Dual voting and second-order effects in the quasi-simultaneous 2019 Spanish regional and national elections

INTRODUCTION[Up]

The 2019 Spanish regional elections were held on Sunday May 26^th, in twelve out of the seventeen autonomous communities that make up Spain^‍‍[1], on the same day as the local and the European elections. They took place against the backdrop of a fragmented and unstable political landscape in the country, and barely one month after a snap general election held on April 28th. This is the first time that regional elections have happened so soon after the national poll in Spain, just 28 days later. In the literature, it is well established that the timing of regional elections in the national political cycle matters: regional voting patterns tend to deviate less from voting patterns in national elections when regional and national elections are held simultaneously (Jeffery, Charlie and Dan Hough. 2006. “Devolution and Electoral Politics: Where Does the UK Fit In?”, in Dan Hough and Charlie Jeffery (eds.), Devolution and Electoral Politics. Manchester: Manchester University Press.‍Jeffery and Hough, 2006) or close in time (Schakel, Arjan H. 2013. “Congruence Between Regional and National Elections”, Comparative Political Studies, 46 (5): 631-662. Available at: https://doi.org/10.1177/0010414011424112‍Schakel, 2013). In addition, regional election outcomes may display more or less “second-order features’ depending on the time that has elapsed between national and regional elections.

This article analyzes the degree to which the 2019 Spanish regional elections may be considered second-order contests by analyzing vote-switching between national and regional elections at both the aggregate and individual levels. Specifically, it first focuses on the degree of (dis)similarity between regional and national election results, and then explores why people voted differently in the quasi-simultaneous 2019 regional and national elections. We argue that voters that split their ticket between the regional and the national elections are susceptible to be classified into two groups: (1) those who chose to support different parties because they consider that the two arenas are independent and (2) those who treat regional elections as second-order elections (Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011).

We contribute to the literature in two ways. First, while most empirical studies of the second-order model draw upon the use of aggregate data such as election results, we rely on individual-level data to complement the aggregate-level analysis. Second, previous research has tended to use congruence of the vote between national and regional elections as an indicator of the “nationalization” of regional elections (Pallarés, Francesc and Michael Keating. 2003. “Multi-Level Electoral Competition. Regional Elections and Party Systems in Spain”, European Urban and Regional Studies, 10 (3): 239-255. Available at: https://doi.org/10.1177/09697764030103005‍Pallarés and Keating, 2003). However, similarities in party vote shares in national and regional elections do not necessarily imply the subordination of the former to the latter. Indeed, the overwhelming majority of voters typically support the same party in both regional and national elections and, in our view, they do so either as a result of partisanship or because they prefer similar policies to be implemented at different tiers of government. In this study, we use a different approach by focusing on dual voters; that is, voters that voted differently across the 2019 quasi-simultaneous regional and national elections.

The remaining of this article proceeds as follows. Section 2 introduces the theoretical framework. Section 3 describes the main characteristics of the Spanish regional governments and the context under which the 2019 regional elections were held. Section 4 provides an aggregate-level analysis of congruence between the 2019 national and regional elections and identifies patterns of second-order voting. Section 5 moves forward to an individual-level analysis of dual voting. Section 6 concludes with a discussion of the results.

THEORETICAL FRAMEWORK: SECOND-ORDER ELECTION EFFECTS IN A QUASI-SIMULTANEOUS ELECTORAL SETTING[Up]

The nationalization of regional elections is a possible consequence of the temporal proximity between regional and national elections. Regional elections may be more influenced by —and more subordinate to— national politics when they are held simultaneously or close to the general election (Jeffery, Charlie and Dan Hough. 2006. “Devolution and Electoral Politics: Where Does the UK Fit In?”, in Dan Hough and Charlie Jeffery (eds.), Devolution and Electoral Politics. Manchester: Manchester University Press.‍Jeffery and Hough, 2006; Schakel, Arjan H. 2013. “Congruence Between Regional and National Elections”, Comparative Political Studies, 46 (5): 631-662. Available at: https://doi.org/10.1177/0010414011424112‍Schakel, 2013). When regional elections are held at the same time or shortly after or before the national election, we may expect greater media attention focused on the national level. In addition, quasi-simultaneity may reduce the autonomy of the regional branches of state-wide parties, and may facilitate coordinated election campaigns, giving room to a greater presence of national issues and actors in regional campaigns (García Hípola, Giselle. 2014. Estrategias de comunicación política en contextos concurrenciales: las campañas electorales de 2008 y 2012 en Andalucía. Universidad de Granada.‍García Hípola, 2014; Cabeza, Laura, Braulio Gómez and Sonia Alonso. 2017. “How National Parties Nationalize Regional Elections: The Case of Spain”, Publius: The Journal of Federalism, 47 (1): 77-98. Available at: https://doi.org/10.1093/publius/pjw030‍Cabeza et al., 2017).

