Original Research Article Party Politics 2022, Vol. 0(0) 1–12 On time and meaningful partisanship: © The Author(s) 2022 Stability, strength, and sway of attachment Article reuse guidelines: sagepub.com/journals-permissions to new parties DOI: 10.1177/13540688221085235 journals.sagepub.com/home/ppq Roberto Pannico  Instituto de Ciências Sociais da Universidade de Lisboa, Lisboa, Portugal Eva Anduiza  Universitat Autònoma de Barcelona, Barcelona, Spain Abstract New parties pose a challenge to the claim that time is an essential element in the construction of partisanship. By definition, new parties have not been around for much time, so the opportunities for the construction of meaningful attachments could be considered limited. In this paper, we test this expectation, unpacking the dynamics and implications of the attachment to new parties. Using panel data collected in Spain during a period of profound party system change, we estimate the extent to which partisanship with new parties is stable and strong, works as a heuristic for preference formation, and predicts vote choice. Our data suggest that attachments to new parties can be as meaningful as those that citizens have with old parties. These findings seem particularly relevant in a context where new parties are on the rise. Keywords partisanship, party system change, new parties, panel data Introduction may be the case in the much-researched US case, we know little about how citizens react to the appearance of new The appearance of new parties has characterized the recent parties in the context of consolidated democracies that political dynamics of many of the world’s most established undergo a significant degree of party system change. New democracies and their party systems (Bolleyer, 2013; De parties, by definition, have not been present during the Vries and Hobolt, 2020; Emanuele and Chiaramonte, 2019). impressionable years of most of the electorate, and there is While the term “new” involves some degree of ambiguity, no record of past behavior that citizens can evaluate. If we its direct reference to time is of crucial significance when take the general argument to its logical consequence, at- trying to understand some of the implications of these party tachments to new parties should be more unstable, weaker, system changes. New refers to something recently created. and less consequential because of the lack of necessary time. Time, in terms of experience and familiarity with parties as In this paper, we explore how citizens relate to parties political objects, has been considered to be a key element that have recently entered the electoral and parliamentary in the construction of partisanship. The fact that age has arenas. We analyze the stability, strength, and consequences been consistently found to correlate positively with of partisanship with new parties in the context of consol- strength of party attachment (Campbell et al., 1960; idated democracies. We use a case, Spain, for which we Shively, 1979; Cassel, 1993) has been interpreted as an have new sizeable parties entering a party system where old indicator that as time goes by, links with parties become stronger, regardless of whether these links are considered to be party identification (Campbell et al. 1960), a form of social identities (Greene et al., 2004; Huddy and Bankert, Paper submitted 5 August 2021; accepted for publication 10 February 2022 2017), or the result of information updates (Achen, 1992; Corresponding author: Fiorina, 1981). Roberto Pannico, Instituto de Ciências Sociais da Universidade de Lisboa, However, a shortcoming of this claim is that it is based on Av. Prof. Anı́bal Bettencourt 9, Lisboa 1600-189, Portugal. evidence that takes parties as given and stable. While this Email: roberto.pannico@ics.ulisboa.pt 2 Party Politics 0(0) parties remain, as well as the necessary longitudinal data to intensifies with age and is relatively immune to changes in track individual variation in partisanship. Our results show the political and economic situation. For the revisionist that citizens can quickly develop partisan attachments, and approach, partisanship is endogenous to political circum- that the attachment to new parties is relatively stable and as stances: government performance, salient issues, policies, strong and influential in terms of policy preferences and leaders. Here, too, individuals need some accumulated vote choice as attachment to old parties. Our findings hence experience to conform to their partisanship. Citizens learn challenge the claim that time is a requirement for the by experience and observation (hence, time) about which construction of a meaningful partisanship. party to trust (Achen, 1992). The article is structured into six sections. The first section Therefore, the implication of the conclusions of previous presents the main ideas on the relationship between time and works is that partisan attachments are reinforced by time. partisanship that can be found in the literature. The second The acquisition of partisanship involves, both for classic section develops the argument that meaningful party at- and revisionist accounts, a learning process that takes time tachment can appear quickly for new parties. The third and of exposure and familiarity (Achen, 1992; Campbell et al., fourth sections present the case and the data, respectively. 1960; Green et al., 2004). Converse (1969) anticipated that We use a 10-wave panel carried out in Spain before, during, several decades are necessary for the development of and after two large new parties, Podemos and Ciudadanos, partisanship. As clearly stated by Greene (2011) regarding appear in the party system. The fifth section presents the the case of new democracies, time does matter for the analysis, which includes four aspects: stability, strength, development of partisanship, regardless of the theoretical party cues, and vote choice. In the last section, we discuss approach: “If party identification is a psychological at- the implications of our results. tachment that crystallizes in early adulthood (Campbell et al., 1960), then few voters would have been socialized over the first few democratic elections. If party identification On time and partisanship is a running tally of preferences (Fiorina, 1981), then limited The relationship between time and partisanship can be knowledge of parties’ platforms makes voters less able to understood in two ways: as a dimension for variation or as a align their policy preferences with the available options” proxy for experience or familiarity. For example, when (Greene, 2011: 401). Flipping the argument, when time is studies show that the stability of partisan attachments in- limited, as is the case for young people, young democracies, creases with age (Achen, 1992; Gerber and