Party Competition in the 2013 Italian Elections: Evidence from an Expert. Survey

Party Competition in the 2013 Italian Elections: Evidence from an Expert Survey Aldo Di Virgilio, Daniela Giannetti, Andrea Pedrazzani and Luca Pinto University of Bologna Abstract In this article, we examine the structure of party competition in Italy in the February 2013 elections. We rely on the spatial approach to party competition to analyse the most salient dimensions of the policy space in the Italian context. Our analysis is based upon quantitative estimates from expert survey data. These data highlight the salience of the socio-economic policy dimension and capture the change in the importance of the EU dimension. Finally, this study provides an analysis of potential coalition governments in the aftermath of 2013 general election that is grounded on the spatial approach to coalition formation. Keywords: expert survey, party competition, coalition formation, Italian politics Paper accepted for publication in Government and Opposition 1

The purpose of this article is to map out the relevant dimensions of party competition and the policy positions of Italian parties in the general election of February 2013. This task is undertaken by using data from an expert survey carried out by the authors which employs the methodology developed by Benoit and Laver (2006). The analysis highlights the importance of two dimensions: socio-economic left right and pro-anti EU policy. In comparison to the 2008 elections, the EU dimension has become much more relevant to parties relative positioning in the policy space. This change may be explained by the fact that the EU played an increasingly important role in Italian domestic politics since the start of the current economic crisis. This was most apparent in the appointment of the technocratic government that ruled Italy throughout the 2012 in order to implement a number of emergency policy measures requested by the EU. The argument presented in this article is structured as follows. Section 1 will provide an overview of the context in which national elections took place. This is followed in Section 2 by a brief summary of the expert survey methodology and the data used in this article. Thereafter, Section 3 will present a two-dimensional account of the Italian party system based on the socio-economic left right and the EU dimension while Section 4 will deal with government formation. Concluding remarks follow in the final section. 1. The Italian general election of February 2013 The Italian general elections held on 24 and 25 February 2013 took place after the resignation of Prime Minister Mario Monti, head of a technocratic government that ruled Italy for about a year (November 16 2011- December 21 2012), followed by the early dissolution of parliament by President of the Italian Republic, Giorgio Napolitano. The Prime Minister had been appointed by President Napolitano at the end of 2011 as head of an emergency government after the fall of the Berlusconi government, which had been unable to guarantee 2

the adoption of policy measures requested by the European Union. The government was supported by a strange majority including the three largest party groups in the parliament, i.e. the centre-right People of Freedom (), the centre-left Democratic Party () and the Centrist Democratic Union (). Three main features marked the Italian political life during this period (Di Virgilio and Radaelli 2013). The first feature was increasing public discontent with government austerity measures, high unemployment rate and lack of economic growth. The second characteristic was the stalemate among political parties who proved unable to implement a long waited reform of political institutions in order to change the electoral rules, cut the costs of politics and fight widespread corruption. The third important feature was the electoral rise of the Five Star Movement () led by the former comedian Beppe Grillo after the regional elections held in Sicily in October 2012, when the emerged as the largest party in the island. Due to internal conflicts, the strange majority collapsed in December 2012, after the withdraw its support for the government. Earlier elections were scheduled for February 2013. Parties and coalitions that contested this general election included the following. The left-wing coalition including the and Left Ecology and Freedom () was led by the secretary Pierluigi Bersani. The right-wing coalition including the the Northern League () and other minor groups such as the right wing splinter of the Fratelli d Italia () was led for the sixth time by Silvio Berlusconi. A coalition including and other centrist forces called Scelta Civica () endorsed the incumbent Prime Minister and life-tenured senator Mario Monti as a prospective prime minister. The led by Beppe Grillo, who did not run as a candidate, was the fourth player in these elections. A fifth option was the extreme left joint list called Civil Revolution (), including a number of candidates from small leftist parties. 3

According to many commentators, the centre left s campaign was lacklustre and mostly targeted towards traditional left wing voters. The incumbent Prime Minister, Mario Monti, championed his reform agenda strongly associated with a pro-europe stance, but lost a great deal of his personal support after entering the political race. The growth in opposition to increased taxation and reduced spending during 2012 created an opportunity for party leaders to seek electoral gains by promoting populist policies such as repealing taxes introduced by the Monti government or re-negotiating Italy s position in the European Union and Euro zone. This populist and Euro-sceptic style of campaigning was most strongly associated with Silvio Berlusconi and Beppe Grillo. For the 2013 elections, Italy used the same electoral system used in 2006 and 2008 elections, i.e. a closed list PR system with a seat bonus which generates a majority of 54% for those party lists or coalitions that gain a plurality of the popular vote. There are important differences in the rules for electing the Chamber and the Senate. The most important one has to do with the allocation of the seat bonus, whereas the seat bonus is assigned on a national basis for electing the Chamber and on a regional basis for electing the Senate. The consequence of this provision is that there is no guarantee for a coalition or party list obtaining the largest number of votes at the national level to obtain an absolute majority of the seats in the Senate. 1 [Table 1 here] Table 1 presents the electoral results both in the Chamber and the Senate. No party obtained more than 26 per cent of valid votes. Nine party lists obtained seats in the Chamber, twelve in the Senate. The centre-left coalition gained a plurality of votes in the Chamber (29.5% of valid votes). The seat bonus ensured the centre left coalition a majority of seats in 4

