Dr. Rossi, Mr. Hu and the politics. The role of immigration in shaping natives political preferences

Dr. Rossi, Mr. Hu and the politics. The role of immigration in shaping natives political preferences Guglielmo Barone (Bank of Italy and RCEA) Alessio D Ignazio (Bank of Italy) Guido de Blasio (Bank of Italy) Paolo Naticchioni (University of Cassino and Southern Lazio, IZA) September 2013 First and very preliminary DO NOT QUOTE OR CIRCULATE Abstract The paper analyses the impact of immigration on the political preferences of natives. We study the Italian case that is particularly interesting because the country has experimented a fast and large migration inflows as well as a right-wing domination for the last 15-20 years. It finds that in municipalities that experienced relatively larger arrivals of immigrants, the electorate has been more willing to vote for the coalition of centre-right (the one headed by the former Prime Minister, Mr. Berlusconi), which had a political platform less favorable to the immigrants. The study also highlights that the positive impact of immigration for the fortunes of the centre-right alliance was more pronounced in areas characterized by high incidence of unskilled labour and by high political competition. Finally, it shows that the gain of votes for the centre-right coalition went hand in hands with a loss of votes for the centre-left parties (and the smaller parties, except those at the extreme right) and a decrease in voter turnout. JEL Classification: P16, J61. Keywords: Immigration, political economy, voting. We thank Francesca Carta and Henri Overman. The views expressed in the paper are those of the authors and do not necessarily reflect those of the institutions they are affiliated with. 1

1. Introduction In Italy immigration has been a recent phenomenon. Still in 1998 the share of immigrants over natives, according to the Eurostat data, was as low as 1.7% (it was 9% in Germany, 5.6% in France). Starting from 1998, however, the country experienced consistent inflows, reaching 8% in 2012, while in Germany and France the shares of immigrants has remained pretty constant. As reported by a national newspaper, 1 among the first ten family names registered in 2012 at the birth office of Milan, three of them are originally from China, the second most recorded being the Chinese Hu. The most recorded name remains Rossi, a typical native surname. Twenty-five years ago no foreign last name (among the fist ten) was recorded at the same office. The effects of immigration for the receiving country have now been extensively investigated. While a number of studies focus on labor market aspects (see, for instance, Friedberg and Hunt, 1995, Card, 2001, Ottaviano and Peri, 2008, Pischkle and Villoso, 1997), others highlight additional channels trough which the presence of immigrants might impact on the host economy, such as the cultural environment (Ottaviano and Peri, 2006), the extent of crime (Moehling and Piehl, 2007, Bianchi et al 2012). Only a few papers deal with the role of immigration for the political preferences of the natives (see: Halla et al, 2012 and Gerdes and Wadensjö, 2008). This lack of attention is quite surprising given that immigration is often related with strong views held by the native population. It seems therefore reasonable to suspect that the political choices of the residents will reflect their opinions with regard to immigration. This paper analyses the role of immigration for the Italian parliamentary elections of 2001, 2006, and 2008. Two of them (2001 and 2008) were won by the centre-right coalition (headed by Mr. Silvio Berlusconi). The election of 2006 was instead won by the centre-left coalition, captained by Mr. Romano Prodi. With respect to immigration, the political platforms of the two coalitions were, consistently over the three elections, very different. In short, the centre-left alliance had a more open stance, stressing the importance of the immigrants for the prospect of the domestic economy, the duty of solidarity for an high-income country, and the benefits of a multi-ethnic society; on the other hand, the political program of the centre-right coalition had a less liberal stance: immigration was considered to be beneficial insofar it was appropriately regulated (allowing access to the country only to the immigrants who have previously found a job). The emphasis was more on the social problems (say, crime) related with immigration and the threat that people with different background could pose for the domestic way of life. Moreover, in all elections the centre-right coalition included the Lega Nord party, whose political stance was not immune from racial prejudices. We are able to identify the impact of immigrants on the political choices of the residents by comparing the voting pattern of the electorates in municipalities differently exposed to migration flows. We refer to about 8,000 small territorial units (the municipalities). This should deliver substantial benefits in the empirical strategy, as the attitudes toward migration is likely to depend on the proximity between natives and immigrants, for instance because there could be competition for 1 See: http://milano.repubblica.it/cronaca/2012/04/15/news/fra_i_cognomi_pi_diffusi_a_milano_il_cinese_hu_scalza_sciu r_brambilla-33351789/ 2