Green, 1998; or new parties, these attachments are expected to be weaker, Stoker and Jennings, 2008), “stability of attachments” refers more unstable, and less consequential. to the first way of understanding time, while “age” refers to The claim that a meaningful partisanship requires time is, the second. however, almost solely based on evidence from contexts Under the first perspective, the analyses and debates with stable party systems, notably the US. New democracies have focused on the extent to which partisanship is rela- have also garnered some attention. Even though, in some tively stable over time, as the original formulation of the cases, partisan attachments have been found to appear fairly concept as an “unmoved mover” would predict, or can be quickly after a transition to democracy (Brader and Tucker, updated with new information, as revisionist formulations 2008; Carlson, 2016; Dinas, 2014; Poertner, 2020), these have suggested (but see Lupu, 2013; for a summary of the analyses have not really challenged the core assumption that extended debate on partisanship stability, see Johnston, partisanship needs time. The main take from the analysis of 2006). partisanship in new democracies is that it is weaker than in Under the second perspective, time is a proxy for factors consolidated democracies (Greene, 2011), presumably be- that may affect the acquisition and consolidation of parti- cause of the lack of sufficient time to develop strong sanship, such as relevant information, experience, or fa- attachments. miliarity with political parties. In this latter sense, time is The analysis of new parties in consolidated democracies considered to be an important ingredient to build partisan has been more limited because, until recently, cases of new attachments. Whether such attachments come from pro- parties were relatively rare or too small to be analyzed. found psychological attachments with social groups Green et al. (2004) studied the case of Italy, finding that (Campbell et al., 1960; Green et al., 2004; Huddy and attachments towards new parties are more instable than in Bankert, 2017) or are the result of evaluating which the UK or the US. However, as the authors acknowledge, party may better suit the individual’s interest (Fiorina, 1981; these previous findings suffer from data limitations (e.g., the Franklin and Jackson, 1983), it is generally considered that panel on which their analysis is based starts after new parties people need time, either to acquire this social identity or to appear). Also, the complete dismantling of the old party gather information for the “running tally.” For the classic system in 1990s Italy is not quite representative of the more perspective, partisanship originates in the early years of frequent situation where new parties coexist with old socialization. It then becomes a stable attachment that parties. Pannico and Anduiza 3 New parties are increasingly frequent in consolidated Third, and relatedly, new parties often appear in contexts democracies (Emanuele and Chiaramonte, 2019). Party of political realignment or, at least, a certain political tur- system innovation has increased significantly in recent years moil. Such contexts of political change and crisis often with new parties entering parliament, sometimes with come with higher levels of political interest and engage- sizeable shares of votes and seats. These new parties have ment. As suggested by Shively’s (1979) model of party received attention, but mostly in terms of why they appear attachment development, a situation where new parties (Lucardie Paul, 2000) or survive (Bolleyer, 2013; Poertner, emerge may actually be prone for the development of 2020; Zur, 2019). We know far less about how individuals partisanship. Anduiza and Pannico (2020) showed that there react to these new political actors. is an association between political interest and developing attachment to a new party. In many realms of life, “new” is often felt as more exciting and attractive than “old,” and this Meaningful attachment to new parties: may also be the case in politics and for parties. Hence, the conditions that generate the appearance of new parties and Our argument the newness itself (Sikk, 2012) can also facilitate the ac- New parties pose a key challenge to the existing conception quisition of a meaningful partisanship. of the relationship between time and partisanship because, In sum, our argument is that, while time may certainly by definition, the relationship between citizens and new reinforce partisanship, it is not a necessary condition for parties is limited to a short time span. Should we expect meaningful party attachments to appear. Therefore, at- partisanship with new parties to be weaker and less con- tachment to new parties may function similarly as attach- sequential because citizens need familiarity and learning ment to old established parties. To test this general that can only come with time? Or can we expect new parties expectation, we need to define what meaningful party at- to be able to rapidly generate attachments that are strong, tachments involve. Following previous works, we identify stable, and consequential? The answer to this question is three dimensions: stability, strength, and sway. ultimately empirical, but while the existing literature has provided support mostly for the former claim, there are at least three reasons why the argument that meaningful Stability partisanship necessarily takes time can be challenged. The first quality partisan attachments must have to be First, political parties are not complex and abstract considered meaningful is a certain degree of stability along constructs that require either familiarity or high cognitive time. The question of partisanship stability has received resources to be understood. In this sense, they are not like considerable attention (see, for instance, Converse, 1969; other elements of the political system, such as, for instance, Green et al., 2004; Johnston, 2006; Neundorf et al., 2011; electoral laws, which indeed require time to be learned Schmitt Beck et al., 2006) because it has been considered a (Gallego et al., 2012). If a new party has clear stances and key element for the resolution of the classic/revisionist appeals to a well-identified constituency, it is not un- debate. While we will neither enter this debate nor argue thinkable that citizens will be able to relate to it as a that partisan attachments need to be any more/less stable relevant political object fairly quickly, particularly if other than other attitudes (see Chen and Goren, 2016; Evans and features of the political system remain stable. Previous Neundorf, 2020; Goren, 2005 for this question), it is in- work on the broader concept of party instability has shown teresting to point