The second-order election model, the dominant approach in the study of regional elections, makes three general predictions about election results (Reif, Karlheinz and Hermann Schmitt. 1980. “Nine Second-Order National Elections. A Conceptual Framework for the Analysis of European Election Results”, European Journal of Political Research, 8 (1): 3-44. Available at: https://doi.org/10.1111/j.1475-6765.1980.tb00737.x‍Reif and Schmitt, 1980). The first and most basic prediction of the model refers to voter turnout, which tends to be consistently lower in regional elections than in national ones (Henderson, Ailsa and Nicola McEwen. 2010. “A Comparative Analysis of Voter Turnout in Regional Elections”, Electoral Studies, Special Symposium: Voters and Coalition Governments, 29 (3): 405-16. Available at: https://doi.org/10.1016/j.electstud.2010.03.012‍Henderson and McEwen, 2010; Schakel, Arjan H. and Charlie Jeffery. 2013. “Are Regional Elections Really “Second-Order’ Elections?”, Regional Studies, 47 (3): 324-341. Available at: https://doi.org/10.1080/00343404.2012.690069‍Schakel and Jeffery, 2013). This is mainly because voters perceive that there is “less at stake” in regional contests. The second prediction of the second-order election model is that national government parties tend to suffer electoral losses in regional elections. This happens because voters often use regional elections as an occasion to express their dissatisfaction with the performance of the national government and to punish the national incumbent party. Finally, the second-order election model predicts a better electoral performance of small, new and radical parties in elections at the regional level, due to the fact that voters are more likely to cast expressive votes in regional elections. Since these elections are not as important as national ones, people are more prone to voting “with their hearts rather than their heads” (Marsh, Michael. 1998. “Testing the Second-Order Election Model after Four European Elections”, British Journal of Political Science, 28 (4): 591-607. Available at: https://doi.org/10.1017/S000712349800026X‍Marsh, 1998).

Yet, when national and regional elections occur within a few weeks some of these regularities may become more or less pronounced. At the aggregate level, and following the first prediction of the second-order election model, we expect to find lower levels of voter turnout in the 2019 regional elections compared to the national one. In this case, the close temporal proximity with national elections may accentuate this specific second-order election effect. When voters are required to vote too often, they could experience fatigue and participation may decline further (Schakel and Dandoy, 2014; Garmann, Sebastian. 2017. “Election Frequency, Choice Fatigue, and Voter Turnout”, European Journal of Political Economy, 47: 19-35. Available at: https://doi.org/10.1016/j.ejpoleco.2016.12.003‍Garmann, 2017). Regarding the second prediction, it has been widely noted in the literature that the electoral losses of the national government party in regional elections tend to be stronger at the middle of the national electoral cycle (Reif, Karlheinz. 1984. “National Electoral Cycles and European Elections 1979 and 1984”, Electoral Studies, 3 (3): 244-255. Available at: https://doi.org/10.1016/0261-3794(84)90005-2‍Reif, 1984; Marsh, Michael. 1998. “Testing the Second-Order Election Model after Four European Elections”, British Journal of Political Science, 28 (4): 591-607. Available at: https://doi.org/10.1017/S000712349800026X‍Marsh, 1998; Romanova, Valentyna. 2014. “The Principle of Cyclicality of the Second-Order Election Theory for Simultaneous Multi-Level Elections”, Politics, 34 (2): 161-179. Available at: https://doi.org/10.1111/1467-9256.12016‍Romanova, 2014). Indeed, national government parties may enjoy a short “honeymoon period” if the second-order election takes place immediately after their electoral success at the national election (Dinkel, Reiner. 1977. “Der Zusammenhang zwischen Bundes- und Landtagswahlergebnissen”, Politische Vierteljahresschrift, 18 (2/3): 348-359.‍Dinkel, 1977; Hix, Simon and Michael Marsh. 2007. “Punishment or Protest? Understanding European Parliament Elections”, Journal of Politics, 69 (2): 495-510. Available at: https://doi.org/10.1111/j.1468-2508.2007.00546.x‍Hix and Marsh, 2007; Jeffery, Charlie and Dan Hough. 2001. “The Electoral Cycle and Multi-Level Voting in Germany”, German Politics, 10 (2): 73-98. Available at: https://doi.org/10.1080/772713264‍Jeffery and Hough, 2001; Marsh, Michael. 1998. “Testing the Second-Order Election Model after Four European Elections”, British Journal of Political Science, 28 (4): 591-607. Available at: https://doi.org/10.1017/S000712349800026X‍Marsh, 1998; Schmitt, Hermann and Ilke Toygür. 2016. “European Parliament Elections of May 2014: Driven by National Politics or EU Policy Making?”, Politics and Governance, 4 (1): 167-181. Available at: https://doi.org/10.17645/pag.v4i1.464‍Schmitt and Toygür, 2016). Thus, in the 2019 regional elections, we do not expect that the national government party (PSOE) losses votes compared to the previous national election, but we expect it to benefit from the “honeymoon” effect. Finally, quasi-simultaneity may also affect the electoral prospects of small and new political parties at regional elections. Although this kind of parties tend to perform better in second-order elections over the whole national electoral cycle, their gains are greatest around mid-term, when the popularity of the national government party tends to be at its lowest point (Jeffery, Charlie and Dan Hough. 2001. “The Electoral Cycle and Multi-Level Voting in Germany”, German Politics, 10 (2): 73-98. Available at: https://doi.org/10.1080/772713264‍Jeffery and Hough, 2001). Thus, at the aggregate level, we expect that the honeymoon period enjoyed by the national government party will mitigate the electoral gains of small and new parties in the 2019 regional elections. Before proceeding with the analysis, the next section briefly describes the main characteristics and development of the regional governments and regional elections in Spain.