the public services provided locally or a native bias in sharing residential areas. Therefore, national or regional-level data mask much of the important variation that is needed to identify causation. To deal with endogeneity issues we use the (now standard in the literature on immigration) instrument proposed by Card (2001), which exploits the fact that immigrants tend to move to area where it is already present a group of immigrants with the same ethnicity. The identifying assumption is that local economic shocks that attracted immigrants in the past are uncorrelated with present political preferences, conditional on the full set of controls. Our outcome is taken to be the voting pattern at the elections held at the national-level. The advantage of using a national voting context is that the political platforms of the two competing coalitions with respect to immigration are clearly identifiable. Moreover, they do not reflect local circumstances as it might happen for local elections. Finally, the electoral rule at the nation-wide level are the same all over the country (while the rules for the local elections differ according to the size of the municipality). Our results are striking. In our preferred specification, with municipality fixed effects and instrumental variable estimates, we find that a 1% increase in the share of immigrants in a municipality entails a 1.2% increase in the share of voting going to the centre-right coalition. The effect also varies very little between the elections of 2001 and those of 2006 and 2008, which were conducted with different electoral rules. The study also highlights that the positive impact of immigration for the fortunes of the centre-right alliance was more pronounced in areas with low share of skilled workers. This suggests that in those areas immigration is perceived as a threat by the natives in the labour market, relatively more intensively than in territories associated with higher share of graduates. This might be due to the fact that in unskilled areas anti-immigrants views find a more fertile soil, probably because the labor market benefits for the natives struggle to materialize while the pressure on the supply of locally provided services increases. We also find that the gain due to immigration for the centre-right coalition is magnified in the municipalities characterized by greater political competition. This occurrence highlights that impact of immigration on political preferences might even be larger than that estimated: in districts where who was going to win was not an issue, the perceived small effect that the individual vote had on the nation-wide immigration policy might have discouraged voters to support the centre-right platform even if in agreement with the proposed stance. Finally, we show that the gain of votes for the centre-right coalition went hand in hands with a loss of votes for the centre-left parties, as expected, and that the results for the centre-right coalition are not driven by vote dynamics of the extreme right party. The paper is structured as follows. The next section briefly summarizes the scant literature on the subject. It also digs a little bit into related studies, mostly in the political science arena. Section 3 illustrates with reference to immigration policy the contents of two competing platforms at the political elections on which we focus in the empirical section. Section 4 explains the data. The identification strategy is illustrated in Section 5 and the empirical results are presented in Section 6. The last section offers some speculative implications of the findings and suggestions for future research. 3

2. Literature review To the best of our knowledge, only two papers deal with the impact of immigration on political preferences of the natives, Halla et al. (2012) and Gerdes and Wadensjo (2010). Halla et al. (2012) investigate the Austrian case, analyzing whether immigration dynamics positively affect the votes for the Freedom Party of Austria (FPO), a party with a clear anti-immigration stance. The FPO has increased its share of votes from less than 5 percent in the early 1980s to 27 percent by the year 1999. Halla et al. (2012) make use of historical settlement patterns of immigrants, in 1971, as a source of exogenous variation in recent spatial distribution of immigrants. Their baseline 2SLSestimate suggests that a one-percentage-point (one standard deviation) increase in the share of immigrants in a municipality increases the percentage of FPO votes in general elections by about 0.4 percentage point (a quarter of a standard deviation). They also show that results are stronger when focusing on the share of low- and medium-skilled immigrants. By contrast, high-skilled immigration has no impact on FPO votes. This is consistent with the hypothesis that high-skilled immigrants are perceived as being associated to better conditions also for the native population, while lower-skilled immigrants pose the greatest threat in the labour market. Their analysis is mainly in cross section, exploiting a pooled sample of six national elections at the municipality levels (from 1979 to 2002), using a time invariant instrument. Their also implement a robustness check in which they estimates the model in differences, investigating how changes in immigration affect changes in FTO, using anyway a time invariant instrument. In this framework, a one-percent increase in immigration rises the FPO votes by 1.41 percentage point, meaning that about two fifth of the increase in the FTO over time can be related to immigration. The second paper is Gerdes and Wadensjo (2010) that investigates how the unprecedented influx of refugees has affected votes for the main political parties at the municipality levels in Denmark. Their analysis covers a period that includes four local government elections and four general elections between 1989 and 2001. They find out that the shares of refugees is positively associated with the two main anti-immigration parties, the Danish People s Party and Progress Party. They also find mixed evidence for the other political parties: for instance they show that even a party with a proimmigration stance, the Socialist People s Party, gains from immigration, while the Liberal party, more on the centre, loses. They carry out OLS and fixed effect estimations. The latter are considered as their preferred specification since they allow controlling for unobserved municipality heterogeneity. Actually, using fixed effects, the relation between immigration and anti-immigration parties is stronger. The main drawback of Gerdes and Wadensjo (2010) of that they do not explicitly deal with endogeneity issues. They claim that the distribution of refugees was more or less random when conditioned on a number of structural and demographic factors and labour and housing market conditions, since from 1995 the policy of the government was to allocate refugees in all municipalities, and not only in medium-big cities, in order to share the influx burden. However, once refugees had been granted asylum, they were free to move wherever they wanted. To reduce mobility, incentives have been provided (housing and child day care). Nonetheless, studies examining the placement policy found that about half of the placed refugees chose to stay, while the other half moved further to another municipality, mainly to areas with a higher proportion of their countrymen. This evidence suggests that location of refugees cannot be assumed as exogenous, and hence estimates have to be taken with caution. 4

