On 21 December 2017, an election to the Catalan regional Parliament took place in a very particular context. On 1 October, the Catalan government organised a referendum on independence for the region from Spain, which the Spanish government and Constitutional Tribunal considered illegal. Over two million Catalans (43% of eligible voters) participated in the vote and over 90% voted in favour of independence (Generalitat de Catalunya 2017a). To try to prevent it, the Spanish government sent riot police and injured over 1000 Catalans who were trying to vote (CatSalut 2017).

Nearly a month after the vote, on 27 October, the Parliament of Catalonia, claiming it had obtained a mandate from voters on 1 October, declared Catalonia an independent state in the form of a republic. In response, the Spanish government invoked Article 155 of the Spanish constitution, which had never been used before, to suspend Catalan autonomy. It also called an election to the Catalan Parliament, hoping to elect a government that would put an end to the independence process. However, it failed and pro-secessionists won another majority of seats (but not of votes), with vote and seat shares that were similar to those they received in the previous regional election, held in 2015 (Generalitat de Catalunya 2017b).Footnote 1 It is in the particular context of this election, that Making Electoral Democracy Work decided to conduct the survey of Catalan voters we present in this article.

In spite of the overall stability in vote and seat shares, there were some minor changes in the outcome. In 2015, the two major pro-independence parties, the Catalan European Democratic Party (PDeCAT) and Catalan Republican Left (ERC), formed a pre-electoral coalition, Together for Yes (JxSí), with the goal of making Catalonia an independent state in spite of resistance by the Spanish government. In 2017, they ran separately. While these two parties nearly evenly split the pro-independence vote on 21 December, PDeCAT, which ran under the name Together for Catalonia (JxC), unexpectedly won more votes than ERC. The far-left pro-independence party Popular Unity Candidacy (CUP), which was not part of the 2015 independence alliance, lost half of its vote share. Moreover, the 2017 leftist coalition Catalonia in Common-We Can (Catalunya en Comú-Podem) won a smaller vote share than its 2015 counterpart, Catalonia Yes We Can (CSQP). Among pro-unity parties, support for the conservative People’s Party (PP), which currently governs Spain, plummeted, while Citizens (Cs), which is also strongly opposed to independence, surged in 2017 and won the most votes and seats. The Catalan Socialists’ Party (PSC) increased its vote share by one percentage point. Finally, it is important to note that turnout in 2017, 79.09%, beat the previous record of 74.95%, set in 2015 (Generalitat de Catalunya 2017b).

Table 1 shows the results of the past two elections.

Table 1 Percentage of Votes Won in 2015 and 2017.
Full size table

This election raises several questions that are central to the study of citizens’ voting behaviour. Most fundamentally, to what extent do long-term factors, such as demographic characteristics and underlying attitudes, on the one hand, and short-term factors, like reactions to political events, on the other hand, influence vote choice? The stability in the vote and seat shares of pro- and anti-secessionist parties between the 2015 and 2017 elections in spite of the major events that took place between the two votes suggests that long-term factors play a role in stabilizing vote choice. Another important question is how did citizens react to the unusual decisions by both the Catalan and Spanish governments? A key aspect of democracy is that elites and citizens accept the result even when their side loses (Anderson et al. 2005; Lago and Martínez 2017). The Catalan government held a referendum on independence that was illegal under the established rules, while the Spanish government took the unusual and unprecedented step of suspending Catalan autonomy and calling an election to the Catalan Parliament. Were citizens willing to accept losing when both the national and regional governments were taking actions that depart from the way democracy usually works?

It is also important to understand the extent to which citizens’ opinions are crystallised, on the one hand, or influenced by parties (Cohen 2003; Guntermann 2017) and question wording (Schuman and Presser 1996), on the other hand. This question is particularly important in the context of the independence crisis. A solution to the conflict over Catalan independence depends partly on how crystallised opinions are. If opinions depend to a large extent on context, it may be possible to find a solution to the conflict in Catalonia. The survey included four experiments that assess the impact of party conflict on secessionist attitudes. Guntermann (2017) found evidence that such conflict leads to polarisation in nationalist attitudes. If conflict is still polarising attitudes, a reduction in the intensity of party conflict may make Catalans more open to a compromise solution.

The Making Electoral Democracy Work (MEDW) project was a seven-year cross-national research project funded by the Social Sciences and Humanities Research Council of Canada and directed by André Blais (Blais 2010). Its goal was to examine how the rules of the game affect both parties’ and voters’ behaviour. The project was innovative in studying these actors’ behaviour at various levels (local, regional, national and supranational). MEDW conducted pre- and post-electoral surveys in Catalonia during the Spanish national election in 2011, the Catalan regional election in 2012 and the European election in 2014.Footnote 2 Wanting to understand how the exceptional political context of the recent Catalan election influenced voting behaviour, we decided to run a survey right before the election. The dataset and documentation are attached to the online version of this article.