the Chamber (340 seats out of 630). In the Senate, the centre-left coalition gained 121 seats: a number far short of the majority threshold required to govern (158). The centre-right coalition gained about 29.2% of valid votes in the Chamber and 30.7% in the Senate. The centre-right coalition lost the majority bonus in the Lower Chamber for a tiny amount of votes (124,407) and gained slightly less seats than Bersani s Pd led coalition (117) in the Senate. The coalition led by incumbent Prime Minister Mario Monti barely passed the 10% electoral threshold set for electing the Chamber. In the Senate, the Monti led coalition won 22 seats, not enough to be pivotal for any potential centre-left or right wing government. To sum up, the 2013 elections resulted in a tie between the centre left, which was predicted to win in the pre-election polls, and the centre-right. The strong performance of the centre right was the result of Silvio Berlusconi s effective personalised election campaign. The coalition led by Mario Monti dropped to fourth place. Finally, the Five Star Movement () was the undisputable winner of this election gaining 25% of the popular vote that translated into 109 seats in the Chamber. In the Senate, the Five Star Movement won 23.3% of the vote and gained 51 seats. In all the Italian general elections between 1996 and 2008, a majority of votes and seats were won by the two main electoral blocs on the centre left and the centre right. As a result, the seat bonus went to coalitions securing about 40% of the popular vote. The fragmentation of party support in 2013 made it possible for the left-wing coalition to obtain the seat bonus in the Chamber with less than 30% of total vote. The fragmentation of party support also had an important impact on the conversion of votes into seats in the Senate, making it impossible for any party to win a majority of seats in the Upper Chamber, and hence to be able to form a government. Voter turnout in 2013 at 75%, although high in comparative terms, represented the lowest level of electoral participation in an Italian general election since 1946; and represents 5

a 5% decline in turnout on 2008. Most commentators interpret the record low level of turnout as reflecting popular disaffection with party politics in Italy. This alienation from contemporary Italian parties is also strongly evident in the changing levels of support for parties between the 2008 and 2013 general elections. All of the main parties lost votes with the right wing parties (i.e. and ) appearing to suffer the highest levels of attrition. One rough index of the level of vote switching between 2008 and 2013 is the fact that the total number of lost votes (11.2 million) constituted a third of the total electorate in February 2013. This remarkably high level of volatility stems mainly from the emergence of Beppe Grillo s political movement, which attracted more than 8.5 million votes in February 2013 to become the second largest party in the Italian Chamber. We will return to the election results in Section 4 when dealing with the process of government formation. 2. Data and methods The spatial approach to party competition is grounded on the assumption that some kind of policy space can be used to describe the preferences and choices of relevant political actors. Policy dimensions can be thought as ways of describing preferences on clusters of related issues. For instance, if preferences over economic issues such as taxation, workers rights, public spending are highly correlated then we may think of these issues as being part of an underlying economic policy dimension. We use policy dimensions to describe different positions taken by political actors on them. For example, a party which favours a strong interventionist role for the state in the economy would be regarded as holding a left-wing position on the economic policy dimension; a party which is in favour of the opposite would be regarded as a right-wing actor on the same dimension. The implementation of spatial models requires the measurement of relative distances between party positions; and their movement in the policy space. In the 6

Western political world, a widely understood left-right dimension has been used to measure position and movement of political parties. In addition, two dimensional maps of policy space have been provided for most European countries (Benoit and Laver 2006). The spatial approach is valuable because it allows us to represent party policy positions and citizens preferences in a common space, in order to test hypotheses about party competition and coalition government formation. In addition, it helps us to understand political change looking for example at how the policy positions of political actors change over time and also how new dimensions structuring the policy space are strategically created by political actors (Riker 1982). A crucial requirement for testing spatial models is accurate estimates of the policy positions of political actors. Several techniques have been used to estimate party positions. Some of them use mass surveys, while others employ elite or expert surveys. Other estimates are based on data derived from hand coded or computerized content analysis of party manifestos (Laver 2001). The most important source of data is the Manifesto Research Group/Comparative Manifestos Project (MRG/CMP) Project that has been coding the electoral platforms of most parties in 25 countries since 1945 (Budge et al. 2001; Klingemann et al. 2006; Volkens et al. 2013). 2 Finally, data about legislative voting behaviour (roll call voting) have been used to infer party positions, especially in the US context (Poole and Rosenthal 1997). All of the above mentioned techniques have both strengths and weaknesses, and it is therefore necessary to make a trade-off depending on the research question being addressed. Here the focus will be on expert surveys data. The expert survey methodology is characterized by the following features: dimensions or scales are pre-defined and parties are located on these scales by country experts. The estimates of party positions are the aggregated results of expert judgements. This methodology has several advantages in 7

