Breaking the cartel: the geography of the electoral support of new parties in Italy and Spain

Abstract

The factors explaining the emergence and electoral success of new parties have been investigated extensively on a cross-national basis, but little is known about why their vote shares differ within a specific country at any given time. This article is an attempt to fill this gap by examining the determinants of electoral support of Podemos and Movimento 5 Stelle across Spanish and Italian municipalities. Using empirical evidence from the 2014 European Parliament elections, we show that a bad economic situation at the local level increases the vote shares of these parties. We also demonstrate that these parties perform better in towns where abstention rates have previously increased, and the population is comparatively younger, and the possible existence of interaction effects between all these factors.

Introduction

The question as to what are the determinants of the emergence and electoral success of new parties in established democracies is a core topic in comparative political science. However, we still lack a parsimonious empirical model explaining the variation of this phenomenon within countries. This gap in the literature is particularly surprising given the huge attention devoted to this topic on a cross-national basis (e.g. Hug, 2001; Tavits, 2006). We aim to address this issue by offering new empirical evidence on the differences in electoral support obtained by a couple of new parties in recent elections within their respective countries.

The analytical literature on the electoral performance of new parties has traditionally revolved around the idea of the different permissiveness of electoral systems initiated by Duverger (1964 [1954]): new parties will be more successful if the rules of the game are hospitable enough to the small parties that are already in the system (Harmel and Robertson, 1985; Lucardie, 2000). Given that running in an election is costly and new parties usually lack organizational resources, their probability of success will tend to be associated with the incentives the electoral system generates to cast a tactical vote. New parties seldom emerge where the rules of the game confine electoral competition to few parties. By contrast, new parties frequently appear as viable competitors in national contexts with already high levels of party system fragmentation. Hence, these studies explain variation in the electoral success of new parties by resorting to the differences in the institutional setting governing the conduct of elections.

In this article, we argue that these explanations are often insufficient to account for variation in vote shares of new parties within countries. Using the case of Podemos (Spain) and Beppe Grillo’s Movimento Cinque Stelle (M5S, Italy) at the 2014 European Parliament (EP) elections, where the incentive structures for tactical voting do not vary within each country due to the single national electoral district, we show, first, that these new parties performed particularly well in municipalities where the economic crisis was particularly deep and, second, that the degree of voters’ discontent with established parties and electoral habituation play a crucial role in accounting for the success of these new parties. The 2014 EP elections in Spain and Italy constitute the perfect context for testing our arguments given the particular acuteness of the recent economic crisis in these two countries and the extraordinary performance of Podemos and M5S in that occasion. Moreover, although these parties share to a large extent a critical position on the European Union, the euro, and the austerity policies imposed by the Troika, they have ended up belonging to different groups in the EP. Finally, the recent success of these two new parties is particularly puzzling because it does not fit any of the three patterns of party replacement posited by Shin and Agnew (2007), neither they have taken over the electoral support enjoyed by or the social world represented by a pre-existing party nor they have shared its votes with other new parties.

The rest of the article proceeds as follows. In the second section, we present the previous literature and our theoretical arguments on the electoral performance of new parties, while in the third, we develop our working hypotheses. In the fourth and fifth sections, we briefly describe our cases and data, and discuss the results of the empirical analysis, respectively. Finally, in sixth section we present our conclusions.

A frame for the analysis of voting behaviour in Spain and Italy nowadays

In explaining the determinants of electoral politics, there have traditionally been three groups of studies that are intimately connected but largely independent: models of stable cleavage-based or partisanship-based behaviour, research focussed on short-term factors, and ideological theories of vote choice. ¹ Bearing this in mind, in this brief literature review we will highlight the role played by four different explanatory factors: class and religious divisions, party identification, state of the economy, and ideology.

Early electoral research argued that during the 1950s and the 1960s voters relied on group cues (i.e. cleavages) and partisan loyalties (i.e. party identification) to guide their individual behaviour at the polls. According to the first of these approaches, structural cleavages like national identity or religious denomination were considered a major determinant for political attitudes and voting behaviour since the 1920s (Lipset and Rokkan, 1967). However, voting would be above all an expression of citizens’ social position and the well-established values and interests associated with it. Within this framework, the persistent importance of cleavage voting would be explained by the presence of middle-structures (e.g. trade unions, churches, etc.), which develop and consolidate these cleavages. But, starting in the late 1960s, the social structural basis of party support seemed to come to an end (Franklin et al., 1992).