We recruited a representative sample of 1500 respondents through Survey Sampling International (SSI). While no hard quotas were set, SSI sought to ensure the sample was representative on age, sex, and province of residence (Catalonia is composed of four provinces). Respondents participated in an online survey programmed in Qualtrics. Distributions of key variables are similar to those in the population. The survey was conducted in the eight days leading up to the election (from 17 December to 21 December 2017). We allowed respondents to select in which of the official languages of Catalonia, Catalan and Spanish, they preferred to answer the survey.

Figure 1 compares the distribution of vote choice in the survey data with the actual election result. We can see that, in spite of some minor biases, the sample broadly represents actual vote choices. Users who want to correct for sampling biases can use one of two types of weights that are included in the dataset. The first weights by demographics (age, sex, education, place of residence) and turnout. The second weights by vote choice in addition to these variables.

Fig. 1
figure 1

Comparison of sample vote choice with the election result

Full size image

To our knowledge, it was the only academic survey run so close to the election. It was also the only survey run in the context of the 21 December election that included questions about independence. The fact that the survey was run so close to the election is important, because vote intentions within the pro- and anti-independence blocs changed considerably in the month preceding the election.Footnote 3 In particular, support for Together for Catalonia, the pro-independence coalition led by deposed leader Carles Puigdemont, and for anti-independence Citizens increased in the weeks leading up to the election.

The survey asked basic demographic questions (age, sex, education, income, first language, and place of residence; see Appendix for a full list of variables) as well as standard questions on political attitudes, turnout, and party preferences (see lists of questions in Appendix). Demographic questions notably allow scholars to see how strongly independence support is rooted in long-term factors, especially language. Previous research has shown that demographic variables are strong predictors of vote choice, of support for independence, and of nationalist attitudes more generally (Miley 2007; Muñoz and Guinjoan 2013; Orriols and Rodon 2016; Rodon and Guinjoan 2018; Verge et al. 2015).

Figure 2 shows vote choice by first language. It distinguishes those whose first language was Catalan from those who grew up learning Spanish, both languages or some other language. Parties are ordered from left to right from the most anti-independence to the most pro-independence party. We can see that Catalans whose first language is Catalan are much more likely to support pro-independence parties than other Catalans. Conversely, non-Catalan speakers are much more likely to support anti-independence parties.

Fig. 2
figure 2

Vote choice by first language

Full size image

The study included numerous questions on nationalism and independence. It asked respondents to place themselves on two measures of national identity, the Linz–Moreno scale (Linz et al. 1981; Moreno et al. 1998) as well as separate scales assessing identification with Catalonia and Spain. The Linz–Moreno scale is included in most surveys about voting behaviour and political attitudes in Catalonia. It asks respondents to describe their identity by placing themselves in one of five categories: ‘Only Catalan’, ‘More Catalan than Spanish’, ‘Equally Catalan and Spanish’, ‘More Spanish than Catalan’, and ‘Only Spanish’. While positions on the Linz–Moreno scale are strongly associated with attitudes towards independence in Catalonia (Serrano 2013), Guinjoan and Rodon (2015) have shown that separate national identification scales better account for identities, particularly Spanish identity, in Catalonia. Figure 3 shows the distributions of responses to the Linz–Moreno question as well as to the Spanish identification scale in the survey. The scale runs from 0 to 10, where 0 indicates no attachment to Spain and 10 indicates strong attachment. Responses to the former (panel a) suggest that Catalans overwhelmingly have ambivalent dual identifications, while a minority identify strongly with Catalonia. The Spanish identification scale shows that a significant minority of Catalans have a strong identification with Spain. This result helps us understand why the most anti-secessionist party, Citizens, won the most votes and seats in the 21 December election. Otherwise, it would be hard to account for the intense polarisation that currently exists in the region.

Fig. 3
figure 3

Distribution of measures of national identification. a Linz–Moreno and b scale of identification with Spain

Full size image

Respondents were also asked to indicate their support for independence on a scale from 0 to 10 where 0 means they totally oppose and 10 that they totally support it. This is a novel measure of support for secession in Catalonia that will allow scholars to understand how crystallised support for independence is. This question shows how divided Catalans are about this issue. Figure 4 shows the distribution of responses to the independence scale question. While conventional dichotomous measures show that the population is close to evenly divided on the issue of independence, the 11-point scale shows how intense the divide is. About a third of Catalans very strongly oppose independence, another third very strongly support it, and the rest of the population is somewhere in between. Thus, this question provides rich information that conventional questions fail to pick up on.

