Microsoft Word - PDF_Issue_13_3_Mancosu.docx Italian Political Science, VOLUME 13 ISSUE 3, DECEMBER 2018 © 2018 Italian Political Science. ISSN 2420-8434. Volume 13, Issue 3, pp. 44-60 Contact Author: Moreno Mancosu, Collegio Carlo Alberto, Turin. E-mail address: moreno.mancosu@carloalberto.org The deep roots of populism: Protest, apathy and the success of Movimento 5 Stelle in the 2013 Italian Elections Moreno Mancosu COLLEGIO CARLO ALBERTO, TURIN Abstract Previous studies have argued that voters’ attitudes of apathy and protest towards the Italian party system, trig- gered by the economic crisis of 2011, were exploited by Movimento 5 Stelle to increase its support in the 2013 Elections. However, little attention has been paid to the determinants of dissatisfaction that voters showed even before the sovereign debt crisis. In addition to providing a thorough geographical exploration of the party’s geo- graphical scattering (with the aid of global and local indicators of spatial autocorrelation), this article tests the relationship between the strength of the Movimento in 2013 and apathy and protest indicators in the pre-crisis period. SAR (Spatial Auto-Regressive) regression models will be employed in order to avoid biased estimates due to spatial autocorrelation. Results show that indicators of apathy or protest represent significant predictors of the Movimento 5 Stelle’s geographical distribution. 1. Introduction he Italian national election of 2013 represented a sudden discontinuity with re- spect to the balance of power that ruled in the so-called Italian Second Republic (ITANES 2013; Vegetti, Poletti and Segatti, 2013). Election results analyses showed that about 40% of voters changed their preferences with respect to previous elec- tions (ITANES, 2013). Scholars also witnessed a dramatic decrease in support for the two main parties, albeit with their names changed, that had ruled the Italian political land- scape in the previous 20 years: the centre-left Partito Democratico (Democratic Party, PD) experienced a loss of almost 3.5 million votes from previous elections, and the right- wing competitor, Berlusconi’s Popolo della Libertà (People of Freedom, PdL), reported a loss of more than 6.2 million votes. A large number of these votes were collected by a new populist party that ran for the first time in a national election and gained 25% of valid votes: Movimento 5 Stelle (Five Star Movement, M5S). As has been underlined many times in the recent literature (Corbetta and Gualmini, 2013), one of the main characteristics of the ideological apparatus of the Movimento is the rejection of traditional political cleavages, starting from the difference between ‘left’ and ‘right’. The main cleavage that the Movimento theorises is that between the ‘people’ and the ‘caste’ of politicians, considered as a close, self-referenced group of corrupt people, unfit to govern the country (Biorcio and Natale, 2013). A crucial element that intersects both the electoral earthquake of 2013 and the claims of Movimento 5 Stelle is the economic and political context in which the 2013 elections took place, characterized by a harsh economic crisis (triggered by the sovereign debt crisis T MORENO MANCOSU 45 of 2011), which led the Italian electorate to develop much more critical attitudes towards politicians and the traditional party system (ITANES, 2013; Bellucci and Maraffi, 2014). As previous research has stressed, the ability of the Movimento and its leader, the former comedian Beppe Grillo, to exploit the disenchanted and anti-political feelings of part of the electorate (Tronconi, 2013; Vegetti, Poletti and Segatti, 2013) contributed to increas- ing electoral support for the party. Several studies (Mosca, 2014; Bordignon and Ceccarini, 2014; Mosca and Quaranta, 2017; Biorcio, 2014) underline the relationship be- tween indicators that measure protest and propensity to vote for the Movimento 5 Stelle. In particular, Bellucci and Maraffi (2014) show quite clearly that the 2011 sovereign debt crisis and the related voters’ feelings of disenchantment towards the political system largely affected Italians’ choices in 2013 (for similar results, see Vezzoni, 2014; ITANES 2013). The elections of 2013 and the emergence of Movimento 5 Stelle are thus important for understanding today’s political trends in Italy because they represent a ‘unicum’, a re- aligning election (D’Alimonte, 2013), happening in a context of profound economic crisis. If previous studies allow us to assess the relationship between protest and the Mo- vimento’s support in a cross-sectional fashion, they do not tell us much about the roots of this protest. Feelings of protest, indeed, did not start with the 2011 crisis and can be traced back to the First Republic era (Mannheimer and Sani, 2001; Tuorto, 2006). This paper aims to account for these ‘protest’ roots, by arguing that pre-crisis elements of discontent might be relevant in explaining the Movimento’s success in 2013. Consistent with previ- ous literature (see Martini and Quaranta, 2015), it is argued that a part of the electorate was already dissatisfied with democracy and the party system way before the debt crisis unfolded, and that this same quota of voters was already looking for a political actor that