Economic Crisis and Attitudes Towards Democracy: How Ideology Moderates Reactions to Economic Downturns

This paper studies the extent to which economic crises bust different reactions depending on individual ideological identity. Our argument is that individual left-right identification shapes how citizens evaluate the functioning of democracy, and consequently, it moderates the impact of economic crises on democratic discontent. We argue that left-wing individuals’ views on democratic performance would be more heavily affected by economic results and strongly shaped by the (in)capacity of a democracy to represent its citizens in its reaction to the economic crisis. After testing our argument with data from the ESS, we provide evidence from a survey experiment conducted in Spain. Both analyses show that economic crises are relevant in explaining democratic dissatisfaction, but the strength of their impact depends on ideological identities. Economic results have a stronger effect on the satisfaction with democracy of left-wing individuals by negatively affecting their perception of the functioning of democratic representation.


INTRODUCTION
The European financial crisis, besides being a tsunami for national economies, brought into being political movements demanding more and better democracy. In response to the economic slump, governments implemented austerity policies that led in several cases to anti-austerity movements (Flesher Fominaya 2017). Consequently, what started as an economic crisis evolved to what some authors have identified as a crisis of political representation (Kriesi 2015). While initially for some citizens their discontent had its origin in the perceived inefficiency of the political system in providing economic prosperity, there was widespread contestation against the lack of representation. This would be particularly relevant in the case of centre and left-wing citizens who found out how the only recipes to overcome the crisis were, in most cases, severe cuts in social expenditure, labour market liberalisation and, in short, policies that limited the expansion of the welfare state (Turnbull-Dugarte and Stuart, 2020). Despite the abundance of literature on how context affects attitudes towards democracy Guthmann, 2014: Magalhães, 2014;Cordero and Simón, 2016: or Christmann and Torcal, 2017 among others), there is a lack of research into why citizens react in different ways to the same scenario. Here we argue that these different reactions towards the economic crisis and its impact on satisfaction with democracy are moderated by left-right ideology.
Left-right orientations are not only a widespread political metaphor, they are also a political compass that guides voters' decisions and structure party competition in Europe (Bobbio 1996;Freire 2006;Carmines and D'Amico 2015). While there is ample research into the relationship between left-right positions and how people vote, the impact of ideology on attitudes towards democracy has largely been ignored (Ceka and Magalhaes 2016). Citizens' ideological selfplacement determines not only their political choices but also, as put by Jost, Federico, and Napier (2009:324), "a wide range of attitudes, preferences, judgments, and behaviors". Therefore, it should also be expected that left-right orientations are associated with how attitudes towards the political regime are shaped in reaction to its context.
Likewise, ideology is the basic template for understanding events and making political assessments, so it should have an impact on the dimensions we give more weight to when assessing the political regime. In this regard, there is empirical evidence according to which citizens on the right are more willing to accept social and economic inequality than those on the left (Jost et al., 2003). As from the last financial crisis less responsive governments and egalitarian societies emerged, economic hardship would have more impact on leftist citizens. This would be the case even for those left-wingers not personally affected by the economic slump, because the social impact of the crisis would affect their expectations of and attitudes to the democratic system. Consequently, we should expect citizens on different sides of the ideological scale to evaluate their satisfaction with the political system under different terms.