comparison to other methods, as it provides more confidence in the accuracy of estimates; it is a relatively quick and costless way of gathering data; and it allows us to place single parties even when pre-electoral coalitions are competing (Benoit and Laver 2006). This is especially important for the Italian case, whereas the Manifesto Research Group since the 1994 elections has coded pre-electoral coalition platforms and not single parties documents which were unfortunately not available. For comparative purposes, expert survey data present some limitations. Some are related to comparison across countries, whereas the substantive meaning of dimensions (i.e. left-right) can change from country to country. Moreover, expert surveys are conducted at specific time points and usually do not provide time series data (MacDonald and Mendes 2001). 3 In the context of this research, these limitations are less relevant for two reasons. First, the analysis is limited to a single country. Second, previous expert survey data about Italian elections have been gathered since 2001 thereby allowing comparison across time. 4 Following the research methodology developed by Benoit and Laver (2006), a survey among Italian experts was fielded in February 2013. We asked political experts to locate political parties on nine substantive policy dimensions, as well on the general left-right dimension using a twenty-point scale. The dimensions in the survey measure parties support for public spending vis-à-vis lower taxes (Taxes vs. spending), state regulation of the market (Deregulation), liberal policies on matters such as abortion, gay rights and euthanasia (Social policy), integration of immigrants (Immigration), environmental protection (Environment) and territorial decentralization of decision making (Decentralization). The survey includes also three dimensions dealing with parties positions on specific aspects of European politics: the scope of EU intervention (EU authority), its peacekeeping role (EU security) and the relative powers of European institutions (EU accountability). 5 8

Taking for instance the Taxes vs. spending dimension, a party that is strongly in favour of increasing public spending is located closer to point 1 on the scale, where the opposite holds for a party that is in favour of cutting taxes. Country experts were also asked to locate each party on a scale measuring the importance of the policy dimension to the party in question. This scale ranges from 1 (not important at all) to 20 (very important). [Table 2 here] The survey was fielded in February 2013, during the campaign for electing the Italian parliament. Following Benoit and Laver (2006: 125-126), we included just the politically relevant parties. In particular, we selected only the 12 parties which in February 2013 were expected to win at least one percent of the votes. We sent an e-mail invitation to 379 experts, 95 of whom completed the questionnaire, with a response rate of about 25 percent. 6 Table 2 presents some summary data from the survey reporting the mean and the standard error of the expert placements for each party on each policy dimension. 7 In the next section, we will make use of these data to map out the relevant dimensions of party competition and the policy positions of Italian parties in the general election of February 2013. 3. A two-dimensional account of the Italian party system, 2013 In the previous section we presented the data and methods used to measure the positions of Italian parties on a set of nine distinct policy dimensions, as well on a general left-right scale. In this section, we will analyze this full set of substantive issues in order to describe the spatial structure of party competition in Italy in the 2013 elections. We will undertake this task in two steps. First, we will use expert judgements to identify which policy dimensions are more politically salient in Italy. Second, we will analyze the patterns of correlation 9

between party positions on different policy dimensions revealed by the experts in order to identify the underlying axes of political competition in Italy. The expert survey methodology is based on a priori decisions that identify the policy issues of potential importance in a given context. Parties attach different degrees of importance to each of these issues, and party saliency scores enable us to understand which dimensions are the most relevant. We measured the overall importance score for each policy dimension in 2013 general election by computing, for each issue, the mean of the partyspecific saliency scores, and weighting it by the vote share received by each party. The first column of Table 2 reports the overall importance score of each dimension, as well as the related standard error. As Table 2 shows, the EU authority dimension, which measures parties propensity to increase/reduce the set of areas of European intervention, was been judged by our sample of experts to be the most important dimension in Italian politics during the 2013 general election. The fact that EU authority is the top-rated policy dimension indicates the importance attached by Italian parties to EU-related themes during the campaign. The policies adopted by the Monti government during 2012 created increasing tensions both between and within parties because of the austerity measures demanded by the EU. Moreover, the involvement of European institutions in the Italian domestic affairs was one of the dominant themes of the 2013 election campaign. Some parties declared themselves ready, in the event of victory in the February elections, to re-negotiate Italy s position in the European Union and Eurozone; and even to propose a referendum for abandoning the Euro and returning to the Lira. Remarkably, the EU authority dimension has received on average higher saliency scores than the dimensions dealing with economic policy ( taxes vs. spending and deregulation ), which are the next most important policy dimensions. This indicates a strong change in the relative weight given by Italian parties to several themes in the election campaign. Indeed, 10