REGIONAL GOVERNMENT AND REGIONAL ELECTIONS IN SPAIN[Up]

For much of its recent history, Spain has been a hyper-centralized state, where territorial diversity was not recognized. The decentralization process began with the transition to democracy after four decades of dictatorship and the proclamation of the Constitution in 1978, which ensured “the right to autonomy of the nationalities and regions” that make up the country (Art. 2). In a relatively short period of time, Spain became a highly decentralized country to the extent of being widely considered a “federation in all but name” (Elazar, Daniel J. (ed.). 1991. Federal Systems of the World: A Handbook of Federal, Confederal and Autonomy Arrangements. Harlow: Longman Group UK Limited.‍Elazar, 1991: 227; Hueglin, Thomas O., and Alan Fenna. 2015. Comparative Federalism: A Systematic Inquiry. 2nd edition. Toronto: University of Toronto Press.‍Hueglin and Fenna, 2015: 9). Today, Spain is made up of seventeen regions called Autonomous Communities. Although asymmetry has been a persistent element in the development of the Spanish Statute of Autonomies, nowadays all regions enjoy a considerable degree of autonomy (Novo, Ainhoa, Sergio Pérez and Jonatan Garcia. 2019. “Building a Federal State: Phases and Moments of Spanish Regional (de)Centralization”, Italian Political Science Review / Rivista Italiana di Scienza Politica, 49 (3): 263-277. Available at: https://doi.org/10.1017/ipo.2018.23‍Novo et al., 2019). Each region has a single-chamber Parliament, whose members are elected for a four-year term by universal suffrage. Since 1979, there has been a total of 177 regional elections in Spain. Each region has held at least ten elections, but Catalonia leads the ranking with twelve. A few regions, such as Andalusia, Catalonia, Galicia and the Basque Country, have their own electoral calendar. The others hold their elections simultaneously on the same day (horizontal simultaneity), and in concurrence with local elections in all municipalities throughout the country (vertical simultaneity)^‍‍[2]. Occasionally, local and regional elections have also been held together with the elections to the European Parliament. So far, this has happened only three times: in 1987, 1999 and 2019. Regional and national elections in Spain rarely coincide. Indeed, regional and national elections have been held on the same day on only six occasions: five times in Andalusia (1986, 1996, 2000, 2004 and 2008) and one in the Valencian Community (2019).

Regional parliaments vary in size from 33 members in La Rioja to 135 members in Catalonia. The number of parties that make it to the regional parliaments also varies greatly between regions (Gómez, Braulio, Sonia Alonso and Laura Cabeza. 2019a. Regional Manifestos Project Dataset (Version 11/2019). Available at: www.regionalmanifestosproject.com‍Gómez et al., 2019a). For instance, while in the Canary Islands (1983-‍1987) or Navarre (1979-‍1983) as many as nine different parties obtained representation, only two parties won seats in the parliaments of Castile-La Mancha (from 1983 to 1987 and from 1999 to 2015) or Extremadura (from 2007 to 2011). In this regard, the results of the 2019 regional elections analyzed here reflect Spain’s increasing political fragmentation: on average, roughly six parties entered regional parliaments, the highest number ever recorded (Figure 1). The most significant consequence of the fragmentation brought about by this new multiparty system is the increasing complexity of the formation of governments. The emergence of new parties in the aftermath of the Great Recession has led to a profound transformation of the Spanish party system at both national (Simón, Pablo. 2016. “The challenges of the new Spanish multipartism: government formation failure and the 2016 general election”, South European Society and Politics, 21 (4): 493-517. Available at: https://doi.org/10.1080/13608746.2016.1268292‍Simón, 2016; Orriols, Lluís and Guillermo Cordero. 2016. “The Breakdown of the Spanish Two-Party System: The Upsurge of Podemos and Ciudadanos in the 2015 General Election”, South European Society and Politics, 21 (4): 469-492. Available at: https://doi.org/10.1080/13608746.2016.1198454‍Orriols and Cordero, 2016) and regional levels (Gómez, Braulio and Ignacio Urquizu. 2015. “Political Corruption and the End of two-party system after the May 2015 Spanish Regional Elections”, Regional and Federal Studies, 25 (4): 379-389. Available at: https://doi.org/10.1080/13597566.2015.1083013‍Gómez and Urquizu, 2015; Rodon, Toni and María José Hierro. 2016. “Podemos and Ciudadanos Shake up the Spanish Party System: The 2015 Local and Regional Elections”, South European Society and Politics, 21 (3): 339-357. Available at: https://doi.org/10.1080/13608746.2016.1151127‍Rodon and Hierro, 2016; Scantamburlo, Matthias, Sonia Alonso and Braulio Gómez. 2018. “Democratic Regeneration in European Peripheral Regions: New Politics for the Territory?”, West European Politics, 41 (3): 615-639. Available at: https://doi.org/10.1080/01402382.2017.1403148‍Scantamburlo et al., 2018; Leonisio, Rafael and Matthias Scantamburlo. 2019. “La competición política en el País Vasco, 1980-2016. El equilibrio entre la dimensión económica y la nacionalista”, in Braulio Gómez, Laura Cabeza and Sonia Alonso (eds.), En busca del poder territorial. Cuatro décadas de elecciones autonómicas en España. Madrid: Centro de Investigaciones Sociológicas.‍Leonisio and Scantamburlo, 2019). Since the beginning of the 1980s, Spanish politics has been dominated by competition between the two big mainstream parties of the center-left and center-right, the People’s Party (PP) and the Spanish Socialist Workers’ Party (PSOE). This has also been the case in most Spanish regions, albeit the presence and strength of regionally-based parties (Pallarés, Francesc and Michael Keating. 2003. “Multi-Level Electoral Competition. Regional Elections and Party Systems in Spain”, European Urban and Regional Studies, 10 (3): 239-255. Available at: https://doi.org/10.1177/09697764030103005‍Pallarés and Keating, 2003). Spain’s two-party system began to shatter in 2014 with the rise of two new parties on the left and center-right, Podemos and Ciudadanos. The emergence of these parties conferred an exceptional character to the 2015 regional elections, which marked the starting point of a new electoral cycle in Spain (Lagares Diez, Nieves, Carmen Ortega and Pablo Oñate. 2019. “La relevancia de las elecciones autonómicas de 2015 y 2016 en el contexto de un sistema multinivel en crisis”, in Nieves Lagares Diez, Carmen Ortega and Pablo Oñate (eds.), Las elecciones autonómicas de 2015 y 2016. Madrid: Centro de Investigaciones Sociológicas.‍Lagares Diez et al., 2019; Scantamburlo, Matthias. 2019. “Who Represents the Poor? The Corrective Potential of Populism in Spain”, Representation, 55 (4): 415-434. Available at: https://doi.org/10.1080/00344893.2019.1669692‍Scantamburlo 2019). Yet, the newest arrival has been the radical right party Vox, that in December 2018 secured 12 seats in the Andalusian Parliament and five months later entered the national parliament for the first time. These achievements represented the end of Spanish exceptionalism in Europe as a country free from the populist far right (Alonso, Sonia and Cristóbal Rovira Kaltwasser. 2015. “Spain: No Country for the Populist Radical Right?”, South European Society and Politics, 20 (1): 21-45. Available at: https://doi.org/10.1080/13608746.2014.985448‍Alonso and Kaltwasser, 2015).