would be able to transform that dissatisfaction into actual preferences and support. The debt crisis, together with the appearance of Movimento 5 Stelle, provided the ‘perfect storm’ for a situation that had been simmering below the surface for a long time.1 To test our theoretical argument, election results (before and after the 2011 sovereign debt crisis) at the municipal level will be employed. These data, indeed, allow us to measure the char- acteristics of the contexts in which people are embedded over time, providing regression coefficients in which the independent variable is, by design, exogenous.2 2. Background 2.1. Movimento 5 Stelle: a new Italian populism? As stressed above, the national elections of 2013 were held in a general climate of – eco- nomic and political – turmoil. The political system instability period started in November 1 An additional indirect proof of this argument is the national elections of 2018 that showed that, even if the economic crisis was not as harsh as in 2013, the Movimento succeeded in increasing its share of votes (gaining almost 33% of valid votes). 2 This work aims at identifying evidence of the connection between contextual political protest and apa- thy before the start of the economic crisis and contextual levels of support for the Movimento during the zenith of the political and economic crisis in which the party emerged. For our aims, thus, subsequent results obtained by the party (such as the European elections of 2014 and general elections of 2018) would lead us far away from the topic of the paper, since they were held in qualitatively different moments (namely, in contexts of economic and post-economic crisis). The deep roots of populism 46 2011, in which the sovereign debt crisis and an increasingly weak majority forced Ber- lusconi’s government to resign. After a year and a half ruled by the technocratic government headed by former EU commissioner Mario Monti (and a set of austerity re- forms), the 2013 elections presented several surprises: in a partially unexpected way, Movimento 5 Stelle became the largest party in Parliament with about 25.5% of valid votes, mainly at the expenses of traditional parties (ITANES, 2013). Also, results showed the second most important vote-swing in the post-war electoral history of the country, with an index of aggregate volatility of 39.1% (Bellucci, 2014; Emanuele, 2015). The electoral campaign of the M5S in the 2013 election was based on accessible and captivating claims, such as, for instance, the institution of a basic income for the unem- ployed or the fight against corruption in public administration (ITANES, 2013). Also, the image of the Movimento was promoted by means of the selection process of MPs: in or- der to signal the distance between the Movimento and the old political elite, a troop of young citizens who had never experienced militancy in classic parties were selected by means of web-based polls (Biorcio, 2015; Caruso, 2015). The strategic choices of the Mo- vimento, taken together, caused analysts to borrow Taggart’s (1995) classification and to define it as a populist (or neopopulist) party (Corbetta and Gualmini, 2013; Biorcio and Natale, 2013; Diamanti, 2014). Two primary characteristics, typical of a populist party or movement, can be detected for M5S: first, the explicit reference to the struggle between ‘the people’ and a (corrupt and unfit) political elite and second, the evocation of direct contact between the charismatic leader and his people (Corbetta and Gualmini 2013, 202–205). 2.2. Political dissatisfaction and abstention in Italy: between apathy and protest Dissatisfaction with the political system, acknowledged as one of the main determinants of the Movimento’s success, is not a phenomenon born during the debt crisis of 2011. Despite Italy being, in the post-war period, one of the countries with the highest turnout (more than 9 Italians out of 10 went to the polling station between 1948 and 1976, see Mannheimer and Sani, 2001; Corbetta and Tuorto, 2004), starting from the 1979 national election, and for all the 1980s, the tradition of high levels of turnout started to fade. During the 1990s and the 2000s, the diminishing trend became even stronger: in 1996 turnout was about 83% and in 2008 the percentage of voters was even lower (80.5%). The 2013 elections presented the low- est turnout level ever in a national election in Italy until then: only 75.2% of the electoral body went to the polls. This 30-year-long trend, an 18 percentage point drop from 1979 to 2013, is partly consistent with what happened in other European countries (Franklin, 2004). Schol- ars argue that one of the determinants of the negative trend could be due to changes that hit European societies in general, such as the crisis of the mass parties, the disappearance of the classic cleavages that had ruled European politics since the end of World War II, and the increasing individualization of the political masses (Corbetta and Tuorto, 2004; Steinbrecher, Huber and Rattinger, 2007, Powell, 1986; Pattie and Johnston, 1998). The lit- erature identifies two main elements as determinants of abstention, apathy and protest (see Tuorto, 2006). As regards apathy attitudes, voters do not go to the polls because parties fail