In this paper we explore this moderating effect of ideology by looking at satisfaction with democracy (SWD) and reactions to economic downturns. Satisfaction with democracy is one of the main attitudinal variables that we should pay attention to in democratic systems, as it is commonly considered to be a key measure of regime performance and democratic quality (Morlino 2004). Taking into account what political psychologists have found about how ideology is associated with individual happiness (Butz, Kieslich and Bless 2017) and basic values (Caprara et al., 2006;Piurko, Schwartz and Davidov 2011;Aspelund et al., 2013), we develop here that rightists' conservatism and justification of the status quo might be relevant in levering their democratic discontent. This would explain why their overall assessment of democracy responds to changes in the economic context without necessarily questioning the extent to which the system represents their ideals. Conversely, left-wing voters have a more representation-driven view of democracy and they are more willing to demand reforms of the political system. 1 This way, leftists' satisfaction with democracy is not only directly affected by economic hardship but also by how this makes them question whether democratic systems truly represent their ideals. In this regard, the European debt crisis should have been a critical moment to leftist parties. The frugal agenda that was introduced in most countries and sometimes forced by EU institutions as a way to overcome the economic slump made them much less useful in the eyes of leftwing voters in pursuit of policy representation. Consequently, centre-left-and left-wing citizens would be much more dissatisfied with the political system as the economic crisis would have limited their chances of having their preferred policies implemented (Turnbull-Dugarte and Stuart, 2020).
After assessing the relationship between democratic discontent, economic distress and individual ideology in Western Europe with data from the ESS, we provide evidence from an experiment run in Spain in January 2018. The case of Spain is relevant to study the impact of economic and political factors on democratic discontent because it is a country that has recently experienced both deep economic and political representation crises (Vidal 2018). Our empirical tests support our hypothesis of the differential effects of economic hardship on citizens depending on their ideological identity. Moreover, we explain that the economic crisis had a major effect on democratic discontent among the left because it negatively affected their perception of quality of representation, which in turn affects SWD.

BACKGROUND
The link between attitudes towards democratic politics and government performance has attracted the attention of scholars since the 1970s. The cornerstone of our understanding of how satisfaction with political regimes -and more specifically with how democratic systems work-is rationalized in Easton (1965), Easton (1975). He distinguished two dimensions: specific and diffuse political support. The former would be associated to the results achieved by the government and the political system, and necessarily linked to the (in)capacity of political institutions to respond to their citizens' demands. At a broader level, diffuse support would be related to principles of the political system. By definition, diffuse support should be more immune to the outputs of political performance. The literature, however, has shown that short-and medium-term events such as economic crises and downturns affect public perceptions of the functioning of democracy and can affect satisfaction with democracy and endorsement of the political regime (Magalhães, 2014;Cordero and Simón, 2016).
A key question, however, is which dimensions of an economic crisis are more important in explaining satisfaction with democracy. The dire performance of most Western economies after the 2008 Great Recession promoted not only widespread dissatisfaction with economic outcomes, but also growing discontent towards politics in general and protests against the political management of the economic system. This indicates that citizen satisfaction with a regime's performance will not only stem from the outcomes achieved, but also on the extent to which citizens perceive that, due to economic crises, the political values of the system are respected. If citizens endorse the principles of a political system, like democracy, their satisfaction with the political system will be vulnerable to the extent to which political process is conducted without eroding those principles. Moreover, in the case of the European financial crisis, the presence of supranational agents and constraints would have conveyed the idea of an imperfect democracy in which decisions and policies are not decided by national governments and would no longer consider their citizens demands (Cordero and Simón, 2016:306;Hobolt and de Vries 2016). This has been analysed by Devine (2021) who showed that supranational interventions and a perceived limitation of the autonomy of the domestic political system is not necessarily associated with a decrease of political support but the latter is significantly affected by economic evaluations. Devine's study, however, focuses on satisfaction with democracy and trust in the country's parliament without considering how different ideological groups might be more frustrated in a context in which their policy preferences and ideological predilections are not an option.
To our knowledge, no study to date has analyzed how individual long-established preferences towards both the political and the economic system, such as ideology, moderate the way individuals respond to economic hardship. Our argument is that the economic slump that began in 2008 had a major impact on left-wing individuals, on the one hand, because they were less willing to accept its social consequences and the austerity measures taken in most countries to overcome the crisis. On the other, because they are more likely to accept changes to improve the political system, they are also more demanding in their assessments of how the system should work. In the next section we develop this argument and explain how left-right ideology is a political shortcut that summarizes differing views towards representation and the role of the national government in the management of the economy. Consequently, this will shape the reaction to economic and political factors and their impact on democratic discontent.

