Identity, instrumental calculations and political radicalism: On the determinants of support for secession in Catalonia

Transcription

1 Identity, instrumental calculations and political radicalism: On the determinants of support for secession in Catalonia Jordi Muñoz Universitat Autònoma de Barcelona Raül Tormos Centre d Estudis d Opinió rtormos.ceo@gencat.cat To be presented at the XXII nd IPSA World Congress of Political Science, Madrid July Abstract How can we explain that a part of the Catalan population gives support to secession while others reject it? Is it only a matter of identity, or do instrumental calculations related to the economic consequences of a hypothetical secession play an important role as well? In this paper we try to answer these questions analyzing the Political Opinion Barometer (July 2011) of the Centre d Estudis d Opinió. The results show how, beyond identity, economic factors, among others, have an important role in explaining support for secession, especially among those in favor of secession who have a weak preference for it. Introduction Available survey evidence consistently points to an increase in the support for secession among Catalan population in the last years. Despite we only have partial time series data with diverse question wordings, if we pay attention to the evolution of constitutional preferences in the Barometers of the Centre d Estudis d Opinió between 2006 and 2012, support for an independent state has more than doubled, going from 13.9 to 29% 1. This situation has been profoundly debated and filled ample space in the media, given its evident political relevance. Nevertheless, the causes of this growth in support for independence remain unclear. The majority of explanations presented in the mass media are related to the effects of the political situation: some hypothesis refer to the debate over the reform of the Catalan autonomous treaty (Estatut d Autonomia), the entry of an explicitly secessionist political party to the government of the Generalitat (the autonomous government of Catalonia), the sentence of the Constitutional Tribunal about the Estatut, etc. Others have pointed, more timidly, to a generation effect: the emergence of a generation who grew under democracy and autonomy, and therefore, consider the current level of autonomous government as a point of departure and not as a culmination of a long political process, has made the segment of secession supporters to become more important. Even though it seems complex to determine with precision the causes of this increase, a fundamental step is to understand the explanatory factors of individual support for independence. Why do some citizens say they would support independence while others declare they would abstain, or vote against in a hypothetical referendum? Is it an identity issue, or do rational calculations on the economic consequences of a hypothetical secession play an important role? Which influence do political parties have on the position of their 1 Data from the Catalan government quarterly public opinion barometer ( 1

2 electorates? And the mass media? Which other additional factors influence on the probability to support independence? In this paper we try to answer these questions. To do it, we discuss and empirically test a model of support for independence which, departing from previous literature and available data, takes into consideration different explanatory factors and, above all, it emphasizes the heterogeneity of support for independence and explores to what extent we have to consider different explanatory factors for different groups of supporters of independence. Support for secession in consolidated democracies: a theory review A great deal of academic work on support for secession has been performed in the context of a particular case: the Quebec one. The explanation is relatively simple: only in the context of a consolidated democracy it is reasonably feasible to obtain reliable public opinion data and, among these countries, the Quebec is, until now, the case in which this option has revealed more clearly, in the form of referenda. This is why, the theoretical basis of our explanatory model of support for independence in Catalonia are founded, to a great extent, in the research based on this case, though we will have to adapt some of the propositions to the Catalan context. National identity is often considered the main driving factor of support for sovereignty. Most of the research on Quebec, as well as Scotland (McCrone and Paterson 1992), Catalonia (Serrano-Balaguer 2010) or the Basque Country (Costa i Font and Tremosa 2006) indicates that identification with the sub-state unit has a strong effect on support for its secession. More precisely, Blais and Nadeau (1992) report that it is identification with Canada what matters the most, because identification with Quebec displays little variation within the province. However, although the models consistently show its strong predictive power, the role of national identity on influencing support for secession has received scant theoretical attention. This is probably so for two reasons: on the one hand, this link is considered commonsensical and thus does not puzzle the researchers. On the other hand, national identity is commonly regarded as a highly stable political attitude that does not change easily, at least in the short term. Therefore, identification with the sub-state unit and the state would only explain the stable patterns of support for secession but not changes (Howe 1998, Blais and Nadeau 1992). But there are reasons to question both arguments. First, the existence of many non-secessionist nationalists that just aim for more power for their territories within the state but not full secession implies that the link between identification and pro-independence attitudes is by no means necessary. Moreover, in Catalonia there is some evidence of intrapersonal changes even in the short term- of national identification (Hierro 2010, Martínez Herrera 2000) that raise doubts about its purported stability. There are exceptions to this lack of theory. The effect of national identity on constitutional preferences has been treated as an analogue of party identity (Blais and Nadeau 1992): it summarizes socialization and past experiences and conditions citizens attitudes and reactions towards certain political issues. Costa-i-Font and Tremosa (2006), following Akerloff and Kranton (2000), propose a model in which identity determines preferences, basically because individuals attain symbolic utility to act in accordance with their identity. 2