to mobilise them. Their vote is thus a mere non-action (Corbetta and Tuorto, 2004) rather than an ‘active’ refusal of the political supply. On the contrary, with protest attitudes, voters do not vote to signal dissatisfaction towards alternatives that official politics offers them MORENO MANCOSU 47 and, to some extent, refuse to be part of the electoral game (Mannheimer and Sani, 2001; Corbetta and Tuorto, 2004; Tuorto, 2006). It is common opinion (Corbetta and Tuorto, 2004; Tuorto, 2006) that, in general, the sentiment of apathy might be the primary driver of the vast majority of Italian abstainers. Socio-demographic characteristics can determine sentiments of apathy or protest: as pointed out by Mannheimer and Sani (2001), detachment from politics and abstention propensity can be increased by gender or age. The antipolitical protest, on the other hand, has been demonstrated to be positively correlated with age: young voters, indeed, tend to protest more than their parents and grandparents (Mannheimer and Sani, 2001). Another way to signal protest attitudes through voting behaviour is to void the ballot (Knack and Kropf, 2003). Previous literature has empirically shown that ballot spoiling is consistent with a form of protest voting. More precisely, it is a clear and relatively cheap way of signalling discontent with all the political alternatives (Damore, Waters, and Bow- ler, 2012). However, a void ballot does not only signal political discontent, rather, it can be an involuntary consequence of the voter’s incapacity to follow voting instructions correctly (Stiefbold, 1965; Knack and Kropf, 2003; Mannheimer and Sani, 2001). The history of void ballots in Italy follows an entirely different path compared to that of abstention (Tuorto, 2008). The number of blank and void ballots increased during the First Republic and reached 2.9 million votes in 1996 – 6% of the electoral body (Tuorto, 2008). In the 2006 and 2008 elections, the number of void and blank votes rapidly decreased, passing from more than 2.5 million between 1994 and 2001 to just 1.2 million in 2006, probably because of the simplification of the voting procedure introduced with the new electoral law and the removal of preference voting (Tuorto, 2008). However, it is beyond doubt that a compo- nent of protest, independent of involuntary annulment, remains in the void votes’ stock (Knack and Kropf, 2003; Damore, Waters, and Bowler, 2012). 2.3. Hypotheses The aim of this work, in addition to providing descriptive insight into the Movimento 5 Stelle geographical scattering in 2013, is to explain the geographical trends of the Mo- vimento’s results in Italy using indicators of apathy and protest. In particular, we want to test whether pre-crisis protest and apathy indicators can explain the 2013 achieve- ment of the Movimento. This would be consistent with the argument according to which exogenous ‘pockets’ of discontent, already present in the country before the crisis, were able to contribute to the Movimento’s success. The first hypothesis connects apathy, measured with the levels of turnout in the na- tional election of 2008, the first preceding the debt crisis of 2011, with Movimento 5 Stelle’s performance. The hypothesis, thus, reads as follows: Hp1. The lower the turnout in 2008, the better the performance of Movimento 5 Stelle in 2013.3 3 This expectation has also been tested with different data by Riera and Russo (2016), who demonstrate that the larger the drop in turnout in previous elections, the higher is the prevalence at the local level of the Movimento in the 2014 European elections. The deep roots of populism 48 As regards protest, we argue, drawing upon previous literature, that the void ballot level can be partly connected with protest attitudes. The second hypothesis can be writ- ten as follows: Hp2. The higher the percentage of void ballots in 2008, the better the perfor- mance of Movimento 5 Stelle in 2013. 3. Data, measures and models 3.1. Data and measures The data employed to test our expectations are drawn from official data provided by the Italian Ministry of the Interior and the National Institute of Statistics (ISTAT). Analyses will be performed at the municipality level, the lower geo-referenced level available. In sum, we will deal with 8,020 cases (Valle d’Aosta has been expunged from the analysis given that, as an autonomous region, it has a different party system).4 The geographical scattering of the Movimento 5 Stelle will be measured by the per- centage of the Movimento at the municipality level in the 2013 elections. The two measures presented above – that is, abstention and void ballots before the crisis – will be treated as possible determinants of the success of the Movimento in the Italian geographical space. As pointed out above, abstention and void ballots are meas- ured in the elections held in 2008. In this way, in addition to the exogeneity of our independent score, we have another advantage, namely, that the 2008 elections were held before the crisis started. By employing 2008 aggregate measures, we can provide insights into the relationship between the Movimento’s support and a set of explanatory variables that measure the level of apathy/protest in non-exceptional conditions, that is, a physiological