The Regionalization of Regional Elections in Italy 1

The regionalization of regional elections in Italy 1

Emanuele Massetti, Gediz University and Giulia Sandri, University of Oxford

Abstract: The impact of the institutional regionalization process on regional party systems in Italy is rather controversial in the literature (Chiaramonte and Di Virgilio, 2000; Loughlin and Bolgherini, 2006; Baldi, 2006; Tronconi, 2009). Since the first regional elections held in 1970, regional party systems have been traditionally built, with only few exceptions, upon the format and the competition structures present at national level (De Mucci, 1987; Grilli di Cortona, 2002; Caciagli, 1988 and 2006) Recent studies have come to apparently contradicting conclusions: on the one hand, they point to the emergence of a pattern of progressive ‘denationalization’ or ‘territorialization’ (as opposed to ‘nationalization’, Caramani 2002) of regional party politics, regional political competition and voting behaviour in regional elections (Baldini and Vassallo, 2000; Tronconi, 2005; Wilson, 2006; Tronconi and Roux, 2009); and, on the other hand, they highlight the presence of more evident signs second-order electoral behavior (Tronconi and Roux, 2009). This paper combines multi-level electoral and party system analysis, thus providing a comprehensive overview of the level and causes of ‘regionalization’ of politics in Italy since 1970. The multi-level electoral analysis makes use of three indexes for measuring dissimilarity (between national and regional voting) proposed by Schakel (2011) and adopts the analytical framework developed in Schakel and Dandoy (Schakel and Dandoy, 2013 forthcoming). Based on this model and using an original dataset, we present and discuss the sociological, institutional and political determinants of multi-level dissimilarity in voting. The factors considered are: the strength of the centre-periphery cleavage, the level of regional authority, the electoral cycles, the voting systems, the level of responsibility of parties (government/opposition), and the structure of competition in multi-level party systems. The developed quantitative models allow comparing individual regional cases across time and space and verify the competing hypotheses of regionalization or nationalization of regional voting in Italy.

Paper presented at the XXVI Convegno SISP 2012, Roma, 13-15 September 2012 Panel 9.3 Elezioni, territorio e de-nazionalizzazione del voto Chairs: Alessandro Chiaramonte; Vincenzo Emanuele

1 A modified version of this paper is going to be published as a chapter in the edited book ‘Regional and National Elections in Western Europe. Territoriality of the Vote in Thirteen Countries’, by R. Dandoy and A. Schakel (eds), Palgrave, Basingstoke, (forthcoming 2013) (http://www.arjanschakel.nl/).

1. Introduction

Since the mid-1990s, a number of relevant political and institutional reforms have been carried out in Italy in the sense of a stronger decentralization of power towards subnational units. While regional autonomy has been gradually growing in Italy in the last two decades, the interpretations of the evolution in regional politics rather differ in the literature. According to several authors, before the 1992-1994 political turmoil, the consequent collapse of the main parties and the regime shifts that ensued, Italian regional politics was argued to be a minor replication of national themes and trends (De Mucci, 1987; Tronconi and Roux, 2009). The patterns of voting behavior and the format of party competition in regional elections were considered to be completely dependent on national dynamics, the only differences being a lower level of turnout and slightly higher vote swings between elections. Given the political and institutional changes carried out in the 1990s, we could logically expect that that regional politics would gain greater autonomy, show an increased diversification among regions and thus make the regional vote less dependent on national dynamics. Nevertheless, the hypothesis that regional elections started to follow a generalized pattern of ‘second order’ electoral behavior precisely during the Second Republic has also been developed (Chiaramonte and Di Virgilio, 2000; Loughlin and Bolgherini, 2003; Tronconi and Roux, 2009). Two main questions then need to be answered: first, at empirical level, to what extent can we find evidence that regional politics has become enfranchised from national dynamics? Secondly, does second-order election theory provide a satisfactory analytical framework to investigate changes in the level of ‘second orderness’ of regional elections?

2. Indicators and Determinants of ‘second orderness’ in Italy

The basic way to study the level of ‘nationalization’ (or ‘regionalization’) of regional elections is to calculate dissimilarity between electoral results at national and regional level. Here, we follow Schakel (2011) where he distinguished between three indicators capturing three different dissimilarities. We calculate the incongruence between the results of a national election in the whole country (NN) and the results of a regional election in each region (RR). This index (NN- NR) represents, so to speak, the difference between the national and regional party systems. This is determined by two sources of incongruence: first, the different preferences of the regional electorate vis-à-vis the national electorate even in the same national election (NN-NR); and

secondly, vote swings within the same regional electorate between a national and regional elections (NR-RR). 2

The analysis of these three indexes allows us to evaluate not only the level of dissimilarity but also the source. Indeed, the basic thesis of the second order election model is that all non- national elections, which obviously include regional ones, are subordinate to first order national elections (Reif and Schmitt, 1980). The main empirical argument is that turnout tends to be lower in second order elections, and voters who do turnout tend to support opposition or small/fringe parties at the expenses of government parties. The symptoms of ‘second orderness’ must, therefore, be searched through a deeper analysis which includes turnout difference, government/opposition parties’ performances, government incongruence between national and regional level, and vote shares for non-state-wide (i.e. regional) parties.