out that previous works have remained that citizens are indeed able to cope with changing parties mostly silent on whether the stability of attachments differs (Marinova, 2016), and the argument could be extended to by party characteristics. Schmitt-Beck et al. (2006) referred new parties. to smaller parties being more likely than larger parties to Second, in the same way as time does not necessarily lose supporters in Germany. Tranter and Smith (2018) found guarantee strong attachments with parties, partisanship may that partisanship with the younger Australian Green Party is crystalize fairly quickly if the experience with the party is less stable than the partisanship with older parties. However, intense and concentrated in a short time span. Time is as- the AGP was almost 20 years old by the time of data collection. sumed to be a proxy for experience, knowledge, and fa- It is difficult to consider it a new party. In fact, the authors miliarity, but the relationship between time and party partly explain the higher instability of the Greens’ partisanship attachment may be more complex and varied than it has been with its ideological proximity with the Labor party and the assumed. Time spans of the same length in different contexts consequent fluidity between Labor and Greens partisanship. or for different people can contain very different levels of From the perspective of this paper, our reference benchmark is political experience conducive to partisanship. In other the degree of partisanship stability that we find among old words, intense experiences can happen in a short time span, parties’ partisans, and we explore whether partisanship with and this could also apply to experiences with new parties. new parties has lower or similar levels of stability. 4 Party Politics 0(0) Strength between Greens and Labor and the possibility of the stra- tegic vote. The second quality meaningful partisan attachments should To sum up, our hypotheses state that the attachment to a have is strength. Although strength was one of the key new party can be as stable (H1), strong (H2), and influential dimensions of the initial conceptualization of partisanship, it on policy positions (H3) and vote choice (H4) as attachment has deserved much less empirical attention. Again, here to old parties. explanations for differences in strength of party attachment refer to contextual characteristics, such as campaign periods (Dassonneville and Grieb, 2018) or individual character- The Spanish case istics such as personality (see Bakker et al., 2015), length of We test our expectations in Spain, a case that allows us to residence in the country (Cain et al., 1991), or age (Lupu, compare new and old parties’ partisanship. After 30 years of 2013), but little is known about the differences across types substantial party system stability, the mid-2010s were of parties. As with stability, we simply take old parties as a characterized by the fast rise of new parties (Rodrı́ guez- benchmark to assess whether new parties are equally or Teruel et al., 2016). During the period of stability, the less likely to generate strong attachments among their Spanish party system was composed by the two large state- supporters. wide parties, Partido Popular (PP) and Partido Socialista Obrero Español (PSOE); a third small state-wide coalition, Izquierda Unida (IU); and several regional parties. In 2011, Sway a handful of small new parties competed in the general Finally, meaningful partisanship should act as a cognitive elections, foreshadowing the major changes in the party shortcut for preference formation and as a robust predictor system that would take place in the following elections. In of vote choice. Many works have addressed the effect of 2015, Podemos and Ciudadanos entered the parliament with party attachment on policy preferences (Bartels, 2002; 69 and 40 seats, respectively, out of the total 350. Podemos Bullock, 2011; Carsey and Layman, 2006; Jacoby, 1988; was born in 2014 from the opposition to the austerity Klar, 2014; Lavine et al., 2012; Pannico, 2020; Torcal et al., policies applied during the economic crisis by both main- 2018). What is of interest here is not so much the differences stream parties. It is a left-wing party with an initial populist in preferences between people that feel attached to different discourse. Ciudadanos was born as a Spanish nationalist parties, but rather to what extent people follow the position party in Catalonia in 2006 and then started to compete in the of their party. Within the large literature of party cues, not Spanish arena when the Catalan conflict escalated. It has many studies have looked at party longevity as a moderator. liberal positions on cultural issues and is center-right in Brader et al. (2013) found a non-significant effect. Coan terms of economic policy. Old parties experienced the et al. (2008) found that party cues are more effective when consequences of a severe crisis of representation (Vidal, voters feel higher levels of familiarity with and trust in the 2018), but survived. Since then, the Spanish party system party. In their analysis, however, these two characteristics has been characterized by the presence of both old and new are conflated in a single factor, making it impossible to party organizations. It is therefore a suitable case to test our disentangle the two effects. Therefore, there seems to be no hypotheses on partisanship for new parties in consolidated established ground to expect that new parties may be less democracies. influential than old parties in terms of conditioning policy preferences among their supporters. Regarding vote choice, Bartels (2000) has shown that the Data predictive power of partisanship is still significant. Several In 2002, Green et al. complained that, “It is a sad com- authors have argued that the relationship between these two mentary on the prescience of the discipline that political variables is reciprocal (Dinas, 2014; Franklin and Jackson, scientists have never conducted a multiwave panel study 1983). The behavior may certainly affect the attitude, and spanning a period of party realignment” (Green et al., 2004, more so in the case of old parties, since, for new parties, 194). Luckily, the POLAT online panel (Hernández et al., there have been fewer opportunities to vote for them. 2021) covers this period of party system transformation in However, this should not be a problem, as it would, in any Spain, running from 2010 to 2018 with a total of 10 waves case, work in the sense contrary to our expectations, re- of online surveys (yearly since 2013). The sample of the first inforcing the association between partisanship and vote wave included 2100 Spanish citizens between 16 and choice for old parties. In the already cited study in Australia, 44 