THE ARGUMENT
Satisfaction with democracy (SWD) is not only the result of citizens' general evaluation of institutional performance, as it is also evaluated through the lens of more stable values and ideology. Why should we assume that the impact of the crisis on the assessment of economic performance and the quality of representation is the same for people with different ideological positions? To our knowledge, the political science literature has merely highlighted the relationship between SWD and left-right self-placement (Christmann and Torcal 2017;Anderson and Just 2013) without discussing which factors affect how citizens on the left (right) side of the scale evaluate the functioning of democracy differently to those on the right (left).
In a piece of research conducted with data from 20 countries, Anderson and Singer (2008) found out that an individual's political attitudes were differently affected by changes in their surrounding context depending on their ideological attachments. They concluded that in countries with high levels of inequality, left-wing citizens evaluate democracy much more negatively, whereas right-wing citizens' attitudes were influenced much less by income inequality.
More recently, Ceka and Magalhaes, (2016) briefly explored the way left-right orientations are associated with different conceptualizations of democracy. They found that those on the left side of the ideological continuum are more prone than right-wingers to emphasize social justice and redistribution as key elements for democracy.
This research makes it clear that an individual's SWD would be differently affected by the political and economic contexts depending on whether they are left-wing or right-wing. In political psychology, the association between left-right ideology and certain individual characteristics and political attitudes has been a main topic since the beginning of the century. Thanks to recent studies in this field, we know that leftists and rightists differ not only in their political preferences but also in their basic values (Piurko et al., 2011) and personality traits (Jost 2006;Caprara and Vecchione 2018). Considering this, it is reasonable to think that leftists and rightists have different expectations about the political system. According to Caprara and Vecchione (2018), studies have shown that political opinions and preferences of those on the left tend to be more affected by their concern for other people (selftranscendence), while those on the right side of the ideological scale put more weight on values of conservation of the status quo, this is, resistance to social change and social stability. At the same time, these values constrain different preferences towards the role of government and shape opposed ideal societies; while selftranscendence is associated with social justice, equality and higher support for civil liberties, conservation relates to free markets, stability, conformity and tradition (Goren, 2016;Caprara and Vecchione 2018). This way, in the right we find the prevalence of their resistance to change and risk aversion over the prospects of social change (Jost, Federico and Napier 2009: 312). Considering this, it can be expected that right-wing citizens will be discontented with the political system in case of economic slump but they would refrain to demand changes as they would prefer to preserve the political system over the uncertainty that would convey changing it. Also, because they are more willing to accept inequality, their assessment of the functioning of democracy will be less affected by the social consequences of the crisis as they would find them more acceptable than leftists citizens. This, in short, would lead to argue that rightists' SWD is more resistant to economic slumps. Conversely, the political cosmovision of the left is sustained by the idea of equality and equal access to political representation, making them more reactive to deficiencies in political representation created by economic hardship, which, in turn, will have a greater impact on their satisfaction with democracy. This would be in line with what Ceka and Magalhaes, (2016) found in their research and what they identified as a more demanding concept of democracy of those identified as left-wing leftists.
We can further develop this expectation concerning the effects of the surrounding context on those on the left with regards to the last financial crisis. The European financial crisis has had a distinctive impact on the ideological left because, beyond economic results, the presence of transnational actors and the accompanying perception of a loss of sovereignty affect citizens' satisfaction with a political system because this would be seen as not representative of what citizens demand (Cordero and Simón, 2016;Hobolt and de Vries 2016;Sánchez-Cuenca, 2017). Moreover, the measures taken as a remedy to the economic crisis in most countries might be seen as a neo-liberal turn that was characterized by serious cuts to the welfare state and, consequently, had a different impact in each ideological group, left and right, and one might see the crisis as a backlash of the welfare state, while the other would have seen how their preferred policies were implemented. Related to this and especially in countries that were intervened by the EU, left-wing citizens might see themselves as political orphans as the electoral choices available would have limited capacity to shape policy alternatives to the austerity (Turnbull-Dugarte and Stuart, 2020). Thus, we theorise that the European debt crisis impacts the satisfaction with democracy of those on the left by damaging their perceptions of quality of political representation.
