel that the elites and institutions are functioning well, discontent will not undermine the legitimacy of the system (Levi et al. 2009). Often this situation is more complex than either being supportive or not supportive of the democratic system. If large parts of the population do support the system, but hardly participate, there is a legitimacy problem. If there is support, this should be expressed by political behaviour like voting. If there is discontent, like during an economic crisis, this should activate citizens to protest against the austerity policies. If there is an increase in disaffection K and this is translated into protest, the legitimacy will be less affected then when disaffection goes hand in hand with apathy and lack of interest. These different ways of expressing support imply that the impact of the economic crisis on the legitimacy of advanced democracies in Europe may vary. It may be at danger in particular regions or countries, but it may also remain largely unaffected by the crisis in other countries (Hernández and Kriesi 2016). It is not only the impact of the crisis itself that matters, but also the behaviour of international actors (such as international financial institutions and the EU) and of national actors (such as parties and interest groups) and the institutional infrastructure (such as the type of welfare regime) that affect the capacity of governments to cope with the economic crisis.
It is not obvious how to conceptualize the consent of the people. Often it is conceptualized as support of the democratic system and the government. Easton (1975) has made the distinction between specific and diffuse support of democracy. Specific support refers to the support and evaluations of the perceived decisions, policies and actions of actors and institutions. Diffuse support refers to a more common feeling of 'generalized attachment' to the democratic regime (Easton 1975, p. 444). Specific support relies on the evaluations of particular institutions, parties or leaders and the perceived performance of policies. It is based on a rational calculation of costs and benefits that are derived from evaluations of performance. Diffuse support, on the other hand, is the firm belief and trust of citizens that the democratic institutions are put in place on their behalf and are basically part of their way of living. Whereas specific support is inherently volatile, diffuse support is more stable as its determinants are related to political values and attachments. Diffuse support is important for the stability of democratic systems because it helps citizens to accept decisions to which they are opposed (Easton 1975). This does not mean that diffuse support will remain stable in all circumstances. It may change due to sudden shocks, a deep crisis or continuous malfunctioning of the system (indicated by stalemates, cabinet crises, polarisation etc.). The economic crisis starting in 2008 is certainly a development that might have shocked both specific and diffuse support of national democracies. The degree of economic misery matters for citizens' trust and evaluations of actors and institutions (Armingeon and Guthmann 2014). More misery often implies more dissatisfaction which may (ultimately) affect legitimacy. Many researchers have shown that people in democratic regimes tend to support democracy when governmental performance is good (Magalhães 2014;Keman 2014). The core argument is that discontent with the way the regime works and/or its policy outcomes erodes the citizens' support of it (Pharr and Putnam 2000). Thus, frustration with government performance and/or an economic recession can affect the core support of democracy (Cordero and Simón 2016).

Hypotheses
Since specific support is closely linked to government performance, it is likely that this type of support is most affected by the economic crisis (H1). The diffuse support is expected to be less affected because the non-democratic alternatives for democratic regimes are by most citizens perceived to perform much worse than democratic systems given the circumstances (as is argued by Easton's theory of political support (Easton 1975) (H2)). Diffuse support is mainly affected in countries where the crisis did strike relatively hard (indicated by the Economic Crisis Index = GDP Growth -Unemployment + Deficit-Surplus, see Kriesi 2013, p. 307) and where governments imposed far-reaching restrictive policies that meet strong resistance among interest groups and the population at large.