past waves of expert surveys fielded in Western European countries revealed that economic issues were most often the top-rated dimensions (Benoit and Laver 2006). 8 The fourth most salient issue in the 2013 Italian elections is immigration, while EU security and decentralization are ranked last in overall importance. 9 The relatively low political salience of the dimension dealing with the EU s peacekeeping role, which ranks just ahead of decentralization is somewhat surprising. Despite the recent political crises in a number of non-democratic regimes in North Africa and the Middle East, with the potential military involvement of European countries, the issue of peacekeeping operations remained one of the least salient electoral campaign themes in Italy. As far as decentralization is concerned, the very low importance accorded by Italian experts to this issue probably depends on the marginal role played in the campaign by the Northern League (), a territorial party which has traditionally been the main promoter of decentralization of decision making. The internal scandals occurred in 2012 did not allow the to place decentralization as a central theme in the public debate; and to challenge the other parties on this issue as it had occurred in the past. We now turn to analysing patterns of correlation among the full set of issues in order to describe the underlying dimensional structure of Italian party competition. A quick glance at the positions of the four largest Italian parties along some salient dimension suggests that a single axis or policy dimension is not enough to adequately characterize the policy space in 2013 general election. As Table 2 shows, Bersani s and Grillo s M5s parties on the taxes/spending and the deregulation issues are on the left and very close to each other: roughly one or two points apart with the M5s slightly toward the centre. On the same dimensions, and the are both on the right and have similar positions: their distance apart is just one or two 11

points, with Berlusconi s party judged as more rightist by Italian experts than the party led by Mario Monti. This picture changes if we consider the EU authority issue. On this dimension, the is quite far from the M5s, and the is well away from. In both cases, the distance is as much as 11 points. In particular, the are on the pro-integration side and very close to Monti s (approximately a two point difference), the party that is most in favour of increasing the European intervention in domestic politics. In turn, the is placed just three points away from the M5s, which is among the most anti-eu parties. In other words, these three policy dimensions, which prove to be the most salient for Italian parties in 2013, are not related to the same underlying axis of competition. Although two of them deal with economic policy and are clearly correlated, they do not seem associated to the EU authority issue, and this implies a need for a multidimensional description of the Italian policy space. We identify and measure the key axes of political competition in Italy by using Principal Component Analysis (often known more informally as Exploratory Factor Analysis), a statistical data reduction technique that allows us to describe the variability among a large set of observed variables in terms of few unobserved underlying factor. 10 Each of the extracted factors can then be substantively interpreted by looking at those original variables that correlate (or load ) on the factor. If some of our nine policy dimensions load highly on a single factor, we can say that the factor summarizes well parties placements on those dimensions, and hence we can replace that set of policy dimensions with a more fundamental or latent axis of political competition. The results of an exploratory factor analysis applied to the experts placements of Italian parties indicate that the full set of the nine policy dimensions identified for Italy can be reduced to two primary axes of competition. As the top panel of Table 3 shows, for only two latent factors the eigenvalue is greater than one, and this implies that the competition among 12

Italian parties in the 2013 elections can be described in a two-dimensional political space. 11 From the last column of the top panel, we see that these two factors account for about 66% of the total variance in Italian party positions on all the nine original policy dimensions. [Table 3 here] The first and most important factor emerging from the analysis explains as much as 43% of the variance in the nine policy issues, and can be interpreted as reflecting socio-economic left-right. As reported in the bottom panel of Table 3, this factor is associated with a wide range of policy dimensions: the two dimensions dealing with economic policy ( taxes/spending and deregulation ), the two dimensions tapping social liberalism vs. conservatism ( social policy and immigration ), as well as environmental policy. All these dimensions exhibit high loadings on Factor 1. Therefore, we may conclude that in Italy parties on the left tend to prefer public spending rather than tax reduction, are in favour of state intervention in the economy, prioritize environmental protection, are liberal on moral issues and promote the integration of immigrants. In contrast, parties on the right tend to have anti-taxes positions, oppose state intervention, support economic growth even at the cost of damaging the environment, are conservative on moral issues and are against immigration. The second factor identified in the analysis is associated to the three dimensions related to European politics, and accounts for approximately 23% of the total variance observed. This factor can thus be easily interpreted as capturing the attitude of Italian parties towards the EU and can be labelled pro-anti EU policy. Pro-EU parties are in favour of greater EU authority over domestic politics and greater involvement of Italy in European military security, and would like to strengthen the European Parliament vis-à-vis national 13