years old and residents of Spain. The choice of this Tranter and Smith (2018) showed that fewer Greens sup- specific age range is due to the limited internet access for porters vote for their party’s candidate compared to older people over 45 in 2010 in this country.1 For the recruitment parties’ supporters. However, also in this case, the authors of the respondents of the first wave, two crossed quotas were explain this difference with the ideological proximity applied: gender/age and region/size of municipality. Along Pannico and Anduiza 5 the years, several fresh samples were introduced.2 The total Analysis number of individuals included in the dataset is 4216 (total number of observations is 17,228). Further information Stability about the POLAT panel survey and its representativeness can be found in the on-line appendix with supplementary We test H1 on the similar levels of stability in partisanship information (Suppeymentary Tables A11-A13). Suppeymentary for new and old party supporters using two multinomial Table A1 provides the coding for all the variables used for the logistic models in a strategy similar to Tranter and Smith different models of the paper. (2018). The dependent variable for both models indicates To measure party attachment, we use the question, “For the partisanship of the respondent (categories are: “PSOE,” which of the following parties do you feel most sympathy or “PP,” “Podemos,” “Ciudadanos,” “Other,” and “No parti- which one do you consider closest to your own ideas?” sanship”). The main independent variable for both models is asked consistently through all 10 waves of the panel. Re- the lag of the dependent variable, indicating respondent’s spondents were presented with a list of parties and asked to partisanship at time “t-1.” The models also include controls choose one of them or the “none of them” option. We for the respondent’s gender, education, income, age, region consider as new all the parties that appeared during the of residence, and ideological self-placement. The standard duration of the POLAT panel. In other words, we catego- errors are clustered by respondent.3 rized as new the nine parties that ran for the first time in the To estimate the stability of the attachment to different national elections in 2011 or after: Podemos, Ciudadanos, parties, we estimate two models. The first one only includes and the smaller Catalunya en Comú, Partido X, VOX, observations from waves 6 to 10 (2014–2018). This model Bildu, Compromis, Equo, and Foro Asturias (see allows us to compare all the parties in the same time period, Supplementary Table A2 in the on-line appendix). These are that is after the appearance of the new parties. However, we completely new parties, not simply old organizations that also estimate the stability of the attachment to old parties changed their name. All the other parties were categorized (PSOE and PP) with a second model that only uses data as old parties. Figure 1 shows the distribution of attachment from waves 1 to 5 (2010–2013). This second model allows to old and new parties in each wave of the POLAT panel. us to put H1 to a harder test, comparing the stability for old The new parties were only included in the option list of the parties estimated during a period of fewer parties and more partisanship question starting from the sixth wave (2014). party system stability with the one for new parties estimated Therefore, from waves 1 to 5, respondents could only report during a period with more parties and electoral change. attachment to an old party. Table 1 provides the predicted probability to have a specific Figure 1. Distribution of partisanship in the POLAT panel. 6 Party Politics 0(0) partisanship at time t, given partisanship at time t-1 (full PSOE and PP were not more able to keep their supporters results in Supplementary Table A3 and Supplementary during the party system stability period (waves 1–5) than Table A4 of the on-line appendix). During the 6 to 10 Podemos and Ciudadanos were during the period of party wave period, a respondent that felt attached to PSOE at time system instability (waves 6–10). t-1 had a 68% probability to have the same partisanship at time t and a 1% probability to change to PP. Overall, the Strength differences found do not seem to be systematically in favor of the old parties. New parties have different degrees of The partisanship question is followed by an item aimed to ability to keep their supporters but not systematically less measure the intensity or strength of the respondent’s party than old parties. PSOE (old) and Ciudadanos (new) have attachment: “How close do you feel to this party?” The very similar stability rates (68% vs. 67%). Podemos (new) response categories are (0) “A little close,” (0.5) “Quite has a lower stability rate (52%), but higher than the rate of close,” and (1) “Very close.” To test our H2 on strength, we PP (old), the lowest in our sample. Therefore, supporters of first compare the answers to the intensity question for old new parties do not seem to be more unstable over time than and new parties. The left-hand panel of Figure 2 shows the supporters of old parties. This is the case also when average of party attachment’s strength for old and new comparing old and new parties in different time periods. parties in waves 6 to 10, that is once new parties have Table 1. Predicted probability of partisanship at time t. Partisanship time t PSOE PP Podemos Ciudadanos Other No partisanship Partisanship (t-1) Waves 6–10 PSOE 0.68 0.01 0.05 0.05 0.08 0.15 PP 0.04 0.30 0.05 0.22 0.16 0.23 Podemos 0.07 0.00 0.52 0.04 0.23 0.13 Ciudadanos 0.03 0.02 0.04 0.67 0.09 0.14 Waves 1–5 PSOE 0.68 0.01 0.15 0.16 PP 0.03 0.54 0.17 0.26 Note: Bold values indicate probability of respondents keeping the partisanship they had at time t-1. Figure 2. Self-reported strength of party attachment (waves 6–10). Pannico and Anduiza 7 appeared. In this period, new parties seem to generate a self-placement. The standard errors are clustered by stronger attachment than old parties (two-tailed t-test, p < respondent. 