In democratic systems, party identification, that was defined as ‘the individual’s affective orientation’ to a party as a group (Campbell et al., 1960: 121), was at the centre of the explanation of electoral choices from 1960 onwards. The concept of party identification was seen as important for four main reasons. In the first place, party identification affects how individuals perceive and process political information. The authors of The American Voter used the metaphor of ‘a perceptual screen’ through which voters see what is favourable to their partisan orientation (Campbell et al., 1960: 133). The second and probably most relevant role played by party identification is its explanatory power of vote choice. Third, although other forms of political participation were not mentioned in this book, it is also expected that this attitudinal variable affects individuals’ propensity to demonstrate, contact their representatives or simply sign a petition. Finally, party identification is important for the Michigan School because it is a long-term stable affinity with a party. According to Converse (1969), party identification reflects accumulated voting experience and, as individuals get older, they have participated in more elections and, hence, their party preferences consolidate. This means that party identification is more intense and mainly more stable for old rather than young voters. This last point will be crucial to justify one of our hypotheses below.

Even when the results of numerous works suggested that party identification mattered to explain vote choice, this model came under strain in the 1980s as well (Dalton et al., 1984). Partly as a consequence of this increasing criticism, the study of economic voting has flourished in the last three decades with dozens of pieces published on the topic (e.g. Blais et al., 2004; Kosmidis and Xezonakis, 2010; Kayser and Wlezien, 2011; Lewis-Beck and Stegmaier, 2013). The basic idea behind economic voting refers to the association of the incumbent’s electoral fate to the state of the economy (Key, 1966). In other words, voters would provide or withdraw support to the party in government depending on whether they see economic prosperity or decline. The precise nature of economic voting has generated many fundamental questions. The first of these puzzles concerns the actual existence of this relationship. The idea that voters cast their vote based on the state of the economy is even challenged by some authors (Cheibub and Przeworski, 1999). A second question deals with the direction of the electorate’s scrutiny. One may wonder whether voters take into account the past economic performance of the incumbent government (Fiorina, 1981), or whether they compare the (expected) future utility they would obtain from having each of the competing parties in office (Downs, 1957). In other words, the question is whether they are retrospective or prospective voters. A third question tackles which types of economic conditions are assessed by economic voters, and, more specifically, whether voters tend to be egotropic and only consider their pocketbook when casting an economic vote (Butler and Stokes, 1974 [1969]), or whether national conditions that matter for electoral behaviour encourage sociotropic considerations and behaviours (Kiewiet, 1983). Finally, there is the fundamental question of which aspects are relevant when voters judge the state of the economy, as they may look at an array of macroeconomic indicators such as GDP growth, inflation, or unemployment.

Despite these controversies, scholars in the field of economic voting agree on two general points: the conditioning effect of the context and the endogeneity of economic evaluations. We examine each of these in turn. First, citizens need to consider the government in charge of the economy before casting their vote on this basis. That is, the importance of economic voting will be higher in a context in which the incumbent is clearly responsible for the state of the economy (Powell and Whitten, 1993). Single-party majority governments in parliamentary and unitary systems are, among others, examples of high-clarity contexts (Fisher and Hobolt, 2010). Besides the political sources of lack of clarity, there are the economic ones, of which globalization is the most relevant example (Fernández-Albertos, 2006). With respect to endogeneity, suffice it to say that the more than likely impact of partisanship on economic perception variables forces us to cast doubts about the real magnitude of the estimated economic voting (Evans and Andersen, 2006).

Finally, ideology can affect voter behaviour for causes similar to those relating to party identification (Enelow and Hinich, 1984). The original spatial models of electoral competition mainly argue citizens vote out of ideological closeness to parties or candidates (Downs, 1957). However, the analysis of the impact of the ideology on voting behaviour has become more complex in recent times. On the one hand, according to the directional model (Rabinowitz and MacDonald, 1989), voters prefer parties ‘on their side’ of the ideological spectrum, and the more on their side the better, provided they are not too extremist and fall, hence, outside of what they call the ‘region of acceptability’. On the other, under Kedar’s (2005) compensation model, voters are concerned with policy outcomes and can, hence, end up supporting parties whose positions differ from their own insofar as these parties pull policy in a desired direction.