In addition to support of democracy, the trust in politics and politicians is important, i. e. the belief that political authorities and institutions do not seek to benefit themselves but produce outcomes that benefit large parts of the population (Newton 2008;Sztompka 1999). The concept of political trust is multi-facetted. It can be divided into forms like thick, interpersonal, and systemic or institutional trust (Newton 2008). Given these diverse forms of trust there is overlap with the concepts of specific support and diffuse support. The main commonality is that trust links ordinary citizens to institutions that represent them (Fagerland Kroknes et al. 2015). If citizens trust these institutions this will enhance both the legitimacy and the perceived effectiveness of democratic government. Both political trust and diffuse support are essential for the stability of democracy and both are a basic prerequisite for the legitimacy of those being in power. If trust in political authorities and institutions is declining this may endanger their legitimacy since low levels of trust hamper the government to perform its tasks because citizens are less likely to obey the law or to accept decisions. Some degree of distrust towards political authority is healthy for democracy because it puts a pressure on elites to perform better (Sztompka 1999;Nye 1997). However, an extended period of distrust could have negative consequences, especially if the reason for distrust is corruption. The trustworthiness of political authorities is important for the legitimacy of elites because the democratic process of representation assumes that people trust that political authorities are honest and not corrupt and that they will deliver on their promises (Beetham 2013). Economic prosperity has been identified as especially important in this regard because it signals the degree to which elites do deliver as promised (Keman 2014). The more the economic crisis has had a negative impact on people's lives, the more they may lose trust in their national parliament, which happened most pronounced in Ireland, Spain, Portugal, and Greece (Roth et al. 2011). In order to assess the impact of the crisis on citizens we need to take into account the socio-economic differences, in particular education. The higher educated will remain more satisfied and will keep more institutional trust than the lower educated (Schäfer 2013). We hypothesise that the Great Recession will lead to lower levels of trust, in particular of the lower educated and in countries with that are hit hard by the crisis (H3). Some research has signalled an increase in dissatisfaction and a decline in political trust (e. g. Armingeon and Guthmann 2014;Dalton and Wattenberg 2000). There is one important exception. If austerity measures are imposed on governments by external actors like the EU and the IMF, governments have far less room to manoeuvre to cope with the crisis. In that case the blame is likely not on the government but on the external actors. We hypothesize that countries in which austerity measures are imposed by external actors the support of the national system may increase (H4). This statement is not supported by all researchers. Some authors agree that the support of democracy is gradually weakened, but this trend is unrelated to the Great Recession because politics itself is becoming more and more discredited due to the K acceptance of neo-liberal ideas by both parties and voters that do not regard the state as competent problem-solver (Hay 2002). Other researchers state that democracy is not at stake at all because parties succeed in fulfilling their functions in ways that congruent with the median voter (Budge et al. 2012;Sanders et al. 2014;Dalton et al. 2011). Finally, a group of researchers arrives at a more nuanced conclusion that the trends are not uni-directional. The cross-national and longitudinal variations show examples of both rising and declining satisfaction with governments and democracy (Norris 2011;Thomassen 2015). According to them there is no systemic decline in the degree of satisfaction with democracy, neither before, during or after the economic crisis. On average the levels of satisfaction remain at a high level.
There has been a long-standing debate in the academic literature on the different types of welfare regimes in Europe and beyond. Although there is no consensus on these types, it is generally accepted that most countries in the main regions of Europe have similar welfare state regimes. This is reflected in the names which are being given to these clusters, like Nordic welfare states, Southern welfare states, East European welfare regimes etcetera (see for an overview of the debate on welfare regime differences: Arts and Gelissen 2002). These countries resemble each other in the degree of universalism, egalitarianism and de-commodification (Esping-Andersen 1990). Countries belonging to the same welfare regime have in common that they provide a similar type and degree of welfare to citizens. These differences matter for the degree of social inequality and the support of democracy since the more social protection is provided, the less threatening are the effects of the economic recession (Albano and Barbera 2010). The more gentler and kinder a society is regulated and the more welfare and well-being is achieved, the more legitimate governance is expected to be.
We do not stand alone in making this causal claim. In the literature on welfare states it has often been argued that the rise of social provisions and services intends to compensate for the risks and inequalities that are involved in capitalism (Van Kersbergen and Manow 2014). Moderate levels of inequality are necessary for democracy to thrive. Democracy and the welfare state have become intertwined. Most established democracies are also advanced welfare states. Citizens living in welfare states wholeheartedly support democracy because it offers the conditions for safeguarding the level of social welfare and income, especially in times of economic crisis. Now that economic growth has slowed down or even halted or decreased, we expect that this has affected the satisfaction with democracy. It is assumed that dissatisfaction with democracy is growing relative fast in countries that are hit hard by the crisis and in which the welfare states cannot compensate for that. Dissatisfaction will not increase in countries that are hit less hard by the crisis and/or in which the welfare state is able to provide sufficient levels of welfare, work and income for all social groups despite the Great Recession.