level of apathy/protest. As an additional control variable, we will insert the number of eligible voters (in logarithm), the sex ratio and the percentage of over-85 voters (those two latter variables extracted from the 2011 Italian census). Measures such as those proposed above are not immune to drawbacks. The first, and most important, is related to the validity of our measures. As pointed out before, stocks of abstainers in each election, as well as void votes, can be related to different individual behaviours and motivation: people can abstain by signalling a protest against the politi- cal system or, instead, because they have not been mobilised. Moreover, people can abstain because they are unable to go to the polls – for instance, because of disability or illness (Mannheimer and Sani, 2001). At the same time, people who cast a void ballot might signal protest or, rather, may not be able to cast their vote correctly. However, once the age and gender composition and the dimension of the municipalities have been con- trolled, the geographical scattering of the involuntary quota of invalid ballots should be randomly dispersed. As pointed out before, gender and age can affect the involuntary be- haviour prevalence in the electoral population. 4 Since we will employ measures of two different elections (2008 and 2013), some small administrative adjustments to the structure of the municipalities (some municipalities merged, others split) have been made to harmonize the data. These adjustments, however, involve few municipalities and have been han- dled by finding a least common geographical arrangement that could preserve the structure of the municipalities in the 2013 elections. MORENO MANCOSU 49 In addition, while bias related to involuntary behaviour can be kept under control, the issue related to the validity of abstention stocks is much more difficult to solve. The abstention measure, although proven by scholars to be mainly related to individual apa- thy behaviours, represents, indeed, a measure of two different types of behaviour, that is, the aforementioned apathy and protest. Here, however, we have to remember that we are working with aggregate data. Thus, it is possible to state that, even if the stock of ab- stention represents different individual behaviour, at the municipality level, it just represents an environment more prone to protest or apathy and, we hypothesize, more prone to support Movimento 5 Stelle (in this respect, see Agnew, 2002; Goodin and Til- lie, 2008; Pattie and Johnston, 1998). Before investigating the relationship between the Movimento and the measure of protest and apathy in Italy, it is useful to assess whether these measures are scattered over the territory.5 The most straightforward way of measuring the non-randomness of the geographical distribution of a variable is Moran’s I. It allows us to calculate the corre- lation between the level of a specific variable in an ecological unit and its nearby locations, the so-called autocorrelation (Moran, 1950, Anselin, 1988). Moran’s I is a standardized measure, in which 1 means perfect autocorrelation (that is, given the level of a variable in a point in space, all nearby points present the same level as the variable), 0 means that the levels of the variable in space are distributed randomly and -1 means perfect negative autocorrelation. Moran’s I presents a local variant –the local Moran’s I (Anselin, 1995). Local Moran’s I identifies outliers of ‘pockets of nonstationarity’ (An- selin, 1995, 93). In this way, combining local Moran’s I with descriptive results of the variable, it is possible to assess where clusters of support (or lack of support) are located. In order to assess the relationship and the magnitude of the effect that links Mo- vimento 5 Stelle support and measures of apathy and protest, we will employ multivariate regression models. The first model that will be fitted is a simple OLS model. This type of model, however, presents the main drawback, that is, the aforementioned spatial autocorrelation. OLS models assume independence of observations and the pres- ence of spatial autocorrelation will lead to biased estimates. In order to solve this problem, and take into account the geographical scattering of the dependent variable, a SAR (Spatial Auto-Regressive) model will be fitted (for more information, see Anselin, 1988; Ward and Gleditsch, 2008). 4. Results 4.1. Descriptive analysis A simple analysis of the strength of the party in different geo-political zones, such as that presented in Table 1, shows performances of the Movimento during 2013 elections in part of the country. The table shows that the Movimento is stronger in the centre and in the islands of the country. 