The determinants of electoral success of new parties

The emergence of successful new parties has mostly been explained by examining the impact exerted by the elements of the inter-party dimension of the electoral system such as the electoral formula, district magnitude, legal thresholds, and assembly size (Lago and Martínez, 2011). According to this idea, the number of parties (old, new, or both) depends on the incentives offered by the electoral system. These incentives determine the level of electoral coordination between parties and voters alike. Within this framework, Harmel and Robertson (1985) distinguish between the formation and the success of new parties. According to these authors, only the latter is associated with the type of electoral system. Hino (2012) revisits Harmel and Robertson’s evidence and also urges us to differentiate between the factors that explain a new party’s emergence and the conditions that affect its subsequent electoral performance. A second key contribution of her book is that we should not assume that the fate of all new parties is equally determined by the electoral system. In a similar vein, Bolleyer and Bytzek (2013) show that the permissiveness of the electoral system only increases a new party’s chance of repeated re-election once it has been re-elected for the first time.

Numerous other studies have focussed on the impact of the electoral rules on the formation and subsequent electoral success of new parties. Lucardie (2000), for example, concludes that first-past-the-post systems will in general prevent the proliferation of new parties by reducing their political opportunities, and Tavits (2006) argues that new parties are more frequent where it is easier to win a seat because the cost of entry into party competition is lower. Likewise, Bollin (2007) finds that the electoral system is an important factor in the explanation of the entrance of new parties into the legislature. Finally, some authors have shown the conditioning effect of the electoral system. For example, Hug (2001) gives credence to the idea that the institutional framework of elections mediates the impact of political issues on the emergence of new parties; and Lago and Martínez (2011), in a study very similar to ours, find evidence of a positive effect of electoral market failure on the probability of successful entries of new parties that decreases as the restrictiveness of the rules of the game increases.

The problem with these accounts is that they are inherently comparative and while they provide insights into factors that affect the entry of new parties into national party systems, they cannot help us to explain different magnitudes of a new party’s breakthrough within the same country. For this reason, we still lack an explanation accounting for the actual variation in the electoral success of new parties at the subnational level. This is only partially true in the case of Spain, where the combination of low and high district magnitudes allows Lago and Martínez (2011) to test the effect of the electoral system on the entry of new viable competitors in regional party systems. Although valuable, these authors’ approach is in our view not completely satisfactory for two reasons: the neglect of the potential role played by economic outcomes; and the main focus on the explanatory power of institutional factors. In order to illustrate the latter, let us take the case of the general elections in the Netherlands.

As probably known, the seats of the House of Representatives in that country are distributed on a nationwide basis among party lists that obtain at least one full electoral quota. ² So, electoral incentives are constant and exert similar pressures on parties countrywide. Even more importantly, variation in vote shares of new parties across different parts of the Netherlands, like in Italy and Spain at the EP elections, cannot be understood by resorting to the operation of the electoral institutions. By contrast, in this article we will argue that the variation in the levels of electoral support of new parties registered within one institutional context depends on three key factors: the depth of the economic crisis, voters’ levels of discontent with established parties, and their degree of electoral habituation. It is to the detailed explanation of these factors that we turn now.

The first factor is the depth of the economic crisis. ³ As we have previously noted, bad economic outcomes lead the electorate to vote for the opposition. However, which parties are the challengers in these cases? With respect to Italy, the government on the eve of the 2014 EP elections was supported by an oversized coalition including all the largest party groups in the parliament but M5S, that is, the centre-left Partito Democratico, the centre-right Il Popolo della Libertà, and the centrist Scelta Civica (di Virgilio and Giannetti, 2014). As a result, the most viable option for those citizens unwilling to support the government was to vote for the M5S. The same can be said grosso modo about Spain. Although the 2011 general election held in that country saw the centre-right Partido Popular win an absolute majority and form a single-party government, it would be wrong to think that the Spanish electorate exonerated the Partido Socialista Obrero Español, the main opposition party in power until 2011, from the responsibility of producing bad economic outcomes. This fact first and foremost explains why voters especially hit by the economic crisis should be more likely to vote for Podemos as well (Fernández-Albertos, 2015).

The second factor relates to the level of voters’ discontent with established parties. According to the conceptual framework provided by Hirschman (1970), voters can react to such discontent in one of two possible ways. ⁴ The easiest solution is that of exit. Some discontented voters may decide to ‘exit’ by simply staying home on election day. By contrast, some of the voters discontented with the operation of the party system can opt for voicing their dissatisfaction by voting for minor parties. In this article, we argue that the amount of previous exit accounts for the dimensions of the current success of new parties. There are at least two mechanisms explaining this relationship. First, it could be that voting for a party is self-reinforcing, as some authors argue (Dinas, 2010). In this context, once voters have already supported an established party in the past, changing to a new one would be more difficult. Second, previous abstention rates could be related to some prior covariates (e.g. antiparty sentiments) that would explain not only lower participation rates at the previous elections but also support for new parties in the current ones.