0.05). At the same time, focusing only on the main parties, In Figure 3 (fixed-effect models in Supplementary Table the right-hand panel of the figure shows that there is some A5 of the on-line appendix), the vertical axis indicates the in-group heterogeneity. Consistently with H2, partisanship predicted party attachment strength for respondents that for the two new parties appears at least as strong as par- switched from an old to a new party. The horizontal axis tisanship for the two old parties. However, while the values indicates the time points from the change, with 0 indicating for PP, PSOE, and Ciudadanos are indistinguishable from the wave in which the respondent made the switch. The left- each other, the average partisanship strength for Podemos is hand panel of the figure refers to respondents that switched higher than for all three (two-tailed t-test, p < 0.001). In any from any old to any new party. As the figure shows, the case, these first results show that partisanship generated by strength of the partisanship is pretty stable in the waves old parties is not stronger than partisanship generated by before the change. For this period, the line is mostly flat and new ones.4 the coefficients in Supplementary Table A5 show that for no As a second strategy of the analysis of the partisanship time point the strength of party attachment is different from strength, we take advantage of the panel structure of our the one predicted for the reference category t( 1). However, data to look at how a switch of partisanship from an old to a at the time of a switch to a new party, the party attachment new party affects the strength of the party attachment. Using strength has a significant increase of 10% of the scale and data from all 10 waves of the POLAT panel, we run fixed- keeps the same level for at least the two following waves effect models that rely on within-individual variation of (note that there is the possibility that after the change the both partisanship and partisanship strength. This type of respondent does not keep the new partisanship). model compares observations over time for the same re- Interestingly, this increase in the strength of party at- spondent. It is therefore possible to estimate whether and to tachment does not seem to happen to the same extent for all what extent the strength of the respondent’s attachment the new parties. The center and right-hand panels of Figure changes when she moves her partisanship from an old to a 3 show, respectively, the results for respondents that switch new party. Moreover, relying on within-individual varia- from any old party to Ciudadanos and the results for re- tion, the models control for all possible time-invariant in- spondents that switch from any old party to Podemos. In the dividual characteristics. Differently from the case of the first case, there is no increase in the strength of the at- stability analysis, we estimate linear models and therefore tachment. On the contrary, making a change of partisanship we can rely on within-individual variation without losing from an old party to Podemos produces a rise of 20% of the observations. Given they analyze within-individual change, scale of party attachment strength. the models only include respondents that switched from an Overall, the analysis of partisanship strength shows that old to a new party during the panel. new parties seem well capable of generating strong at- The dependent variable is the same we used in the tachment. In fact, new parties’ partisans seem to have previous step (i.e., the answers to the intensity question slightly stronger feelings of attachment than do partisans of coded from 0 to 1). The main explanatory factor is a variable old parties. Furthermore, becoming the supporter of a new that indicates the time distance from the moment the re- party actually increases the strength of partisanship. Im- spondent switched from a partisanship for an old party to a portantly, however, not all parties seem to be able to secure partisanship for a new one. The panel wave in which the these effects. individual shows a partisanship for a new party for the first time is coded as the crucial point, t(0). The time point t(1) indicates the following wave, t(2) indicates the second wave Sway after the switch, and so on. Negative values indicate the To test H3 about partisan influence on policy positions, we panel waves prior to the change. The time variable ranges use data from a survey experiment embedded in the seventh from t(-9), indicating the first wave of the panel for people wave of the POLAT panel (2015). Out of the 1014 Spanish that switched from an old to a new party in the 10th wave, to citizens that completed the survey in this wave, only the 410 t(4), indicating the 10th wave for people that made the respondents that reported to have a partisanship for PP, change as soon as the new parties appeared in the sixth PSOE, or Podemos participated in the experiment (see wave. The use of this time variable allows us to investigate online appendix, Section 2). not only how partisanship strength changes when the re- Participants were randomly assigned to a control or a spondent switch partisanship, but also to assess the trend of treatment group. In both groups, they were presented with partisanship strength before and after the change (Prior four different political issues and asked to choose, for each 2018). The models also include control variables for the of them, their preferred policy option among three different following time-variant individual characteristics: age, proposals. In the treatment group, each proposal was labeled education, income, region of residence, and ideological with the name of one of the three parties (i.e., PP, PSOE, or 8 Party Politics 0(0) Figure 3. Within-individual change in the strength of party attachment. Podemos), while in the control group the proposals were issue, the party cue from Podemos seems to have a larger unlabeled. The policy options listed for each issue were the effect than the party cues from PP and PSOE (see actual positions of the three parties, taken from their Supplementary Table A6). For the “TTIP” issue, the effect electoral manifestos or public speeches. The four political for Podemos is also slightly larger than in the case of the issues used in the experiment varied in terms of governance other two parties, even though in this case the differences level (European Union level or national level) and com- are not statistically significant. For the “Austerity” and plexity (easy or hard). For the EU level, the issues were “Energy” issues, the treatment effect among the supporters “European austerity policies” and “Transatlantic Trade and of Podemos is smaller than in the case of supporters of one Investment Partnership (TTIP).” For the national level, the of the old parties, but similar to the effect for respondents issues were “Management of families’ mortgage debt” and that feel close to the other. In sum, our experiment shows “Development of alternative energy sources.” The re- that the effect that a new party cue has on the policy po- spondent’s membership in the treatment or the control group sitions of its supporters is not smaller than the effect exerted was the same for all the issues, meaning that she received the by old parties, and, in some cases, it may even prove larger. policy options with party labels for the four issues if in the However, it could be argued that our experiment un- treatment group or for none of them in the control