Party Ideology and Policies Matteo Cervellati University of Bologna Giorgio Gulino University of Bergamo March 31, 2017 Paolo Roberti University of Bologna Abstract We plan to study the relationship between party ideology and the policies parties enact, using data on Italian municipalities. We will take advantage of a peculiar feature of the Italian electoral system: the position of parties on the ballot is determined by a random draw. The peculiar arrangement of parties on the ballot creates a focal position within coalitions supporting a candidate for mayor. More specifically, votes often concentrate on parties listed on the same line as the name of the candidate for mayor. This ballot order effect, and the random distribution of parties on the ballot, provide exogenous variation in the percentage of votes and number of seats obtained by the party in the focal position. To the best of our knowledge, this characteristic of the Italian electoral system allows, for the first time, to clearly answer a series of open research questions in the field of political economy. First, what happens to policies enacted by a ruling coalition when a party with a specific ideology increases its seats on the municipality council? This empirical exercise can help us understand whether parties adhere to their original political manifestos and electoral promises, e.g. does Rifondazione a (a Communist Party) increase spending on social services if it randomly receives more seats on the city council? Second, we will study the influence of populist parties on policy, in order to determine whether they can be characterized by a specific set of policies or are rather identified solely by their political narratives. Finally, we will be able to investigate the role of political polarization within a city council in shaping policies. 1 Introduction Increasing effort has been devoted to investigating the links between parties, candidates, and the policies they implement. This has been at the core of research in political economy since the seminal work of Downs (1957) and recognition of the tremendous importance of government decisions in every phase of economic life. While there exists a vast theoretical literature that investigates Department of Economics. University of Bergamo. Via dei Canania 2, 24127 Bergamo, Italy. (e-mail: giorgio.gulino3@unibo.it) Department of Economics. University of Bologna. Piazza Scaravilli 2, 40126 Italy. (e-mail: paolo.roberti3@unibo.it) 1
the incentives of candidates and parties and how they affect policies, the empirical relationship between party ideology and policies is still unclear. One of the main obstacles has been the lack of a clear and convincing identification strategy. Indeed, a number of confounding factors can affect this relationship, from voters preferences and idiosyncratic features of party candidates, to the direct effect of policies on preferences and party structure and ideology, which could lead to a problem of reverse causality. A relatively recent literature has used a regression discontinuity design (RDD) to address this issue. Pettersson-Lidbom (2008) has shown that left-wing ruling coalitions in Sweden spend and tax more than right-wing coalitions, by comparing municipalities where a left-wing coalition won by a small margin with municipalities where a right-wing coalition won by a small margin. Ferreira and Gyourko (2008) use the same empirical approach to show that whether a mayor is a Democrat or a Republican does not affect the size of US city governments or the allocation of local public spending. Folke (2014) implements a modified RDD tailored for proportional systems to show that small parties matter for secondary policies, such as immigration and environmental policies. While the results of these works shed important light on the link between parties and policy, they also suggest that this relationship is complex and influenced by the presence of multiple parties in ruling coalitions, which are heterogeneous in terms of size and ideology. The goal of this project is to contribute to this literature by investigating the link between parties and policies, using a large dataset on Italian municipalities and a new identification strategy that allows for a comprehensive understanding of the behaviour of parties in coalition governments. By using the ballot order effect and the random allocation of parties on the ballot paper, we can exploit random shocks in the share of votes and party seats on the municipal council. The Italian municipal electoral system incentivizes the formation of coalitions of parties. More specifically, parties usually form coalitions that have a broad common political objective, but are also often highly heterogeneous with respect to their political manifestos and their electoral promises. Local parties frequently exist at the national level as well, although a sizable number of parties are present only locally (liste civiche). Moreover, it has been suggested that the range of local parties in Italy possibly covers one of the largest ideological spectra in the world. One interesting institutional feature is that the coalition of parties that wins the election has a majority premium and gains at least 60 per cent of the seats on the municipality council. These seats are distributed proportionally to the parties within the ruling coalition. The presence of multiple parties in a coalition government, and their marked ideological heterogeneity makes this an ideal setting for understanding how different party features interact in shaping policy. Finally, the presence of a large majority granted by the law to ruling coalitions helps isolating the role of ideology from possible confounding factors related to the risks of government stability. 