Four sets of factors have been identified in the literature as shaping the relationship between regional and national elections: the strength of the centre-periphery cleavage, the level of regional authority, the electoral cycles and the voting systems. To this list, we add a country specific factor, which is also linked to the national electoral system but whose scope is wider and its nature more political. We refer to the passage from the so called First Republic to the so called Second Republic in 1993, which consisted primarily in the collapse and disappearance of the historical parties that dominated Italian politics since 1946, a substantive change of the voting system (which will be changed again in 2005), and a slow re-building of a new party system characterized by an increased saliency of the territorial dimension of party competition (Massetti and Toubeau, forthcoming 2013); the structuring of political offer around both individual parties and coalitions of parties; and by the emergence of a bipolar dynamics of competition. We argue, also on the basis of recent studies (Tronconi and Roux, 2009), that the latter characteristic might have and important influence in the way in which a ‘second order’ perception of regional elections is translated into actual voting behaviour. More precisely, since

2 For a detailed explanation on the index calculation, see Schakel, 2011. The data used for calculating the three indexes have been collected by Arjan Schakel (Maastricht University) and by the authors. In order to calculate the three dissimilarity indexes we adapted the Lee index, usually used for measuring the dissimilarity of the vote across different government levels (Hearl, Budge and Pearson, 1996; Pallares and Keating, 2003; Jeffery and Hough, 2003, 2009; Skrinis and Teperoglou, 2008; Tronconi and Roux, 2009; Floridia, 2010). The dissimilarity index is calculated by taking the sum of the absolute differences between regional and national vote shares for each party and subsequently dividing the sum by two, in order to avoid double counting (See Schakel and Dandoy, 2013: 13). The n =1 − formula is given by: Dissimilarity∑ | XiN X iR | where XiN is the percentage of the vote won by 2 i=1 party i in a given national election N, and X iR is the percentage of the vote won by party i in the closest (in time) regional election R to the national election in question. Scores may vary from complete congruence/similarity (0%) to complete incongruence/dissimilarity (100%). Therefore, in order to calculate the three different versions of the adapted Lee index, we plugged into the index the following shares of votes: - General electoral results at state-wide level, per party and per election (NN) - The vote share in a particular region for the general election, per party and per general election (NR) - The vote share in a particular region for the regional election, per party and per regional election (RR) 3

second order election theory has rested primarily on the assumption of a bipolar (or even two- party) competition, we think that a growing manifestation of second order symptoms in the Second Republic might not necessarily be the consequence of decreased importance of regional elections in the eyes of voters but simply a reflection of a change in the dynamics of electoral competition in the Italian multi-level political system.

2.1 Sociological factors: the centre-periphery cleavage

Like in other states that have been established relatively recently, the centre-periphery cleavage has always been of a certain importance in Italian politics. The definition of state borders has undergone several changes from 1861 to 1954 – inclusion of Veneto in 1866; inclusion of Lazio in 1970, inclusion of Trent, Bolzano/Bozen , Trieste and Istria in 1919; the loss of Istria and Trieste in 1945; and the final re-inclusion of Trieste in 1954 – and such borders mainly concerned the Alpine region, widely re-known for its ethno-cultural variety and its tradition of local/regional self-government (Caramani and Meny, 2005). As a result, some northern bordering regions include several ethno-linguistic minorities, the most important being the German speaking community in the region Trentino-Alto Adige , which has always constituted a majority within the Bolzano/Bozen province ( Südtirol ); the French speaking community in the small region Val d’Aosta, whose once important size has considerably decreased in the last decades due to immigration form other regions and Italianization of the regional population; and the Slovenian community in the region Friuli-Venezia Giulia , which has always been a small minority. Regionalist parties have, therefore, been very active and successful in Trentino-Alto Adige and Val d’Aosta , much less so in Friuli-Venezia Giulia .

Linguistic minorities that have been present for centuries in southern Italy, primarily Albanian and Greek communities, have always been extremely small in numbers and not politically active. In addition, southern Italy has always been economically dependent on the central state, a characteristic that, coupled with the absence or weakness of ethno-cultural specificities, tends to hinder the electoral growth of regionalist parties. As a result, in the south, the only two regions traditionally characterized by the presence of a centre-periphery cleavage have been the two big islands, Sicily and Sardinia. And even here, the importance of this cleavage and the strength of regionalist parties have never come close to the levels of Val d’Aosta and Trentino-Alto Adige. In the latter two regions the Union Valdôtaine (UV) and the Sudtiroler Volks Partei (SVP) have been very strong, or even dominant, regionalist parties (Sandri, 2012; Massetti and Sandri, 2012; Pallaver, 2007). In contrast, the main regionalist parties in the two islands, the Movimento per l’Indipendenza della Sicilia (MIS) and the Partito Sardo d’Azione (PSdAz) have been respectively a meteoric and a minor force in regional politics. Only in recent years southern regionalist (or regional) parties, such as the Udeur, the Movimento per le Autonomie (MpA), Noi

Sud (NS) and Forza Sud (FS), have started to emerge, but still with rather limited electoral scores.

The birth of these parties, however, has less to do ethno-cultural peculiarities of the regions in which they are created then they have with the development of national politics. In particular, they are the result, on the one hand, of the disintegration or crisis of state-wide centre-right parties (respectively the DC in the early 1990s and the PDL from in the late 2000s) and, on the other hand, they are the response to the emergence of the Lega Nord (LN) and its attempt to politicize the north-south economic divide to the advantage of northern regions (especially the northern RSOs). Indeed, the LN surely represents the most important ‘interpreter’ of territorial politics in the Second Republic (Cento Bull and Gilbert, 2001). The party, which has its electoral stronghold in Lombardy and Veneto, has penetrated all northern and, to a lesser extent, central regions, becoming since the 1992 general election a relevant actor in the national party system. Given the nation-wide relevance of this party, its vote does not always constitute a source of regional deviance from the national average. For instance, regions such as Liguria and Emilia- Romagna tend to have electoral results that, as far as the LN’s performances are concerned, come very close to the national average.

Having provided an overview of centre-periphery politics in Italy, we can now discuss its impact on specific ‘dependent variables’ by formulating some hypotheses. A region characterized by a strong ethno-territorial identity, especially if this translates into the presence of regionalist parties, can be expected to follow to a comparatively lesser extent a ‘second order’ logic. Therefore, all symptoms envisaged by second order theory (low turnout, losses for government parties, gains for opposition/small parties) should feature less prominently in these regions. In addition, the presence of strong regionalist parties competing at all levels of government represent the most important source of dissimilarity both between the regional and national electoral preferences (NN-NR), and between the formats of the regional and national party systems (NN-RR). The exception to this rule is represented by those few regions where the LN’s scores are very similar to its national average.