(.) 3. Immigration regulations and coalition platforms Given its position in the centre of the Mediterranean, and with over 8,000 kilometres of coastline, Italy is considered the most accessible entrance to Europe by migrants. For this reason, immigration has become an issue that strongly concerned the public opinion in the last decades. In 1999 a survey form Censis (an Italian public centre for socio-economic research) revealed that the fear of immigrants was the fourth item in the ranking of the worries of the Italian population (the first three were, in order, unemployment, mafia, and drugs). Also, almost an half of the household interviewed were convinced that the multiethnic society, toward which the country was leaning, increased the risk of social conflict. This Section documents that the two main political coalitions, that competed to run the country in 2001, 2006 and 2008, had two very different platforms as regard to immigration. The antiimmigration stance of the Berlusconi coalition has been one of the constant argument before political elections, also because of the presence in the coalition of both Alleanza Nazionale (National Alliance), headed by Gianfranco fini, which was the successor of the Movimento Sociale Italiano (MSI - Italian Social Movement) -the neo-fascist party in post-wwii Italy, and Lega Nord (Northern League), headed by Umberto Bossi, which has also taken the role that nationalistic parties usually play in the political arena of other European countries. Reconstructing the political debate in the parliament, the first attempt to regularize immigration goes back to 1990 through the Law 39/90 (the so-called 'Martelli law'). The main objective was to plan migratory flows in cooperation with the countries of origin. In 1998, the centre-left government headed by Romano Prodi issued the Law 40/98 (the so-called 'Turco-Napolitano law') which was mainly an attempt to regularize the position of non-eu immigrants and improve their integration, introducing regulations in the areas of fundamental rights, such as employment, health and education. The three main goals of the Turco-Napolitano were to implement more effective planning to deal with persons entering for employment reasons, preventing illegal immigration, and integrating foreign citizens who had legal residence permits. The law was considered a liberal, and even proimmigrant law compared with the closed-door policies of other European countries at that time. The law included equal rights for workers, recognition, and preservation of the language and cultural heritage of immigrants, and even provided for the eventual participation of these immigrants on a local level. The Turco-Napolitano set anti-discrimination norms within the framework of entry and residence rules, forbidding direct and indirect discrimination on the basis of race, ethnic origin, nationality, or religion. An important feature of the Turco-Napolitano law was the identification of procedures for the deportation of illegal immigrants who, once arrested by the police, could eventually be deported only after their case was judged by a magistrate. After the magistrate issued an order of deportation, the illegal immigrant had two weeks to appeal against the decision. One of the critique for this law was actually that in many cases immigrants often used this time to go underground. This issue was criticized by the centre-right coalition during the campaign for the May 2001 general election. Just after having won the election, the centre right coalition headed by Mr. Berlusconi passed the Law No. 189 of 30 July 2002, known as the 'Bossi-Fini law', the leaders of the two parties with a more accentuated anti-immigration stance within the centre-right coalition. This law amended the 1998 immigration law and introduced some new stringent clauses. According to the Bossi-Fini 5