Figure 1.

Average number of parties that have entered regional parliaments

Note: Average in the regions that hold elections on the same day.

Source: Own elaboration with data from Archivo histórico electoral, Argos.

The 2019 regional elections, which are the focus of this article, are characterized by both horizontal simultaneity (twelve regions went to the polls) and vertical simultaneity with other second-order elections (they were held on the same day as the local elections, as well as the elections to the European Parliament). Indeed, because of that concurrence of elections on the same date, many have labelled the election day as a “Super Sunday”^‍‍[3]. Both vertical and horizontal simultaneity may have significant implications for election results. On the one hand, turnout tends to increase when regional elections are held concurrently with other elections (Schakel, Arjan H. and Régis Dandoy. 2014. “Electoral Cycles and Turnout in Multilevel Electoral Systems”, West European Politics, 37 (3): 605-623. Available at: https://doi.org/10.1080/01402382.2014.895526‍Schakel and Dandoy, 2014). On the other hand, second-order effects are more pronounced when regional elections are held on the same date in several subnational units of a country since this gives them nationwide resonance (Schakel, Arjan H. and Charlie Jeffery. 2013. “Are Regional Elections Really “Second-Order’ Elections?”, Regional Studies, 47 (3): 324-341. Available at: https://doi.org/10.1080/00343404.2012.690069‍Schakel and Jeffery, 2013). However, if there is one thing that distinguished the 2019 regional elections, it is their unprecedented close proximity in time with respect to the national election. They were held only 28 days away from the Spanish general election, while the average number of days between regional and national elections in the Spanish ordinary regions is 267. Does the proximity with the national election had consequences for the 2019 election results? The next section explores the effect of this vertical quasi-simultaneity on the level of congruence between national and regional election results, and to what extent the results of the 2019 regional elections fit the predictions of the second-order election model.

AN AGGREGATE-LEVEL ANALYSIS: CONGRUENCE OF THE VOTE AND SECOND-ORDER EFFECTS[Up]

Election congruence[Up]

According to previous research, vertical and horizontal simultaneity increases the level of congruence between regional and national election results (Schakel, Arjan H. 2013. “Congruence Between Regional and National Elections”, Comparative Political Studies, 46 (5): 631-662. Available at: https://doi.org/10.1177/0010414011424112‍Schakel, 2013). For the 2019 regional elections, then, we should find the lowest dissimilarity scores in the Spanish democratic history. Not only were these elections held in twelve different regions, but they also took place on the same day as two other second-order elections (local and European elections) and quasi-simultaneously with the national election. One of the most common indicators to measure the level of nationalization of regional elections is the dissimilarity index (Jeffery, Charlie and Dan Hough. 2003. “Regional Elections in Multi-Level Systems”, European Urban and Regional Studies, 10 (3): 199-212. Available at: https://doi.org/10.1177/09697764030103002‍Jeffery and Hough, 2003; Pallarés, Francesc and Michael Keating. 2003. “Multi-Level Electoral Competition. Regional Elections and Party Systems in Spain”, European Urban and Regional Studies, 10 (3): 239-255. Available at: https://doi.org/10.1177/09697764030103005‍Pallarés and Keating, 2003). Taking absolute values, this index subtracts the vote shares of a party participating in different elections, subsequently sums them up and divides the sum by two in order to avoid double counting (the loss of one party is the gain of another). Following Schakel (Schakel, Arjan H. 2013. “Congruence Between Regional and National Elections”, Comparative Political Studies, 46 (5): 631-662. Available at: https://doi.org/10.1177/0010414011424112‍2013) dissimilarity is calculated as follows:

where iN stands for the percentage of the vote won by party i in the national election N and iR stands for the percentage of the vote won by party i in the closest regional election R. Scores may vary from 0 per cent, indicating complete congruence or similarity (i.e. nationalization), to 100 per cent, indicating complete incongruence or dissimilarity (i.e. regionalization). While the index allows different types of analytical combinations with respect to the type of election (national or regional) in conjunction with the territorial unit of analysis (national or regional), what matters for our analysis is the comparison between the national and the regional vote in a given region. Such a combination keeps the regional electorate constant but varies the type of election so that the effects of dual voting are incorporated.