governments. On the other hand, anti-eu parties oppose further European integration and involvement in EU security, and would give more powers to national governments rather than to a supra-national institution like the European Parliament. The emergence of a distinct pro/anti EU axis of political competition is remarkable and certainly deserves further discussion. Although it is not surprising if we consider the current crisis in the Euro zone and its dramatic consequences on Italian politics, the relevance of proanti EU policy for party competition in 2013 represents a major change from the past. Neither in the 2006 nor in 2008 general Italian election estimates from expert surveys revealed a distinct pro/anti EU axis. In both cases, EU authority and EU accountability were highly correlated with other policy dimensions and incorporated into the socio-economic axis, which constituted the first factor. EU security formed a single axis on its own in 2001, when it represented the third factor together with the factor dealing with socio-economic issues and the factor based on decentralization and deregulation, but not in 2008, when it was associated with decentralization and deregulation; and contributed to form the second axis of competition. 12 Therefore, while in the past policy issues concerning the EU were subsumed into other principal axes of competition, in 2013 attitudes towards the EU became a primary political axis on their own, orthogonal to socio-economic policy. Such a change can be explained mostly by the presence in 2013 of large parties opposing the EU on both the left () and on the right () of the socio-economic scale. [Figure 1 here] Having identified the underlying structure of the policy space in the 2013 Italian election, we estimated parties positions on the two fundamental axes of competition obtained through factor analysis. Figure 1 presents the two-dimensional map of the policy space among Italian 14

parties in the last February elections. 13 In this graph, the horizontal axis corresponds to the socio-economic policy, while the vertical axis represents pro-anti EU policy. The size of the bubbles depends on each party s vote share. As Figure 1 shows, the two largest parties, i.e. and, are very close on the socio-economic left-right, but really far apart on the pro/anti EU- axis. Although both parties are quite to the left, the is rather pro- EU, while the strongly opposes further European integration. The third largest party, the, is between the other two parties on EU policy, but far to the right on socio-economic issues. Monti s party is on the right on the socio-economic axis and clearly pro-eu on the second dimension, whereas and are located on the extreme left and slightly opposed to European integration. The remaining small parties are on the right in socio-economic terms. Some of them have a radically anti-eu stance ( and ) while others such as, and clearly support the European integration project, while seems rather neutral on the EU issue. In the next section we will examine how estimates of parties policy positions can be used to analyze the process of coalition government formation using a spatial approach. 4. A spatial analysis of potential coalition governments after the 2013 general election Earlier formal models of coalition formation assume that political parties and their leaders are rational actors that try to maximize their share of office rewards. Starting from these assumptions, office-seeking models elaborate a series of propositions about coalition formation in multiparty systems which are essentially based on the relative size (seats) of the parties involved in coalition bargaining. Relevant examples are the minimal winning coalition proposition (Von Neumann and Morgenstern 1944); the size principle proposed by Gamson (1961) and Riker (1962); and the bargaining proposition suggested by Leiserson (1966, 1968). In light of the high number of real world government coalitions that 15

appear to contradict these propositions, scholars have taken into account the policy motivations of parties. According to policy seeking models, government coalitions should be formed between actors with similar ideological background (Leiserson 1966, 1968; Axelrod 1970 and De Swaan 1973). When parties compete on a single ideological dimension, policy seeking models predict that the party controlling the median legislator will have a bargaining advantage in government negotiations, since there are no other proposals in the policy space that can defeat the median legislator s position through a majority vote. In a unidimensional political space, a party that includes the median legislator always exists regardless of its size. This account however does not help to explain why minority government which do not include the party controlling the median legislator are often formed (Warwick 1996), suggesting the possibility that coalition formation may be analysed using a two-dimensional rather than a unidimensional account of the policy space. Within the two-dimensional approach, the political heart model developed by Norman Schofield (1986, 1993, 1995, 1996) in a series of studies takes into account both parties size (i.e. seats) and their policy positions. The model gives specific predictions about coalition formation based on the notion of political heart. The political heart is the union of the core and of the cycle-set. A core party is a party that occupies a position in the policy space that cannot be defeated in a majority vote. In a two dimensional space, a core party will exist only when all the median lines (i.e. lines defining a majority in both closed half spaces created by each line) intersect at one party s ideal point. In general, only the largest party in the parliament can be a core party. When the median lines do not intersect at one point in the policy space the core is said to be empty. In this situation, the possibility of instability and chaos in coalition government formation emerges. However, cycles do not span over the entire political space, but are confined to the area enclosed by the intersection of the median 16