3 Whilst identity is generally found by the models as being the primary driving force of support for secession, it is by no means the end of the story. Other variables have been consistently found to play a role in determining constitutional preferences, and one that has received more attention from the literature is the perceived economic costs and benefits of the secession (Blais et al. 1995, Howe 1998, Martin 1994, Mendelsohn 2003, Bélanger and Perrella 2007, Blais and Nadeau 1992). A so far consistent finding estates that cost-benefit calculations on the potential economic consequences of the sovereignty do play a role in explaining support for independence in Quebec. Costa-Font and Tremosa-Balcells (2006) for the Basque Country case, did not find any effect, however they used a very indirect measure of costs and benefits. Indeed, the role of instrumental calculations, together with the degree of relative development with respect to the rest of the state, in aggregate claims for secessionism has been quite convincingly established (Sambanis and Milanovic 2009, Sorens 2004). Whilst identity is generally found by the models as being the primary driving force of support for secession, it is by no means the end of the story. Other variables have been consistently found to play a role in determining constitutional preferences, and one that has received more attention from the literature, especially in the Quebec case, are the perceived economic costs and benefits of the secession (Blais et al. 1995, Howe 1998, Mendelsohn 2003, Bélanger and Perrella 2007, Blais and Nadeau 1992). The so far consistent finding states that cost-benefit calculations on the potential economic consequences of the sovereignty do play a role in explaining support for independence in Quebec. Costa-i-Font and Tremosa (2006) fail to identify such an effect in the case of the Basque Country, although they use a much more indirect measure of cost-benefit calculations. Indeed, the role of instrumental calculations in explaining aggregate claims for secessionism has been quite convincingly established, for instance in the degree of development of the region with respect to the rest of the state, and the secessionists claims (Sambanis and Milanovic 2011, Sorens 2004). However, there is a fundamental objection that makes these estimates unreliable: economic evaluations of the consequences of the sovereignty might well be a rationalization of individuals preferences on sovereignty itself and, therefore, the relationship would be endogenous (Howe 1998, Mendelsohn 2003). If citzens prospective economic evaluations of sovereignty are a function of their current preferences on the issue, therefore we cannot establish the causal role of cost-benefit calculations on support for independence. Howe (1998) provides compelling evidence that indeed rationalization of preexisting preferences is the main mechanism that explains prospective economic assessments on an independent Quebec, and suggests new indicators to be designed that should be able to capture the exogenous component of these assessments. But, as Howe himself acknowledges, there is a high degree of uncertainty with respect to the economic consequences of separation that, considered together with cognitive biases such as status quo effects (Costa-i-Font and Tremosa 2006) raises serious doubts on the feasibility of an observational estimation of the causal relationship between cost-benefit calculations and support for independence. There are two a priori feasible strategies to solve the endogeneity problem: on the one hand, an instrumental variable estimation as proposed by Howe (1998) that, however, would require the identification of exogenous factors affecting people s perceptions on the economic consequences of independence; and on the other hand a purely experimental setup, in which respondents were randomly presented with different credible scenarios to 3

4 see how they affect attitudes toward secession (Muñoz and Tormos, forthcoming). Nevertheless, in this paper we will use an observational approach that, apart from the problems already mentioned, will provide information about the effects of economic considerations in support for independence. Beyond the issue of endogeneity, the literature has also considered heterogeneous effects of economic calculations. It has been shown that they have a much stronger impact among citizens with weaker Québécois identities than among those with strong attachment to Quebec vis à vis Canada (Howe 1998). Therefore, the economy seems to play an important role in shaping support for sovereignty among those citizens with ambivalent identifications, while strong Quebeckers do not seem to rely on this sort of reasons, or do so much less than weak identifiers. Another interesting interactive effect has been identified for the various political generations in Quebec (Martin 1994). According to this analysis, the impact of economic considerations is clearly weaker for the Québécois baby-boomers, socialized during a period of rapid economic growth than for the pre-war generation, whose attitudes seem to be much more driven by material factors. The role of economic considerations in shaping support for sovereignty has also been argued to depend on a general predisposition towards risk-taking (Nadeau, Martin and Blais 1999): while for risk-acceptant individuals cost-benefit calculations seem to apply, for highly risk averse citizens might give more weight to worst-case scenario type of considerations. Related to this, Duchesne et al. (2003) tested with ecological data some implications related to the risk taking hypotheses. They argue that ceteris paribus those citizens living in areas that could be more deeply harmed by the negative consequences of a hypothetical separation of Quebec from Canada would show less support for secession. Therefore, they argue, dependence on inter-provincial trade and living in an economically homogeneous area hinder support for secession, since diversification is regarded as reducing risks. Although they provide compelling evidence of such a community-level relationship, the specific mechanisms remain to be uncovered through a more fully specified individual or multilevel analysis. In the case of Quebec also costs and benefits with regards to the French language have been identified as a factor driving support for secession (Nadeau and Fleury 1995, Blais et al. 1995, Mendelhson 2003). However, this orientation is probably even more affected by endogeneity than the previous one. Clarke, Kornberg and Stewart (2004), alongside a critique of Nadeau et al. (1999) model, propose and provide evidence for an additional explanation that had already been advanced in earlier works (Pammett and LeDuc 2001, Clarke and Kornberg 1996): the role of intermediating agents such as parties, leaders and governments. According to this argument, identification, closeness or support for specific parties and leaders might have an independent effect on citizens choices with regards to sovereignty. The reason is that, since the consequences of secession are highly uncertain, citizens will resort to available heuristics to overcome the difficulties in evaluating the costs and benefits of the various options. Leaders and parties provide such reliable and available cues (Bartels 1996, Sniderman et al. 1991), and it is reasonable to expect that citizens will use them in a highstakes and high uncertainty decision. Most of the research has so far been cross-sectional, and thus our knowledge on the dynamics of support for independence is very limited. However, in a recent piece, Yale and Durand (2011) analyze data over a long time-span ( ) to show that in the trends 4