Legitimacy is thus systematically related to the variations in socio economic performance and the type of welfare state. Most countries with similar welfare states have similar levels of social and economic welfare, cultural heritage and traditions and phase of democratic development (Bodor et al. 2014). The vast amount of literature of welfare state classifications provides an excellent source of the main socioeconomic similarities and differences between these regimes (see for an overview: Arts and Gelissen 2002). These characteristics may make countries more or less vulnerable to the effects of the Great Recession. The Nordic countries are expected to have the highest developed welfare regimes and levels of socio-economic welfare and are least affected by the crisis. The Southern countries are expected to have an intermediate level of collective welfare provisions and are strongly affected by the crisis. The Eastern countries have the lowest level of socio-economic welfare, but the degree to which they are affected by the crisis varies (see for an overview: Honkapohja and Korhonen 2013). The continental countries have intermediate levels of socio-economic welfare and most of them are modestly affected by the crisis. Finally the two countries with liberal welfare regimes (the UK and Ireland) have been hit relatively hard by the crisis (Kriesi 2013). We hypothesise that the type of welfare state and socio-economic performance have a moderating effect on the degree to which citizens satisfied with democracy during the Great Recession (H5). This does not mean that the welfare regime approach offers the only explanation of the support of democracy. Alternative explanations do exist like those that stress the impact of austerity policies by the EU, the IMF and the ECB (e. g. Armingeon and Guthmann 2014;Cordero and Simón 2016). In addition there are approaches that stress the individual-level variation within welfare states (see for an overview of this argument: Kumlin 2002). These alternative explanations will be examined by incorporating the most relevant indicators into the analysis (namely whether a country is bailed out and EMU-membership).

Data and operationalisation
We divide Europe into five regions that each have similar welfare regimes: Nordic, Continental, South, East and Anglo-Saxon (Ferrera 1996). 1 There is a longstanding ongoing debate on the proper operationalization of welfare state regimes. It started with Esping-Andersen who introduced three types of welfare states consisting of Social Democratic, Conservative and Liberal regimes (Esping-Andersen 1990). In his typology Italy is part of the family of the corporatist welfare state regimes, whereas Spain, Portugal and Greece are not included in the analysis. This has been criticized by many authors who have pointed at the clientelistic character of the Southern regimes (Arts and Gelissen 2002). According to many commentators Southern European countries form a separate cluster of welfare regimes (Ferrera 1996;Bonoli 1997). A similar argument also applies to Eastern Europe. This cluster is not included in Esping-Andersen's typology, whereas according to many observers these countries form a distinctive cluster that does not fall into this trilogy (Aidukaite 2009;Kaariainen and Lehtonen 2006). The East European or post-communist regime has similarities with both the liberal and conservative corporatist welfare regimes 1 The countries are part of five regions which are Scandinavia (Norway, Finland, Denmark, Sweden), Anglo-Saxon Europe (Ireland, United Kingdom), Continental Europe (Switzerland, Belgium, Netherlands, France, Germany), Southern Europe (Portugal, Spain, Italy) and Eastern Europe (Estonia, Slovenia, Slovakia, Bulgaria, Hungary, Czech Republic, Poland). Excluded are the smallest countries (Iceland, Cyprus, Luxembourg) and the most peripheral countries (Turkey, Ukraine). Greece is part of the longitudinal analysis, but not of the cross sectional analysis because it did not participate in the sixth wave of the ESS. as well as some distinct post-communist feature such as a high take-up rate of social security but relatively low benefit levels; the identification of the social policy system with the Soviet past, and a low level of people's trust in the state institutions (Aidukaite 2009). Thus, it could be argued that although the East European region is very diverse, it has similar macro-level institutional an cultural features that justify the clustering of these countries into the ideal type post-communist regime.
Since support of democracy is multi-facetted, it is measured by a variety of indicators. Norris (2011) has distinguished five levels of political support with accompanying empirical indicators. We will focus on four of them, namely regime principles (democracy better than any other form of government), regime performance (satisfaction with the functioning of democracy), political institutions (trust in political institutions) and political authorities (e. g. satisfaction with incumbent government).
To find out whether the crisis has weakened satisfaction and support, we need information on the longitudinal variations in order to cross-validate the cross-sectional analysis. Since many empirical studies contradict each other on the degree to which there is a decreasing support of and trust in democracy and government (Hurrelman et al. 2005) it is important to adopt an inclusive research design with ample variation between countries and time periods. The required data are available in the European Social Survey (ESS) which includes questions on the satisfaction with democracy and government. Some governments are better able to deal with the economic crisis than others. This variation can help to understand why the Great Recession has different effects on satisfaction of citizens living in European welfare regimes.
Specific support is operationalized by means of two indicators: the degree of satisfaction with democracy (stfdem) and the degree of satisfaction with government (stfgov). These are indicators of the degree of satisfaction of citizens with the way democracy works and how the government is doing its job at the moment of the Survey. Satisfaction with government is a relevant indicator since legitimacy is affected by the goals and achievements of the government (Keman 2014;Beetham 2013).