An alternative explanation for the lower turnout rates in the EP elections lies on their characterization as second-order elections (Reif and Schmitt, 1980; Norris, 1997). According to this theory, some elections are largely perceived by the electorate as inconsequential for their lives, and, as a result, they are less likely to participate in them. We expect a weaker impact of previous turnout rates on vote shares of new parties if this ‘less at stake’ argument explains why individuals vote at lower rates in EP elections. However, we think that using changes in abstention rates rather than levels as explanatory factor can contribute to (at least partially) solve this problem. As Galais and Blais (Forthcoming) have recently shown, a weaker feeling of duty of voting in EP elections is responsible for lower turnout rates in this type of elections. And this feeling is arguably sticky and barely changes over time. This is why we think that increases of abstention rates in EP elections across municipalities occur when political demands are not satisfied.

The third factor that could affect voters’ likelihood of supporting new parties is the degree of citizens’ electoral habituation. Let’s imagine, for example, two Italian voters that are 40 and 25 years old, respectively. As a consequence of her age, voter A has been eligible to vote for more than 20 years. On this basis, she has developed a strong loyalty towards one party, and voting for it has become a habit (McPhee and Ferguson 1962; Converse, 1969, 1976; Butler and Stokes, 1974 [1969]). By contrast, voter B is a relatively inexperienced voter. According to the habituation hypothesis, she is not anchored to a particular voting pattern yet. For this reason, she should be more likely to switch to a new party when it emerges. To put it in Hirschman’s words again, older people are more reluctant to defect from the party they are identified with because they have higher levels of loyalty towards it.

That electoral habituation is unthinkable without established party systems is now conventional wisdom in political science. When party systems are young and/or in flux, parties are not well rooted in the electorate (Converse, 1969) and voters are, as a result, more likely to be volatile (Dalton, 2013). Using panel data from Brazil’s 2002 presidential elections, Baker et al. (2006) show that social networks play a major role in shaping voters’ party preferences in case of weak partisan attachments among the electorate. Unlike many ‘third wave’ democracies, rapid preference changes and unpredictable election campaigns had been quite uncommon in Italy and Spain until recent elections. With the remarkable exception of the ‘political earthquake’ of the early 1990s, the Italian electorate used to be quite stable until 2013 and vote transfers were largely taking place within ideological blocs (Russo, 2014). Likewise, the Spanish party system suffered a profound restructuration as a consequence of the critical 1982 elections, but had stayed considerably stable since then (Gunther and Montero, 2001).

Finally, it would be possible to posit two slightly different variants of the economic voting story. These two variants differ in terms of what they have to say about the effect of the economic crisis. The first variant argues that the impact of the economic crisis depends on the volume of previously discontented voters. Increases of abstention rates in the past also matter because economic voting is believed to have its strongest effect where there has been an important increase in the pool of dissatisfied voters who did not participate in the previous elections. The smaller the increases of the abstention rates in the past, the less likely it is that economic conditions will significantly influence the vote shares of new parties. The second variant of conditional argument is very similar except that it argues that the electoral performance of new parties will be more closely tied to economic conditions if the average voter has not developed stable voting patterns yet. One interpretation of this latter scenario is that it implies a poorly institutionalized party system where voters do not support ‘their party’ irrespectively of considering its performance ‘right or wrong’. Hence, Hypotheses 4 and 5 require the concurrence of two types of factors (i.e. economic and political) in order to see a new party performing electorally well. On the one hand, economic motivations are present when voters have been particularly hit by the financial crisis. On the other hand, political reasons lead voters to support new parties because it is less costly to do so for them either because they did not vote in the previous election or their partisan outlooks have not consolidated yet.

In sum, given that the electoral system is a constant within Italy and Spain in the 2014 EP elections, it cannot explain the differences in new parties’ electoral performance across municipalities. Hence, we need a new theoretical framework to understand why new parties obtain more votes in some places than others. On the basis of the arguments presented above, we can formulate the following hypotheses on the factors that explain the variation in Podemos and M5S’s results in the 2014 EP elections across municipalities:

Hypothesis 1 The deeper the economic crisis is, the higher the vote shares of new parties.