group. The derestimates the strength of party cues for old parties due to order of presentation of the issues in the survey was ran- a pre-treatment effect (Slothuus, 2016). Policy positions of domized, as was the order of the policy proposals for each old parties are likely to be better known than positions of issue. This experimental design allows the estimation of the new parties. In such a case, supporters of old parties in the party cue effect for different groups of party supporters. In control group would be more likely to identify their party’s other words, it is possible to discover whether the presence of position than supporters of the new parties, affecting party labels affects the choices of the respondents and downward the treatment effect. To exclude this possibility, whether the size of this effect is different for different parties. we looked at the share of respondents that in the control We run four logit models, one for each issue. The de- group chose their party’s policy option. Only in the case of pendent variable is a dummy that distinguishes participants the “Mortgage” issue are Podemos supporters significantly that selected the policy option of their preferred party from less likely than PP and PSOE supporters to pick their party’s respondents that chose any other option or answered, “I option in the control group. For two other issues (Austerity don’t know.” The main independent variable of the models and Energy), Podemos supporters are actually more likely identifies the experimental treatment (i.e., the presence of than PSOE supporters to choose the party’s option. For the party labels). This variable is interacted with a categorical TTIP issue, the three parties are not distinguishable ac- one that identifies the participant’s partisanship. Figure 4 cording to conventional level of significance (see shows the marginal effects of the experimental treatment for Supplementary Table A7 in the on-line appendix). the three parties and the four issues (detailed results in Finally, we test our H4 about the relation between party Supplementary Table A6 of the on-line appendix). attachment and vote to see whether attachment with a new The figure does not show evidence of a weaker party cue party predicts vote to the same extent as attachment with an effect for the new party, Podemos. For the “Mortgage” old party. To this aim, we again run a multinomial logit Pannico and Anduiza 9 Figure 4. Cue effect for old and new parties. Marginal effect of the treatment. where the dependent variable is vote intention if general been often assumed. By meaningful attachments we denote elections were to be held on the following day (“PP,” “PSOE,” attachments that are relatively stable, strong, and influential “Podemos,” “Ciudadanos,” “Other,” “Abstention/Blank/Null/ over policy preferences and vote choice. Although time is DK”). The independent variable identifies the partisanship of short for the development of new parties’ partisanship, our the respondent (“PSOE,” “PP,” “Podemos,” “Ciudadanos,” analyses show that new parties can generate links with “Other,” “No partisanship”). The model includes waves 6 to 10 citizens that are at least as stable, strong and influential as and controls for gender, age, education, income, region of those of old parties. residence, and ideological self-placement. In this case, we also A number of reasons may account for this finding. New added controls for waves to account for the possibility that the parties need not be difficult to grasp and can be exciting. relation between partisanship and vote changes over time. They are not complex, abstract institutional structures, but Standard errors are clustered by respondent. rather organizations with visible leaders that can quickly Table 2 provides the predicted probability for the re- connect to specific groups or demands that people can spondent to vote for a specific party given her partisanship recognize rapidly. Moreover, in a context of accelerated (full results in Supplementary Table A8 of the on-line information flows, the necessary knowledge and familiarity appendix). These probabilities seem to be very similar required for party attachments to form may take less time for all parties. No pattern emerges that points to differences than expected, because what is new typically generates between old and new parties. PSOE partisanship seems to more interest and attention. New parties appear embedded be slightly more consequential for vote choice than for other in processes of political change that happen in periods of parties, but at the same time, PP partisanship seems to be the particular intensity. Therefore, time as the “fourth dimen- least consequential for the vote.5 sion” gives an impression of linearity that may not ade- Attachments to new parties, therefore, do not seem to be quately represent the rhythm of political processes, which less consequential than attachment to old parties. The label may intensify at times, generating periods with a rapid of Podemos, a new party in Spain, is able to affect voters’ succession of political events that generate interest and policy preferences as much as the labels of PP and PSOE. At excitement and, as consequence, knowledge and familiarity. the same time, the attachments to old parties do not correlate Of course, the specific chain of events leading to partisan with vote choice more than attachments to new parties. attachments remains to be tested, and further research should devote more specific attention to how time is related to interest, knowledge, familiarity, and experience, as well Discussion as to the specific mechanisms that channel the formation of In this article we argue that time may be less relevant for the such attachments. Is this due to individuals developing development of meaningful partisan attachments than it has affective and social identity ties with these new political 10 Party Politics 0(0) Table 2. Predicted probability to vote for old and new parties. Vote PSOE PP Podemos Ciudadanos Other Abstention/Blank/Null/DK Partisanship PSOE 0.82 0.00 0.01 0.03 0.02 0.11 PP 0.02 0.75 0.01 0.05 0.03 0.15 Podemos 0.04 0.00 0.79 0.02 0.06 0.09 Ciudadanos 0.02 0.04 0.01 0.78 0.03 0.12 Note: Bold values indicate the probability of respondents to vote for a party when feeling attached to it. objects? Or is it because citizens evaluate the added value consolidated democracies is not necessarily temporary or that these parties bring into the system in terms of policies? precarious. New parties can rely on stable and strong at- Our findings also suggest that the ability to generate tachment that enables them to shape policy preferences and attachment varies across different new parties and, thus, translates into electoral support. appears more as a potentiality (new parties