5 we can easily identify certain period effects of relevant political events that have a temporary impact on the survey results. Therefore, support for independence would have a conjunctural component linked to the political context which cross-sectional models cannot properly capture. Finally, as an additional explanation not accounted in the literature, we shall consider political radicalism or extremism. Secession is a highly consequential political move, that bears a degree of uncertainty and might have unintended consequences. It represents a radical break with the status quo and affects one of the core objects of the political system (the political community). Therefore, it might be easier for individuals with a stronger tendency to hold more radical ideological stances to position themselves in the pro-independence side, whilst moderated or centrists might feel more comfortable with other options, everything else being equal. The case Given that the literature has been disproportionally focused on the analysis of the Quebec case, we need to make an additional effort to adapt these theoretical contributions to our case study. Logically, the main difference refers to the fact that while in Quebec two referenda have been held, in Catalonia this is a much more distant possibility and, therefore, when citizens express their vote intention they refer to a hypothesis more than to a plausible reality. It is possible that this difference would have an effect on the role of instrumental calculations and risk-aversion attitudes, diminishing its relative importance. With respect to the economic calculations of the consequences of a hypothetical secession, it has to be kept in mind that in the Catalan case the public debate is just beginning, although there have been some recent contributions (Guinjoan and Cuadras 2011) which have received a certain echo in the Catalan and international press (see Ghemawat 2011, and the response by Cuadras and Guinjoan 2011 in the Financial Times). This is why it seems less plausible to think that cost-benefit calculations about a hypothetical independence would have a key role in citizens preferences. However, this does not mean that considerations of an economic sort would not play a relevant part: in Catalonia, the public debate over the economic consequences of the relationship with Spain is intense, and therefore we could expect that it would enter in the equation of support for secession. Though it will do it in a specific format linked to the particular issues of the Catalan political debate, which has mainly revolved around the concept of fiscal deficit. Therefore, we expect that attitudes toward fiscal deficit and the solutions that have been suggested (fiscal sovereignty) would be important in determining support for independence. Another specificity of the Catalan case could refer to the role of ideology and political radicalism. The fact that secessionism has traditionally been supported in an explicit manner from leftist positions could make that, other things held constant, citizens with a more leftist ideological stances would have more probabilities to support independence. Additionally, we think that the framing effects exerted by the mass media should be taken into consideration. As Fernández-i-Marín and López indicate (2010, from a revision of the literature), the concept of framing comes from the theories of communication and refers to the cognitive frame the mass media produce with regard to a particular issue, which has the effect on public opinion of defining a set of conceptual links when thinking 5