5 Appendix 1 presents descriptive statistics of all the variables taken into account in the paper. The deep roots of populism 50 Table 1. Distribution of Movimento 5 Stelle’s strength in 2013 by geo-political zone. M5s % N municipalities North-west 22.4 2,987 North-east 24.0 1,132 Red belt 25.0 966 Centre 27.4 683 South 22.8 1,485 Islands 31.2 767 Total 24.7 8,020 Note: Percentages calculated on voters A more thorough analysis of the geographical distribution of the Movimento in 2013 is plotted in Figure 1.6 The distribution of the party is significantly autocorrelated over the ter- ritory (Moran’s I is equal to .70) and presents original characteristics: traditional electoral Italian studies (see Galli, 1968; Diamanti, 2003) used to split the Italian territory into dif- ferent ‘sub-cultures’, in which one of the two main parties of the so-called Prima Repubblica (the Communist Party and the Christian Democrats) was dominant. For instance, the ‘red zone’, a strip of the region in the central part of Italy, used to present long-standing and strong electoral support for the Communist party (and saw high levels of support for centre- left parties, such as the Partito Democratico della Sinistra, lately Partito Democratico). The north-east part of the country, on the contrary, used to be ruled by the Christian Democrats, which was subsequently substituted by the Forza Italia/PdL-Northern League alliance. As pointed out by other contributions (see Cataldi and Emanuele, 2013) the pattern of Mo- vimento 5 Stelle is new; that is, no other party in the Italian Republican history seems to behave geographically like the Movimento. In general, several characteristics of this geographical pattern can be stressed: in the centre-north of the country, the Movimento’s pattern seems to break the existing sub-cul- tural zones. In the north-western part of the country (which encompasses Piedmont, Lombardy and Liguria), the Movimento is strong in the four provinces of Liguria, in the province of Turin (but the average level is high in the whole of Piedmont) and is poorly sup- ported in west Lombardy. This is surprising, since west Lombardy and Piedmont, during the first and the second Republic, tended to behave similarly. The north-east of the country, which encompasses east Lombardy, Veneto and Friuli- Venezia Giulia, used to vote solidly for centre-right parties, shifted and supported the Mo- vimento in 2013. Also, the centre of the country witnessed an upheaval. The so-called red zone – which encompasses Tuscany, Emilia-Romagna, Marche and Umbria and has, for al- most a century, seen a robust left-wing majority – was broken by the new Movimento’s rise of support. In Umbria and Marche, the Movimento has, almost everywhere, over 28% of valid votes, while in Emilia-Romagna and Tuscany the average level of support is lower. Moreover, the northern part of Lazio shows high levels of support for the Movimento. The other two clusters of strength are situated in the south of Sardinia and almost all of Sicily. 6 The classes in all the maps were obtained by dividing the distributions into 5 quantiles. MORENO MANCOSU 51 Figure 1. Movimento 5 Stelle percentage of valid votes, geographical distribution (Moran’s I: 0.70). The deep roots of populism 52 Figure 2. Movimento 5 Stelle 2013 results - Local Moran’s I and statistical significance. MORENO MANCOSU 53 Figure 3. 2008 Turnout geographical distribution. The deep roots of populism 54 Figure 4. 2008 Void ballots geographical distribution. MORENO MANCOSU 55 Figure 2 gives us results for the local Moran’s I and shows a very similar situation (the parts of the country in which local Moran’s I is not significant to 10% are coloured grey). A high level of autocorrelated, statistically significant support is in south and west Sicily (in particular the province of Trapani), in south and north-west of Sardinia, the coast of Marche (especially Ancona), in northern Lazio. By comparing Figure 2 with the previous one, it is possible to detect also clusters of low support, such as that in the south- ern part of Campania, north Lombardy and Südtirol. Summarising, the geography of the Movimento, in its first electoral competition, is entirely different from that of other parties, present or past. As regards specific local out- breaks, however, it is possible to hypothesise determinants given by the particular situation of those places. For instance, in the province of Turin, and in particular in Susa Valley, the public work of high-speed rail (the TAV) to connect Turin with Lyon, has pro- duced, for about 20 years, a movement against the construction of the line. Movimento 5 Stelle and Beppe Grillo himself have repeatedly underlined solidarity towards the No- Tav movement (Tronconi, 2013, Biorcio and Natale, 2013). Thus, it is possible to expect that citizens of these zones, massively against the public works, would support the only party that was explicitly hostile to it. Similarly, the cluster of support in southern Sar- dinia can be related to the struggle of Sulcis miners who, again, received Beppe Grillo’s solidarity (Tronconi, 2013). These local peculiarities should not challenge our main argument: if localised pro- tests can account for localised outbreaks, our aim is to find a general, spatially coherent relationship between high (or low) levels of pre-crisis protest/apathy and high (or low) levels of Movimento 5 Stelle support. Before examining the regression models, it seems useful to assess the descriptive situation of void ballots and turnout (that is the hundred- complement of abstention) in the 2008 Elections. Figures 3 and 4 present the spatial dis- tribution of the two measures. For instance, we can say that both measures present a clear north-south difference: in most of the south, the prevalence of void ballots is higher, and turnout is lower. Piedmont, Liguria, Friuli-Venezia Giulia, and the province of Belluno, moreover, present lower levels of turnout and higher levels of invalid ballots. Some differences, also, can be found by comparing the two apathy/protest measures. For instance, there is a relative absence of void ballots in the north-west, where there is a cluster of low turnout. Sardinia, also, presents medium-high levels of abstention (under 76% of turnout) and low levels of void ballots. In general, we can state that void ballots scattering is less clear than that of ‘turnout. 