2.2 Institutional factors:

Regional authority

The Italian Republic is formed by the (central) state, the regions and the local governments (provinces, metropolitan cities and local councils). There are formally twenty regions. All regions, except small Val d’Aosta, are divided in at least two provinces, which are further divided in many local Councils. The level of regional authority in Italy has changed through time and is different across regions. The politicization of the above mentioned ethno-cultural specificities in the three bordering regions in the Alps and in the two big Mediterranean islands 5

have translated into the adoption of an a-symmetric form of regionalised structure. In the 1948 Constitution, five regions were granted a special status (RSSs) and were allowed to elect their regional assemblies since the late 1940s (with the exception of Friuli-Venezia Giulia which remained ‘frozen’ until the solution of the Trieste question). The other fifteen ordinary regions (RSOs) were recognized by the 1948 Constitution but were given less powers and were left in waiting for the necessary legislation to implement their actual establishment. Such legislation arrived only in the late 1960s and the first ‘ordinary’ regional elections were held simultaneously in 1970.

In 1972 the Constitution was changed to declare the shift of most special powers from the region Trentino-Alto Adige to its two constituent parts, the two autonomous provinces (APs) of Trent and Bolzano/Bozen (or Südtirol ). The full implementation of this constitutional amendment took several decades but, eventually, led to the separation of the two entities that, although still called ‘provinces’, are now treated as being two separate regions (for instance when the Constitution refers to the special status entities it does not mention the five RSOs but the four RSOs and the two APs). Indeed, the last proper regional election in Trentino-Alto Adige was held in 1998. Since then there have only been separate provincial elections, and the regional assembly has become a mere meeting of the two provincial assemblies. In general, regional authority has grown considerably in all regions during the Second Republic. In particular, the reforms of the late 1990s gave the regions some fiscal powers: a new tax, regional tax on producing activities (IRAP), set by the central government but fully collected by the regions was introduced; and a fixed share of the income tax, again set by the central government, was left to the regions. In addition, the constitutional reform approved in 2001 extended considerably the areas under exclusive competence of the regions or under shared-competence. More controversial is the outcome of the latest ‘fiscal federalist’ reform (2009-2011), which enhanced the fiscal powers of the regions but, surprisingly, compressed their overall financial capacity (Massetti, 2012). While these reforms have considerably reduced the gap between RSSs and RSOs, the former still enjoy a privileged treatment, especially in terms financial arrangements with the sate (Baldi, 2010; Barbero and Zanardi, 2011). The Regional Authority Index (RAI) of the RSOs scored 7 in the 1970s, 9 at the end of the First Republic, and 14 by 2006; while the RAI of the RSSs and APs scored 11 in the 1950s, 12 at the end of the First Republic, and 18 by 2006 (Hooghe et al., 2010, Appendix B, Table B4). So, we expect all symptoms of ‘second orderness’ (low-turnout, losses for governmental parties, gains for opposition and/or small parties) to be stronger in RSOs than in RSSs-APs, and to be stronger in the period 1948-1992 than in the period 1994-2010.

Electoral cycles

The electoral cycles have to be discussed in terms of both vertical and horizontal congruence/incongruence. As far as horizontal simultaneity is concerned, RSOs have always 6

voted together, with the minor and rather recent exceptions of Molise (that held anticipated elections in 2001 and has voted a year after all the other RSOs since) and Abruzzo (that held anticipated elections in 2008, thus skipping the last round of RSOs elections in 2010). RSOs elections, therefore, represent a strong national test for political parties and, for this reason, should subject to ‘second order’ logics. In addition, they have never been held in vertical simultaneity and always more than a year away from the closest general election (5 out of 9 times two years away from the closest general election; and 5 out of 9 times in the second half of the term; no evident differences between First and Second Republic). In contrast, RSSs-APs elections have always followed their own calendar, avoiding horizontal (and vertical) simultaneity. According to electoral cycles, therefore, we expect ‘second order’ tendencies to be more evident in RSOs than in RSSs-APs elections.

Voting systems

Italy has been characterized by an overall congruence between the electoral systems deployed at national and regional level. During the First Republic (1946-1993), all of them - with the exception of Val d’Aosta in the short period 1946-1963 – were proportional, with no formal barriers. The new national voting system, introduced in 1993, was a mixed one, predominantly majoritarian: 75% of seats allocated in single-member constituencies through FPTP and 25% of seats allocated in multi-member constituencies through PR, with a 4% national threshold. Regional voting systems, as far as the fifteen RSOs are concerned, were also changed in 1995 and 1999, introducing a majoritarian logic, with the provision of a majority bonus for the winning coalition; and a personalistic logic, with the direct election of the regional ‘first minister’ (called President of the region). Although the individual regions have the power to amend their own voting system, virtually all RSOs have remained loyal to a common scheme. Voters cast two votes, one for the candidates for president (each supported by a coalition of parties) and one for a party list (not necessarily of the same coalition) 3. Seats are distributed proportionally between parties but the coalition supporting the winning candidate for president (who is elected in a second round if s/he has not reached 50% plus one vote in the first round) is awarded a bonus of seats ensuring a majority in the regional assembly. The bonus is then redistributed amongst the parties of the winning coalition. Moreover, compulsory voting has

3 Since 1995, in all the RSOs, the directly elected President of the region nominates the regional ministers. The regional executive is not subject to a confidence vote by the regional parliament, but the latter can vote a no- confidence motion to remove the executive from office. The plenary size varies between 90 seats in the Sicilian parliament and 30 in the Basilicata legislature.

been enforced (but only through administrative sanctions) both in national and regional elections from 1947 until 1993 4.