The diffuse support is operationalised by institutional trust which is measured by the mean of the values on three variables: Trust in Parliament, Trust in the legal system and trust in the police. These institutions represent the rule of law and the democratic representation of the will of the people. Trust is the second indicator for diffuse support as existing research has shown that it is both conceptually and empirically strongly related to the structural support for democracy as a systemic feature of the political system in which citizens live (Kumlin 2002).
Although the indicators for specific and diffuse support have been criticized because they do not mean exactly the same across countries and social groups (Ferrín and Kriesi 2016), existing research shows that they are distinct concepts that do relate to the more specific (or ad hoc) and the more systemic aspects of the functioning of democracies (e. g. van Van Beek 2010; Norris 2011). Fig. 1 shows the relationship between the change in GDP per capita (Worldbank, 2007(Worldbank, -2015 and the change in the support of democracy (ESS, 2006(ESS, -2012. The pattern shows a modest positive relationship between the two variables (r = 0.541, . Most countries enjoying a growth in GDP also show a growth in the support of democracy. Most countries that are confronted with a decline in GDP can be divided into two groups. In one group this decline coincides with a decline in the support of democracy (Italy, Spain. Portugal, Slovenia, Ireland). Several of these countries have been hit hard by the crisis and have been bailed out. In the other group there is no decline in the support of democracy despite a decline in GDP. The plot also shows whether countries are clustered regarding their belonging to a particular welfare state regime. Such clusters do exist as most countries belonging to the same regime are positioned in the same quadrant. The plot shows a continental cluster (Switzerland is the only exception due to its high GDP-growth), a Nordic cluster, an Eastern cluster (Slovenia is the only exception due to its decline in GDP). The three Southern countries which are included in the respective ESS-rounds also fall into the same quadrant. The plot confirms the basic assumption made earlier that the impact of economic growth on the support for democracy is shaped by the type of welfare regime.

Trends and cross-sectional variations
One major claim made in some recent research articles is that "Support for national democracy" -operationalised as satisfaction with the way democracy works and as trust in parliament -declined dramatically during the crisis (Armingeon and Guthmann 2014). However, other scolars argue that there is no such trend towards less satisfaction with democracy across the whole of Europe (Thomassen 2015). In addition they argue that dissatisfaction is not necessarily related to economic problems. It can also be a sign of increasing emancipation of critical citizens which one would expect in the most advanced democracies (Norris 2011). The values are differences in averages of being either satisfied (scored 1) or dissatisfied (scored 0) between the years 2006 and 2012. The distinction between satisfied and dissatisfied is based on the median score. The F-score is based on the aggregated scores at the country-level and indicates the ratio between the within-group variation and between-group variation. i.e. how diverse the countries within the regions are. All variables (except "Institutional Trust") are ordinal and range on a scale from 1 to 10. The measurement level of Trust is interval and ranges from 0 (no trust) to 10 (complete trust). The cases are weighted by the design weights (DWEIGHT) and the population size weights (PWEIGHT) in the ESS data. Appendix 1 includes additional information on the within and between country variation (ANOVA analyses) The explorative analysis focuses on the cross-sectional variations in the dependent variable, i. e. specific and diffuse support. The goal of this analysis is to find out whether regions with different welfare regimes are characterised by different levels of support for democracy. Since there may be differences between the countries per region, another goal of the analysis is to determine the within-region and betweenregion variation. If the between-region differences are much stronger than the withinregion differences this supports our argument that regional differences matter due to different welfare regimes. This is reflected by the F-score in Table 1 which is based on the aggregated scores at the country-level and indicates the ratio between the within-group variation and between-group variation, i. e. how diverse the countries within the regions are. The F-scores confirm our assumption that the between-group variation is larger than the within-group variation. Table 1 shows the change in the average support of democracy between 2006 and 2012 on the basis of three indicators. The values are differences in averages of being either satisfied (score 1) or dissatisfied (score 0). The distinction between satisfied and dissatisfied is based on the median score. Negative signs indicate that there is a decrease in satisfaction. Overall the table shows that there is no general decrease in satisfaction as is often assumed in the literature. Table 1 shows that the specific support has been strengthened in all regions except in the South. The South has been most affected by the economic crisis (H1) since the satisfaction has dropped from 0.56 to 0.27. A similar pattern is shown in case of the satisfaction with government. The indicator of diffuse support, institutional trust, shows a small rise, except in Southern Europe where it drops from 0.52 to 0.39. These results confirm hypothesis 3 which assumes that the Great Recession will lead to lower levels of trust, in particular in countries with that are hit hard by the crisis and/or with a lean welfare system.