In summary, our argument is that left-wing and right-wing citizens have different expectations pertaining what a democracy should be, what outcomes it should deliver, how economic crises affect them and what policies should serve to remedy them. Thus, their satisfaction with democracy would be shaped by different factors and economic hardship would more heavily affect the democratic discontent of those who place themselves on the left side of the ideological scale. Accordingly, we develop the following hypothesis: H1-Economic hardship impacts satisfaction with democracy of left-wing citizens by eroding more strongly their perceptions of the quality of political representation.

TESTING OUR HYPOTHESES WITH OBSERVATIONAL DATA
To analyse the validity of our hypothesis we firstly test it with observational data from the European Social Survey. As the political terms left' and "right" are not so commonly used in Eastern countries, we use only Western countries in the sample.
In this first part of the analysis, we explore the impact of economic crisis in Europe on satisfaction with democracy and other indicators of political satisfaction and the differential impact of economic outcomes depending on ideology. We use a sample formed by the first seven rounds of the European Social Survey, spanning from 2002 to 2014.
Our argument is that ideology moderates the impact of the economy on satisfaction with democracy. Our main dependent variable is the standard variable measuring satisfaction with the political system. This variable is the self-placement in a scale from 0 to 10 to the question "How satisfied are you with the way democracy works in your country?", where 0 is extremely dissatisfied and 10 means extremely satisfied. Secondly, as a robustness test, we account for the impact of the interaction between economic performance and ideology on three other indicators of political trust: trust in politicians, trust in political parties, and trust in the parliament. These are a basic indicator of citizens' support for their political system (Newton and Norris 2000;Montero, Zmerli and Newton 2008;Devine 2021). The three variables are also measured in a 0 to 10 scale, where 0 means "no trust at all" and 10 "trust completely". This will allow us to tease out the extent to which democratic discontent due to the crisis runs parallel to the deterioration of feelings of political representation conditional on ideology. In the next section, as satisfaction with democracy and with political representation might be endogenous to each other, we use an experimental approach that allows us to address issues of potential endogeneity, as well as any omitted variable bias.
As a measure of ideology, we use the usual individual selfplacement in a left-right scale that ranges from 0 (extreme left) to 10 (extreme right). As measures of economic performance, we use the two most common measures in the economic voting literature: annual GDP growth and annual increase in the unemployment rate. 2 These two indicators of the economic situation are also relevant because citizens often see them as the result of government's policies (Lewis-Beck and Stegmaier 2000). We focus on the interaction between the left-right selfplacement and these two indicators of economic performance to assess whether the impact of contextual factors such as the economic situation on democratic attitudes can be moderated by individual ideology. Descriptive statistics presented in Table 1 show the range of values that these two variables can take in our analysis.