The result is that the pattern of competition for regional government is the same as in Rome: two main coalitions fighting for office and a relatively high number of parties gaining representation. In 2001, the voting systems of the RSSs were changed too, and in the same direction as the RSOs. In addition, the change in 2005 of the national electoral system, bringing back a formally PR system but strongly affected by the majoritarian logic of a bonus for the winning coalition (Massetti, 2006), has made the voting systems for the two levels even more similar (Calise, 2006). Therefore we do not have specific expectations based on the voting systems (it is incongruence between the two levels that generates hypotheses).

2.3 Political factors: the passage from the First to the Second Republic

During the First Republic, voting behaviour was characterized by strong party loyalty, due to widespread ideologization of the electorate and the presence of strong political subcultures - Christian democratic, Communist, Socialist, laic (Liberal and Republican), Neo-fascist and, up to the early 1950s, Monarchic. As far as dynamics of competition are concerned, the First Republic featured a sort of ‘blocked democracy’, with extremely partial alternation in office. The Christian Democratic Party (DC) was always in government, in coalition with some (or all) centre-right, centre and centre-left parties, leaving the Communists and the Neo-fascists (and the Monarchists) permanently in opposition (Sartori, 1976). The change of the national electoral system in 1993, together with the collapse of these traditional parties, sanctioned the beginning of the Second Republic (Katz, 1996; Morlino, 1996). The new voting system incentivized parties to form broad coalitions in order to have better chances to win seats in single-member constituencies (Di Virgilio, 1994). In the first two elections (1994 and 1996) it became clear that the electoral space for centre coalitions (heirs of Christian Democracy in 1994, and the Lega Nord in 1996) was being squeezed between the left and right coalitions, which effectively competed for office.

By the 2001 election, centre coalitions had virtually disappeared, sanctioning the strict relationship between the new voting system and a two-coalition pattern of competition (Pasquino, 2001). This polarized, fragmented and bipolar party system was characterized, on the side of political offer, by a double unit of structuration - the party and the coalition (Bartolini et al. 2004; Diamanti, 2007) – and, on the side of voting behaviour, by what has been labelled as ‘light loyalty’ (Mannheimer and Natale, 2009), that is by a tendency of voters to remain loyal to

4 This provision, in art. 48 of the Constitution, was abrogated with the adoption of a new electoral law in 1993 (and by the D.P.R. 534/1993).

a coalition but also to switch between voting and non-voting as well as between different parties within the same coalition. As a result, if analysed with the parameters of the coalitions, regional elections may follow more a second order logic in the Second Republic (Tronconi and Roux, 2009), with the coalition in office at national level losing control of most regional governments (especially in RSOs). However, because of the growing number of parties and the instability of the political offer (Newell and Bull, 1997; Di Virgilio, 2006; Massetti, 2009), this pattern can be lost following the results of individual parties.

We also expect the instability of political offer in the Second Republic to produce an increase in NN-RR incongruence. Finally, the personalization of electoral competition induced by the new regional voting systems tends to manifest itself in the presentation of a list (not a coalition, but a list competing for proportional votes like all other parties) formed around the candidate for president and, therefore, peculiar to that individual region. We expect that this could result in an increase of both NN-RR and NR-RR incongruence.

Having presented many hypotheses, it is important to say something about the relationships between them, both in terms of hierarchy of importance and in terms of compatibility. First, the pairs of expectations based on ethno-cultural distinctiveness of the regions and on the level of regional authority (RAI) tend to reinforce each other. In contrast, the expectation based on the growth of RAI in the Second Republic clashes with the one based the emergence of bipolar competition in the same period. This means that they could weaken or neutralise each other.

3. Incongruence of the vote

One way to measure the interactions between the regional and national levels in terms of electoral patterns is to look at the degree of congruence of the vote. The general idea here is that the greater the difference in voting patterns between the two political arenas, the lower the second order effects. In order to understand the distinctiveness of voting across territorial units we use the three indexes of incongruence presented in chapter 1 (Schakel, 2013). Figures 1A and 1B display the averages values for the three measurements of congruence of the vote discussed in this article, and this for RSSs (Fig. 1B) and RSOs (Fig 1A) respectively. The measurements are averaged over regions per election. The results of each national election held between 1945 and 2010 are compared to the results of regional elections held closest in time. More precisely, the index of party system incongruence (NN-RR) compares the national elections results for the country as a whole with the regional election results for a particular region. NN-RR conflates cross-level (national and regional elections) and cross-electorate (national and regional voters) dissimilarities. In order to disentangle them, we also use the index of cross-level (same electorate) incongruence (NR-RR), which tackles the difference between the national vote in the region and the regional vote in the region, and the index of cross-electorate (same level)

incongruence (NN-NR), which represents the difference between the national elections’ vote in the whole country and the national elections’ vote in one region.

[Figure 1A about here]

Figure 1A reveals two main trends. First, the index of party system incongruence (NN-RR), showing the level dissimilarity between the national and the regional party systems, increases considerably during the Second Republic or more, precisely, from the breakthrough of the LN at the 1992 general elections onwards. In the period 1970-1987, it remains rather constant just over 10%, while in the following period it fluctuates between about 20% and 40%. Secondly, it emerges very clearly that the main source of an increase in NN-RR is the impressive growth of cross-level incongruence (NR-RR), that is the shifting choices of regional voters between national and regional elections. During the period 1970-1987, the NR-RR index shows a declining trend, from below 8% to below 5%. This means that regional voters used to vote overwhelmingly for the same party in both regional and national elections. In fact, in this period, most of NN-RR incongruence in RSOs is due, as captured by the higher level of cross-electorate index NN-NR (around 10%), to the presence of territorially concentrated subcultures, namely the Communist one in the regions dominated by the PCI (the “red belt”: Tuscany, Emilia-Romagna and Umbria) and the Catholic one in the regions dominated by the DC (the “white belt”: Abruzzo, Basilicata and Veneto). However, in the following period, vote swings across the two levels of elections (NR-RR) have increased dramatically, fluctuating between about 15% and 30%. This finding indicates the emergence of a more marked ‘dual voting’ or (classic) ‘second- order’ pattern of voting behavior. In contrast, the index of cross-electorate incongruence (NN- NR) shows a much more modest increase in the second period, from substantive stability around 10% to fluctuations within the range 10-18%. At first sight, this finding is rather contradictory vis-à-vis to the breakthrough of a strong regionalist party like the Lega Nord . However, as mentioned before, the multi-regional presence of the LN, its consequent relevance nationwide, and its penetration in non-core regions (central Italy) end up mitigating the dissimilarities. For instance, although the LN achieved better results in 1992 and 1996 than in 1994, the NN-NR index results higher in 1994. This is mainly due to the concentration of votes in its core-regions in 1994, as opposed to wider penetration of non core-regions in 1992 and 1996. In addition, the contribution of the LN to overall incongruence is rather complex because of its marked fluctuations in electoral scores, also across levels of elections, throughout its history. In short, while the emergence of the LN and other minor regionalist parties have had a limited impact on the dramatic growth of party system incongruence (NN-RR) in RSOs, the main cause has to be found, as it emerges from the considerable increase in NR-RR, in the instability of the political offer during the Second Republic and in a change in ‘second order’ dynamics. 10