There are significant regional differences in the satisfaction with government and democracy. In the Nordic countries the satisfaction is around 0.70 to 0.80. This situation is exceptional. The relatively high level of social welfare is likely to contribute to this. In the Liberal countries the degree of satisfaction is much lower, namely around 0.50. There is not much increase in satisfaction in these countries. In the continental countries the levels of satisfaction are similar to those in liberal countries and they are also rising. Within this region there is a high degree of satisfaction with government in Switzerland and a low level of satisfaction in France. Apparently the satisfaction with governments in majoritarian systems (France, UK) is relatively low. In Southern Europe the satisfaction is at the lowest level and also the degree of trust is relatively low in 2012. This regional pattern indicates that the depth of the economic crisis plays an important role in the degree of satisfaction in combination with the external intervention by the EU and the IMF which has taken place (to varying degrees) in Southern European countries. The observation by Cordero and Simón (2016) that, as a reaction to these interventions, there is a rise in satisfaction with national democracy in this region is not confirmed.
In Eastern Europe the satisfaction with government and democracy in 2006 is much lower than in Southern Europe, but in 2012 it has become higher. The trust in Eastern Europe is lower than in Southern Europe in both years. These patterns reveal that both the change and the degree of satisfaction and trust are related to regional variations in the intensity of the economic crisis, the type of welfare state and the phase of democratic development (Lijphart 2012). Although there is ample variation within the regions (as is specified in Online Appendix 1), the differences between the regions outweigh the differences within the regions (which is indicated by the F-statistic). In 2006 the between-group variation is rather low and barely significant. In 2012 the between-group variation has increased significantly. This means that the differences between the regions in the degree of satisfaction and trust have become more stronger. The main cause of these increased differences is Southern Europe that has encountered a dramatic decrease in satisfaction with democracy and government and of institutional trust. Finally Table 1 confirms Lipset's expectation that specific support is more volatile than diffuse support as the changes in diffuse support between 2006 and 2012 are relatively low compared to the changes in specific support.
In general, the patterned variations in Table 1 reveal that the type of welfare regime matters for the degree of satisfaction with government and democracy (H5). We may roughly distinguish between two groups, namely the South and East of Europe in which the levels of trust and satisfaction are lowest (in particular in 2012) and Nordic and Continental regions where these levels are relatively high (both in 2006 and in 2012). Southern Europe is unique in the severity of the economic crisis, whereas Eastern Europe witnesses an exceptional low level of trust among its citizens. In the scholarly literature the low levels of trust in Eastern Europe are linked to the presence of corruption and the relatively short duration of democratic development. Additional analysis (reported in Online Appendix 2) shows that in Eastern countries the lowly educated have (on average) nearly as much trust as the highly educated and the decline of trust of this group is lower than in case of the higher educated. Hence, in case of Eastern Europe having more knowledge of the system coincides with having less trust. In this respect Eastern Europe differs from the other regions. One explanation might be that most Eastern European democracies are still developing so that their deficiencies and inequalities are obvious for the highly educated observer (Schäfer 2013). Table 1 confirms our central argument that there are regional differences in the degree and direction of change in dissatisfaction and trust. If we examine these trends on a year by year basis (Online Appendix 3) it turns out that the increase of dissatisfaction often started well before the Great Recession. This indicates that it is not the crisis that leads to distrust or disaffection, but the different socioeconomic and institutional conditions in the European regions lead to systemic differences in the satisfaction with and trust in democracy. Most prominent are the differences in welfare statism and democratic performance. If countries are hit by the crisis, but they can cope with the consequences using social safety nets and by means of accountable policy measures, the effects on satisfaction are weak or even absent. But if such safety nets are absent or partly inadequate and the accountability of government is low, the effects of the crisis are felt much stronger and hence disaffection with democracy will be higher.