law, each year, before 30 November, the Prime Minister have to lay down the number of non-eu workers who can be admitted in the country in the next calendar year. In order to require a residence permit, it needs a 'residence contract' (contratto di soggiorno) i.e. a contract of dependent employment signed by an employer (a firm or a family) and the immigrant worker. When the contract expires, the immigrant worker must either renew it or return to the country of origin. Residence permits issued for employment reasons can last for a maximum of two years, even if the worker has an open-ended contract of employment. The new law also addressed issues of family reunification. Only non-eu immigrants with a regular residence permit can be entitled to ask for reunification. Permits have to be renewed when the old one expiries. After six years of regular residence in Italy, non-eu citizens will be able to ask for a form of permanent permit instead of a temporary residence one. Furthermore, with this law Italy has been one of the first European country to require immigrants asking for residence permit to provide their fingerprints. Another important difference with the previous law is that in case of deportation, it will be immediate and will not be suspended even if the immigrant appeals to the courts. Further, if illegal immigrants return to Italy, they will be arrested and tried by the courts. For immigrants who were already in Italy in 2002, a decree-law issued on September 2002 provided the terms for regularization for two types of irregular immigrant workers, those employed as domestic workers and home-helper, and those employed as dependent workers. Immigrants whose residence permits have expired will also be able to regularize their situation, provided that they have not received a deportation order. The anti-immigration stance of the centre-right coalition can be detected in other government decisions. For instance, in 2008 the Berlusconi s government issued an emergency decree that allows volunteers in organized groups to patrol and identify suspicious individuals to report to the police. Only an amendment of the centre-left opposition Democrats managed to prevent the patrols from being armed and to limit their activities to the reporting of suspicious activity. In these years Romanian were under the spotlight, especially because of specific criminal episodes, so much that future Prime Minister Silvio Berlusconi in 2007 menaced to ban Romanian workers from Italy (The International Herald Tribune, November 4, 2007) and then, once in power, to call for expulsion of groups of migrants (The Economist, January 29, 2009). That gained the Italian government a reprimand from the EU, but the move was very popular amongst the Italian public opinion. It is interesting to note that both for the 2001 and 2008 elections the centre-right alliance focused strongly on immigration in the electoral campaign, and that just having won the elections the coalition had to propose in the short run, in 2002 and in 2009, some legislative measures to accomplish the expectations of voters. It is also worth underlying that even centre-left parties had to tackle with issues related to feeling of insecurity with respect to immigration. For instance, the candidate of the centre-left coalition for general election in 2008, Walter Veltroni, declared repeatedly during the campaign in that public order was a universal right, whose protection was beyond partisan politics (Corriere della Sera, 26/III/2008). Apart from the official documents, mainly national political programs and laws, the different stance between centre-right and centre-left has been even more evident in everyday life, in interviews in newspaper and television, in political meetings etc. 6

4. Data Data on election outcomes at the municipality levels are taken from the Atlante storico elettorale Zanichelli, for the elections in 2001, 2006, 2008. Data on immigration flows by provinces are taken from the Italian Istitute of Statistics (ISTAT) and from the Ministry of the Interior. (TO BE WRITTEN IN MORE DEPTH) Immigrant residents in Italy, 1998-2011 7,000,000 6,000,000 5,000,000 4,000,000 3,000,000 2,000,000 1,000,000 0 Immigrants IT POP/10 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 [ ] [Table 1] 7

5. Empirical strategy We exploit the geographical variability of the distribution of immigrants as well as its variation over time. The estimating equation is: (1) where the dependent variable y mt is the share of voting for the centre-right alliance, 2 defined at the municipality level m for each of the three elections for which we have data (t = 2001, 2006, 2008). 3 Our variable of interest is, which is the share of immigrants over population; X at is a matrix including a number of controls taken at the local level. From 2001 Census, we use the following municipality-level controls: demographic structure (aging index), human capital (share of graduates), social capital (share of employees in the no profit sector), economic development (proxied by the employment rate), which are available at the beginning of the period and time-varying population density; we also include GDP growth rate at NUTS1-level to take into account local business cycle that may simultaneously affect both immigrants settlement and political preferences. λ a is a set of spatial fixed effects that are, depending on the specification, defined at a = region (NUTS2), province (NUTS3), or municipalities; finally, µ t includes time fixed effects and ε mt represents the regression error. Table 1 reports the descriptive statistics for the covariates. The OLS estimates of equation (1) can not be intended in a causal way. They might be flawed by the usual drawbacks. First, an omitted variables bias may be at work. For instance, an unobserved positive productivity shock affecting firms located in the municipality could lead to an increase of both labor demand for immigrants and political preferences for the centre-right alliance, to the extent that the latter better serves the need of the rich. This would entail a positive bias. Second, we are also worried that our results might reflect reverse causation, as the immigrants might refrain to go where live a bulk of people with views against multiculturalism. This would imply a negative bias. Finally, there could be also measurement error in the independent variables, as the tracking of the non-natives on the national territory is admittedly imperfect, both because of illegal immigrants and because those who leave omit to signal their departure. In this case the estimates would be biased downward. To address all these issues, we use instrumental variable estimation. Following Card (2001), we exploit the fact immigrants tend to move to areas where a group of immigrants with the same ethnicity has already settled in the past. The identifying assumption is that local economic shocks that attracted immigrants in the past are uncorrelated with present political preferences conditional on the full set of controls included in the equation (1). In detail, we construct our instrument as: 2 Note that the extreme right parties are included in the centre-right alliance, parties that however in Italy are negligible in terms of votes. Our results do not change when excluding the extreme right parties. 3 Note that in the 2001 there was a mixed electoral system (25% of members under a proportional rule and 75% under a majoritarian rule). We consider as dependent variable the party vote shares under the proportional system. In 2006 and 2008 elections the system become basically proportional with a winning bonus, and hence our dependent variable is the party share. 8