[Figure 1B about here]

Figure 1B shows that the evolution over time of the voting patterns in the RSSs follows rather different dynamics, as can be expected at least on the basis of the Italian asymmetric federalism and of the ethno-linguistic cleavages present in at least three of these regions ( Trentino-Alto Adige, Sardinia and Val d’Aosta), but also when taking into account the horizontal and vertical differences in the electoral cycles. In particular, party system incongruence (NN-RR) is much higher compared to the average scores of the RSOs throughout the period. In addition, like for RSOs, we can see a clear increase during the Second Republic. Taking aside the foundational 1948 election, we observe: a first period (1953-1987) in which the NN-RR index results rather stable just above 20%; a second period (1992-2001) in which the same index grows and stabilizes itself around 30%; and a third period (2006-) in which it rockets at around 40%. These results are not surprising given that in two of these five RSSs, Trentino-Alto Adige and Val d’Aosta, the party systems are completely separate from the statewide party system and this separation has constantly increased over time. The vote dissimilarity has reached 70% in Val d’Aosta in the last decade and lately has topped 50% in in the Bolzano/Bozen province. During the Second Republic, in Sardinia, it constantly hovers 20%, while the party system can be considered less distinctive in Sicily (especially before the breakthrough of the MpA in the 2005 regional election) and FVG. In the latter region, the biggest regionalist party is, like in other northern regions (mainly RSOs), the Lega Nord .

In contrast to RSOs, it seems that the main source of voting distinctiveness relies in the extent to which the regional electorates of the RSSs are different from the statewide electorate, rather than in ‘second order’ effects. Indeed, the cross-electorate incongruence index (NN-NR), which captures precisely the degree of dissimilarity between a given regional electorate and the statewide one, remains much higher than the cross-level index (NR-RR) and even slightly higher than the party system index (NN-NR) almost throughout the whole period. This is in line with our expectations on the salience of the centre-periphery cleavage and the presence of strong regionalist parties that perform well also in national elections. The only evident exception is represented by the very last two elections (2006-), when the strong increase in NN-RR appears to be mainly fostered ‘dual’ or ‘second-order’ voting, as evidenced by high levels of NR-RR scores. This might be due to different factors in different RSSs, such as the emergence of MpA in Sicily, the creation of regional-only lists linked to the candidates for regional presidents in the Trento AP and in FVG, and the growth of relatively new regionalist challengers in the Bolzano/Bozen AP (and to a lesser extent in Sardinia) that score better in regional than in national elections.

[Table 1 about here]

[Figure 2A about here]

In terms of government congruence, the analysis is limited here to an exploratory overview of the evolution over time of regional governmental coalitions and of their congruence with national ones. Table 1 describes the party compositions of the government coalitions in all the 20 regions from 1970 (first regional elections in the RSOs) to 2010 and compares them to the composition of national governments over time. Figure 2A focuses on RSOs and shows the degree of dissimilarity between seat shares of government parties in national and regional governments 5. From the very raw data showed in Table1, at least concerning the RSOs, we can see that national and regional government incongruence has slowly and constantly increased in the passage from the First to the Second Republic, and has continued to do so since the 2000s. The government congruence figures showed in Figure 2A confirm this pattern, with the degree of dissimilarity between regional and national coalitions remaining rather high and stable between 1995 and 2005 (and then soaring at the 2010 regional elections). Besides, in the RSSs the governmental congruence has always and constantly been rather low, even in the 1970s and the 1980s, mainly due to the presence of locally dominant regionalist parties such as the SVP and the UV and to the stronger local electoral roots of the Christian Democrats in FVG and Sicily. Thus, the governmental coalitions remain rather distinctive in the RSSs. On the contrary, in the RSOs the pattern of competition for regional government has become (slightly) more similar to the one for national government: two main coalitions fighting for office and a relatively high number of parties gaining representation, but the dissimilarity between seat shares of government parties in national and regional governments has constantly increased during the Second Republic.

5 Government congruence is indicated by a dissimilarity index but, in contrast to congruence in the vote, there is only one operationalization, namely the national government (NN) is compared to the regional government (RR). Another difference is that seat shares instead of vote shares for the governing parties are taken. For each region, the regional coalition formed after each regional election is compared to the closest in time national government in terms of seat shares. The seats shares are subsequently plugged into the dissimilarity index. This means taking the absolute difference per party between the shares of seats gained within the national coalition and the regional coalition, dividing by two, and summing the absolute differences.