Of course, one may wonder which aspects of democracy so many citizens are negative about. The sixth wave of the ESS has a included a battery on questions about the degree to which the democratic system is performing well on a number of criteria like freedom of speech, minority protection, accountability and policy outcomes like reducing income differences. These items enable us to get more specific information about which aspects of democracy are underdeveloped according to the respondents. Of particular importance are income inequality and poverty which affect the degree to which citizens do trust and support public institutions (Anderson and Singer 2008). A simple descriptive analysis (reported in Online Appendix 4) turns out that economic government policies (i. e. reducing income inequality and fighting poverty) in this respect are regarded by 54% of the citizens as being very important for democracy in general. There are significant regional differences in this respect. In Southern and Eastern Europe 67% of the respondents who are dissatisfied with government find this important, whereas in Western and Northern Europe this is only 51%. In case of the respondents that are satisfied with government these percentages are 58% and 42% respectively. Similar patterns do also apply to satisfaction with democracy which is also lower in case of respondents that find it important that government fights poverty and reduce income inequality.
The results so far indicate that the main problem that most respondents have with the functioning of democracy concerns the lack of accountability (input-related) and the poor performance on achieving equality, in particular the degree to which the government protects all citizens against poverty (output-related). Thomassen (2015) has accounted for this pattern by claiming that the minority of highly educated citizens care most about input whereas the majority of low educated citizens care more about output. This pattern is partly confirmed if we compare the lowly educated with the highly educated on items that are input-related and items that are outputrelated. It turns out that the lowly educated are less satisfied with the output features of democracy. An example is this item: "In country the government protects all citizens against poverty" with a mean score 3.96 for lowly educated and 4.44 for highly educated (a higher score indicates more satisfaction). The highly educated are less satisfied with the input features which is indicated for example by this item: "In country national elections are free and fair" with a mean score of 2.60 for lowly educated and 2.27 for highly educated. However, the group differences in both examples are rather small. This is an indication that the fear of some scolars, like in the debate on the relationship between democracy and capitalism (Merkel 2014), that the Great recession has led to a larger divide between social strata and their satisfaction with democracy is not fully justified.

Multi-level analysis
The explorative analysis has described the variations in the support of democracy. The next step is to explore the causal connections between support of democracy on the one hand and some potential determinants that have been discussed in the theoretical section. The focus will be on whether the change in economic growth (due to the economic crisis) has affected the support of democracy. Fig. 2 shows the relationship between the change in GDP in the period 2007-2015 and the degree of satisfaction with democracy for all countries within each welfare regime. The pattern indicates that the level of satisfaction with democracy is not affected by economic growth (r = 0.11, p = 0.630, n = 21). How economic growth impacts on satisfaction mainly depends on the type of welfare state. A decline of economic growth solely leads to less support of democracy in case of particular welfare regimes. Economic decline (i. e. a score lower than zero) corresponds with a high level of satisfaction in Nordic countries and with a low level of democratic support in Southern and Eastern countries. A similar pattern also applies to economic growth. A high level of economic growth (i. e. a score higher than zero) coincides Fig. 2 The relationship between satisfaction with democracy (Y) and change in GDP per capita (X) per country and type of welfare regime K with a low level of satisfaction in Eastern Europe and with a high level of democratic support in Switzerland. Although there is some variation within the regions, the main pattern shows that the welfare regime moderates the impact of the economic recession on satisfaction. This effect goes in two directions: an economic decline does not translate into less satisfaction (like in the Nordic countries) and an economic growth does not lead to more satisfaction (like in the Eastern countries). The rationale is that economic growth solely leads to more social and economic welfare for all citizens if the welfare regime effectively redistributes wealth. If the regime lacks the capabilities to do this, economic growth will not benefit the mass public that consequently will not express more support for the democracy that they live in. Economic decline will mainly lead to less support for democracy if the welfare regime is not capable to compensate for the economic losses by means of social transfers and services. If there is economic decline and the welfare regime is able to offer adequate social protection to all citizens, the support for democracy will not be affected by the economic problems or will even increase. In sum, the relationship between economic tide and support of democracy depends on the type of welfare regime. The great recession did erode the support of democracy in case of lean welfare regimes that were not capable to compensate the citizens for the financial losses.