Figure 2 represents the average dissimilarity and number of days between regional and national elections since 1983, when the first round of regional elections took place in the Spanish ordinary regions. Against the predictions of the literature, the temporal proximity to the national elections did not translate to an increase in election congruence in the 2019 regional elections in Spain. Indeed, it is precisely in 2019, when the time that elapsed between the national and regional elections is the shortest, where we can observe the highest score on the dissimilarity index. According to the data in Figure 2, in the case of Spain, the level of election congruence does not seem to depend at all on the proximity or distance between regional and national elections. In fact, dissimilarity has remained virtually constant between 1983 and 2011, despite the variation in the number of days elapsed between national and regional elections.

Figure 2.

Average dissimilarity and number of days between regional and national election (1983-‍2019)

Note: Dissimilarity is measured by an index that ranges from 0 per cent (complete congruence/similarity) to 100 per cent (complete incongruence/dissimilarity). Average in the regions that hold regional elections on the same day.

Source: Own elaboration with data from Archivo histórico electoral, Argos.

Figure 3 reports the dissimilarity index in 2019 by region, showing a high level of cross-regional variation in election congruence. This variation is mainly due to the presence and strength of regionalist parties. Regionalist parties are those whose political program is the defense of a distinctive territory within the state; they are all rooted in the center-periphery cleavage and organize exclusively in their peripheral territory (Alonso, Sonia, Laura Cabeza and Braulio Gómez. 2017. “Disentangling Peripheral Parties’ Issue Packages in Subnational Elections”, Comparative European Politics, 15 (2): 240-263. Available at: https://doi.org/10.1057/cep.2015.15‍Alonso et al., 2017; Vampa, Davide and Matthias Scantamburlo. 2020. “The ‘Alpine Region’ and Political Change: Lessons from Bavaria and South Tyrol (1946-2018)”, Regional and Federal Studies. Online first. Available at: https://doi.org/10.1080/13597566.2020.1722946‍Vampa and Scantamburlo, 2020). Since the democratic transition, regions with strong regionalist parties in Spain are characterized by a high degree of dual voting, whereby sectors of the electorate opt for a regionalist party in regional elections but switch to a state-wide party in national elections (Riba, Clara. 2000. “Voto dual y abstención diferencial. Un estudio sobre el comportamiento electoral en Cataluña”, Revista Española de Investigaciones Sociológicas, 91: 59-88. Available at: https://doi.org/10.2307/40184275‍Riba, 2000; Pallarés, Francesc and Michael Keating. 2003. “Multi-Level Electoral Competition. Regional Elections and Party Systems in Spain”, European Urban and Regional Studies, 10 (3): 239-255. Available at: https://doi.org/10.1177/09697764030103005‍Pallarés and Keating, 2003). Looking at Figure 3, those regions at the lower end of the dissimilarity scale (more congruence between regional and national election results) are those without regionalist parties or with weak ones, like La Rioja or Castile and Leon. On the other hand, those regions located above the average (less congruence) are characterized, with few exceptions, by the presence of strong regionalist parties, like in the case of Cantabria or Navarre.

Figure 3.

Dissimilarity between regional and national election results by region (2019)

Note: Dissimilarity is measured by an index that ranges from 0 per cent (complete congruence/similarity) to 100 per cent (complete incongruence/dissimilarity).

Source: Own elaboration with data from Archivo histórico electoral, Argos.

The level of congruence between national and regional election results not only depends on the presence and strength of regionalist parties. Conjunctural or region-specific factors also play a role. For instance, Madrid is among the regions with the highest dissimilarity index in 2019 (see Figure 3) but there are no regionalist parties as such. The high score on the dissimilarity index in this case is due to the emergence of the new political platform Más Madrid, which did not participate in the April 2019 national election. On the contrary, despite the presence of relevant regionalist parties, Aragon stands out for its very low score on the dissimilarity index. The low dissimilarity score in Aragon can be explained by the idiosyncrasy of this region. According to Spanish political scientists, Aragon is the “Spanish Ohio” (Fernández Albertos, Joséet al. 2015. Aragon es nuestro Ohio. Así votan los españoles. Barcelona: Malpaso Editorial.‍Fernández Albertos et al., 2015). Similarly to what happens in United States with the midwestern “swing state”, the first party in Aragon has always been the first party in Spain. Like Ohio, which contains a bit of everything American, Aragon is a kind of Spain in miniature: it combines a large urban center with large rural areas, it has a very similar party system characterized by the presence of mainstream parties and both center-left and center-right regionalist parties, and even reproduces regionalist parties’ weight at national level.