Fig. 4
figure 4

Distribution of independence scale

Full size image

Two question-wording experiments also assess responses to different questions on independence. The survey includes an experiment for a binary independence question, one asking about support for Catalonia becoming an independent state, the other asking about separation from Spain. It also includes an experiment providing respondents with four options for Catalonia’s relationship with Spain. One group was presented with the options of an autonomous community with less autonomy than it has now, an autonomous community with the same autonomy it has now, an autonomous community with more autonomy than it has now, and an independent state. The other group had the same options except that instead of an autonomous community with more autonomy, they were given the option of a state in a federal Spain.

Various questions have been asked about independence in surveys in recent years. Our experiments allow researchers to see how levels of support for independence vary based on the question. Furthermore, they allow scholars to determine which groups’ support for independence depend more on the question asked as well as which groups’ independence positions are more crystallised and are invariant to question-wording effects. Prior research has shown that answers to survey questions depend on the way questions are asked (Schuman and Presser 1996). Our project allows scholars to test these effects on independence attitudes using experiments. Four other experiments also allow researchers to assess the impact of party cues on attitudes towards independence and on perceptions of objective factors. Previous experimental research in Catalonia has shown that cues influence nationalist attitudes (Guntermann 2017).

The survey also asked several standard questions about evaluations of the economy and of various political actors. In addition to a standard question about retrospective economic evaluations, the survey asked respondents to evaluate each party’s level of corruption, a major issue in Spanish politics. Previous research has shown that such economic and political evaluations influence vote choice in Catalonia and in Spain more generally (Bosch 2016; Fraile and Lewis-Beck 2010; Orriols and Rodon 2016). For a better understanding of how the literature on voting behaviour and elections applies to multi-level contexts like Spain, we encourage readers to consult the recent book Multi-Level Electoral Politics (Golder et al. 2017).

The survey also asked questions assessing perceptions of Catalonia’s economic relationship with the rest of Spain as well as the state of the economy in an independent Catalonia. Research by Muñoz and Tormos (2015) has shown that such considerations influence support for Independence in the region. We also asked respondents to evaluate the role of various actors’ (the Catalan and Spanish governments, the European Union, the main parties) in the recent independence crisis.

Of particular interest in the context of the conflict over independence are questions about respondents’ reactions to four major events that took place in fall 2017. People were asked to evaluate four decisions made by the Catalan and Spanish governments: the Catalan government’s decision to call the illegal referendum, the Spanish government’s decision to send the police to prevent the referendum, the Catalan government’s decision to declare independence, and the Spanish government’s decision to suspend Catalan autonomy. As Table 2 shows, Catalans were strongly divided on these issues. A large majority were critical of the Spanish government’s actions. A minority of them though supported the Spanish government’s decision to send the police to prevent the referendum as well as the decision to suspend Catalan autonomy. The table also shows that these attitudes are largely, but not entirely, rooted in attitudes towards independence. A non-negligible minority of pro-independence voters did not agree with the actions taken by the Catalan government and many anti-independence Catalans failed to support the actions of the Spanish government.

Table 2 Support for government decisions by independence position (% supporting each decision)
Full size table

The survey also shows that many Catalans who had voted for the left-wing Catalonia Yes We Can coalition, which was ambivalent about the independence issue, in 2015, shifted to a pro-independence party in 2017 if they adopted pro-independence positions on these issues (not shown). Thus, these questions allow us to see that, while long-term factors play a major role in explaining public opinion and voting behaviour, short-term factors can lead people to change their minds.

Finally, one particularity of the survey is that respondents were asked whether they would be willing to accept a government formed by the other side of the independence divide. The objective was to determine to what extent voters are willing to accept losing an election in such a polarised political context. As Table 3 shows, less than half of Catalans who voted for parties that did not support independence stated that they would be willing to accept a government formed by pro-independence parties even if the latter had won a majority of seats in the Catalan parliament. A similar proportion of anti-independence party supporters said they would accept a government formed by the pro-independence parties even if they won a majority. These results suggest that, in a crisis that questions the established rules of the game, many losers may not accept that they have lost the election. Since a key aspect of democracy is losers’ willingness to accept the result, these findings are worrying for democrats. We believe that an interesting avenue for research would be to analyse in depth the attitudes of Catalans who stated that they would refuse to accept that they have lost in order to determine what distinguishes them from those who would accept a government formed by the other side.

Table 3 Willingness to accept a government formed by the other side (%)
Full size table

The survey also asked a variety of other questions, notably on populist attitudes and on obeying the law even when one considers it unfair. These and other questions should allow scholars to answer a large number of important research questions.

In short, we have prepared this survey to answer some key questions about voting behaviour and public opinion in such an exceptional context. The codebook and the dataset were entirely translated from Catalan and Spanish to English in order to allow researchers from both Catalonia and elsewhere to use them. Nevertheless, it is still possible to identify the language in which respondents answered the survey. The dataset and codebook are attached to the online version of this article. It is our hope that numerous scholars will benefit from the dataset to produce research that will help enhance our understanding of the Catalan situation as well as voting behaviour and public opinion more generally.