Hypothesis 2 The more voters discontented with established parties are, the higher the vote shares of new parties.

Hypothesis 3 The more electorally habituated voters are, the lower the vote shares of new parties.

Hypothesis 4 The depth of the economic crisis has a positive effect on vote shares of new parties that becomes more pronounced as the size of voters’ discontent with established parties increases.

Hypothesis 5 The depth of the economic crisis has a positive effect on vote shares of new parties that becomes less pronounced as voters’ electoral habituation increases.

Data and methods: why Podemos and M5S at the 2014 EP elections?

As previously mentioned, we will conduct an intensive comparative study of two cases in order to properly answer the question on where new parties are electorally more successful. Given the presence of unobserved heterogeneity in cross-national comparisons, within-country studies are particularly suitable for our purposes: they allow us to go slightly beyond mere correlations by controlling for the effect of covariates that could have an impact on our dependent variable (Przeworski, 2007).

We rely on data at the municipal level from the 2014 EP elections in Spain and Italy to show that the magnitude of the economic crisis, the existence of voters discontented with established parties, and the absence of habituated voters improved the performance of Podemos and M5S. There are two main methodological reasons for this case selection. On the one hand, despite sharing their newness, these are two quite diverse parties that allow us to draw some general conclusions about the determinants of electoral support of new parties. More specifically, M5S is some kind of populist movement not easily definable on ideological grounds (Russo et al., 2015), while Podemos is clearly a left-wing party (Fernández-Albertos, 2015). For instance, in the EP the M5S is part of the Europe of Freedom and Direct Democracy group along with far-right parties such as the UK Independence Party or the Swedish Democrats. By contrast, Podemos is part of the European United Left/Nordic Green Left along with most left-wing non-Socialist parties of Europe. In addition, there was a second crucial difference between these two parties in 2014: the M5S had participated in conventional ⁵ politics since 2009, while the examined EP elections were the first test for Podemos.

A second methodological reason for this case selection lies in the three features that these two cases have in common that allow us to test the validity of our hypotheses while controlling for other variables (mainly, institutional factors) that might also explain the variation in the electoral performance of new parties across countries and over time. First, in both countries the allocation of EP seats to parties is proportional to the total number of votes obtained at the national level. This implies that the effects of our independent variables are tested not only in the same electoral context (i.e. the 2014 EP elections) but also under the operation of a similar institutional framework. However, resemblances stop here. In Spain, the electoral formula is the D’Hondt method and there is no threshold. Party lists are closed and national. By contrast, in Italy seats are allocated according to the Hare formula. Parties have to obtain at least 4% of the national vote in order to be eligible to receive seats. Citizens can cast up to three preferential votes. Once the number of seats that correspond to each party has been calculated on the basis of its national vote, seats are allocated to the regional lists of each party. In the 2014 EP elections, Spain and Italy elected 50 and 72 MPs, respectively.

Second, the 2014 EP elections in Spain and Italy were characterized by the outstanding performance of these two new parties. Despite the tardy official decision to launch the party (only 4 months before the elections were held), the list of Podemos headed by its leader Pablo Iglesias obtained an impressive 8% of the total vote share, beating estimations by the official polling institution (Centro de Investigaciones Sociológicas) by more than 6 percentage points. ⁶ Likewise, although the M5S in its first participation in an EP election performed slightly worse than in the 2013 Italian general election (from 25 to 21% of the national vote), it still managed to be the second most-voted list just behind the Partito Democratico of the incumbent Prime Minister Matteo Renzi. If, hence, we aim to address the determinants of the electoral performance of new parties, the 2014 EP elections seem to be the perfect scenario to do so.

The third point shared by these two cases is the depth of the economic crisis registered in these countries since 2008. In Spain, and according to Eurostat, there was a remarkably severe decrease of the volume index of GDP per capita in Purchasing Power Standards that passed from 102 in 2008 to 93 in 2014, ⁷ and the number of people at risk of poverty increased from 8,161,000 in 2008 to 8,517,000 in 2014. ⁸ The unemployment rate also considerably augmented, reaching its peak of 26.1% by 2013, ⁹ and being particularly high for young people. This incredibly bad economic situation had for sure political consequences: the ruling Socialist Party since 2004 was forced to call early elections in 2011 and was swept from power with its worst result ever since the democratic transition. The main opposition force, the People’s Party, obtained a record number of seats and votes, and managed to form a single-party majority government (Chari, 2013).