Be that as it may, to look exclusively at differences or similarities between national and regional vote shares is not enough to determine whether regional elections are nationalized or not (Schakel, Arjan H. 2013. “Congruence Between Regional and National Elections”, Comparative Political Studies, 46 (5): 631-662. Available at: https://doi.org/10.1177/0010414011424112‍Schakel, 2013; Dandoy, Régis. 2013. “Belgium: Toward a Regionalization of National Elections?”, in Régis Dandoy and Arjan H. Schakel (eds.), Regional and National Elections in Western Europe: Territoriality of the Vote in Thirteen Countries. Houndmills: Palgrave Macmillan. Available at: https://doi.org/10.1057/9781137025449‍Dandoy, 2013). A high score on the dissimilarity index, like the one that we have found for the 2019 regional elections, may be a sign of regional distinctiveness. However, it may equally indicate that the regional vote is driven by national dynamics. Indeed, high dissimilarity can be the result of second-order election effects, such as the lower voter turnout in regional elections (that may change the composition of the electorate) or it can reflect the electoral losses of national government parties in regional elections (voters may switch from national government parties to opposition parties induced by cues that they receive from the national electoral arena). Before moving to the analysis with individual-level data, the second half of this section focus precisely on whether the aggregate-level predictions of the second-order election model hold for the 2019 Spanish regional elections.

AN INDIVIDUAL-LEVEL ANALYSIS: DUAL VOTING IN THE 2019 REGIONAL AND NATIONAL ELECTIONS[Up]

Analysis based on aggregate election results are useful to quantify the overall differences between regional and national vote shares. However, only survey data can provide evidence on the real amount and the specific direction of vote-switching at the individual-level. This section uses data from the post-election survey conducted by the Centro de Investigaciones Sociológicas (CIS) after the 2019 Spanish regional and local elections (Centro de Investigaciones Sociológicas (CIS). 2019. Estudio 3253 Postelectoral elecciones autonómicas y municipales 2019. Madrid.‍CIS, 2019). Although the main focus of this survey is on voting behavior at the regional and local elections, respondents were also asked about their vote choice in the April 2019 national election. This allows us to identify the transition between the national and the regional vote for each individual.

According to these data, a large proportion of the electorate voted the same way in the 2019 Spanish national and regional elections. Yet, those who vote differently represent a 23.7 per cent. This is not a negligible amount especially if we consider two things: (1) the short time lapse between the two elections, and (2) the fact that we are analyzing only the ordinary Spanish regions, not the “historic” nations such as Catalonia, the Basque Country and Galicia. In any case, what is relevant to our purpose is to examine whether dual voting respond to the prevalence of a second-order election logic or not. Voters that split their ticket between the regional and the national elections are susceptible to be classified into two groups: those who change their vote because they treat regional elections as second-order elections and those who vote for different parties at different levels because they consider that both arenas are independent (Railings, Colin and Michael Thrasher. 1998. “Split‐ticket Voting at the 1997 British General and Local Elections: An Aggregate Analysis”, British Elections and Parties Review, 8 (1): 111-134. Available at: https://doi.org/10.1080/13689889808413008‍Railings and Thrasher, 1998; Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011). The transitions concerning the first group of voters have been labelled as SOE vote transitions and categorized according to the main characteristics of the second-order election model. We include in this category voters who cast their vote in national elections but abstained in regional elections. This phenomenon has been called differential abstention (Riba, Clara. 2000. “Voto dual y abstención diferencial. Un estudio sobre el comportamiento electoral en Cataluña”, Revista Española de Investigaciones Sociológicas, 91: 59-88. Available at: https://doi.org/10.2307/40184275‍Riba, 2000; Riera, Pedro. 2011. “La abstención diferencial en el País Vasco y Cataluña”, Revista de Estudios Políticos, 154: 139-173.‍Riera, 2011; Riera, Pedro. 2012. “La abstención diferencial en la España de las autonomías. Pautas significativas y mecanismos explicativos”, Revista Internacional de Sociología, 70 (3): 615-642. Available at: https://doi.org/10.3989/ris.2010.10.07‍2012; Liñeira, Robert and Josep M Vallès. 2014. “Differential Abstention in Catalonia and the Community of Madrid: A Socio-Political Explanation of an Urban Phenomenon”, Revista Española de Investigaciones Sociológicas, 146: 69-92. Available at: https://doi.org/10.5477/cis/reis.146.69‍Liñeira and Vallès, 2014). A sector of the electorate that turns up at the polls in national but not in regional elections may perceive the latter as contests where there is “less at stake”. Second, we include those voters who chose a mainstream state-wide party at the national election and changed to a new or a small state-wide party at the regional election. This would be in line with the predictions of the second-order election model: voters are more prone to support these kinds of parties at the regional level because they consider these elections less important than national ones.