can generate these attachments rapidly) rather than as a regularity (new Acknowledgements parties do generate attachments rapidly). Not all new parties Earlier versions of the article have been presented at the Electoral are equal and, of course, contextual variations may be Psychology Observatory Workshop (London School of Eco- important. Our case, Spain, provides necessarily limited nomics), the Democracy Elections and Citizenship Seminar evidence. Our benchmark for this assessment is the at- (Universitat Autònoma de Barcelona), and the Social and Political tachment to old parties, analyzed both in the period of party Attitudes: Resilience and Change Seminar (Instituto de Ciências system stability, before the appearing of the new parties, and Sociais da Universidade de Lisboa). We wish to thank all the in the following period of volatility. However, our data, participants for their helpful comments and suggestions. Any error however rich and unprecedented, only date back to 2010, a in the article remains our sole responsibility. few years before new parties appeared. This means we could not account for previous changes in the nature of parti- Declaration of conflicting interests sanship and the consequent processes of dealignment The author(s) declared no potential conflicts of interest with re- (Dalton, 2013). Similarly, the time span of the data only spect to the research, authorship, and/or publication of this article. allows for a comparison between old and new parties during the period the latter first appear. Other dynamics may prevail Funding once the turmoil of the party system has stabilized. In spite of these limitations, the evidence is quite ex- The author(s) disclosed receipt of the following financial support traordinary in the sense that it tracks individual variation of for the research, authorship, and/or publication of this article: This partisanship in a context of profound party system change. It research is part of the “Political Change in Spain: Populism, provides sufficient proof for the argument that new parties Feminism and New Dimensions of Conflict” project (CSO2017- can quickly generate meaningful partisan attachments. The 83086-R) financed by the Spanish Ministry of Science, Innovation conditions under which new parties are successful in and Universities. generating these new partisan attachments will be a relevant question for further comparative research because, impor- ORCID iDs tantly, we also find significant cross-party differences in the Roberto Pannico  https://orcid.org/0000-0001-5079-4249 quality of the links with citizens. This means that party Eva Anduiza  https://orcid.org/0000-0003-0924-8064 characteristics in terms of organization, communication strategy, or identification of a well-defined constituency Supplemental material probably influence – more than their newness – the extent to which they are able to generate meaningful attachments, and Supplemental material for this article is available online. indirectly, their chances of surviving. For now, however, our study has shown that, while new Notes parties may have fewer supporters than old parties, the kind 1. By the end of the panel, the sample was up to 55 years old, and of attachment they generate can be comparable to that so the panel gains representativeness in terms of age as times generated by established organizations. This suggests that goes by. However, it cannot claim to be representative for the the presence of these newcomers in the party systems of whole population. This means that we have a younger sample, Pannico and Anduiza 11 expected to be more supportive of new parties, but with more Bartels LM (2002) Beyond the running tally: partisan bias in instability and less strength, which if anything would work political perceptions. Political Behavior 24(2): 117–150. against our expectations. Bartels LM (2000) Partisanship and voting behavior, 1952-1996. 2. A fresh sample of 620 respondents with only compulsory American Journal of Political Science 44(1): 35–50. education levels was added in the second wave to correct for the Bolleyer N (2013) New Parties in Old Party Systems: Persistence overrepresentation of the college-educated population. The and Decline in Seventeen Democracies. New York, NY: OUP ninth wave also included a fresh sample of 996 individuals that Oxford. was added applying quotas for education, region, size of Brader T and Tucker JA (2008) Pathways to partisanship: evidence municipality, and gender/age to compensate for attrition. Fi- from Russia. Post-Soviet Affairs 24(3): 263–300. nally, an additional sample of 504 individuals was added in the Brader T, Tucker JA and Duell D (2013) Which parties can lead tenth wave with quotas for education, region of residence and opinion? experimental evidence on partisan cue taking in gender/age. The quotas for region of residence and gender/age multiparty democracies. Comparative Political Studies were designed to achieve the representativeness of the final 46(11): 1485–1517. sample for the Spanish population between 18 and 55 years old. Bullock JG (2011) Elite influence on public opinion in an informed 3. It could be argued that multinomial fixed-effect logits would be electorate. American Political Science Review 105: 496–515. more suitable for our purpose. However, logit fixed-effect Cain BE, Kiewiet DR and Uhlaner CJ (1991) The acquisition of models exclude from the analysis all the individuals that are partisanship by Latinos and Asian Americans. American constant on the dependent variable (i.e., respondents that al- Journal of Political Science 35: 390–422. ways have the same partisanship) because, for these cases, the Campbell A, Converse PE, Miller WE, et al. (1960) The American regressors are perfect predictors of the outcome, and the co- Voter. Chicago: University of Chicago Press. efficients would be infinite in magnitude. As a result, if we used Carlson E (2016) Finding partisanship where we least expect it: multinomial fixed-effect logit models, we would lose about evidence of partisan bias in a new African democracy. Po- 45% of the observations. litical Behavior 38(1): 129–154. 