The Italian situation is very similar. According to Eurostat, the volume index of GDP per capita in Purchasing Power Standards shrank from 106 in 2008 to 97 in 2014, ¹⁰ and the number of people at risk of poverty went up from 9,157,000 in 2008 to 9,201,000 in 2014. ¹¹ The unemployment rate almost doubled going from 6.7% in 2008 up to 12.7% in 2014, ¹² reaching a peak of 55.4% for young people (15–24 years). ¹³ The severe financial crisis also had direct consequences in the political domain and, more specifically, for the incumbent government: as a result of the extreme deterioration of the economy, the government’s credibility decreased so much that Prime Minister Silvio Berlusconi was forced to resign in November 2011 (Magalhães, 2014). However, due to the extremely bad economic situation, new elections were not held and the main centre-left (PD) and centre-right (PdL) parties reached an agreement to install and support a technical government led by the economist Mario Monti. New elections took place in February 2013, but due to the outcome (M5S collected 25% of votes and refused to form a coalition government, as largely announced during the campaign), ¹⁴ a new Grand Coalition with both centre-left and centre-right parties was formed.

As is standard in the economic voting literature, we use the change in the municipal unemployment rate between May 2008 (i.e. before the economic crisis started) and May 2014 (i.e. the month in which the EP elections took place) for Spain. ¹⁵ Due to the lack of equivalent data for Italy, we use the change in the unemployment rate at the provincial level (110 provinces) between the same dates. ¹⁶ Our proxy for voters’ discontent with established parties is the change in abstention rates between the previous two EP elections (Lago and Martínez, 2011). New competitors that enter the electoral fray for the first time in 2014 are expected to perform better where the results of the previous elections suggest that there is an opportunity window for them. And sharp drops in electoral participation between 2009 and 2004 seem to be a good indicator of this opportunity window. ¹⁷ Finally, to capture the effect of habituated voters we use citizens’ average age in each municipality. ¹⁸ As we have previously seen, the level of voters’ habituation is a function of their age.

Our aggregate-level controls are female and foreigner rates, ¹⁹ the log of the number of eligible voters, ²⁰ and the mean ideology of the municipality. We decided to opt for a parsimonious strategy with few and very exogenous controls. The first two are measured at the municipal level and the provincial level for Spain and Italy, respectively. We take the log of eligible voters in order to capture possible non-linearities. In order to calculate the mean ideology, we follow Fernández-Albertos’ (2015) empirical strategy and use survey data from the last general elections held in each country. ²¹ First of all, we place all parties with parliamentary representation on the traditional left-right axis by considering the self-declared ideology of their reported voters at the national level (0–10 scale). We then weight this ideology by parties’ vote shares in each municipality. ²² The descriptive statistics for all the variables included in the analyses are displayed in Tables 1 and 2.

Table 1 Descriptive statistics, Spain

Table 2 Descriptive statistics, Italy

M5S=Movimento Cinque Stelle.

With respect to modelling, we have taken every municipality in each country as a distinct case. The number of municipalities is similar, being slightly higher than 8000 in both countries. Our dependent variable, the vote share of each new party, is a continuous one. ²³ For this reason, we have run ordinary least squares regressions. All models include fixed effects at the regional level to avoid unobserved heterogeneity biasing our estimates, and clustered standard errors by region to tackle heteroskedasticity problems. We prefer to avoid the use of province-fixed effects in order to consume fewer degrees of freedom. As most of our explanatory factors are computed at the municipal level, we do not see it necessary to implement hierarchical linear models. ²⁴

Results

Before conducting the multivariate analyses, Figures 1 and 2 show the distribution of vote shares of the two examined new parties across regions. When comparing the two countries, the percentages reported allow us to reach a first tentative conclusion: the variance of the electoral performance across regions is much higher for Podemos than for the M5S. However, it is very difficult to establish which factors are systematically associated with higher levels of electoral support for these two new parties. In Spain, for example, Podemos obtains its best and worst results in the two regions with the most important capitals: Madrid and Catalunya, respectively. Although interpreting the results for Italy is not much easier, a closer visual inspection of the data summarized in Figure 2 seems to reveal the existence of a slightly positive correlation between the regional unemployment rate and the vote shares of M5S.

Figure 1 Podemos’ vote shares across municipalities by region, 2014 European Parliament elections. Official results from the Ministerio del Interior.