Concerning the second group, labelled as No SOE transitions, we include electoral transitions that deviate from second-order expectations. To this group of voters pertain those who switched from a state-wide party in the national election to a regionalist party in the regional election. Extensive research has been conducted in Spain analyzing this kind of vote transition, mainly focusing on the Catalan case (Font, Joan. 1991. “El voto dual en Cataluña: lealtad y transferencia de votos en las elecciones autonómicas”, Revista de Estudios Políticos, 73: 7-34.‍Font, 1991; Riba, Clara. 2000. “Voto dual y abstención diferencial. Un estudio sobre el comportamiento electoral en Cataluña”, Revista Española de Investigaciones Sociológicas, 91: 59-88. Available at: https://doi.org/10.2307/40184275‍Riba, 2000; Fraile, Marta and Santiago Pérez-Nievas. 2000. Is the Nationalist Vote Really Nationalist? Dual Voting in Catalonia 1980-1999. Estudio/Working Paper, 2000/147. Madrid: Instituto Juan March.‍Fraile and Pérez-Nievas, 2000; Riera, Pedro. 2013. “Voting Differently across Electoral Arenas: Empirical Implications from a Decentralized Democracy”, International Political Science Review, 34 (5): 561-81. Available at: https://doi.org/10.1177/0192512112467216‍Riera, 2013). According to previous studies, patterns of voting change from state-wide parties at the national level to regionalist parties in regional elections are persistent and would reflect that voters make their vote choices based on level-specific factors (Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011; Liñeira, Robert and Jordi Muñoz. 2014. “Voto dual y abstención diferencial: ¿Quién se comporta de manera diferente en función del tipo de elección?”, in Francesc Pallarés (ed.), Elecciones autonómicas 2009-2012. Madrid: Centro de Investigaciones Sociológicas.‍Liñeira and Muñoz, 2014). We also consider as a No SOE transition those vote transitions inversely related to the second-order model. These include voters who abstained at the national level but did cast their vote at the regional one, and those who switched from a new or small party to a mainstream party. While the former may plausibly consider the regional level to be more important, the latter can be conceived as national switchers, i.e. voters who make strategic considerations in their general election vote (Lago, Ignacio and André Blais. 2020. “El voto estratégico en las elecciones generales”, in Carmen Ortega, Juan Montabes Pereira and Pablo Oñate (eds.), Sistemas electorales en España: caracterización, efectos, rendimientos y propuestas de reforma. Madrid: Centro de Investigaciones Sociológicas.‍Lago and Blais, 2020). Finally, we also include in this group those who voted for a different mainstream party (for instance, PSOE and PP) or a different new party (for instance, VOX and Ciudadanos). This kind of transition is not very frequent, but it clearly reflects the preference of the voter for different parties at different levels.

Table 2 focuses on the people that voted differently and shows the direction of vote change between national and regional elections. The vote transitions that we have theoretically defined as No SOE transitions (62.1 per cent) are more frequent than those that we have considered as SOE transitions (37.9 per cent). Indeed, many individuals casting their votes for different parties switched from a state-wide party (either mainstream or new) in the national election to a regionalist party in the regional election. This is clearly the most frequent form of dual voting (29.6 per cent), and the most well documented by previous research on the Spanish case (Font, Joan. 1991. “El voto dual en Cataluña: lealtad y transferencia de votos en las elecciones autonómicas”, Revista de Estudios Políticos, 73: 7-34.‍Font, 1991; Riba, Clara. 2000. “Voto dual y abstención diferencial. Un estudio sobre el comportamiento electoral en Cataluña”, Revista Española de Investigaciones Sociológicas, 91: 59-88. Available at: https://doi.org/10.2307/40184275‍Riba, 2000; Fraile, Marta and Santiago Pérez-Nievas. 2000. Is the Nationalist Vote Really Nationalist? Dual Voting in Catalonia 1980-1999. Estudio/Working Paper, 2000/147. Madrid: Instituto Juan March.‍Fraile y Pérez-Nievas, 2000; Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011; Liñeira, Robert and Jordi Muñoz. 2014. “Voto dual y abstención diferencial: ¿Quién se comporta de manera diferente en función del tipo de elección?”, in Francesc Pallarés (ed.), Elecciones autonómicas 2009-2012. Madrid: Centro de Investigaciones Sociológicas.‍Liñeira and Muñoz, 2014). According to this literature, this behavior would be a sign of autonomy between electoral arenas. Many voters prefer different parties in different arenas: a regionalist party in the regional government and a state-wide party at the national level.

Why people vote differently between the 2019 Spanish regional and national elections? Which factors drive SOE and no SOE voting transitions? In order to answer these questions, we proceed to a multivariate analysis. The dependent variable takes on three values: No vote change (when the individual voted for the same party or opted for abstention in both the 2019 regional and national elections), SOE transitions and No SOE transitions. We use multinomial logistic regression and set No vote change as the baseline group to which the other two outcomes are compared. Standard errors are clustered at the regional level to account for potential correlation of individuals within each region.

We consider in our analysis the three fundamental factors that previous research has shown to be the main predictors of dual voting: national identity, assessments of national government or leaders, and assessments of regional government or leaders (Fraile, Marta and Santiago Pérez-Nievas. 2000. Is the Nationalist Vote Really Nationalist? Dual Voting in Catalonia 1980-1999. Estudio/Working Paper, 2000/147. Madrid: Instituto Juan March.‍Fraile and Pérez-Nievas, 2000; Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011; Riera, Pedro. 2013. “Voting Differently across Electoral Arenas: Empirical Implications from a Decentralized Democracy”, International Political Science Review, 34 (5): 561-81. Available at: https://doi.org/10.1177/0192512112467216‍Riera, 2013). On the one hand, a predominantly Spanish sense of identity has been associated to the phenomenon of differential abstention: those individuals who feel only Spanish or more Spanish than from their region are more likely to participate in national elections but to abstain in elections at the regional level (Fraile, Marta and Santiago Pérez-Nievas. 2000. Is the Nationalist Vote Really Nationalist? Dual Voting in Catalonia 1980-1999. Estudio/Working Paper, 2000/147. Madrid: Instituto Juan March.‍Fraile and Pérez-Nievas, 2000). Therefore, we expect Spanish identity to be associated with SOE transitions. On the other hand, previous research has found that dual voters often share a strong sense of regional identity (Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011). Thus, we expect that regional identity would increase the likelihood of a No SOE transition. National identity is measured with the so-called “Linz-Moreno question”, a categorical variable with five values.