4.2. Testing the relationship between void ballots, turnout and M5s geographical distribution Two models will be treated in order to test the relationship between Movimento support and measures of apathy/protest. The second model presents the same parameters but is fitted as a SAR model. Coefficients for both models can be seen in Table 2. It can be seen that in both models, the parameters are significant and follow our expectations. The SAR model presents smaller coefficients, once corrected for autocor- relation. The expected change in the Movimento’s support with respect to a one-unit increase in turnout is -.06 percentage points for the SAR model. Although significant, it is not big. If we think that the central 90% of the distribution of turnout is around 18 The deep roots of populism 56 points (from 69% to 89%), the difference in the predicted values of the dependent varia- ble of a 5th-95th percentile-change of the turnout is something more than 1.2 percentage points. More interesting is the effect of invalid ballots. Keeping other variables constant (included age and gender controls) in the SAR model, the predicted rise of Movimento 5 Stelle’s support with respect to a rise of 1 point in the level of void ballots is of .61 (the parameter is significant). Table 2. OLS and spatial lag regression model OLS model SAR model Predictors Coef. S.E. Coef. S.E. 2008 turnout (in %) -0.10*** (0.01) -0.06*** (0.01) 2008 void ballots (in %) 1.51*** (0.08) 0.61*** (0.05) Eligible voters (n. logarithm) 1.33*** (0.06) 0.42*** (0.04) Sex ratio 0.08*** (0.01) 0.04*** (0.01) Over-85 persons (in %) 0.10* (0.05) -0.21*** (0.03) Rho 0.77 *** (0.01) Constant 9.82*** (1.90) 1.24 (1.17) Observations 8,020 8,020 R-squared (Or Nagelkerke) 0.10 0.60 Notes: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 The predicted difference for the 5th percentile to the 95th one is of 1.7 points (a bit less than 10% of the dependent variable’s central 90% range). Thus, even if we are not dealing with big effects, we can say that part of the geographical variance of Movimento 5 Stelle is explained by measures of political apathy or protest.7 5. Discussion In 2013, Movimento 5 Stelle, a new populist party, ran for the first time in a national elec- tion, gaining 25% of valid votes - one of the most striking successes of all times in mature democracies. The aim of this work was twofold. Firstly, except for few contributions (Cataldi and Emanuele, 2013), the literature lacked an accurate geographical analysis of this unique phenomenon. In this article, by employing local and global measures of au- tocorrelation, and with the help of municipality-level maps, we now have a clearer idea of the scattering of the territory of the Movimento in that crucial election. Second, we stressed that the relationship between apathy/protest and the strength of Movimento 5 Stelle was only tested with reference to the exceptional situation of 2013, 7 There are two additional alternative hypotheses that might undermine the results presented here. First, apathy, in addition to other individual mechanisms, might be due to the medium/long-term local tradi- tions of political engagement/disengagement that are typical of a certain area. Second, the economic conditions of the local context might change the structure of the coefficients that we see in Table 2. In Appendix 2 there are two additional models: the first model measures the gap in turnout between 2006 and 2008 (in a way that negative values of the variable represent a decrease in turnout). The second model adds an economic control variable (the average income for each municipality in 2012 - Source: Italian Revenue Agency). As can be seen, the two models do not show significant differences with respect to those shown in the main text, further confirming the overall pattern of the results. MORENO MANCOSU 57 implicitly addressing the success of the party led by Beppe Grillo as a fortuitous combi- nation of events that led to its achievement. Little evidence was committed to finding more profound relationships between what we can call a ‘traditional’ level of apathy/pro- test and the result of the Movimento in 2013. Our basic expectation was that in places where measures of protest and apathy were stronger in non-exceptional moments, the support for the Movimento would have found fertile soil. As concerns the first aim, and consistent with previous results (Cataldi and Eman- uele, 2013), results showed entirely new electoral geopolitics. Geographical cleavages that lasted during the First and Second Republic are not traceable in any way in the Mo- vimento’s geographical scattering. Instead, old sub-cultures seem to be broken by the