where is the share of immigrants from country c in municipality m in 1991 and Immigrants ct is the country-level number of immigrants from country c in year t (t = 2001, 2006, 2008), and N stands for the number of top foreign nationality in the host country. We set N equal to 15 nationalities: Albania, Romania, Morocco, China, Philippines, Tunisia, Serbia and Montenegro, Macedonia, Poland, India, Peru, Senegal, Egypt, Sri Lanka, Ecuador. 6. Baseline Results Table 2 displays the OLS results for specifications that include both the time-variant and timeinvariant controls described above. 4 Our dataset is a panel of roughly 7,850 municipality observed over the three election years (2001, 2006, 2008). Therefore, OLS estimates exploit both crosssection and time variability. In Column (1) the cross-section variability used to estimate the impact of immigration for the fortunes of the centre-right coalition is taken to be that observed within each NUTS2 area (20 regions, regioni). As we are comparing municipalities that are close to each other, it is likely that a number of confounding omitted variables are differentiated away. In the case of Italy, this seems to be a sensible strategy as the country is featured by large social and economic differences between regions and little discrepancies within regions. Estimates suggest a positive (and statistically significant) correlation between the share of immigrants and the percentage of voting going to the centre-right coalition. The magnitude of the effect is not negligible: a 1% increase in the share of immigrants is associated with an increase of 0.373% at the polls. Column (2) presents the results for an exercises that is more demanding: we introduce NUTS3 (103 provinces, province) fixed effects, therefore the variability used for the estimate is that coming from municipalities that share a very limited piece of land. 5 The estimated impact of immigration remains similar to that of Column (1), i.e. 0.296%, reinforcing our identification strategy. Finally, in Column (3) we take an even more conservative stance and include municipality fixed affects. In this case, we are killing all the cross-section variability and using only the time variation available in the data. In this case, identification comes from the overtime correlation between the variation in immigration and the variation in the share of the centre-right preferences, experiences by each municipality over the 2001-2008 period. 6 Notwithstanding the strong requirements, which imply that we might be 4 All specifications include time variant covariates, while the time invariant covariates (which are defined in 2001 at the municipality level) cannot be part of the specification of Column (3). Note that all estimates are weighted by using the log of voters in each municipality. 5 In Italy, there are 103 provinces, with an average population of 588,600 habitants, an average surface of 2,925 Kmq, and an average number of municipalities of 78.5. 6 Note that in this specification, a possible sources of bias might come from time-varying omitted at the municipality level that are not adequately captured by the time varying control that we have at that level of aggregation (density). 9

throwing away much of the variation needed to identify the impact, the estimated coefficient slightly reduces (0.293) while remaining highly significant. [Table 2] Table 3 includes the IV estimates. It comes out that the impact on the share of preferences for the centre-right alliance is now much larger (around 1%) and highly significant, so implying an overall downward bias in the OLS estimates. In column (1), including regional dummies, the coefficient is equal to 0.781; the coefficient does not change much when controlling for provincial dummies (0.771) in column (2). Interestingly, when using municipality fixed effect, exploiting the within variability, the coefficient is higher: an increase in 1% of the immigrant shares entails an increase in votes for the center-right party of 1.235%. We consider the latter as our preferred specification, since it allows controlling for unobserved municipality heterogeneity. Note that the instrument fits well the actual distribution of immigrants across Italy s municipalities: it enters with the expected sign and it is highly significant. In all specifications the F-statistics is higher than 10, meaning that our estimates do not suffer from issues of weak instruments. [Table 3] [Table 4] 6.1. Robustness checks. Estimates in Table 2 and 3 are weighted by the log of voters, since we want to take into account differences in municipality size. We did not use number of votes since so doing the estimates would be driven basically by big cities such as Rome and Milan. We are in fact concerned that weighting observations by the number of voters in 2001 gives too much importance to large cities, given the very large positive skewness of the distribution of population in Italian municipalities. In 2001 the mean value was 7,166, the median was 2,394, the 75 th percentile 5,788, the 90 th 13,009, the 95 th 22,811, the 99 th 70,188; the skewness equaled 41.8. As robustness check, we present in column (1), (2), (3) of Table 4 the IV estimates derived trimming the sample at the 99% of the distribution of the municipality population and weighting by number of voters. Basically, this sample excludes cityyear observations with more than 70,188 inhabitants. Results are basically confirmed, with coefficients being still close to 1 using region, province, and municipality dummies, respectively. 7 Columns (4), (5), (6) instead includes all municipalities, still weighting by the number of voters. In this case, estimates are positive and statistically significant only when using provincial dummies. This evidence suggests that in big cities the relation between increase in immigrants and votes for the centre-right coalition is much less pronounced, and because of the high associated weight it makes estimates not significant. This might be explained by the fact that in big cities it is easier to 7 We experiment extensively with different thresholds for the trimming (90%, 95%), and the results reported in the table differed only little. Results are available upon request. 10