Unfortunately, the CIS survey does not contemplate the evaluation of the national government in the questionnaire; but we account for the influence of the national arena by including in our model the evaluation of Pedro Sánchez, who was the Spanish Prime Minister when the 2019 regional elections were held. It is measured by a continuous variable with 10 points, where 1 stands for very bad and 10 for very good. In order to assess the effect of regional assessments, we include in our model the evaluation of the regional government. The original variable is categorical and ranges from 1 (very good) to 5 (very bad). To facilitate interpretation, we have inverted the order of the scale and grouped the responses into three categories: (1) bad, (2) regular and (3) good. The CIS survey does not ask for the evaluation of regional leaders. As control variables, we include interest in the regional election campaign, level of education and ideology, all of them as categorical variables. Table 3 shows the results of the multinomial logistic regression.

The effect of national identity is as expected. Individuals with exclusionist feelings are more likely to cast a dual vote. Specifically, Spanish identity increases the chance of carrying out a SOE transition, while those individuals who feel exclusively attached to their region are more likely to carrying out one of the transitions that we have defined as No SOE transitions. This is consistent with previous research on dual voting that points towards the influence of national identity (Liñeira, Robert and Jordi Muñoz. 2014. “Voto dual y abstención diferencial: ¿Quién se comporta de manera diferente en función del tipo de elección?”, in Francesc Pallarés (ed.), Elecciones autonómicas 2009-2012. Madrid: Centro de Investigaciones Sociológicas.‍Liñeira and Muñoz, 2014) and more generally, with previous studies that have found national identity to be a determinant factor to explain whether individuals vote in regional elections based on national or level-specific issues (Cabeza, Laura. 2018. “‘First-Order Thinking’ in Second-Order Contests: A Comparison of Local, Regional and European Elections in Spain”, Electoral Studies, 53: 29-38. Available at: https://doi.org/10.1016/j.electstud.2018.03.004‍Cabeza, 2018).

The evaluation of the performance of the regional government also affects voters’ dual behavior. A positive retrospective evaluation of the regional government decreases the likelihood of carrying out a No SOE transition, while it increases the chance of carrying out a SOE transition. However, evaluations of the performance of the Spanish Prime Minister do not affect the likelihood of voting differently in the 2019 national and regional elections. The coefficients of this variable are not statistically significant, either for SOE transitions or for No SOE transitions. Previous research on the Catalan and the Scottish cases has not found any perceptible impact of the evaluation of the national incumbent on the likelihood of casting a dual vote, either (Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011; Riera, Pedro. 2013. “Voting Differently across Electoral Arenas: Empirical Implications from a Decentralized Democracy”, International Political Science Review, 34 (5): 561-81. Available at: https://doi.org/10.1177/0192512112467216‍Riera, 2013). This not only confirms that voters take regional government performance into account when making voting decisions, but also, and more fundamentally, it shows that dual voters have a substantial autonomy from national considerations (Liñeira, Robert. 2011. “‘Less at Stake’ or a Different Game? Regional Elections in Catalonia and Scotland”, Regional and Federal Studies, 21 (3): 283-303. Available at: https://doi.org/10.1080/13597566.2011.578790‍Liñeira, 2011). Dual voting is not due to a bad or a good image of the national incumbent, and, thus, it does not respond to the desire to punish or reward the national government in regional elections.

DISCUSSION[Up]

The 2019 regional and national elections in Spain were separated by less than a month. Taking advantage of this unprecedented quasi-simultaneous electoral setting, this article has analyzed the degree to which the 2019 Spanish regional elections may be considered second-order contests, looking at vote-switching between national and regional elections at both the aggregate and individual levels. The contribution to the literature is twofold: on the one hand, it closes an empirical gap by complementing aggregate with individual-level data; on the other hand, it proposes a new approach to the study of second-order election effects by focusing on a subgroup of the electorate, dual voters, and theoretically classifying their different vote transitions into two categories: those who vote differently because they treat regional elections as second-order elections and those who do so because they consider that the two arenas are independent.

At the aggregate level, we have observed an increasing incongruence between regional and national election results since the Great Recession. First, quasi-simultaneity did not translate into an increase in election congruence. On the contrary, it was precisely in 2019, when the time that elapsed between the national and regional elections was the shortest, where we observed the highest score on the dissimilarity index. Differences between party vote shares in regional and national elections were relatively large. Second, we have shown that in the Spanish case, the degree of election (in)congruence does not depend on the proximity or distance between elections at different political levels, but on the presence of regionalist parties. Together with increasing dissimilarity, we also observed second-order election effects when analyzing election results, with a persistent lower turnout in regional elections compared to national elections, and the increase in vote share for the PSOE, confirming the thesis of the “honeymoon period” of governing parties in temporally close elections. Contrary to the expectations of the second-order election model, however, all smaller and new parties have experienced electoral losses at the regional level compared to their vote share on the national election.

Yet, a quite different picture emerged from the analysis at the individual level. Our findings suggest that most of the dual voting at the individual level does not respond to the logic of the second-order election model but rather to voters’ autonomous considerations of the regional arena. In fact, we found no perceptible impact of the evaluation of the national incumbent on the likelihood of casting a dual vote. In this sense, our multivariate analysis has shown that two main determinants of dual voting: (1) national identity, which increases the likelihood of voting differently across national and regional elections; and (2) the evaluation of the performance of the regional government. These findings gain particular significance given that our analysis has been circumscribed to the regions that have held elections on the same date, the so-called ordinary Spanish regions. Even in these regions, that could be considered the most-likely cases for the prevalence of a second-order election vote logic, regional elections are becoming increasingly decoupled from national politics.