Movimento’s support: the example of low Movimento support in Tuscany and high sup- port in neighbouring Umbria – which used to be part of the same sub-culture - is enlightening. Geography, also, helps us to recognise clusters of support that are mainly due to particular local situations (such as the aforementioned Susa valley and the south of Sardinia). At the same time, by means of regression models, we have seen that the engine of high-level support for the Movimento is not only related to a number of local issues but also has a systematic nature: in places where, even before the economic crisis, the level of protest and apathy (measured with abstention and void ballots percentages) was high, the level of support for the Movimento has been high too. Moreover, it seems that a clearer protest measure – invalid ballots percentage at the municipal level – has higher effects with respect to measures that do not tap precisely the concept of protest but are in-between to measure protest or apathy (that is, abstention). This could be indirect evi- dence that protest attitudes were activated in the 2013 national elections and that the success of the Movimento has deep roots, partly founded in traditional levels of apathy and protest. M5S success, thus, seems not to be produced entirely by a fortuitous series of circumstances (the crisis of representation and the economic crisis) that crystallised in more structured partisan support during the 2013-2018 electoral cycle. Instead, it seems that the routes of success were (at least in part) based on pockets of discontent that existed even before the intricate pattern of crises broke out. These pockets, geograph- ically scattered, contributed to form the first kernel of the Movimento’s support. This paper presents at least three shortcomings: first of all, the variables employed to measure apathy and protest – though being the only ones available that, according to the literature, can tap the two concepts – present sizeable potential distorsions. Although we have tried to refine the analysis by adding possible confounders to the relationship (the percentage of female and over-85 voters, as well as other measures of apathy and economic wealth measures, see Appendix 2), we must take into account that our coeffi- cients might be biased because of the fact that the levels of void ballots and abstention do not overlap completely with the concepts of protest and apathy. Second, as concerns the relationship between protest attitudes and the Movimento’s support, it must be stressed that the analysis investigated just the aggregate levels of this support. This does give us clues as to individual-level mechanisms. An analysis that encompasses both the geo- graphical and the individual levels would be able to present more explicit evidence on the past (and future) of the Movimento 5 Stelle. Third, the paper only takes into account the 2013 elections: although the choice is justified by the relevance of this election as a The deep roots of populism 58 turning point in Italian political history (in which profound economic and representa- tive crises happened at the same time), future research should be devoted to assessing how the relationship between apathy/protest feelings and anti-system parties - and, in particular, the Movimento – has changed. Has the increasing institutionalization of the Movimento during the 2013-2018 cycle (see Ceccarini and Bordignon, 2018) led to a low- ering of the protest component of its quota of support, or, rather, is the party still considered a viable option for expressing protest and discontent? Future research will be able to answer this crucial question for the future of the party. Appendix 1 Table A1. The variables involved: descriptive statistics. Variables Minimum Maximum Mean Std, dev. 5 perc 95 perc M5S 2013 results 0.1 56.5 22.7 6.9 11.9 33.9 PdL 2013 results 0.3 57.1 20.8 6.8 10.9 32.8 PD 2013 results 0.9 56.6 22.9 7.7 12.4 37.5 2008 void ballots % 0.0 11.1 2.1 0.9 1.0 3.7 2008 Turnout % 17.8 100.0 81.2 6.3 69.1 89.1 Eligible voters (logarithm) 3.4 14.6 7.6 1.3 5.5 9.9 Sex ratio 67.8 190.9 97.1 6.3 88.8 106.8 Over-85 (in %) 0.5 18.5 3.4 1.7 1.5 6.5 Appendix 2 Table A2. Alternative SAR models. Original SAR model (see Table 2) Alt. model (with turnout gap) Alt. model (with income) Predictors Coef S.E. Coef S.E. Coef S.E. 2008 turnout (in %) -0.06*** (0.01) -0.09*** (0.01) 2008 void ballots (in %) 0.61*** (0.05) 0.71*** (0.05) 0.69*** (0.05) 2006-08 turnout gap (in %) -0.12*** (0.01) Eligible voters (n. logarithm) 0.42*** (0.04) 0.42*** (0.04) 0.35*** (0.04) Sex ratio 0.04*** (0.01) 0.04*** (0.01) 0.04*** (0.01) Over-85 persons (in %) -0.21*** (0.03) -0.14*** (0.03) -0.19*** (0.03) Average income in thousands € (2012) 0.11*** (0.02) Rho 0.77 *** (0.01) 0.77 *** (0.01) 0.77 *** (0.01) Constant 1.24 (1.17) -4.19*** (1.06) 2.05** (0.96) Observations 8,020 8,020 8,020 Nagelkerke R2 0.60 0.60 0.60 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 MORENO MANCOSU 59 References Agnew, J. A. 2002. Place and Politics in Contemporary Italy. Chicago: University of Chicago Press. Anselin, L. 1988. Spatial econometrics: methods and models. Dordrecht: Kluwer. Anselin, L. 1995. ‘Local indicators of spatial association – LISA.’Geographical Analysis 27(1): 