6 or mentioning this phenomenon, and discarding others. According to these authors, in Catalonia two distinct cultural frames of reference coexist: the catalanocentric, and the hispanocentric. The catalanocentric cognitive frame transforms the information and cultural stimuli produced exclusively or predominantly in Catalonia into basic references of the social universe of the individual, while in the hispanocentric frame it is Spain and Spanish related issues the reference from which the world is observed and understood (Fernández-i- Marín and Lopez 2010). In this respect, our hypothesis is that the media with a catalanocentric scope will establish a frame of reference in which the links between concepts and the type of issues considered to be relevant will favor support for independence, conversely to what will happen with hispanocentric media. Dependent variable Our dependent variable is support for independence. However, its measurement is not as straightforward as it might seem. In Catalonia, until very recently, the question of secession has not usually been asked directly in publicly available surveys, but embedded within a multi-option question on constitutional preferences in which respondents were to choose among independence, federalism, autonomy (the status quo) or a centralized state. More recently, the official institute for public opinion research in Catalonia, the Centre d Estudis d Opinió, has introduced a dichotomous question on the voting behavior in a hypothetical referendum on its political barometers 2. The exact wording of this question is: P39. And more precisely, if a referendum to decide about the independence of Catalonia was to be held tomorrow, what would you do? Read response options 1. Vote in favor of independence 2. Vote against independence 3. Abstain / would not vote 80. Other responses (DO NOT READ, WRITE THE LITERAL RESPONSE) 98. Do not know 99. Do not answer P39a. For what reason would you <LITERAL P39>? BASE: ANSWER 1, 2 or 3 to P39. In particular, this question is placed just after another one about the constitutional preferences of the respondent. Yale and Durand (2011) have shown that asking for vote intention vis à vis a general orientation is consequential, at least in Quebec. According to their analysis, asking about voting intention rather than a general attitude elicits a weaker support for secession. However, asking about intended behavior seems more politically relevant than a generic attitude. In any case, in the data set we use 3, we have a dichotomous measure on voting intention in a hypothetical referendum together with a multi-option question on the 2 The Institut de Ciències Polítiques i Socials has also included a question in its annual barometers. 3 In this research we use the July 2011 public opinion barometer of the Centre d Estudis d Opinió (CEO). This barometer is a telephone survey representative of the Catalan population above 17 years of age 6

7 preferred constitutional arrangement. At first sight, we observe huge differences in the support for independence elicited by both questions (table 1): while in the voting intention question a 43% of the sample declared that it would support independence, in the multioption question only a 25% of the respondents choose an independent state as their most preferred option. In other words: just a 55% of those that say would vote independence in the referendum, choose secession when given the federal option. 35% of the proindependence would choose a federal solution given the option, while more surprisingly, a 7.4% seem to prefer the current arrangement. This puzzling result might indicate that a part of the sample understands the preferred constitutional arrangement question in terms of the current situation rather than their ideal solution. Table 1: Support for independence and preferred constitutional arrangement Vote in favor Vote against Abstention Total A region An autonomous community A federal state An independent state DK/NA Total Nonetheless, in this article we take advantage of this discrepancy, since, we argue, it captures a distinction between core and weak supporters of independence: those that choose independence in both questions constitute the core support group, while those that support secession but that, when given the option, choose federalism are probably the weak group. In other words, while the first group comprises the hard core of Catalan secessionism, the second one is built up by disenchanted federalist, who in the current political context would support independence. We have considered this distinction in the formulation of our hypotheses and our strategy of analysis since we can probably expect different explanations to hold for these two groups. The argument we pose here is that, while for the group with a stronger preference for independence identity motivations would be comparatively more important, for the group with a weaker preference instrumental calculations would have more relevance. The logic of this argument is twofold: first, given that we can expect national identity to be a rather stable attitude, it is expected that it would be more tightly bound to a congruent and strong preference for independence. In turn, instrumental considerations can foster support for independence, however it would be more likely that they would promote a less intense and nuanced type of support for secession than that grounded on identity motives. The second reason to formulate this hypothesis is related to the particular history of Catalan secessionism which, though traditionally based on identity, in the last years has incorporated more clearly a discourse of an instrumental character. Therefore, identity factors are expected to have more weight for more loyal and traditional secessionists, while (N=2,500). A stratified sampling procedure (by province and city size) was applied and the respondents were selected randomly according to sex and age quotas. The margin of error is ± 2.47 per cent, for a 95 per cent confidence interval (p=q=0.5). 7

8 economic factors will have more capacity to explain support for secession of other segments of the population. Hypotheses Departing from the literature reviewed above, and adapting the arguments and findings of the research so far, we can formulate a set of hypotheses on the determinants of support for independence in Catalonia: H1: National identity will be the first, and strongest predictor of support for independence. Those citizens identifying as primarily or exclusively Catalan will have higher probability of being in favor of independence. H1B: However, national identity will play a stronger role among the core supporters than the weak supporters. Identity-driven support for independence will lead to a strong attitude and therefore it will be more important for core supporters. H2: Economic considerations will also play a role in explaining support for independence. Those citizens foreseeing economic benefits from secession will be, everything else being equal, more in favor of independence. H2B: Contrarily to what we expect for identity, economic considerations will play a more important role in determining support for independence among weak supporters than among core supporters. H3: Indeed, identity mediates the role of economic calculations: they will be more important for citizens with ambivalent identity positions. H4: Everything else being equal, sympathizers of pro-independence parties will be more prone to support independence. This will hold especially for weak supporters, since core supporters will be basically identity-driven and therefore will rely less on partisan cues. H5: Everything else being equal, citizens with a general orientation towards ideological radicalism will have higher odds of supporting independence, especially in the case of core supporters. Independent variables Our main independent variables are national identity, economic considerations, party preferences, ideology and political radicalism, and additionally the framing effect of the mass media. In Catalonia, national identity is usually measured by means of a bipolar scale ranging from only Spanish to only Catalan, through more Spanish than Catalan, both Catalan and Spanish, and more Catalan than Spanish. As reference category, we have defined the central one: both Catalan and Spanish, given that it accumulates a higher amount of response. 8