2 Research questions By exploiting random variation in the share of votes for parties, we are able to answer a set of important research questions, for the first time with a causal interpretation. First, how do policies change when a specific party sees an increase in its share of votes? To answer this question we will compare ruling coalitions where a specific party (e.g. Forza Italia) is in the focal position (treated) with ruling coalitions where the same party is not in the focal 2
position (control), where belonging to the treated vs. control group is random. In other words we will isolate the effect of an exogenous increase in the political weight of, e.g. Forza Italia, within all center-right coalitions in which this party participates. This exercise can be performed for any coalition and party. This empirical exercise will help answer a number of important, but essentially unaddressed, research questions. In particular: do parties adhere to their political manifestos and electoral promises? To this end, we will carefully analyse the political manifestos and electoral promises of each party, where possible at a municipality level, in order to relate the policies implemented by a party with its promises and manifestos. We will also analyse both aggregate spending and taxation decisions, as well as issue-specific policies such as spending on the environment, security and police, education and culture, and business services. This same empirical exercise will also shed light on competing theories of policy making in ruling coalitions (Martin and Vanberg (2014)), that is whether policies implemented by coalition governments react to the ideology of each party in the coalition, to the median party, or the median voter in each municipality. Second, we will study the effect of party size on policy. As shown by Folke (2014), small parties influence secondary policies. But are only large parties able to affect primary policies, such as the tax rate? This empirical exercise will contribute to the literature on the specialization of parties in more or less salient dimensions of the policy space. Third, we will investigate the effect of so called populist parties on policy. We will focus in particular on Lega Nord, a party that has lately taken a political stance that is both antiimmigration and uses a populist narrative, identifying the Italian people as united against a corrupt elite. Is Lega Nord a more extreme version of other moderate right-wing parties (e.g. Forza Italia) or does it differ in some relevant policy dimensions, such as social spending? This exercise will shed light on debate concerning the nature of populism; whether it is simply a political discourse or can be identified by a specific set of policies (Mudde (2004)). Finally, we will investigate the impact of political polarization on policies. Political polarization is on the rise in the US and in Europe (McCarty et al. (2008)), and while there has been a systematic effort to identify the causes of political polarization (see for example Ortoleva and Snowberg (2015)), it remains unclear what happens to policies when political polarization increases. This is primarily due to the challenges of finding a good identification strategy, given that many confounding factors influence both political polarization and policies. In order to solve this problem, we will consider random shocks to both parties in the ruling coalition and those in the opposition. In particular, if extremist parties in both the ruling coalition and opposition receive a boost in votes thanks to the ballot order effect, polarization on the city council increases. We will relate this random shock to policy and show the causal effect of polarization on policies. 3 Methodology and Data The ballot order effect refers to the relation between the order of symbols of parties (or names of candidates) on a ballot paper and the distribution of votes. The existence and relevance of this relation has previously been addressed in the political science literature by, for example, Miller and Krosnick (1998), Ho and Imai (2008) and Meredith and Salant (2013). The empirical literature has shown that when parties are (vertically) listed on the ballot, the first and the last parties receive 3