Due to the hierarchical structure of the data, we run multilevel models with varying intercepts by country-waves. This means we have 108 level-2 units, except for the models on trust in political parties, where we have 92 level-2 units because the question was not asked in the first wave of the European Social Survey. In all models, we include a set of individual-level covariates that allow us to control that other variables relevant in shaping citizens' attitudes are affecting our results. In this regard, we build on the work by Christmann and Torcal (2017) and include the same covariates they use to replicate their models. Specifically, the models include a dummy for gender that takes value "1" when the respondent is a male and "0" for female. We have mixed expectations with regards to the impact of gender. Although Hansen and Goenaga (2021) find that women are less satisfied with democracy, Bol et al. (2021) show that women were more satisfied with government during the recent COVID-19 lockdown. Models also control for age -as older people tend to be more satisfied with democracy (Dahlberg et al., 2015)-and education-as educated citizens tend to be more dissatisfied (Norris 1999). The former is coded as (coded as how many years old the respondent is and the variable of education measures the number of years that the respondent has been enrolled in an education programme. We also include as Christmann and Torcal (2017) a variable for marital status that takes value 1 if the respondent is married and a variable that captures the most vulnerable occupational status, this is, whether the individual is unemployed (1) or not (0), as those worse off will tend to be less satisfied with how the democratic system is working. Finally, we add a variable on satisfaction with the government, measured in a 0-10 scale, to account for the fact that voters' assessment of democracy in the country is endogenous to whether a preferred party is in government and 2 Data on the GDP growth and changes in the unemployment rate by country and wave included in the analysis can be found in Supplementary Table S1 in the Appendix. whether voters perceive themselves as losers or winners in the political system (Anderson and Guillory 1997). Table 2 displays the results for economic growth. Before getting to our main variables of analysis, we can comment on our control variables. We find, consistent with Bol et al. (2021), that male respondents are more dissatisfied with democracy. The coefficients of age and education years are contrary to our expectations. However, in alternative models we explore whether the effects are not linear. When modelled in a curvilinear fashion, we do find that the oldest are more satisfied with democracy and the most educated are more dissatisfied.
The first models run the analysis without the interaction. The last four (Models 2.5 to 2-8) include the interaction between ideology and GDP growth. In the first set of models it is perhaps surprising that we tend to find not significant effects of the economy on satisfaction with democracy and the three political trust measures. This is basically because the variable satisfaction with the government absorbs all its impact. 3 The introduction of the interaction between the GDP growth and ideology allows us to assess the differential impact of the financial crisis for citizens with different ideologies. Importantly, and as expected, the interaction shows that as a result of the    financial crisis, left-wing voters reduce more their satisfaction with democracy than right wing voters (model 2.5). The negative and significant sign in the interaction indicates that when GDP increases, the increase in satisfaction with democracy will be lower for right wing individuals. This already tells us that leftwing voters' assessment of the political system will be more conditional on the state of the economy. To account for the magnitude of these effects, in Figure 1 we simulate the impact of the economy on satisfaction with democracy for voters with different ideological profiles. It can be observed, more specifically, that when the economy is doing well, the predicted satisfaction with democracy of a left-wing and a right-wing individual are virtually the same. 4 However, once there is an economic slump, the satisfaction with democracy of a left-wing voter will significantly deteriorate more than a right wing voter's. The magnitude of the effects is not large, which indicates that an attitudinal variable such as satisfaction with democracy is possibly explained by many other underlying attitudinal cofounders, but it runs in the expected direction.
This result is reinforced in models 2.6 to 2.8. where we use as dependent variable the three different political trust variables, we find the same results. Economic crises erode more strongly trust in different dimensions of the political system if the voter has a left-wing ideology. The positive sign of the interaction shows that the reduction in political trust is less strong as the individual has a more right-wing ideology.
As a robustness check, in Table 3 we replicate the same models with the same sample, but introducing increase in unemployment as a measure of economic performance. The expected signs are now the opposite as now higher values of the economic performance variable mean worse economic results (larger increases in unemployment). The results are consistent with the previous models. Increases in unemployment reduce satisfaction with democracy as well as political trust. However, as the interaction in models 3.5-3.8 shows, the negative effect of economic performance is weaker for right wing voters (and stronger for left-wing individuals).
In summary, we find results consistent with our theoretical approach. Economic crises reduce levels of satisfaction with democracy for all citizens. However, its effect is more negative and significant for left-wing individuals. This is consistent across the three indicators of political trust: trust in political parties, parliaments or politicians. The impact of the crisis on satisfaction with democracy operates more strongly in left-wing individuals. However, whether these differences between ideological groups are the result of the deterioration of perceptions of the quality of political representation must be tested with a different approach. To provide more evidence of our proposed mechanism we run a survey experiment in the next section.