lines (i.e. the cycle-set). Assuming that no policy proposals will be made that make all members of a majority coalition worse off, the cycle-set contains all the points in the policy space that are Pareto optimal for every majority coalition. The political heart model has proved to be a powerful analytical tool in the study of coalition politics. 14 Here we will use the political heart model in order to shed light on the government formation process following the Italian general elections of February 2013. As noted earlier in Section 1, the electoral rules allowed the centre-left electoral alliance to gain an absolute majority in the Chamber of Deputies, while they generated almost a tie between the two main electoral alliances in the Senate. Panel A of Figure 2 presents the spatial map of parties policy positions. Party positions are represented as points in this space, while in parenthesis are reported the seats each party holds in the Senate. 15 We focus on the Senate because no coalition in the Upper Chamber has a majority of seats required to form a government. As we can see, the core is empty, since the median lines do not intersect at a single party ideal point. However, theory allows us to predict that the solution for the government formation game should be found in one of the combinations of parties delimiting the cycle-set. Therefore, the, the and the M5s (i.e. those parties that bound the heart) can be considered dominant players in government negotiations, while the parties outside the region delimited by the median lines (, and ) are peripheral in the coalition formation game. [Figure 2 here] Following the political heart model, the Italian government after the general elections of February 2013 should be formed by one of the following party combinations: 1) and M5s, 17

possibly with the support of ; 2) and, possibly with the inclusion of ; 3) and M5s, with the support of. Hence, our theory depicts three scenarios. The first one, to which we can refer to as the government of the change, is the path taken by the Secretary of the, Pierluigi Bersani, appointed as a formateur by the President of the Republic, Giorgio Napolitano, a few days after the elections. The project of Bersani, failed because of the resistance of the M5s, was to build a government majority based on eight programmatic points aiming at a big change in the social and public life of the country. The second scenario, to which the press gave the name of governissimo, was the revival of the majority that supported the technocratic government of Mario Monti. This was the solution preferred by the President of the Republic Napolitano, who never concealed in public debate his preference for a government based on the agreement between the two largest parties on the left and on the right. The third option was a coalition composed by and M5s, with the support of the. This unprecedented combination includes the parties that contested most the authority of the European Union. Hence, we can refer to this scenario as the government of the Euro sceptics. Predictions from the political heart model are based exclusively on party strength and the distribution of parties policy positions in the policy space. However, the analysis of coalition politics might take advantage from the combination of office and policy aspects with the behavioural and institutional ones (Debus 2008, 2009). The emphasis here is on the role of different types of institutions in structuring the outcome of coalition formation process. Among the list of possible constraints that should be taken into account there are pre-electoral commitments to govern together and anti-pact rules such as party combinations that are rejected a priori by at least one party that participates in the potential coalition (Debus 2009: 47). Due to the importance given to credibility, these statements, which are usually public, constitute a powerful constraint on coalition bargaining (Golder 2006; Martin and Stevenson 18

2001; Strøm et al. 1994). Taking into account such behavioural constraints to coalition formation, we can easily rule out our third scenario. Given the institutional design of the Italian political system based on symmetric bicameralism, requiring an investiture vote in both Chambers, all coalitions must necessarily include the, which is the largest party in the Chamber of Deputies, holding about the 47% of the seats. Hence, the government of the Euro sceptics does not survive as a feasible option. The same fate is reserved to the government of change. A coalition between and M5s is not viable because of the refusal of the party led by Beppe Grillo to take part in any coalition including parties linked to old politics. This statement was made public before the elections and it works as an anti-pact extended to all the parties in parliament. Once excluded M5s from any potential coalition, the only feasible option suggested by the political heart model remains the governissimo between and as Panel B of Figure 2 shows. Excluding the M5s from the set of potential negotiators gives as result, in fact, only one median line connecting the ideal points of the other two largest parties in parliament. This is the coalition government that actually formed in Italy on April 30 2013. Such outcome supports the spatial model predictions, complemented with institutional and behavioural considerations. Party leaders reached an agreement over government formation only after 61 days. 16 Both the and the have been the leading parties of the two opposing coalitions on the centre left and the centre right in 2013 as well as in all the elections held in Italy the past two decades. These two parties have been engaged in harsh electoral competition and parliamentary battles, alternating in government, with the only exception of the joint support given to the Monti government. This might explain why the party leader Bersani, appointed as a formateur, explored for several days the possibility of obtaining the support of 19

the Five Star movement or even forming a government lacking a majority of seats in the Senate. It should be noted that the process of government formation was compounded with the election of the President of the Republic, scheduled on March 15. The process of electing the head of state ended up with a reappointment for a seven year second term of the incumbent President Napolitano, after other candidates that had been proposed jointly or unilaterally by parties failed to get enough votes in parliament. After re-election, Napolitano nominated the vice-secretary, Enrico Letta, as a formateur. Letta worked to build an alliance among the and the and was able to form a government including ministers from the, the,, plus several independents. The government obtained the full support of Monti s party and for this reason can be categorized as a surplus or oversized coalition. The inclusion of this party, not strictly predicted by the political heart model, might be understood in terms of intra-party politics considerations, not taken into account in the previous analysis (Giannetti and Benoit 2009). First, the newborn formed by Monti just two months before the elections included, among others, candidates from both the and the. Second, the party was deeply divided about coalitional strategies in the aftermath of 2013 election. 17 This implies that the full parliamentary support for government legislative provisions was not guaranteed from the start, suggesting the prospective prime minister had to rely on a wider parliamentary majority. The government passed the investiture vote with 453 yes, 153 no and 7 abstentions in the Chamber. In the Senate the government passed the investiture vote with 233 yes, 59 no and 18 abstentions. The, the and voted in favour and the and parties voted against. The abstained both in the Chamber and in the Senate where abstention was equivalent to voting against the Letta government. 20