4. Similarly to what was done for the stability test, we also Carsey TM and Layman GC (2006) Changing sides or changing compared partisanship strength for old parties during the 2010- minds? Party identification and policy preferences in the 2013 period (waves 1 to 5) with the partisanship strength for American electorate. American Journal of Political Science new parties during the 2014-2018 period (waves 6 to 10). The 50(2): 464–477. results in Supplementary Figure A1 of the appendix show that, Cassel CA (1993) A test of converse’s theory of party support. The overall, the latter is still higher than the former (two-tailed t-test, Journal of Politics 55(3): 664–681. p<0.001). However, in this case the partisanship for Ciuda- Chen PG and Goren PN (2016) Operational ideology and party danos is weaker than the partisanship for PP and PSOE identification. Political Research Quarterly 69(4): 703–715. (p<0.001 and p<0.05, respectively). Coan TG, Merolla JL, Stephenson LB, et al. (2008) It’s not easy 5. Although our analysis is not aimed to establish a causal effect of being green: minor party labels as heuristic aids. Political partisanship on vote, as robustness check we run a second Psychology 29(3): 389–405. multinomial logit model using lagged partisanship as main Converse PE (1969) Of time and partisan stability. Comparative independent variable. The results from the model and the Political Studies 2(2): 139–171. predicted probabilities are provided in Supplementary Table A9 Dalton RJ (2013) The Apartisan American: Dealignment and and Supplementary Table A10 of the on-line appendix. In this Changing Electoral Politics. Thousand Oaks, CA: CQ Press. case also no pattern emerges that would suggest that parti- Dassonneville R and Grieb A (2018) Partisanship and political sanship for old parties is more consequential for vote than sophistication during and outside of election times: a lon- partisanship for new parties. gitudinal analysis using the German SOEP survey 1984- 2012. German Politics 27(1): 44–65. De Vries CE and Hobolt SB (2020) Political Entrepreneurs: The References Rise of Challenger Parties in Europe. Princeton: Princeton Achen CH (1992) Social psychology, demographic variables, and University Press. linear regression: breaking the iron triangle in voting research. Dinas E (2014) Does choice bring loyalty? Electoral participation Political Behavior 14(3): 195–211. and the development of party identification. American Anduiza E and Pannico R (2020) Attitudinal consequences of Journal of Political Science 58(2): 449–465. partisanship for new parties. In: Oscarsson H and Holmberg S Emanuele V and Chiaramonte A (2019) Explaining the impact of (eds). Research Handbook of Political Partisanship. Chel- new parties in the Western European party systems. Journal tenham, UK: Edward Elgar Publishing, 281–293. of Elections, Public Opinion and Parties 29(4): 490–510. Bakker BN, Hopmann DN and Persson M (2015) Personality traits Evans G and Neundorf A (2020) Core political values and the and party identification over time. European Journal of long-term shaping of partisanship. British Journal of Political Political Research 54(2): 197–215. Science 50(4): 1263–1281. 12 Party Politics 0(0) Fiorina MP (1981) Retrospective Voting in American National Neundorf A, Stegmueller D and Scotto TJ (2011) The individual- Elections. New Haven: Yale University Press. level dynamics of bounded partisanship. Public Opinion Franklin CH and Jackson JE (1983) The dynamics of party Quarterly 75(3): 458–482. identification. American Political Science Review 77: Pannico R (2020) Parties are always right: the effects of party cues 957–973. and policy information on attitudes towards EU issues. West Gallego A, Rico G and Anduiza E (2012) Disproportionality and European Politics 43(4): 869–893. voter turnout in new and old democracies. Electoral Studies Poertner M (2020) The organizational voter: support for new 31(1): 159–169. parties in young democracies. American Journal of Political Gerber A and Green DP (1998) Rational learning and partisan Science 65(1). attitudes. American Journal of Political Science 42: 794–818. Prior M (2018) Hooked: How Politics Captures People’s Interest. Goren P (2005) Party identification and core political values. Cambridge, UK: Cambridge University Press. American Journal of Political Science 49(4): 881–896. Rodrı́guez-Teruel J, Barrio A and Barberà O (2016) Fast and Green DP, Palmquist B and Schickler E (2004) Partisan Hearts furious: Podemos’ quest for power in multi-level Spain. South and Minds: Political Parties and the Social Identities of European Society and Politics 21(4): 561–585. Voters. New Haven: Yale University Press. Schmitt-Beck R, Weick S and Christoph B (2006) Shaky at- Greene KF (2011) Campaign persuasion and nascent partisanship tachments: individual-level stability and change of parti- in Mexico’s new democracy. American Journal of Political sanship among West German voters, 1984-2001. European Science 55(2): 398–416. Journal of Political Research 45(4): 581–608. Hernández Pérez E, Galais Gonzàlez C, Rico G, et al (2021) Shively WP (1979) The development of party identification among POLAT project. Spanish Political Attitudes Panel Dataset adults: exploration of a functional model. American Political (Waves 1-6). DOI: 10.5565/ddd.uab.cat/243399 Science Review 73: 1039–1054. Huddy L and Bankert A (2017) Political partisanship as a social Sikk A (2012) Newness as a winning formula for new political identity. Oxford research encyclopedia of politics. Oxford, parties. Party Politics 18(4): 465–486. UK: Oxford University Press. Slothuus R (2016) Assessing the influence of political parties on Jacoby WG (1988) The impact of party identification on issue public opinion: the challenge from pretreatment effects. attitudes. American Journal of Political Science 32: 643–661. Political Communication 33(2): 302–327. Johnston R (2006) Party identification: unmoved mover or sum Stoker L and Jennings MK (2008) Of time and the development of of preferences? Annual Review of Political Science 9: partisan polarization. American Journal of Political Science 329–351. 52(3): 619–635. Klar S (2014) Partisanship in a social setting. American Journal of Torcal M, Martini S and Orriols L (2018) Deciding about the Political Science 58(3): 687–704. unknown: the effect of party and ideological cues on forming Lavine HG, Johnston CD and Steenbergen MR (2012) The Am- opinions about the European Union. European Union Politics bivalent Partisan: How Critical Loyalty Promotes Democ- 19(3): 502–523. racy. Oxford, UK: Oxford University Press. Tranter B and Smith J (2018) The mother of all effects? Stability Lucardie Paul (2000) Prophets, purifiers and prolocutors: towards and change in Greens party identification in Australia. En- a theory on the emergence of new parties. Party Politics 6(2): vironmental Politics 28(4): 1–24. 175–185. Vidal G (2018) Challenging business as usual? The rise of new Lupu N (2013) Party brands and partisanship: theory with evidence parties in Spain in times of crisis. West European Politics from a survey experiment in Argentina. American Journal of 41(2): 261–286. Political Science 57(1): 49–64. Zur R (2019) Party survival in parliament: explaining party du- Marinova DM (2016) Political knowledge in complex information rability in lower-house parliaments. European Journal of environments. Acta Politica 51(2): 194–213. Political Research 58(3): 960–980.