9 The survey does not contain any explicit question about the economic prospects of an independent Catalonia. However, it includes an open-ended question just after the vote intention at the referendum which asks the respondents what are the motives of their opposition or support to independence. The literature has distrusted introspective formulations, in which the respondent is requested in an open-ended format to provide the reasons of their behavior, to unveil the causes of political behavior. The grounds of it are two: first, voters are not necessarily aware about the true reasons which drive their behavior, and second they could rationalize and refuse to report emotional motives. Blais et al. (1998), show how in the referendum on the sovereignty of Quebec in the year 1995, the response to open-ended questions about the reasons of voting yes or no were good instruments to unveil the relevance of different considerations in the vote. Therefore, we have used the open-ended question, classifying respondents into three groups: those who express economic/instrumental motives of their vote, those who found their vote intention on identity, and those who express other reasons or no reasons. To be able to test hypothesis 3 we also consider the interaction between identity and economic considerations. Testing the hypothesis of the instrumental motivation of support for independence is central to our research, and therefore we have introduced an additional indicator to the introspective question. As mentioned earlier, in Catalonia the debate over the economic consequences of the relationship with Spain has principally revolved around the concept of fiscal deficit and the available alternatives. In that respect, the survey from the CEO incorporates a question to measure attitudes toward a hypothetical fiscal sovereignty. In particular, the respondent is requested to tell how much in favor is s/he of the possibility that the Catalan public administration should collect and decide upon the distribution of all the taxes which are paid by citizens and companies in Catalonia. If the attitude toward financial sovereignty still has a significant effect after being controlled by the rest of the variables in the model, it would mean that economic motivations per se have the capacity to promote support for secession. In the models this variable is included in its ordinal format with four categories. The vote in a hypothetical referendum is embedded in a more general political context related to factors such as political parties, leaders, or the assessment of government performance. As Clarke et al. (2004) note, in a situation of high uncertainty about the possible consequences of an action, as in the case of a referendum, voters look for help in heuristic mechanisms. Political parties and leaders play a key role in providing cues to the electorate about each of the options in a referendum. In this respect, we include party identification or preference as an independent variable. In the survey, party preference is measured with a standard open-ended question which requests the respondent what party does s/he feels more sympathy for. In the models, we use a set of dichotomous variables for each party in the Parliament of Catalonia, keeping as a reference category the preference for CiU. In our models we also include the role of ideology and political radicalism. In Catalonia, secessionist stances have traditionally been held by parties in the left hand side of the political spectrum. For that reason, we could expect a relationship between being leftist and opting for secession. We measure ideology using a 10 point scale, ranging from 1 which is extreme left to 10 which means extreme right. We also take into account the hypothesis of political radicalism. Holding extreme ideological positions can reduce the 9

10 costs of breaking up with the status quo and opting for secession. To measure radicalism we use for each person in the sample, the distance to the ideological centre. To measure the framing effects of mass media, we use a dichotomous variable that splits the sample in two groups: those who follow the news broadcast of Televisió de Catalunya to receive political information (catalanocentrist) and those who follow the rest of TV channels (hispanocentrists). As sociodemographic control variables we include sex, age, education level, parents place of birth, and the language used in daily life, in particular, the language the respondent uses to talk with friends. These two last indicators are of special interest, given their potential effect in support for secession. The place of birth of the respondent or that of the parents of the respondent has shown to have consequences in the patterns of political behavior in Catalonia (Viròs, Canals and Pallarés 1991, Riba 2000 among others). Sons of parents born in Catalonia tend to feel more identified and involved with Catalan institutions than those with parents born in other parts of Spain. Pallarés and Font (1998) indicate that individual and family s place of origin plays a similar role to that of national identification attitudes on party preferences and vote choice in Catalonia. Something similar happens with the language used in daily life. As Ferrer (2008) points out, those who engage in social relations using mainly the Catalan language with family and friends tend to get more involved in the issues of the Catalan political arena, such as voting in elections to the Catalan Parliament. In our models we include the language usually spoken with friends. This is an indicator of peer-group socialization, which also refers to the idea of frame of reference mentioned before. Getting involved in social relationships in which the Catalan language is the main vehicle of communication alludes to a catalanocentrist emphasis, instead of a hispanocentrist one. The connections among concepts, as well as the main issues that may arise in peer-networks in which Catalan is the predominantly spoken language are potentially different to those in which the dominant language is Spanish. Analysis: Explanatory model of support for independence Table 2 presents the outcome of the model that contains the main explanations of support for independence. The results confirm our basic hypotheses. Support for independence is mainly grounded on national identity, but also on instrumental considerations, and on the effect of intermediary agents. The explanatory capacity of the whole model is quite high, with a pseudo-r² of Table 2. Main model of support for independence Coefficients (standard errors) Average marginal effect Confidence interval (95%) Only Spanish A -1.64*** (0.53) More Spanish than Catalan (0.73) More Catalan than Spanish 1.94*** (0.23) Only Catalan 3.11*** (0.30) 10