have segmented neighborhoods, with natives and immigrants areas that are far away one another, in such a way reducing the impact of the possible negative sides related to proximity to immigrates (crime, competition of local public services etc), while having the chance to exploit the positive sides of immigration, such as maids and nannies. Another possible explanation might be due to the fact that in big cities there is a higher share of skilled workers, usually more in favor to immigration integration. [Table 4] Another robustness check takes into account the fact that in 2005 Italy s experienced a change in the electoral rules at the national elections (legge n. 270-2005, also named legge Calderoli). The new rules were approved by the coalition of centre-right, while the centre-left coalition and the other small parties strongly opposed to them. Basically, the law introduced a substantial winning bonus (the party/coalition with the highest number of votes could be overwhelmingly represented in the Parliament); a threshold (4% of the votes) for the smaller political entities in order to get seats, and the abolishment of the possibility to pick the candidate (voters were only allowed to choose the party/coalition). For our purposes, it is important to note that the approval of the new rules was interpreted as an attempt by the coalition headed by Mr. Berlusconi to maintain the power in the likely scenario of a substantial decrease of votes in the coming election of 2006. 8 Therefore a general theme of elections of 2006 was that of the democratic emergency, as the new rules were considered to be in favor of the incumbent. Even though the hot issues in 2006 campaign were different from those of 2001, the impact of immigration on the voting for the centre-right coalition is quite similar (Table 5) and so it was for the elections of 2008, held with the new rules approved in 2005 (and won this time by the coalition of Mr. Berlusconi). Note also that in such a case we cannot introduce municipality fixed effects, since there is no time variation in the data. Table 5 also shows that our results are not driven by a specific election, and that the immigration impact on centre-right votes are rather constant over time. [Table 5] 6.2 Heterogeneous impacts Recent research on the effect of immigration in the host economy has highlighted that important heterogeneities could be related to the degree of economic development of the area in which the immigrants settle. For instance, developed areas might have more likely an excess of demand for immigrant labor. 9 Additionally, they might increase the competition for the locally provided public services. Table 6 takes a first cut at this issue and provides sample-split estimates, for the immigration impact on municipalities with a share of graduates greater (lower) than the median national share. Using our preferred estimation, with municipality fixed effects and IV, we find that 8 The scenario was only in part accurate. In 2006 the centre-right coalition won and benefited from the winning bonus. 9 See, among others, Cortes and Tessada (2009) on the link between female labor market participation and immigrant maids and nannies. 11

the impact on centre-right votes is much stronger (1.708) in municipalities below the median of graduates, while it is still positive but not statistically significant in municipalities above the median. [Table 6] Another interesting refinement concerns the degree of political competition at the municipality level. We define as high (low) political competition a situation in which the difference between the first and second party vote share in the 1996 general elections is below (above) the median. We find that the impact of immigration for the former case is overwhelmingly higher than that for the latter. Table 6 shows that, controlling for municipality fixed effects and using IV, the impact of immigration is driven by municipalities where the political competition is high, while it is still positive but not statistically significant in municipalities with low political competition. This evidence highlights that the impact of immigration on political preferences might even be larger than that estimated: in municipalities where who was going to win was not an issue, the perceived small effect that the individual vote had on the nation-wide immigration policy might have discouraged voters to support the centre-right platform even if in agreement with the proposed stance. [Table 7] Finally, Table 8 extends the analysis considering as dependent variable the shares of votes of the other political party as well as vote turnout, using our preferred specification, i.e. IV with municipality fixed effects. In column (1) we separately consider the votes for the extreme-right parties, that were included in the centre-right coalition in previous estimates. Interestingly, we find that the coefficients is much smaller, i.e. 0.228 instead of 1.235, suggesting that in the Italian case immigration has boosted mainly the centre-right party, the one headed by Mr. Berlusconi. This contrast with the results obtained for Austria by Halla et al. (2012). As expected, the higher share for centre-right coalition comes at the expense of votes for the Centre, the Centre-left and the extreme left, even if only for the Centre-left party the coefficient is significant. As last remark, note that there has been a negative impact on voter turnout. Even if the coefficient is not statistically significant, it suggests that left-wing voters that were in some sense against immigration might have chosen not to vote instead of directly voting for the Centre-right coalition. 7. Conclusions Caveat: the impact is that of a country that has recently experience a surge in immigrants inflows. To the extent that the aversion to immigration decreases over time (because people feel less threatened or start to appreciate the benefits of immigration that might take time to materialize) our estimates might be on the upper bound. Implication: Time-consistency. The centre-right should have a platform against immigration (to attract votes); at the same time the coalition should refrain from taking actions that discourage immigration, if it wants to keep the electoral benefits documented in this paper. 12