Until now the explorative analysis has shown the patterned variations in the dependent variables, i. e. the levels of trust and satisfaction with democracy. In the following analysis we will combine them into a multi-level model that is used to assess the causal impact of the relevant independent variables in order to explain these patterned variations. Most existing approaches predict that on the demand side, individual characteristics (such as education, age, life satisfaction) do affect individual level democratic orientations. In addition, on the supply side, several contextual level factors also have an important effect on these orientations, in particular the type of welfare regime, the external intervention into the economy and EMU-membership. The models involve measurement at two distinct levels: the first level includes a sample of individual respondents which is nested within a second regional level. The models in this study use restricted maximum likelihood techniques (REML) to estimate direct and cross level effects (Goldstein 1995). This technique produces similar results as the ML technique (Maximum Likelihood). In hierarchical linear models, as is customary, all independent variables were centered, by subtracting the grand mean (which becomes zero). The standardized independent variables all have a standard deviation of 1.0. The independent variables were treated as fixed components, reflecting the weighted average for the slope across all groups, while country was treated as a random component, capturing the national variability in the slope. The strength of the beta coefficients (slopes) can be interpreted intuitively as how much change in the dependent variable is generated by a one percent change in each independent variable.
The dependent variable in the multi-level model is satisfaction with democracy ( Table 2). The same model has been applied to satisfaction with government which yields similar results. This is an indication that the used model is robust, i. e. the results will not change if the dependent variable is operationalised differently. The most commonly used variables at the individual level have been inserted into the  son and Guillory 1997;Armingeon and Guthmann 2014). Some previous research shows that older, highly educated and wealthier citizens are more positive about democracy. Other researchers do not find consistent empirical evidence that demographic and social status variables really matter (Anderson and Guillory 1997). The results of our analysis are mixed. Respondents that are not satisfied with government are on average lower educated. Sex is significant: females are slightly more satisfied than males. Age is not significant. Included in the table are satisfaction with the present state of the economy, coping with income and satisfaction with life as their causal impact is highly relevant for the answer to the research question. Satisfaction with the economy is the most important determinant of the satisfaction with democracy. This is an important finding as it confirms the Lipset-thesis (Lipset 1959) that there is an intrinsic relationship between economy and democracy. Satisfaction with life also matters since more satisfaction with the living conditions enhances the satisfaction with democracy.
At the country level a mix of economic and institutional variables is included. At the economic level the unemployment rate (unempc) is included in the model, but this indicator does not yield a significant result. The same goes for economic growth (GDPCH) that does not yield a significant effect. Political-institutional structures matter as they mediate the relationship between a person's socio-economic status and his or her satisfaction with the way the system works. Several researchers have claimed that citizens in consensual systems display higher levels of satisfaction with the way democracy works than citizens with a similar socio-economic position in systems with majoritarian characteristics (Lijphart 2012). The reason is that inclusive systems that take into account the rights of minorities and vulnerable groups invoke a higher level of satisfaction among the population. Majoritarian systems that are based on the "winner takes all" principle invoke dissatisfaction among the losers of the competition. On this basis it is expected that in consensus systems there is a higher level of satisfaction with democracy and government than in majoritarian systems. We use the indicator which is conceptualized by Lijphart (2012) and operationalized by . This variable turns out to be not significant. Hence, consensual institutions do not enhance satisfaction with democracy. This is not very surprising as Lijphart himself also found very weak relationships between consensus democracy and satisfaction by means of bi-variate correlations.
Three additional institutional variables that figure prominently in the literature are included in the baseline model. The first variable is whether countries are bailed out or not. It is expected that dissatisfaction with democracy is higher in bailedout countries due to the inability of these countries to cope with the effects of the economic crisis. The second variable is the membership of the European Monetary Union (EMU).  have found that the austerity measures imposed on countries in the Eurozone's periphery has weakened the support of democracy of a majority of citizens in crisis countries. Finally, the welfare regime is included. Welfare regime enters the equation as a number of dummy variables. Only the continental group yields significant results. In addition, the interaction effects between welfare regime and economic growth do not yield significant effects.  Table 3 shows the results after removing the variables with insignificant effects step-by-step starting with the most insignificant ones. After removing these variables only one variable remains that has a significant effect, namely welfare regime. This outcome confirms our main hypothesis that higher developed welfare regimes are associated with higher levels of satisfaction with democracy. The only welfare regime that is insignificant results is Southern Europe. This outcome indicates that the degree of satisfaction in these countries varies compared to the reference group (which is Eastern Europe). This is also indicated by Fig. 2 which shows that the degree to which citizens in Southern European countries are satisfied, compared with inhabitants of Eastern countries, varies per country. This is a sharp contrast with the other welfare regimes which uniformly show higher levels of satisfaction than the Eastern countries. This result underlines the distinctiveness of the countries in Southern Europe compared to the countries in the Continental group.