Figure 2 Movimento Cinque Stelle’s vote shares across municipalities by region, 2014 European Parliament (EP) elections. Official results from the Ministero dell’Interno.

Tables 3 and 4 extend the previous analysis into a multivariate setup incorporating important aggregate-level controls. More specifically, Table 3 reports a set of linear specifications predicting vote shares of Podemos. As shown in models 2 and 3, the change in the unemployment rate have a positive coefficient with a very satisfactory level of statistical significance (p⩽0.01) when predicting the electoral performance of Pablo Iglesias’ party. For every unit increase in unemployment rate between 2008 and 2014, Podemos’ vote shares grow between 0.12 and 0.13 percentage points on average. This is a remarkable magnitude given that Podemos obtained slightly less than 8 percentage points in the 2014 EP elections. However, this result is not robust as the coefficient for change in the unemployment rate fails to reach conventional levels of statistical significance, even if it still has the correct sign and a non-negligible magnitude, when we control for average age (models 4 and 6). Overall, the results in Table 3 partially validate our first hypothesis.

Table 3 The determinants of Podemos’ vote shares across municipalities in the 2014 European Parliament elections

Estimation is by ordinary least squares with robust standard errors by region in parentheses. Fixed effects by region are included but not shown.

The levels of statistical significance are *p<0.10, **p<0.05, and ***p<0.01 (two-tailed tests).

Table 4 The determinants of Movimento Cinque Stelle’s vote shares across municipalities in the 2014 European Parliament elections

Estimation is by ordinary least squares with robust standard errors by region in parentheses. Fixed effects by region are included but not shown.

The levels of statistical significance are *p<0.10, **p<0.05, and ***p<0.01 (two-tailed tests).

Consistently with Hypothesis 2, the measure of voters’ discontent with established parties, change in the abstention rate between 2009 and 2004, has a positive, non-negligible, and significant effect on the electoral performance of Podemos. The magnitude of the coefficients ranges from 0.032 in model 5 to 0.04 in model 3. Finally, the coefficient of the voters’ habituation measure (i.e. citizens’ average age) has a larger size and a considerable level of statistical significance (p⩽0.01) in all models. With regard to interactions, the registered positive effect of unemployment change is higher among considerably demobilized electorates (model 5) and relatively young citizens (model 6), even though none of these effects are significant at traditional levels of statistical confidence. In terms of controls, Table 3 shows that women and foreigners rates have a statistically insignificant effect on the electoral performance of the new party across all model specifications. By contrast, population size (in three out of six models) and ideology (in all models) are negatively associated with Podemos’ vote shares. Whilst the sign of the latter was expected (Podemos has been consistently identified as a left-wing party), the former effect is somehow puzzling because new parties should arguably perform worse in small (rural) municipalities. As regards the goodness-of-fit in Table 3, the amount of explained variance ranges from 16% in model 1 to 21% in models 4, 5, and 6.

Table 4 replicates the same analyses shown in Table 3, but using M5S’s vote shares as the dependent variable. The change in unemployment and abstention rates and the citizens’ average age perform well, which suggests once again that the economic conditions and the degree of voters’ political discontent and electoral habituation need to be taken into account. By contrast, none of the hypothesized interaction effects shows up. In terms of controls, Table 4 suggests that women are less likely to vote for M5S, while population size is positively associated with higher levels of M5S’s vote shares. The statistically insignificant effect of ideology corroborates previous findings on the composition of the electorate of the M5S (Russo et al., 2015). As regards the goodness-of-fit in Table 4, the amount of explained variance ranges from 28% in model 1 to 38% in models 3, 4, 5, and 6.

Finally, in Figures 3 and 4 we plot the conditional effect of change in unemployment rates on parties’ vote shares for the whole range of values of the modifying variables (i.e. change in abstention rates between 2004 and 2009, and citizens’ average age). According to Brambor et al. (2006) and Kam and Franzese (2007), the effect of an interaction term cannot be evaluated through the p-value shown in the regression table. Thus, it is necessary to graphically illustrate the marginal effect of the main explanatory factor on parties’ electoral performance over different values of the other constitutive term of the interaction. As predicted in the hypotheses section, increases of the unemployment rate over time has a significant positive effect on Podemos’ vote totals when the increase of the abstention rate in the previous EP election is sufficiently big (i.e. when it is, more or less, >10%) but not when it is lower than that threshold. By contrast, the right panel in Figure 3 does not show any statistically significant effect of increases in unemployment rates and, hence, does not corroborate Hypothesis 5 in the case of Spain. As regards Italy, the evidence displayed by the left panel of Figure 4 runs against our fourth hypothesis and seems to suggest that bad economic outcomes only help new parties when abstention rates have decreased or moderately increased in the past. Finally, according to expectations, positive changes in the unemployment rate at the municipal level increase the vote shares of M5S only when citizens’ average age is sufficiently low (i.e. about lower than 50 years old). To sum up, the results presented here provide mixed evidence about the validity of our interactive hypotheses.