Table 1. Descriptive statistics Variable Units Obs Mean Std. Dev. Min Max Centre-right wing share Percentage 2392 4 0.485 0.128 0.004 0.887 Share of immigr./population Percentage 2392 5 0.04 0.036 0 0.293 Sh. of empl. in the no p rofit sector Percentage 2392 5 0.019 0.045 0 0.936 Ag ing index Ratio 2391 9 1.836 1.564 0.22 41.5 Share of graduates Percentage 2392 5 0.044 0.021 0 0.366 Employment rate Percentage 2392 5 0.371 0.077 0.132 0.609 Population density (# pers./km 2 )/1000 23925 0.291 0.629 0.001 13.087 GDP growth rate At NUTS1 2392 5 0.019 0.014-0.002 0.052 13

Table 2. Immigration and voting for the centre-right coalition. OLS estimates Immigration 0.373*** 0.296*** 0.293*** (0.034) (0.031) (0.033) Social Capital -0.020-0.0386** (0.023) (0.020) Aging index -0.007*** -0.004*** (0.001) (0.001) Graduate share -0.349*** -0.243*** (0.050) (0.046) Employment rate -0.148*** 0.117*** (0.038) (0.034) Density/1000 0.000 0.001*** 0.001*** (0.000) (0.000) (0.000) GDP growth rate -1.070*** -1.076*** -1.081*** (0.075) (0.075) (0.074) Constant 0.577*** 0.470*** 0.503*** (0.015) (0.018) (0.005) NUTS2 fixed effects YES NO NO NUTS3 fixed effects NO YES NO Municipality fixed effects NO NO YES Observations 23,579 23,579 23,579 R-squared 0.461 0.62 0.629 Notes. The dependent variable is the share of voting for the coalition of centre-right at the political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. Time invariant controls (demographic structure, population skill level, social capital endowments, labor market conditions) defined at the level of municipality are included in Columns (1) and (2); they are excluded from Column (3) as in this specification includes municipality fixed effects. Time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by ln(voters) in 2001. Robust standard errors clustered at municipality level in italics *** significant at 1% ** significant at 5% * significant at 10%. 14

Table 3. Immigration and voting for the centre-right coalition. IV estimates. (1) (2) (3) Immigration 0.781*** 0.771*** 1.235** (0.246) (0.255) (0.526) Social Capital -0.022-0.040** (0.023) (0.020) Aging index -0.006*** -0.004*** (0.001) (0.001) Graduate share -0.384*** -0.292*** (0.055) (0.054) Employment rate -0.181*** 0.085** (0.041) (0.038) Density/1000-0.001-0.009*** 0.013 (0.002) (0.002) (0.051) GDP growth rate -1.035*** -1.036*** -1.001*** (0.079) (0.079) (0.089) Constant 0.581*** 0.481*** (0.015) (0.019) NUTS2 fixed effects YES NO NO NUTS3 fixed effects NO YES NO Municipality fixed effects NO NO YES F-stat Excluded instruments 10.643 10.230 27.663 Observations 23,579 23,579 23,579 R-squared 0.461 0.62 0.629 Notes. The dependent variable is the share of voting for the coalition of centre-right at the political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. The instrument is explained in the text. Time invariant controls (demographic structure, population skill level, social capital endowments, labor market conditions) defined at the level of municipality are included in Columns (1) and (2); they are excluded from Column (3) as in this specification includes municipality fixed effects. Time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by ln(voters) in 2001. Robust standard errors clustered at municipality level in parenthesis. *** significant at 1% ** significant at 5% * significant at 10%. 15

Table 4. Immigration and voting for the centre-right coalition. Trimmed samples. IV estimates. Trimmed at 99th All municipalities (1) (2) (3) (4) (5) (6) Immigration 1.016*** 1.213*** 0.965* 0.212 0.975** -0.582 (0.226) (0.268) (0.511) (0.343) (0.399) (0.622) Social Capital -0.023-0.051* -0.066-0.087*** (0.039) (0.029) (0.040) (0.030) Aging index -0.019*** -0.015*** -0.022*** -0.016*** (0.002) (0.002) (0.003) (0.002) Graduate share -0.238*** -0.151** -0.155** -0.241*** (0.069) (0.060) (0.069) (0.077) Employment rate -0.273*** -0.031-0.173*** 0.028 (0.055) (0.053) (0.066) (0.054) Density/1000-0.004*** -0.008*** -0.001-0.007*** -0.009*** 0.062 (0.002) (0.002) (0.026) (0.002) (0.002) (0.042) GDP growth rate -0.911*** -0.887*** -0.869*** -1.017*** -0.881*** -1.210*** (0.115) (0.116) (0.130) (0.201) (0.235) (0.268) Constant 0.629*** 0.546*** 0.620*** 0.540*** (0.022) (0.027) (0.025) (0.028) NUTS2 fixed effects YES NO NO YES NO NO NUTS3 fixed effects NO YES NO NO YES NO Municipality fixed effects NO NO YES NO NO YES F-stat Excluded instruments 4.060 3.754 11.099 7.346 5.376 26.785 Observations 23,558 23,558 23,550 23,797 23,797 23,795 R-squared 0.451 0.603 0.609 0.485 0.630 0.635 Notes. The dependent variable is the share of voting for the coalition of centre-right at the political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. The instrument is explained in the text. Time invariant controls (demographic structure, population skill level, social capital endowments, labor market conditions) defined at the level of municipality are included in Columns (1) and (2); they are excluded from Column (3) as in this specification includes municipality fixed effects. Time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by voters in 2001. Robust standard errors clustered at municipality level in parenthesis. *** significant at 1% ** significant at 5% * significant at 10%. 16