a boost in votes. However, as shown in Figure 1, in our setting parties are not simply listed on the ballot. Rather, the combination of symbols and names on Italian municipal ballot papers is quite detailed. Parties are grouped into coalitions (white boxes), the name of the running mayor is written on the left side of the box, while parties that belong to the coalition are listed on the right side of the box. Figure 1: Facsimile of a Ballot Paper We were able to retrieve information on party names and their positions on ballot papers from 2002 to 2012 (excluding 2003), from the Italian Ministry of Internal Affairs. Thanks to this database we can associate, for all parties, their position on the ballot with their share of votes. Since 1990, in Italy, after approval of the Law No. 53/1990, the vertical list of party symbols within each box representing a single coalition is randomized. The draw, organized by state officials, takes place once, and the same ballot paper is distributed to all voters. Thanks to this feature of the Italian voting system, Cervellati, Gulino and Roberti (2017) show the existence of a ballot order effect in the Italian Municipal elections, and use this exogenous variation to instrument the political fragmentation of ruling coalitions. In particular, the authors identify the causal effect of party position on the percentage of votes obtained by each party. The peculiar placement of parties on the ballot creates a focal position in every coalition supporting a candidate for mayor: votes are likely to concentrate on parties listed on the same line as that of the name of the candidate for mayor (i.e. parties surrounded by red arrows in Figure 1). They observe that parties in the focal position obtain on average 5.25 percentage points more votes than other parties within the coalition. Results are robust to various specifications and controls. Due to the proportional distribution of seats within the coalition, a reallocation of 5 percentage points can significantly change the balance of power within the ruling coalition. The implicit minimum threshold, a consequence of the discrete nature of seat allocation, to get at least one seat is, in many cases, lower than 4 percentage points. Indeed, the party in the focal position sees its probability 4
of getting at least one seat increase by 5-18 percentage points, depending on the number of parties present within the coalition on the ballot paper. This setting allows us to study the effect of party ideology on policy through a natural experiment. First, we select all government coalitions where a specific party, let us denote it by Party A, is present. The treated group is composed by all municipalities in which Party A is in the focal position on the ballot, while the control group is composed by municipalities in which Party A is not in the focal position. It is important to stress that we are not studying the extensive margin of the presence of Party A in the coalition, comparing coalitions with and without Party A. Instead, we are precisely identifying the intensive margin of party ideology, comparing coalitions where the relative power of Party A is lower, as opposed to coalitions where the relative power of Party A is higher. The database is further enriched by the inclusion of all budget details of Italian municipalities from 1997 to 2014. Preliminary evidence underlines two main results. Increases in the relative power of extremist parties, both right-wing and left-wing, have a stronger effect on policies then increases in the relative power of more moderate parties. Moreover, results show that extreme left-wing parties spend more on education and social care, while Lega Nord spends more on police and justice, but also on public schools. Our next step will be to expand the database by digitalizing ballot papers available only in paper format in the archives of the Italian Ministry of Internal Affairs. Expanding the database will be crucial to obtaining more robust and definitive results. In addition, we will collect information on election campaigns. First, we will search and download political manifestos and where possible, the electoral programs of parties at the municipal level, through a data scraping process. We will then extrapolate the sentiment of political manifestos and electoral programs with respect to different topics, so as to build for each topic the policy orientation of a party. In this way, we can identity party ideologies. Second, we will test whether and to what extent parties are committed to their political manifestos and electoral programs. 5