4. Second order elections effects

If we look now at those dependent variables dealing more specifically with the traditional ‘second order’ symptoms, such as turnout and losses/gains by government/opposition parties between regional and national elections, we can see that the variation between the First and Second Republic largely reflects the trends emerged in the analysis of vote incongruence. The idea of two types of ‘second order’ electoral behavior according to the decade appears thus more strongly from the exploratory overview of these indicators. As Tronconi and Roux (2009) also showed, a ‘classic’ second-order pattern of voting behavior in regional elections emerged more strongly since the mid-1990s. In this section, thus, we will briefly look at the data concerning the difference in turnout levels among regional and national elections, the shifts in party vote shares between the regional election and the previous national ones. Finally, we will briefly explore also the evolutions in the support for regionalist parties in regional and national elections in order to assess to what extent the fact that during the Second Republic the regional vote seems to be more dependent on national dynamics may be compensated by an increasing diversification among regions in terms of political supply linked to ethno-regional factors.

In terms of turnout, regional elections are held at the same time only for RSOs (except, as we have seen, for Abruzzo and Molise ), while the RSSs follow a different electoral calendar and in any case regional election never coincide with lower chamber elections. Given the general absence of vertical simultaneity and due to the horizontal simultaneity present only among most of the RSOs, we consider that the RSOs elections, therefore, represent a strong national test for political parties and, for this reason, should subject more to ‘second order’ logics than the RSSs elections. This is because, as explained in the first chapter of this book, the coincidence of regional elections may make the electoral behaviour more dependent on national dynamics in the sense that regional elections become a referendum where voters tend to sanction or reward the national government’s performance rather than express their preferences in terms of regional politics. First of all, Figure 3A displays average turnout figures for national and regional elections with their standard deviations for the RSOs, while Figure 3B concerns the RSSs.

[Figure 3A about here]

[Figure 3B about here]

From Figure 3A we can see that the expectation to see increased second-order election effects during the last two decades is supported by the data. Even considering just the RSOs, the graph shows that turnout is rather high for regional elections before the mid-1990s and then it drops just after the first regional elections held during the Second Republic (1995). Turnout in regional elections attracts slightly less than 90% of the voters during the First Republic, while it drops to around 70% in the 2000s and to 62% in the last elections held in 2010. This could not be explained by compulsory voting rules given that the same rules were in force for both levels. In addition, the difference between the degree of participation in regional and national elections does not exceed 5 percentage points before 1995, while it hovers around 10% during the Second Republic (and has reached 17% in 2010). Although electoral turnout in regional elections has been for several decades very high, showing the extent to which Italian voters do find subnational politics important (especially compared to the case of other Western European countries), this phenomenon has radically weakened in the last two decades. A similar pattern can also be observed for RSSs, as shown in Figure 3B. Nevertheless, in this case the difference between the degree of participation in regional and national elections has always been wider, ranging from 11 % in 1947 to 2% in 1983 and has slightly but steadily increased more recently.

It is important to note that this set of findings contradicts the expectation that the level of turnout should increase with the transfer of more powers and competences to the regions. In this case, institutional factors appear to succumb in front of more powerful political factors in driving voters’ behaviour. As we hypothesized in the first section, the establishment of a clear bi-polar dynamic of competition with alternation in power at national level might have set electoral behavior in regional elections more into a ‘classic’ second order pattern. As far as turnout is concerned, in the new context of ‘light loyalty’ (Mannheimer and Di Natale, 2009), more and more voters have probably chosen non-voting as an instrument to send a negative signal to their party/coalition.

In terms of difference in party vote shares between the regional election and the previous national election for opposition and government parties, the logic that we follow here is similar to the one applied to the analysis of turnout patterns. Anti-government swing is expected to be higher in RSOs than in RSSs and especially in the last two decades. In Figures 4A and 4B are reported the average gains for government and opposition parties for ordinary and special regions respectively. From Figure 4B, we can see that no clear ‘second order’ elections pattern emerges among the RSSs, not even in the last two decades. Except for the first regional elections held in 1948, in the RSSs government and opposition parties tend to obtain similar vote shares in regional and lower chamber elections throughout the whole period covered here. There is, in fact, a clear pattern of loss of vote shares for both opposition and government parties in the RSSs. Identity voting leads to stronger non-state-wide, regionalist parties, and the state-wide parties lose vote share no matter their government status. Moreover, the standard errors are

relatively high for all elections among the RSSs, which indicate that there are relevant differences in the electoral behaviors in the different regions.

[Figure 4A about here]

[Figure 4B about here]

On the contrary, Figure 4A shows that among the RSOs a clearer second-order elections pattern can be observed briefly during the 1970s and then again since 1995. In the 1980s, though, we cannot observe any clear pattern. We can see, conversely, that during the Second Republic government parties lose votes at the gain of opposition and new parties. Besides, the losses in vote shares for government parties are relatively considerable in size, ranging from 2% to 4.5%. Moreover, the standard errors are relatively low for the RSOs for all elections, indicating thus that voters in most regions behaved in a similar way. In the case of the RSOs, a clear distinction between the elections before and after 1995 can thus be observed, with a strong anti-government swing emerging during the Second Republic. We think, as also Tronconi and Roux (2009) have suggested, that this is mainly due to the particular format of the Italian electoral calendar. Since the mid-1990s, in fact, the regional elections have regularly been held about one year before the next parliamentary one and thus the RSOs elections represent rather a national test for political parties and anticipate trends that tend to be confirmed on the following year. For instance, the 1995 regional elections anticipated the electoral success of the centre-left parties in 1996, while in 2001 and 2006 the victories of, respectively, the centre-right and centre-left coalitions were anticipated by the results of the regional elections of the previous year.