11 Mentions economic motivations 2.37*** (0.21) Mentions identity motivations 0.70*** (0.21) Ideology -0.14* ,00 (0.09) Distance to the centre ,04 (0.13) Support for fiscal sovereignty 0.41*** ,07 (0.12) Sympathy for PP B -1.56*** (0.59) Sympathy for ERC 0.88** (0.41) Sympathy for PSC -1.25*** (0.32) Sympathy for ICV (0.38) Sympathy for Cs (1.04) Sympathy other secessionists 3.53*** (1.08) No party sympathies (0.26) Follows TVC news broadcasts 0.38** (0.20) Language spoken with friends: Catalan 0.40* (0.21) Age (0.01) Woman (0.18) Parents born in the rest of Spain (0.23) At least a college degree C (0.34) Secondary education (0.19) Constant -3.85*** (0.77) Observations 1827 log likelihood Degrees of freedom 23 Pseudo-r * p<0.1, ** p<0.05, *** p<0.01 Reference: Both Spanish and Catalan. Reference: sympathy for CiU. C Reference: no formal education/primary education. Subjective national identification is the factor with more explanatory capacity, as pointed out by the average marginal effects and the change in the pseudo R² of entering and withdrawing this variable (table 3). As Nadeau et al. (1999) argue, national identification works analogously as party identification in models of electoral behavior. It can be understood as the main force directing political attitudes toward national issues (see Nadeau et al about Wattenberg 1990). In the model we can see how the more Catalan the 11

12 identity of the respondent, the higher his/her probability to support independence. The effect is monotonic and almost lineal across categories. Those who consider themselves more Spanish than Catalans do not differ in their support for secession from the reference category. Despite the explanatory capacity of national identity, the effect of instrumental motivations is not suppressed. Economic considerations mentioned in the open-ended question are intensely related to voting yes, even tough when they are controlled by the rest of the variables in the model. Identity motives are also positively related to the support for independence. Beyond the introspective question, the other indicator we use to measure economic motivations is support for fiscal sovereignty. This indicator has also an independent impact in the decision to vote in favor of secession. Ceteris paribus, the more favorable the respondent is to the possibility that the Catalan administration should collect all taxes, the more likely for him/her to vote in favor of secession in a referendum. We can therefore conclude that support for independence is in part caused by instrumental considerations. However, the model we are presenting here does not preclude the possibility that instrumental considerations were in fact a rationalization of actual preferences about independence, this is, that we face the type of endogenous effect commented by Howe (1998) for the Quebec case. The political ideology of the respondent does also play a role in support for secession. In Catalonia, the national and the left-right ideological axis are not completely orthogonal. The independentist option has been historically promoted from leftist positions, still leaving a lasting imprint until the present day. When we control for the whole set of predictor variables, especially for party preference, finding oneself in the left hand side of the political spectrum seems to foster support for secession. On the other hand, the hypothesis of radicalism does not find support. There does not seem to be a relationship between the distance to the ideological center and support for secession. Including party preferences in the model produces similar results to Clarke et al. (2004) findings about the impact of intermediating agents in the case of Quebec. In cases of decision making under uncertainty conditions, voters look for the cues parties provide. Controlling for the rest of predictor factors, party identification has a considerable effect on the referendum vote, as revealed by the average marginal effects. Those who feel close to ICV and those with no sympathy for any party are similar in their stances with respect to the reference category, preference for CiU. To have a preference for parties that explicitly include support for independence in their electoral manifesto, such as ERC and other secessionist parties (SI, CUP or Reagrupament), clearly increases the probability to vote favorably in a referendum. Conversely, sympathizing with the PSC or the PP, reduces it. Once we control for the rest of the variables in the model, voters that receive information through the Catalan TV news broadcasts (Televisió de Catalunya) do have a higher probability to support independence. Our interpretation of this effect goes beyond the idea that mass media actively campaign for the cause of secession, as intermediary agents such as political parties and leaders can actually do. We tend to support a possible framing effect, in tune with what Fernández-i-Marín and López (2010) maintain. The fact that some media have Catalonia (catalanocentrists) or Spain as reference (hispanocentrists) has consequences on the type of news that receive more attention, and on the conceptual links established when mentioning the different political issues. Linked to the debate about the effects of the frames of reference, we find that the language spoken with friends has a significant effect on support for secession. Previous 12