Formal models of coalition formation predict instability for those coalitions which are not an equilibrium point. Empirical evidence support such prediction as minimal winning coalitions last longer than minority governments and oversized coalitions (Gallagher et al. 1992). In a policy based account, minority government may be an equilibrium and consequently they are expected to last. The stability of oversized coalitions is less predictable, as it is related to parties size and the overall configuration of policy positions in the policy space as well as to institutional constraints and intra-party politics related features. This last variable is particularly crucial in the Italian case, where institutional incentives to party cohesion are weak. For all these reasons it is hard to tell how long the Letta government will last. To sum up, the theoretical and empirical arguments stated above allowed us to identify the potential coalition governments in the aftermath of the 2013 general election. According to the political heart model, three coalition governments were predicted to form. Institutional and behavioural constraints contribute toward explaining why the coalition government including the two largest parties on the left and on the right was the one that actually formed. Conclusion In this article, we examined the structure of party competition in the 2013 Italian general election. Our analysis relied on original data from an expert survey that was carried out by the authors immediately before the election. By using these data we were able to describe the most salient dimensions structuring party competition in 2013 and to locate parties in the policy space. Our analysis suggests that a two dimensional map where the main axes are the socio-economic left right and the pro-anti Europe policy dimensions are the most appropriate for analyzing party competition in the 2013 Italian general election. 21

The emergence of a pro-anti Europe dimension is hardly surprising if we consider the current economic crisis within the Euro zone and its consequences on Italian politics, including the formation of the technocratic government that ruled in Italy during 2012. The relevance of a pro-anti EU dimension in 2013 represents a major change from the past; and might explain the electoral fate of the two parties that locate themselves on the opposite side along this dimension, i.e. the party list Scelta civica led by Mario Monti and Beppe Grillo s Five Star Movement. Moreover, our work provides an analysis of potential coalition governments in the aftermath of 2013 elections grounded on the spatial approach to coalition formation. The political heart model allowed us to identify three potential coalition governments on the basis of parties size and policy positions estimates, including the government that eventually formed in late April. Finally, we suggested that institutional and behavioural constraints as well as intra-party politics related factors not taken into account in the pure spatial approach may help to understand the process of government formation and the outcomes of coalition bargaining. 22

Endnotes 1 Other differences are related to the different thresholds to be passed in order obtain votes, i.e. 4% and 10% for coalitions and parties respectively in the Chamber, and 8% and 20% for coalitions and parties respectively in the Senate. 2 For more details, see the Manifesto Project Database website: https://manifestoproject.wzb.eu/ 3 However, Benoit and Laver (2006) successfully compared their expert survey results with those in Laver and Hunt (1992) in order to study the extent of party system change in Western European party systems. 4 This comparison is not fully developed in this article. See, however, Section 3. For expert survey data about Italy see Benoit and Laver (2006) and Curini and Iacus (2008). 5 We also asked experts to place parties on a twenty-point sympathy scale indicating the respondent s closeness to each party. See the appendix for the precise wording of the questions and the corresponding scales. 6 Experts were members of the Italian Political Science Association (SISP). We gratefully thank the Association for supporting this enterprise. 7 Standard errors are calculated as the standard deviation of the expert placements divided by the square root of the number of placements minus one. Greater errors indicate higher uncertainty in expert judgements in positioning the party. 8 Benoit and Laver (2006: 110) found that, generally speaking, policy towards the European Union was often the most salient dimension in Eastern Europe, where the issue of joining the EU was obviously a central theme in electoral campaigns. 9 EU authority, taxes/spending, deregulation and immigration are judged as more important than the overall mean calculated across all the nine dimensions identified in Italy. The weighted mean saliency score for those four issues is above the overall mean of the dimension-specific importance scores (11.83). 10 Factor analysis has been applied to extract primary dimensions of party competition both from the policy issues identified by expert surveys (Laver and Hunt 1992; Benoit and Laver 2006) and from the coding categories of party manifestos (Budge et al. 1987; Gabel and Huber 2000). The method we use for factor extraction is principal component analysis. According to this method, the input variables are modelled as linear combinations of a smaller set of factors in order to account for the maximum possible variance in the data. To obtain factors that are more easily interpretable, we used varimax rotation, which is the most common rotation option employed in previous research. 11 In the factor analysis technique, only factors with eigenvalues greater than unity are conventionally retained, since only these factors contain more information than a single one of the input variables. 23