In particular, the underlying idea is that within a ‘second order’ frame the governing parties will perform well when regional elections are held soon after national ones - during the ‘honeymoon’ period with the voters – while they will experience significant vote losses when regional elections are held around mid-term (Tronconi and Roux, 2009: 153). In fact, the Italian case shows that, more than around mid-term, towards end of term (or, anyway, well in the second half) is the period in which classic second-order electoral behavior and, in particular, government parties’ punishment, tends to manifest most. When looking at the number of days between the regional contest and the previous national one when can easily see that since 1995 vote shares changes are related to the electoral calendar: in 1970 the time lag between regional and national elections was 749 days and the vote gains for governing parties reached 2.5 percentage points, in 1980 the time lag was 371 days and governing parties gained 1.4% more votes than in previous national elections. On the contrary, in 2000 a time lag of 1456 corresponded to a 6.6% vote loss for the centre-left coalition in power at national level, while in 2005 a time lag of 1421 was 15

associated to a loss of 7.6% of votes for governing parties. In sum, since 1995 the RSOs elections follow a ‘second order’ pattern.

Finally, we look briefly at the evolutions in the strength of non-state-wide parties in regional and national elections. We have seen in the previous section that, both in the RSOs and the RSSs, regional party systems have become increasingly dissimilar during the last two decades. In order to explore more in-depth these political changes, in Figure 5 are reported the data concerning average non-state-wide party strength for regional and lower chamber elections in the RSOs, while ion Figure 6 are reported the same data but for the RSSs.

It is important to note here that in Italy non-state wide parties are generally both regional and regionalists . In terms of territoriality of the vote, most of the parties classified as non-state wide usually compete in only one region. The main exception is constituted by the Lega Nord , which gains votes bot at regional and national elections in five RSOs and by the Movimento per le Autonomie, which is also a multi-regional party in terms of political supply (it gains votes in 8 southern regions). On the other hand, all the non-state wide parties whose vote shares are shown in Figure 5 and 6 are also regionalist, in the sense that their ideology is based on autonomy claims that question, although to different extents and in different ways, the territorial structure of the Italian state.

[Figure 5 about here]

As far as RSOs are concerned, regionalist parties have clearly emerged only since 1990, with the take-off of the Lega Nord that will subsequently establish itself as an important political actor in both national and (northern) regional elections.

[Figure 6 about here]

In the case of RSSs, it is worth noting that Figure 6 displays a quite different pattern. Non-state- wide parties have always been very important actors in these regions in both elections, sometimes creating completely separate regional party systems. The overall strength of non- state-wide parties in regional elections never dropped below 10% and generally fluctuated between 20% and 36%, while in national elections it has never dropped below 18%. It is interesting to note that, while during the First Republic non-state-wide parties scored generally better in national than in regional elections, since the late 1990s the opposite trend has emerged.

This appears to confirm that the strong increase in the NR-RR index of incongruence in RSSs in recent years is actually due to ‘dual voting, and in particular to the emergence of regionalist parties that perform better in regional than in national elections.

5. Conclusions

The analysis of the data presented above points to some evident trends. First, there are strong differences in voting behaviour between the RSSs and RSOs. All indexes of incongruence are higher in RSSs than in RSOs, with the greatest difference resting on cross-electorate index (NN- NR). The main source of these differences appears to be the saliency of the centre-periphery cleavage and the consequent strength of regionalist parties, rather than institutional asymmetry (i.e. the strongest/wider powers and competences of the RSSs vis-à-vis the RSOs). One institutional feature that might play an important role is the horizontal simultaneity of RSOs’ elections vis-à-vis the individual electoral calendar of RSSs. However, all expectations linked to the level of regional authority where not substantiated by the findings. In particular, levels of turnout where always lower in RSSs than in RSOs, and decreased in both in spite of regional empowerment. Similarly, other classic ‘second-order’ symptoms, such as punishment of government parties in regional elections, emerged more clearly along with the empowerment of RSOs during the Second Republic.

The second most important finding is, indeed, a general increase of incongruence during the last twenty years. There are several causes that can explain this change. The collapse of the old party system and the death of traditional parties that embodied the rooted political sub-cultures determined an increase in electoral volatility and, perhaps more importantly, prolonged instability in the political offer (though more in terms of party labels, splits and mergers than in terms of political personnel). The change of voting system in 1993 (and then again in 2005) favoured the emergence of bipolar competition with alternation in power between two main coalitions. The re-structuration of the political offer not only around (more or less new) parties but also around coalitions contributed to the creation of a lighter loyalty by voters, who showed a stronger inclination towards switching preferences or, even more, to resort to abstention from voting. Finally, the emergence of multi-regional regionalist parties, first and foremost in the north (LN) and then in the south (MpA), has added to the increase in dissimilarity. In addition, institutional/political developments at regional level converged in creating the conditions for an increase in overall incongruence. In particular, the reform of the regional voting systems favoured personalization of electoral competition for the regional presidency and, with that, regionalization of the political offer, through the introduction of ‘president lists’.

These trends have involved both RSSs and RSOs but in slightly different ways. From the data it emerges very clearly that cross-electorate incongruence (NN-NR) keeps playing a more

important role in RSSs than in RSOs. In addition, the growth of cross-level incongruence (NR- RR) in both types of regions seems to be due to different logics. In RSSs it appears to be predominantly linked to ‘dual voting’, with the appearance of new regionalist parties that score better in regional than in national elections. In contrast, in RSOs a classic pattern of ‘second- order’ electoral behaviour, with lower turnout and substantive losses for government parties, (which was absent during the First Republic) appears to have emerged in the last twenty years. Moreover, regional governmental coalitions remain rather distinctive in the RSSs, while in the RSOs the pattern of competition for regional government has become (slightly) more similar to the one for national government, although the dissimilarity between national and regional governments has constantly increased during the Second Republic.

This brings us to the puzzle that inspires the title of this chapter: have RSOs’ elections in Italy moved from a ‘first order’ (high turnout, no systematic punishment for government parties, non systematic gains for opposition or small parties) to a ‘second order’ election pattern, or have they manifested their ‘second orderness’ in two different ways in the two periods? The extant literature on regional elections in Italy and the fact that regional governments have substantively increased their powers would suggest that the second thesis is more plausible. Voting behaviour in RSOs’ elections has mainly been driven by national politics in both periods but, crucially, national party politics was shaped by different dynamics in the two periods: blocked democracy and strong ideological voting vs. bipolar competition, alternation in power and lighter party loyalty. This raises a general question on the so called ‘second-order’ elections theory, one that goes beyond the scope of this individual chapter.