13 research (Fernández-i-Marín and López 2010, Grup d Estudis de les Identitats 2007) has shown that the mass media people use to receive political information is a less useful predictor than language. In our model both variables have a similar effect. Parental place of birth does not have a significant effect in explaining support for independence once we control for the rest of the variables in the model. Actually, different statistical analysis that use stepwise regression procedures (not shown) indicate that it is the language spoken with friends the key sociological variable, more important than self and parental place of birth, and the language spoken with the family and at work. The rest of control variables finally included in table 2 age, sex, and education are not able to retain their explanatory capacity. Table 3. Specific contribution of each explanatory variable in terms of Pseudo-R² Pseudo-R² Full Model Exclusion of one variable Inclusion of one variable National identity % % Economic motivations % % Identity motivations % % Ideology % % Distance to the center % % Support for fiscal sovereignty % % Party sympathy % % TVC news broadcasts % % Language with friends: Catalan % % Age % % Woman % % Parents born in the rest of Spain % % Education % % All identity indicators % % To quantify the contribution of each variable to the explanation of support for independence, we have calculated the effect in the pseudo-r² of the model of excluding only one each time while keeping the rest in. Taking the difference between the R² which includes all the variables in the model, and the R² of each submodel in which only one variable is selectively excluded, we can have an alternative estimate of the total effect of that independent variable. Using this procedure, we see that the variables that produce larger reductions in the R² are national identity and instrumental motivations expressed through the open-ended question. It calls our attention that both seem to reduce the explanatory capacity of the model in a similar fashion (17.7%). We tend to think that subjective national identity has in fact a larger effect than instrumental considerations. It is 13

14 possible that the strong correlation of identity with other indicators of the model such as party preference, parental origin, or language spoken with friends can retain part of the effects of identity, once it is controlled. Therefore, we decide to exclude this whole group of variables tightly linked to national identity. The reduction in the R² is the highest; it means a 41.7% decrease with respect to the full model. It would indicate that identity factors do actually have a larger weight in support for independence. We can verify this assertion if we include only national identity as explanatory variable. With a pseudo-r² of 0.361, it happens to be 70.9% that of the full model with all variables included. Heterogeneity in the explanatory factors In the main model we have seen that both identity and considerations of instrumental nature, apart from intermediating agents play an important role in the explanation of support for independence in Catalonia. We have seen that the model we have specified has a large explanatory capacity. Moreover, the analysis performed so far assumes only one explanatory model for all kinds of supporters/opponents of independence, though we know that this assumption is not realist. For example, identity and economic factors can have a different weight in determining support for secession within particular groups of citizens, as some studies of the Quebec case suggest (Howe 2003). This is why, in this section we explore the heterogeneity in the explanatory factors of support for independence and, to do it, we introduce some additional nuances to the general model presented earlier. As we have noted in the descriptive section, not all respondents who say they will vote in favor of independence in a hypothetical referendum do actually choose the independent state option when they have to decide among the four different alternatives of the constitutional preferences question. Actually, a significant fraction of supporters of independence, in this question, choose the federal state option. This gives us a key element to distinguish between two groups of supporters of independence, those who have a strong preference and those who have a weak preference. The first group always chooses secession, no matter the alternative options they are offered, while the second group encompasses those citizens that will support secession in a referendum if it presents the alternatives in a dichotomous format, however they are willing to consider other options, as a federal articulation. To make a first contrast of hypotheses H1B and H2B, which reflect on the heterogeneity in the decision process of supporters of secession with a strong preference (based on identity) and weak preference (based on instrumental considerations), we enter in the previous model the constitutional preferences for federalism and an independent state (leaving support for the autonomic state or status quo and regionalism as reference categories), in addition to their interaction with identity motivations on the one hand, and support for fiscal sovereignty on the other. 14

15 Table 4. Interactions of support for independence (1) (2) Model 2 Main effects Model 3 Interactions b/se b/se Only Spanish A -1.77** (0.73) (1.32) More Spanish than Catalan (0.68) (0.66) More Catalan than Spanish 1.50*** 1.46*** (0.25) (0.25) Only Catalan 1.91*** 1.95*** (0.35) (0.35) Mentions economic motivations 2.18*** 2.21*** (0.22) (0.23) Mentions identity motivations 0.73*** (0.24) (0.44) Ideology -0.21** -0.22** (0.10) (0.10) Support for fiscal sovereignty 0.49*** 0.34 (0.14) (0.24) Preference for federalism B 1.13*** (0.24) (1.30) Preference for an independent state 3.78*** 3.74*** (0.40) (1.30) Federalism*identity motivations 0.75 (0.53) Independent state*identity motivations 4.59*** (1.22) Federalism*support for fiscal sovereignty 0.66* (0.35) Independent state*support for fiscal sovereignty (0.38) Sympathy for PP C -1.64*** -2.17*** (0.58) (0.70) Sympathy for ERC (0.41) (0.41) Sympathy for PSC -1.19*** -1.24*** (0.34) (0.35) Sympathy for ICV (0.41) (0.43) Sympathy for Cs (1.01) (0.96) Sympathy other secessionists 2.97*** 2.94*** (1.15) (1.07) No party sympathy (0.28) (0.29) Follows TVC news broadcasts 0.41* 0.41* (0.22) (0.22) Language spoken with friends: Catalan (0.25) (0.25) Age * (0.01) (0.01) Woman (0.21) (0.21) 15