EXPERIMENTAL EVIDENCE
We have argued that left-wing voters will place more importance on political representation as a trait of democracies. The previous results support with observational evidence our argument about the differential impact of the failure of the economic system on citizens' SWD and political trust depending on their ideology. The economic crisis, however, was not only seen as a period of economic hardship. Particularly in Southern Europe, the economic crisis has generated a strong dissatisfaction with political representation. As Armingeon et al. (2015) argue, citizens' perceptions about the quality of representation have been eroded as a result of the economic crisis, especially where the response to the crisis was mostly based on austerity and where Eurozone constraints did not allow national governments to define the way they wanted to address the crisis. Also, we have also emphasized that in some countries, especially in those whose economies were intervened by the EU as is the case of Spain, left-wing citizens might have felt politically orphan because the parties that should have represented the policy alternatives to the austerity measures prioritized being responsible in the EU's eyes (Turnbull-Dugarte and Stuart, 2020). Thus, we also test this argument with data from a survey experiment conducted with an online sample in Spain and fielded by Netquest during the last week of January 2018.
The Spanish case is particularly useful for our purposes. On the one hand, despite the fact of the Spanish democracy was born after a 4 decades dictatorship, there are no reasons to think the values and issues associated to the left and to the right in Spain are significantly different to those associated to each ideological side in other West European democracies (Caprara and Vecchione 2018). Also, and to the purposes of this research, Spain is a country where the economic crisis has had a very strong impact. More importantly, the crisis showed that the economic model in Spain was very sensitive to the economic cycle as the country transitioned from a period of economic expansion and job creation to economic recession and the second highest unemployment rate in the European Union in a dramatically short period of time. This will allow us to test the sensitivity of different types of individuals to the economic crisis and its impact on their satisfaction with democracy. In addition, Spain is relevant for studying the impact of the crisis on the perception of flawed political representation. Spain was one of the countries that received a bail-out (in the Spanish case just for the banking system) and where policy-making was restricted firstly, during the leftist government leaded by José Luis Rodríguez Zapatero and also later under Mariano Rajoy's conservative executive. In the case of Spain, the collective Democracia real, Ya (in English, "Real democracy, now") was part of the nucleus of the Indignado 5   Constant and Country and wave varying intercepts not shown. Standard errors in parentheses, ***p < 0.01, **p < 0.05, *p < 0.1. 5 The name "Indignados" was inspired by Stephane Hessel's essay "Time for Outrage: Indignez-vous!" Frontiers in Political Science | www.frontiersin.org August 2021 | Volume 3 | Article 685199 movement that rallied on the streets under the motto "They do not represent us." 6 This movement, mostly composed of leftwingers, could be understood as a natural response to the severe economic crisis in Spain. However, it is worth noting that most of the criticism which emanated from the indignados was against political institutions and how they were no longer able to represent the demands of ordinary citizens. As a result of the crisis, both Prime Minister Zapatero and Rajoy had to abandon their electoral pledges and implement policies imposed by the European institutions. Citizens' contestation though was about flawed political representation and not just the economic outcomes of the crisis. Moreover, because the outbreak of Spanish protest took place under a socialist government, this movement should not be seen as the reaction of the political losers. The source of this democratic discontent could be considered to be rather ideological than partisan.
Our experiment had a sample of 1,200 individuals and respondents were divided into two groups (n 600 in each group). Respondents were sampled using quotas by gender, age and region (Autonomous Community). All respondents had to answer to what extent they agreed (on a five point scale: completely agree, to some extent agree, neither agree nor disagree, somehow disagree, completely disagree) with two statements about the political situation and two more about the economy. The two statements associated with the political situation were: "Political parties do not represent me" and "Political reforms are necessary". The two related to the economy were "Economic reforms are necessary" and "The economic system benefits those with power". As the distribution of the responses is not linear, we recoded the answers to create a dummy variable that takes value "1" when the respondent completely or to some extent agrees with the statement and "0" when s/he neither agrees nor disagrees, somehow disagrees or completely disagrees.