12 The Italian political space described by Benoit and Laver (2006) for the 2001 elections was threedimensional (data were collected in the years 2003-2004). For the 2008 elections, we used factor analysis and built a two-dimensional political space based on Curini and Iacus (2008) data. This analysis is not reported in this article. Unfortunately do not exist data for 2006 elections. 13 We estimated the individual scores of parties using the regression coefficients based on the varimax rotated factors. 14 For applications of the model to the Italian case see Giannetti and Sened (2004) and Curini (2011) and Curini and Pinto (2013). 15,, and did not pass the electoral threshold for electing senators. For this reason, their policy positions are not reported in Figure 2., and presented a joint list in the Senate. Consequently, they are represented by just one ideal point, which is the weighted mean (by their vote share) of their original policy preferences. is currently a member of the Mixed group in the Senate. We represented it as a single group, and redistributed the remaining four Senators between the other parties according to the list they were elected. We did the same for other two small groups. The ten senators from Great autonomies and liberties were divided between and, while the ten members of the group For the autonomies were assigned mainly to. Finally, we did not consider life tenured senators in computing seats. 16 So far, only two governments did worse in the entire history of Italian Republic: Cossiga I (August 1979, 62 days) and Amato I (April 1992, 82 days). 17 The divisions within the party revealed also in the vote for electing the President of the Republic, whereas 101 MPs did not vote for the candidate proposed by the party leader Bersani. 24

References Axelrod, R.M. (1970), Conflict of interest (Chicago: Markham). Benoit, K. and Laver, M. (2006), Party Policy in Modern Democracies (London: Routledge). Budge, I., Robertson, D. and Hearl, D. (1987) (eds), Ideology, Strategy and Party Change. A Spatial Analysis of Post-War Election Programmes in 19 Democracies (Cambridge, Cambridge University Press). Budge, I., Klingemann, H-D., Volkens, A., Bara, J. and Tanenbaum, E. (2001) (eds), Mapping Policy Preferences: Estimates for Parties, Electors, and Governments, 1945-1998 (Oxford: Oxford University Press). Curini, L., (2011), Government survival the Italian way: the core and the advantages of policy immobilism during the First Republic, European Journal of Political Research, 50(1): 110-142. and Iacus, S. (2008), Italian Spatial Competition Between 2006 and 2008: A Changing Party System?, paper presented at the XXII Congress of the Italian Political Science Society (SISP), Pavia, 5-8 September. and Pinto, L. (2013), Government Formation Under the Shadow of a Core Party. The Case of the First Italian Republic, Party Politics, 19(3): 502-522. De Swaan, A. (1973), Coalition Theories and Cabinet Formation (Amsterdam: Elsevier) Debus, M. (2008), Office and Policy Payoffs in Coalition Governments, Party Politics, 14(5): 515-538. (2009), Pre-Electoral Commitments and Government Formation, Public Choice, 138(1): 45-64. Di Virgilio A. and Radaelli, C. (forthcoming 2013) (eds), Italian Politics The Year of the External Podestà (New York, Oxford: Bergham Books). 25

Gabel, M. and Huber, J. (2000), Putting Parties in Their Place: Inferring Party Left-Right Ideological Positions from Manifestos Data, American Journal of Political Science, 44(1): 94-103. Gallagher, M., Laver, M. and Mair, P. (1992), Representative Government in Modern Europe (New York: MacGraw-Hill). Gamson, W.A. (1961), A theory of coalition formation, American Sociological Review, 26(3): 373-382. Giannetti, D. and Sened, I. (2004), Party Competition and Coalition Formation: Italy 1994-1996, Journal of Theoretical Politics, 16(4): 483-515 Giannetti, D. and Benoit, K. (2009) (eds), Intra-party Politics and Coalition Governments (London: Routledge) Golder, S.N. (2006), The Logic of Pre-Electoral Coalition Formation (Columbus: Ohio State University Press). Klingemann, H-D., Volkens, A., Bara, J., Budge, I. and MacDonald, M. (2006) (eds), Mapping Policy Preferences II: Estimates for Parties, Electors and Governments in Central and Eastern Europe, European Union and OECD 1990-2003 (Oxford: Oxford University Press). Laver, M. (2001) (ed), Estimating the Policy Positions of Political Actors (London: Routledge). and Hunt, W.B. (1992), Policy and Party Competition (London: Routledge). Leiserson, M. (1966), Coalitions in Politics. A Theoretical and Empirical Study, PhD thesis, New Haven, CT, Yale University. (1968), Factions and Coalitions in One-Party Japan: An Interpretation Based on the Theory of Games, The American Political Science Review, 62(3): 770-787. 26