Table 5. Single election results. IV estimates Election of 2001 Election of 2006 Election of 2008 Immigration 0.794*** 0.616** 0.613** 0.653*** 0.919*** 0.893*** (0.276) (0.256) (0.243) (0.253) (0.320) (0.345) Social Capital -0.018-0.036-0.023-0.040** -0.024-0.043* (0.026) (0.022) (0.022) (0.020) (0.025) (0.022) Aging index -0.005*** -0.002* -0.005*** -0.004*** -0.008*** -0.006*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) Graduate share -0.283*** -0.175*** -0.263*** -0.183*** -0.606*** -0.504*** (0.056) (0.053) (0.054) (0.054) (0.066) (0.067) Employment rate -0.271*** 0.077** -0.098** 0.121*** -0.176*** 0.060 (0.040) (0.036) (0.042) (0.040) (0.053) (0.051) Density/1000 0.001-0.010*** -0.002-0.008*** -0.000-0.007*** (0.002) (0.002) (0.001) (0.002) (0.002) (0.002) GDP growth rate 3.732*** 2.128 8.495*** -4.614*** 10.792*** -0.746 (0.443) (3.595) (0.702) (1.663) (0.949) (3.551) Constant 0.484*** 0.429*** 0.434*** 0.386*** 0.446*** 0.456*** (0.020) (0.102) (0.015) (0.022) (0.017) (0.053) NUTS2 fixed effects YES NO YES NO YES NO NUTS3 fixed effects NO YES NO YES NO YES F-stat Excluded instruments 4.849 4.657 8.511 8.077 16.891 16.120 Observations 7,937 7,937 7,929 7,929 7,931 7,931 R-squared 0.323 0.577 0.413 0.584 0.400 0.562 Notes. The dependent variable is the share of voting for the coalition of centre-right at each single political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. The instrument is explained in the text. Time invariant controls (demographic structure, population skill level, social capital endowments, labor market conditions) and time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by ln(voters) in 2001. Robust standard errors clustered at municipality level in parenthesis. *** significant at 1% ** significant at 5% * significant at 10%. 17

Table 6. Immigration and voting for the centre-right coalition. Refinements. IV estimates. Tertiary education Political competition Low High Low High Immigration 1.708** 0.469 1.057 1.358** (0.722) (0.350) (0.895) (0.655) Density/1000-0.013 0.023 0.006 0.024 (0.075) (0.028) (0.133) (0.043) GDP growth rate -0.990*** -0.620*** -0.933*** -1.095*** (0.121) (0.153) (0.161) (0.122) NUTS3 fixed effects YES YES YES YES F-stat Excluded instruments 18.694 12.399 14.899 14.392 Observations 19,794 4,001 11,910 11,885 R-squared 0.436 0.627 0.509 0.479 Notes. The dependent variable is the share of voting for the coalition of centre-right at the political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. The instrument is explained in the text. Only time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by ln(voters) in 2001. Robust standard errors clustered at municipality level in parenthesis. *** significant at 1% ** significant at 5% * significant at 10%. Table 7. Winners and looser from immigration. IV estimates (1) (2) (3) (4) (5) Extreme right Centre Centre-left Extreme left Turnout Immigration 0.228** -0.590-0.782** -0.149-0.581 (0.092) (0.487) (0.362) (0.163) (0.427) Density/1000-0.022** 0.045-0.029-0.004-0.180*** (0.009) (0.047) (0.035) (0.016) (0.050) GDP growth rate 0.160*** -0.553*** 1.548*** 0.049-0.816*** (0.030) (0.076) (0.073) (0.031) (0.095) Municipality FE YES YES YES YES YES F-stat Excluded instruments 27.663 27.663 27.663 27.663 27.662 Observations 23,795 23,795 23,795 23,795 23,796 R-squared 0.386 0.117 0.322 0.592 0.192 Notes. The dependent variable is the share of voting for the coalition of centre-right at the political elections of 2001, 2006, 2008. Immigration is defined as the share of immigrants over population. The instrument is explained in the text. Only time variant controls (population density and NUTS1 business cycle) are included in all specifications. Observations are weighted by ln(voters) in 2001. Robust standard errors clustered at municipality level in parenthesis. *** significant at 1% ** significant at 5% * significant at 10%. 18