Importantly, all these beliefs have an impact on satisfaction with democracy. We tested whether having perceptions of both a failed economic system and a deteriorated political representation diminishes satisfaction. We run successive OLS regressions where satisfaction with democracy is our dependent variable and the agreement with these statements are the explanatory variables. As shown in Figure 2, agreement with the statement in all four cases correlates significantly with lower levels of satisfaction with democracy.
Thus, the question now is whether economic hardship affects these perceptions and if so whether it does so differently for rightwing and left-wing citizens. In order to test this, the control group was asked to answer to what extent they agreed with four further statements, while the treatment group was exposed to a framing experiment. Before answering the same four statements, they were asked to think about the last 10 years and the economic crisis. The aim behind this strategy is to assess whether citizens establish links between these statements and the consequences of the economic crisis. Our goal is twofold. On the one hand, we intent to address how citizens understand and interpret the economic crisis, as well as the extent to which they link it to economic and/or political factors. On the other, we want to see if this understanding of the crisis differs across political ideologies.
The wording of the treatments and dependent variables in this case were as follows: Control Group. To what extent do you agree with the following statements? (Completely agree, to some extent agree, neither agree nor disagree, somehow disagree, completely disagree) Treatment Group. Thinking about the last 10 years and the economic crisis, to what extent do you agree with the following statements? (Completely agree, to some extent agree, neither agree nor disagree, somehow disagree, completely disagree).
Percentages of agreement with each of the aforementioned assessments (recoded into dummy variables) differ by citizens' ideology (see Table 4). In addition, and more importantly, the agreement with the statements also changes when citizens are exposed to the framing treatment and think in terms of the financial crisis and the last 10 years.
It is worth mentioning the differences between left-wing and right-wing citizens in the control group concerning the extent to which political parties represent them. While the degree of agreement with the statements about economic problems, for instance, is basically similar for leftist and right-wing voters, the two political statements receive more endorsement from leftwing voters. This already indicates that political dissatisfaction seems to be larger among left-wing voters, confirming our previous descriptive evidence that left-wing voters are more demanding in terms of the role that political representation plays in a democracy. When the economic crisis is mentioned, we see that agreement with the two statements about the political side of the crisis (political parties do not represent me and political reforms are necessary) increases mostly among left-wing voters, while it increases less (or even decreases) for rightwing citizens. This, somehow, indicates a different view of the economic crisis or invites us to think that individuals might understand the economic crisis to have different origins and solutions depending on their ideological identity. Thus, for citizens on the left the crisis might not be all about the economy, but instead they feel that the political side of it (and particularly political representation) are intimately related to the last 10 years of economic hardship and the slow recovery. We do not see, however, differential effects across ideological types for the two economic statements. Both left-and right-wing voters make a similar link between the economic crisis and economic statements. To see whether these differences remain in a multivariate analysis, we ran logistic regression models including as control variables gender, age, education and occupational status. Our dependent variables are each of the dummy variables for each of the four statements. In Figure 3 we display the treatment effects conditional on ideology for the two statements related to political representation and in Figure 4 the effects for the two statements related to the economic measures. 7 Concerning the agreement with the statement "Political parties do not represent me", the marginal effect of the treatment in comparison to the control group by ideology is shown in Figure 3. As seen those on the left are more prone to say they do not feel represented by political parties when the economic crisis is mentioned. However, the effect is not so clear for citizens on the right. For right-wingers there might be no significant differences between control and treatment groups. These results are on a 90 per cent level confidence, though. Differences between leftists and rightists are more evident concerning the statement "Political reforms are needed". As seen in the right-hand panel, not only the magnitude of the effects is higher, the effect of the treatment has a steeper slope when respondents' ideological position moves from left to right which points to a clear division between left and right in what respects to their preferences towards the political system. 8 On the other hand, the level of agreement with the statements about the economy, displayed in Figure 4 seems not to be affected by the treatment or at least the evidence is not as revealing as with the political statements. Although the effect moves in the opposite direction to that of the political system statements, we do not observe that right-wing individuals significantly change their perception of the quality of the economic system as compared to left-wing citizens.