All in all, the outcomes of the multi-level regression confirm the results of the explorative analysis. At the individual level the satisfaction with the economy is the most important determinant of the satisfaction with democracy. At the contextual level the welfare regime is the only statistically relevant factor. In addition, there are clear regional implications of the findings. The countries in the regions that are K not hit hard by the crisis and/or were not confronted with a bail-out, do also show the highest levels of satisfaction with democracy compared with countries that were confronted with these problems.

Conclusions
We have shown that the Great Recession did not lead to a legitimacy crisis of European democracies and that the diffuse support of democracy remains high in most regions. However, in countries where the economic crisis did strike hard, the support of democracy has been weakened dramatically. This outcome takes a middle position between two opposite outcomes in the ongoing academic debate on the support of democracy. On the one hand, some researchers like Dalton (2011) and Dalton and Wattenberg (2000) claim that the degree of satisfaction with democracy has declined across the whole of Europe. On the other hand, several other researchers have found no empirical evidence that the support of democracy as a core value has declined across Europe. They claim that merely the specific support has decreased in some countries (Norris 2011;Thomassen 2015).
Both positions regarding the increase or decrease of support of and satisfaction with democracy are in need of more nuance by taking into account the role of welfare regimes which are clustered in regions which share not only a type of welfare state, but also history, culture, state-society relations, phase of democratic development, degree of international vulnerability and above all a similar level of socio economic welfare. This context shapes the expectations that citizens have about the development of both their own income and the national socio economic welfare. Many existing research assumes a uniform European context that shows either increasing or decreasing levels of satisfaction with democracy. Our research has shown that the effects of the economic crisis on the support of democracy are moderated by the type of welfare state. The dissatisfaction with democracy is lower in countries in which the welfare regime is well equipped to cope with the consequences of the economic crisis.
Most of the hypothesised relationships are confirmed. The research shows that specific support is more affected by the economic crisis than diffuse support which is demonstrated by the moderate change in institutional trust (H1 and H2). On a more disaggregated level the Great Recession did to lower levels of trust, in particular of the lower educated and in countries with that are hit hard by the crisis (H3). Countries that were confronted with austerity measures that were imposed by external actors witness a drop in the support of national democracy and not an increase of this support (H4). The type of welfare state has had a moderating effect on the degree to which citizens have been satisfied with democracy during the Great Recession (H5). A declining economic growth did not lead to less support of democracy in countries with a highly developed welfare state, but it did have this effect in countries with lean welfare states.
Although the empirical trends show no overall decrease of support of democracy across the whole of Europe, but there are significant regional differences that have to be taken into account. In Southern Europe we find a significant increase in citizens that is dissatisfied with democracy and its performance. In Anglo-Saxon Europe we see a similar development but to a lesser degree. The severity of the economic and financial crisis has a significant impact on the support of democracy.
In addition to this main finding, the research has empirically demonstrated the dual character of legitimacy, namely the difference between specific and diffuse support of democracy. The trends and variations in both types of satisfaction with democracy indicate an interesting paradox: notwithstanding the apparent low levels of specific support, the overall trust in representative democracy as a regime remains intact. Given the high degree to which the economic crisis has affected the level of social and economic welfare, the public acceptance and adherence to the democratic systems is surprising. This research has also shown that the Great Recession has lowered the diffuse support of democracy in Southern Europe which is the region that has been hit hard by the Great Recession.
The societal implications of these results are that citizens in European regions live in different socio-economic circumstances which affect their view on what the EU has to offer them. These regional differences may hamper the process of European integration because they strengthen existing socio economic and cultural cleavages in Europe. In addition, in some countries the EU institutions are blamed for the economic crisis and/or more integration is seen as a threat to the national autonomy and level of economic welfare. This sentiment may strengthen the call for renationalisation and hamper the process of European integration. This new development may (in the long run) slow down the recovery of the economic crisis and strengthen the discontent with both the national and European democracy. In order to cope with these regional differences, they have to be taken (more) serious in policy-making and research.
Open access funding provided by VU University Amsterdam.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Appendix: Variables and operationalisation
How satisfied with the way democracy works in country (stfdem): And on the whole, how satisfied are you with the way democracy works in [country]? How satisfied with the national government (stfgov): Now thinking about the [country] government, how satisfied are you with the way it is doing its job? Institutional Trust is measured by the mean of the values on three variables: Using this card, please tell me on a score of 0-10 how much you personally trust each of the institutions I read out: Trust in Parliament, Trust in the legal system and trust in the police The type of welfare regime (Welstate: as specified in note 1)