Figure 3 Marginal effect of unemployment change on Podemos’ vote shares as the value of several modifying variables increases (ordinary least squares regression estimates), 2014 European Parliament (EP) elections. Results are derived from Table 3, models 5 and 6.

Figure 4 Marginal effect of unemployment change on Movimento Cinque Stelle’s vote shares as the value of several modifying variables increases (ordinary least squares regression estimates), 2014 European Parliament (EP) elections. Results are derived from Table 4, models 5 and 6.

Conclusions

Europe’s recent past has shown that a context of economic crisis may lead to the emergence of new parties and a considerable increase, as a result, of the levels of electoral volatility. The cases of Podemos in Spain and the Movimento 5 Stelle in Italy are good examples of these patterns. Despite the tardy official decision to enter the competition (only 4 months before the elections were held), the M5S gained an impressive 25% of the total vote share in the 2013 Italian parliamentary election, making it the most-voted list. Likewise, Podemos obtained a remarkable 8% of the total vote in its first participation ever in an election (i.e. the 2014 EP election). Looking at empirical evidence from the latter elections in these two countries at the municipal level we, first of all, have shown that variation in the electoral performance of new parties is considerable at this level, and, second, we have analysed the determinants of this variation.

There are important commonalities but also remarkable differences in the key factors that explain the electoral fortunes of new parties across countries. With regard to the former, bad economic outcomes (i.e. a considerable boost of the unemployment rate at the municipal level during the crisis years) increase in overall terms the vote shares of new parties. Moreover, the existence of unsatisfied political demands shared by a significant number of individuals, measured as increases in abstention rates between the 2009 and the 2004 EP elections, has a positive impact on vote shares of Podemos and the M5S. Third, the degree of voters’ electoral habituation, measured as the average age of citizens in each municipality, negatively affects the electoral performance of new parties. These similarities are particularly meaningful because they suggest that the same factors explain the electoral performance of two recently formed parties in spite of their important genetic and ideological differences.

Yet, the examined cases significantly diverge regarding the hypothesized interaction effects. On the one hand, only in Spain the positive effect of the change in municipal unemployment rates appears to be conditioned by the increases of non-voters in the previous EP elections. On the other, our economic indicator shows a significant effect on the vote shares of M5S but not Podemos in case of municipalities with a young population. What is the meaning of these differences? Although we will leave for further research a more elaborate interpretation of these results, we think that individuals’ levels of political information and the gap in the age of the two parties could explain these differences: Podemos was running for the first time in 2014 and for this reason it was not particularly successful at attracting votes from young arguably uninformed people in high-unemployment municipalities (Fernández-Albertos, 2015). By contrast, the M5S had already run in previous elections at the national and subnational levels and emerged as an appealing option for those young poorly informed voters specially hit by the economic crisis. ²⁵

Recent developments in the literature on new parties have pointed to the importance of taking into account the institutional context in which they operate to understand their electoral performance. Existing models have helped to explain how parties’ fortunes vary across countries. However, quantitative approaches that examine how contextual variables may determine their electoral results within one country are still missing. The findings presented in this article are only a first step in the framing of a comprehensive theory of the determinants of electoral performance of new parties at the subnational level. Nevertheless, they suggest that we should devote more attention to disentangle the mechanisms that lie beneath the success of these political actors. Future studies of other new parties that operate under different institutional arrangements are necessary to evaluate whether our findings can be generalized. To be more specific, it would be interesting to study the case of Syriza, the only example within the recent wave of new parties that has managed to reach the national government. Moreover, our findings demonstrate the necessity of digging deeper into the determinants of vote shares of new parties by looking at the performance of the parties under consideration here (i.e. Podemos and M5S) in other electoral contexts, and examining whether second-order elections such as the EP elections constitute an important stepping stone in the emergence and consolidation of new parties in Europe.