With these results we can state that priming respondents with statements about the economic crisis has an effect on perceptions of the quality of representation, but this effect differs depending on a citizen's ideology. Thus, when the economic crisis is mentioned, those on the left are more critical of politics and are more willing to agree with statements about the lack of representation and the necessity of political reforms. This behaviour is different to that of those on the right side of the ideological scale whose assessments on the quality of representation would not be linked to the economic crisis. This gives support to our hypothesis and suggests that our FIGURE 4 | Marginal effect of treatment by ideology (continuous variable) on agreement with statements "Economic reforms are needed" and "The economic system only favors the powerful." Note: Models with covariates (90% confidence intervals). Estimates are based on the third and fourth logistic regression models presented in Table A 8 We have also tested in a different experiment for a direct effect of the political representation crisis on satisfaction with democracy, but we do not find any differential impact according to ideology. Thus, we only have evidence of it as being a mediator of the effect of economic crises by negatively affecting perceptions of political representation of left-wing people.
suggested mechanism is correct. The impact of the economic crisis on perceptions of democratic representation is more salient within the left. This means that not only, as we previously showed, do left-wingers tend to be more dissatisfied with democracy than those on the right, but also this discontent is the result of a more demanding concept of democracy in which economic crisis leads to a re-assessment of the political system.
Altogether, the experiment provides evidence on how the mechanisms by which economic crises decrease the general satisfaction with democracy have a different intensity contingent on individuals' ideology. While there are no differential perceptions of the economic dimension of economic crises for left-and right-wing individuals, economic crises do affect more strongly left-wing citizen's perception of democratic representation. These results shed light on the sharp decrease in democratic satisfaction that most European countries have suffered in the aftermath of the financial crisis. The crisis hit both sides of the political spectrum, but the way it has impacted citizens is different according to their ideology.

CONCLUSION
In this paper we argued that, as the literature has traditionally stated, economic hardship has a negative impact on citizens' satisfaction with democracy. However, we have shown that the mechanism by which their satisfaction with democracy decreases differs depending on their ideology. The reason is that individuals' ideology defines different conceptions of democracy. Right-wing citizens give more weight to their resistance to change and risk aversion, while left-wing citizens place more importance on the political representation dimension. As a result, economic hardships affect more left-wingers' satisfaction with the political system by negatively affecting their perception of democratic representation.
This research can be expanded in the future in several ways. First, we have shown that ideology is relevant when defining which dimensions of democracy citizens consider fundamental. In other words, ideology correlates with distinct conceptions of what a democracy is and what it should achieve. Research on this topic, however, has been surprisingly scant. Further research exploring more systematically whether citizens have different perceptions of what a democracy is and what they expect from could help illuminate why left and right also differ in what they await from the political system. Secondly, we have shown how citizens in Spain, more in particular those on the left of the ideological scale, link the 2008 Great Recession with the erosion of democratic representation. To better assess the implications of the crisis, it would be necessary to evaluate whether the deterioration of democratic representation is just a short-term effect or whether the legacy of the crisis is a change in the perception of the extent to which democratic systems are able to represent citizens' political demands. Finally, we studied the effects of economic hardship on satisfaction with democracy by using experimental evidence from Spain. We argue that the Spanish case is very suitable as it is a country where both the economic and political crises have been very pronounced and where citizens established clear links between them both. However, we are aware of the limitations of focusing on such a specific context and cross-country research with more countries in which the economic crisis has affected the economy will contribute to a better understanding of the association between economic results, perceptions of democratic representation and evaluations of the political regime.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.