Bibliographical references. Baldi, B. 2010. ‘I “numeri” del federalism fiscale: un confronto fra le regioni’ Istituzioni del Federalismo 5-6: 495-514. Barbero, M. and A. Zanardi, ‘Quelle regioni ancora più speciali’ 7 June 2011, http://www.lavoce.info/articoli/-istituzioni_federalismo/pagina1002345.html. Bartolini, S., Chiaramonte, A. and R. D’Alimonte. 2004. ‘The Italian party system between parties and coalitions’ West European Politics 27, no. 1: 1–19. Calise, M. 2006. La Terza Repubblica: Partiti contro presidenti . Rome: Editori Laterza. Caramani, D. and Y. Meny (eds.) Challenges to consensual politics: Democracy, identity and the populist protest in the Alpine region . Brussles: P.I.E.-Peter Lang. Cento Bull, A. and M. Gilbert 2001. The Lega Nord and the Northern Question in Italian Politics . Basingstoke: Palgrave. Di Virgilio, A. 1994. ‘Dai partiti ai poli: la politica delle alleanze’ Rivista Italiana di Scienza Politica 24, no. 3: 493–547. Di Virgilio, Aldo. 2006. ‘Dal cambiamento dei partiti allevoluzione del sistema partitico’ in L. Morlino and M. Tarchi (eds.) Partiti e caso italiano. Bologna: Il Mulino. Diamanti, I. 2007. ‘The Italian centre-right and centre-left: Between parties and the party’ West European Politics 30, no. 4: 733–62. Holzer, A. and Schwegler, B. (1998) “The Sudtiroler Volkspartei: A HegemonicEthnoregionalist Party” in De Winter, L. and Tursan, H. (eds) Regionalist Parties inWestern Europe , London: Routledge, p. 158-73. Hooghe, L., Marks, G. and A. Schakel 2010 The Rise of Regional Authority: A Comparative Study of 42 Democracies . Oxon: Routldge. Katz, R. 1996. ‘Electoral reform and the transformation of party politics in Italy’ Party Politics 2, no. 1: 31–53. Loughlin J. and Bolgherini S., 2006.Regional Elections in Italy : National Tests or Regional Affirmation?, in D. Hough and C. Jeffery (éd.), Devolution and Electoral Politics , Manchester, Manchester University Press, , p. 140-156 Mannheimer, R. and P. Natale 2009. ‘Mobilita’ elettorale e fedelta’ di coalizione’ in Legnante, G., Mannheimer, R and P. Segatti (eds.) Il cittadino-elettore in Europa e America: studi in onore di Giacomo Sani . Bologna: Il Mulino. Massetti, E. 2012. ‘Federal Reform: The End of the Beginning or the Beginning of the End’, in Bosco, A. and D. McDonnell (Eds.) Italian Politics 2011. Oxford: Berghahn. Massetti, E. 2009 ‘The Sunrise of the Third Republic? The evolution of the Italian Party System after the 2008 General Election and the Prospects for Constitutional Reform’ Modern Italy 14, no.4: 485-498.

Massetti, E. 2006. ‘Electoral reform in Italy: From PR to mixed system and (almost) back again’ Representation 42, no. 3: 261–9. Morlino, L. 1996. ‘Crisis of parties and change of party system in Italy’ Party Politics 2, no. 1: 5–30. Massetti, E. and Sandri, G. 2012. Francophone Exceptionalism within Alpine Ethno- Regionalism? The Cases of the Union Valdôtaine and the Ligue Savoisienne, Regional and Federal Studies Vol. 22(1). Newell, J. and M. Bull. 1997. ‘Party organizations and alliances in Italy in the 1990s: A revolution of sorts’ West European Politics 20, no. 1: 81–109. Pallaver, G. 2007 “I partiti politici in Alto-Adige dal 1945 al 2005” in Ferrendi, G and G. Pallaver (eds) La Regione Trentino-Alto Adige/Sudtirol nel XX secolo I . Poltitica e Istituzioni (Grenzen/Confini, 4/1), Trient: Historical Museum. Pasquino, G. 2001. ‘Un elezione non come le altre’ Rivista Italiana di Scienza Politica 31, no. 3: 387–98. Reif K. and Schmitt H. (1980). Nine Second-Order National Elections. A Conceptual Framework for the Analysis of European Election Results, European Journal of Political Research, 8, pp. 3- 44. Roux, C. 2006 ‘The Partito Sardo d’Azione: Regionalist Mobilization in Southern Italy’, in De Winter, L., Gomez-Reino, M. and P. Lynch (eds.) Autonomist Parties in Europe: Identity Politics and the Revival of the Territorial Cleavage . Barcelona: ICPS. Sandri, G. 2012. Ethno-linguistic identity and party politics in the Aosta Valley, Ethnopolitics , forthcoming, Vol.11(3). Schakel, Arjan H. (2011) Congruence between regional and national elections, Comparative Political Studies, published online on 18 October 2011: http://cps.sagepub.com/content/early/2011/10/06/0010414011424112 . Schakel, A. and Dandoy, R. (eds.) (2013). Regional and National Elections in Western Europe. Territoriality of the Vote in Thirteen Countries, Palgrave, Basingstoke (forthcoming 2013). Sartori, G. 1976. Parties and party systems: A framework for analysis . Cambridge: Cambridge University Press. Tronconi, F. 2009. I partiti etnoregionalisti. La politica dell'identità territoriale in Europa Occidentale, Bologna, Il Mulino, 2009. Tronconi, F. and C. Roux 2009 ‘The political systems of Italian regions between state-wide logics and increasing differentiation’, Modern Italy, vol. 14, no. 2, pp. 151-166.

FIGURES AND TABLES

Figure 1A: Congruence between the national and regional vote (ordinary regions)