4 Preliminary Results Table 1: The Effect of Being in the Focal Point on Party s Electoral Outcomes (Lega Nord - North League) (1) (2) (3) (4) (5) (6) OLS OLS OLS OLS OLS OLS VARIABLES Seats Share of Votes Share of Seats Lega in the Focal Point 1.5459*** 1.2245*** 7.7359*** 6.0622** 8.6362*** 6.9298** (0.517) (0.467) (2.672) (2.392) (3.133) (2.873) Observations 162 161 162 161 162 161 R-squared 0.2786 0.5794 0.3627 0.6326 0.3237 0.6004 Running Parties Fixed Effect Election Day Variables Geographical Characteristics Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Table 2: The Effect of Being in the Focal Point on Party Electoral Outcomes (Rifondazione a - Communist Party) (1) (2) (3) (4) (5) (6) OLS OLS OLS OLS OLS OLS VARIABLES Seats Share of Votes Share of Seats Party in the Focal Point 0.3397*** 0.3630*** 1.8193*** 2.1935*** 2.1352*** 2.5643*** (0.111) (0.107) (0.556) (0.561) (0.695) (0.697) Observations 475 466 475 466 475 466 R-squared 0.0739 0.2644 0.1668 0.2706 0.0952 0.1898 Running Parties Fixed Effect Election Day Variables Geographical Characteristics Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 6
Table 3: The Effect of Being in the Focal Point on Party Electoral Outcomes (Partito o - Party) (1) (2) (3) (4) (5) (6) OLS OLS OLS OLS OLS OLS VARIABLES Seats Share of Votes Share of Seats Party in the Focal Point 1.0904*** 0.9203*** 5.6900*** 5.1313*** 5.7299*** 5.2918*** (0.420) (0.231) (1.568) (1.050) (1.803) (1.250) Observations 619 609 619 609 619 609 R-squared 0.0368 0.7151 0.2970 0.7079 0.2529 0.6642 Running Parties Fixed Effect Election Day Variables Geographical Characteristics Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Table 4: The Effect of Being in the Focal Point on Party Electoral Outcomes (Forza Italia - Berlusconi s party) (1) (2) (3) (4) (5) (6) OLS OLS OLS OLS OLS OLS VARIABLES Seats Share of Votes Share of Seats Forza Italia in the Focal Point 0.2826 0.5543* 1.8522 2.7509** 2.0952 3.1594* (0.366) (0.284) (1.437) (1.373) (1.708) (1.632) Observations 405 402 405 402 405 402 R-squared 0.0801 0.4970 0.2501 0.4292 0.1727 0.3670 Running Parties Fixed Effect Election Day Variables Geographical Characteristics Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 7
Table 5: The Effect of Being in the Focal Point on Party Electoral Outcomes (Alleanza Nazionale - Right-Wing Party) (1) (2) (3) (4) (5) (6) OLS OLS OLS OLS OLS OLS VARIABLES Seats Share of Votes Share of Seats Alleanza Nazionale in the Focal Point 0.4486 0.3298 2.5149 2.7464** 3.1837 3.4208** (0.330) (0.278) (1.634) (1.373) (1.965) (1.668) Observations 250 247 250 247 250 247 R-squared 0.1022 0.5721 0.1450 0.5027 0.1113 0.4713 Running Parties Fixed Effect Election Day Variables Geographical Characteristics Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 8
Figure 2: The Effect of Increasing Party Representation in the Council on Spending Outcomes (Left- Wing Coalitions) -,04 -,02 0,02,04,06 -,04 -,02 0,02,04,06 (a) Current Spending (b) Current Spending (CV) -,1 -,05 0,05,1 -,15 -,1 -,05 0,05 (c) Police & Justice (d) Police & Justice (CV) -,05 0,05,1,15,2 -,05 0,05,1,15 (e) Education & Culture (f) Education & Culture (CV) -,15 -,1 -,05 0,05,1 -,1 -,05 0,05,1 (g) Business Services (h) Business Services (CV) -,1 -,05 0,05,1,15 -,1 -,05 0,05,1,15 (i) Social Sector (j) Social Sector (CV) 9
Figure 3: The Effect of Increasing Party Representation in the Council on Spending Outcomes (Right- Wing Coalitions) -,15 -,1 -,05 0,05 -,1 -,05 0,05,1 (a) Current Spending (b) Current Spending (CV) -,1 0,1,2,3 -,1 0,1,2,3 (c) Police & Justice (d) Police & Justice (CV) -,1 0,1,2 -,1 0,1,2 (e) Education & Culture (f) Education & Culture (CV) -,2 -,1 0,1,2 -,2 -,1 0,1,2 (g) Business Services (h) Business Services (CV) -,15 -,1 -,05 0,05,1 -,15 -,1 -,05 0,05,1 (i) Social Sector (j) Social Sector (CV) 10
References Cervellati, M., Gulino, G., and Roberti P. (2017). Separated under the same roof: Fiscal inefficiency of party fragmentation and mayor s political power. Mimeo. Downs, A. (1957). An economic theory of political action in a democracy. Journal of Political Economy, 65(2), 135-150. Ferreira, F., and Gyourko, J. (2009). Do political parties matter? Evidence from US cities. The Quarterly journal of economics, 124(1), 399-422. Folke, O. (2014). Shades of brown and green: party effects in proportional election systems. Journal of the European Economic Association, 12(5), 1361-1395. Ho, D. E. and Imai, K. (2008). Estimating causal effects of ballot order from a randomized natural experiment the california alphabet lottery, 1978-2002. Public Opinion Quarterly, 72(2):216-240. Martin, L. W., and Vanberg, G. (2014). Parties and policymaking in multiparty governments: The legislative median, ministerial autonomy, and the coalition compromise. American Journal of Political Science, 58(4), 979-996. Enns, P. K., McCarty, N., Poole, K. T., and Rosenthal, H. (2008). Polarized America: the dance of ideology and unequal riches. Meredith, M. and Salant, Y. (2013). On the causes and consequences of ballot order effects. Political Behavior, 35(1):175-197. Miller, J. M. and Krosnick, J. A. (1998). The impact of candidate name order on election outcomes. Public Opinion Quarterly, pages 291-330. Mudde, C. (2004). The populist zeitgeist. Government and opposition, 39(4), 542-563. Ortoleva, P., and Snowberg, E. (2015). Overconfidence in political behavior. The American Economic Review, 105(2), 504-535. Pettersson-Lidbom, P. (2008). Do parties matter for economic outcomes? A regression-discontinuity approach. Journal of the European Economic Association, 6(5), 1037-1056. 11