93-115. Bellucci, P. 2014. ‘Partisanship and the swing/vote in the 2010s: The Italian case’. Paper pre- sented to the ECPR General Conference, University of Glasgow, September 3-6. Bellucci, P., & Maraffi, M. (2014). Government Performance and Political Attitudes in the Ital- ian Political Cycle 2011-2013. Polis, 28(1), 37-60. Biorcio, R. 2014. ‘The reasons for the success and transformations of the 5 Star Movement’. Contemporary Italian Politics 6(1): 37-53. Biorcio, R. 2015. Gli attivisti del Movimento 5 Stelle. Dal web al territorio. Roma: Franco Angeli. Biorcio, R, Natale, P. 2013, Politica a 5 stelle. Idee, storia e strategie del movimento di Grillo, Milano: Feltrinelli. Bordignon, F., & Ceccarini, L. (2014). Protest and project, leader and party: normalisation of the Five Star Movement. Contemporary Italian Politics, 6(1), 54-72. Caruso, L. 2015. ‘The 5 Star Movement and the end of politics.’Rassegna Italiana di Sociologia 56(2): 315-340. Ceccarini, L., & Bordignon, F. (2018). Towards the 5 star party. Contemporary Italian Politics, 10(4), 346-362. Corbetta, P., Gualmini, E. (eds.). Il partito di Grillo. Bologna: Il Mulino. Corbetta, P. Tuorto, D. 2004. ‘L’astensionismo elettorale in Italia: trasformazioni culturali o smobilitazione dei partiti?’Polis 18(2): 287-311 Cataldi, M. Emanuele, V. 2013. ‘Lo tsunami cambia la geografia e strappa 50 province a Pd e Pdl’, In De Sio, L. Cataldi, M. De Lucia, F. Le Elezioni Politiche 2013. Roma: CISE. D’Alimonte, R. (2013). The Italian elections of February 2013: the end of the Second Repub- lic?. Contemporary Italian Politics, 5(2), 113-129. Damore, D. F., Waters, M. M., & Bowler, S. (2012). Unhappy, uninformed, or uninterested? Understanding “none of the above” voting. Political Research Quarterly, 65(4), 895-907. Diamanti, I. 2003. Bianco, rosso, verde... e azzurro: mappe e colori dell'Italia politica. Bologna: Il Mulino. Diamanti, I. 2014. ‘The 5 Star Movement: a Political Laboratory.’ Contemporary Italian Poli- tics 6(1): 4-15. Emanuele, V. (2015). Dataset of Electoral Volatility and its internal components in Western Eu- rope (1945-2015). Rome: Italian Center for Electoral Studies, http://dx. doi. org/10.7802/1112. Franklin, M. N. 2004. Voter turnout and the dynamics of electoral competition in established democracies since 1945. Cambridge: Cambridge University Press. Galli, G. (Ed.). 1968. Il comportamento elettorale in Italia. Bologna: Il Mulino. Goodin, R. E., Tilly, C. (Eds.). 2008. The Oxford handbook of contextual political analysis. Ox- ford. Oxford Handbooks Online. ITANES (Ed.). 2013. Voto amaro. Disincanto e crisi economica nelle elezioni del 2013. Bologna: Il Mulino. The deep roots of populism 60 Knack, S., Kropf, M. 2003. ‘Voided ballots in the 1996 presidential election: a county-level analysis.’ Journal of Politics, 65(3): 881-897. Mannheimer, R., Sani, G. 2001. La conquista degli astenuti. Bologna: Il mulino. Martini, S., Quaranta, M. (2015). Finding out the hard way: Uncovering the structural founda- tions of political dissatisfaction in Italy, 1973–2013. West European Politics, 38(1), 28-52. Mosca, L. 2014.‘Il Movimento 5 Stelle e i conflitti locali.’ Il Mulino 63(2): 223-230. Mosca, L., & Quaranta, M. (2017). Voting for Movement Parties in Southern Europe: The Role of Protest and Digital Information. South European Society and Politics, 22(4), 427-446. Moran, P. A. P. 1950. ‘Notes on Continuous Stochastic Phenomena.’ Biometrika, 37 (1): 17–23 Pattie, C., Johnston, R. 1998. ‘Voter turnout at the British General Election of 1992: Rational choice, social standing or political efficacy?.’European Journal of Political Research, 33(2): 263-283. Powell Jr, G. B. 1986. ‘American voter turnout in comparative perspective.’ The American Po- litical Science Review. 80(1): 17-43. Riera, P., & Russo, L. (2016). Breaking the cartel: the geography of the electoral support of new parties in Italy and Spain. Italian Political Science Review/Rivista Italiana di Scienza Po- litica, 46(2), 219-241. Stiefbold, R. P. (1965). The significance of void ballots in West German elections. American Political Science Review, 59(2), 391-407. Steinbrecher, M., Huber, S., Rattinger, H. 2007. Turnout in Germany: citizen participation in state, federal, and European elections since 1979. Baden-Baden: Nomos Verlagsgesell- schaft Mbh & Company. Taggart, P., 1995. ‘New populist parties in Western Europe. ‘West European Politics 18(1): 34- 51. Tronconi, F. 2013. ‘Da dove arrivano i voti del Movimento 5 Stelle?’Il Mulino, 62(2): 356-363. Tuorto, D. 2006. Apatia o protesta? L'astensionismo elettorale in Italia. Bologna: Il Mulino. Tuorto, D. 2008. ‘Il primo motore del cambiamento: l’astensionismo.’ In ITANES (eds.) Il ri- torno di Berlusconi. Vincitori e vinti nelle elezioni del 2008. Bologna: Il Mulino. Vegetti, F., Poletti M., Segatti, P., 2013. ‘When responsibility is blurred. Italian national elec- tions in times of economic crisis, technocratic government, and ever-growing populism’. Rivista Italiana di Scienza Politica. 43(3): 329-252. Vezzoni, C. (2014). Italian National Election Survey 2013: a further step in a consolidating tra- dition. Rivista italiana di scienza politica, 44(1), 81-108. Ward, M. D., Gleditsch, K. S. 2008. Spatial regression models. Los Angeles: Sage.