16 Parents born in the rest of Spain (0.26) (0.27) At least a college degree D (0.34) (0.34) Secondary education (0.22) (0.23) Constant -4.67*** -4.05*** (0.86) (1.08) Observations log likelihood Degrees of freedom Pseudo-r * p<0.1, ** p<0.05, *** p<0.01 A Reference: Both Spanish and Catalan B Reference: autonomous community + region C Reference: sympathy for CiU. D Reference: no formal education/primary education. These models, which replicate the general one presented in table 2, show with particular clarity the heterogeneity in the explanatory factors of support for secession existing between the groups we have identified according to their constitutional preferences. The interaction between mentioning identity motivations and preference for an independent state in the multi-option question has a positive and statistically significant coefficient which means that identity motives are more important for the group with a strong preference for independence. Conversely, in the case of those who choose the federal solution as the preferred option, what seems to play a more significant role is the support for fiscal sovereignty: for this second group, the interaction with fiscal sovereignty is the statistically significant and positive one. To facilitate the interpretation of the interactions, we present figures 1 and 2, which show the predicted probabilities of expressing support for independence for the two groups with different constitutional preferences, based on mentioning identity motivations in the first case, and support for fiscal sovereignty in the second. They are simulations performed holding the rest of the variables in the model constant at their means. In figure 1 we can readily appreciate how, while for the federalist group to mention or not identity motivations does not have any effect, for the group with a strong preference it actually does. On the contrary, in figure 2 we can see the opposite effect: support for fiscal sovereignty does not make any statistically significant difference in the probabilities of giving support to independence among those who choose an independent state in the question about constitutional preferences, while it does have a clearly positive effect in the probability of expressing support for independence among federalists. 16

17 Figure 1. Effect of indentity motivations, predicted probabilities 95% CIs Mentions identity motivations Autonomous Community/Region Independent state Federalism ote: Identity motivations are coded as 1 when the respondent mentions them, and 0 when not. Figure 2. Effect of support for fiscal sovereignty, predicted probabilities 95% CIs Support for fiscal sovereignty Federalism Independent state ote: The scale of support for fiscal sovereignty has been inverted with respect to the original one which appears in the questionnaire such that 1 means totally against fiscal sovereignty, 2 rather against, 3 rather in favor, and 4 totally in favor. 17

18 These models give us an important indication about the heterogeneous process by which these two groups establish their support for independence. It seems that while for the group with a strong preference support for secession is mainly based on identity, the group with a weak preference found their support on instrumental considerations, more specifically, in considerations related to fiscal issues. Nevertheless, to properly express this heterogeneity, and see which of the variables in the general model have, indeed, a different role in determining support for secession in the both groups formerly mentioned, it seems more adequate to follow a different strategy, more consistent with the theoretical discussion: we build a typology that distinguishes nonsupporters of independence from supporters with a weak preference, and the group with an intense preference. We use this newly created dependent variable to specify a multinomial logistic regression model to compare each one of these two groups with the non-supporters of independence (those against independence and abstainers). The independent variables we include in this new model are, essentially, the same as those in the general model, with a slight modification to facilitate the interpretation of the outcome: we introduce subjective national identity as a continuous variable (as in the previous models we have seen its effects are monotonic and approximately lineal). Table 5. Typology of supporters of independence. Multinomial logistic regression (reference category: against independence + abstain) Weak pref. Strong pref. Coefficient and Standard Error Average marginal effect Coefficient and Standard Error Average marginal effect Only Spanish A -2.60*** (0.96) (0.65) More Spanish than Catalan *** (0.69) (0.39) More Catalan than Spanish 1.83*** *** 0.13 (0.26) (0.30) Only Catalan 2.08*** *** 0.48 (0.34) (0.36) Support for fiscal sovereignty 0.70*** (0.15) (0.14) Mentions economic motivations 2.27*** *** 0.13 (0.23) (0.24) Mentions identity motivations *** 0.1 (0.25) (0.24) Ideology -0.21** (0.10) (0.10) Distance to the centre ** 0.04 (0.14) (0.16) Sympathy for PP C -2.06*** * (0.68) (0.71) Sympathy for ERC ***