For God s sake. Papal visit and European perception on the recent refugee crisis

Transcription

1 For God s sake. Papal visit and European perception on the recent refugee crisis Claudio Deiana, Gianluca Mazzarella and Elena Claudia Meroni February 1, 2018 Abstract The aim of this paper is to analyse the influence of media exposure on individual beliefs. We study the effect Pope Francis s visit to the Lesbo Island on April 16 th 2016 on the perception about the EU action toward migration of EU citizens. The papal visit to the island hosting thousand of refugees living in precarious conditions and his message about how Europe should make the effort to host and welcome these people was highly reported in the news and the media. We use data on perception taken by a special issue of Eurobarometer Survey and in a difference-in-difference framework exploit the fact that the Pope visit to the Lesbo Island occurred exactly in the middle of the Eurobarometer interviews, therefore some respondents were exposed to the mediatic event and the Pope s message (those interview after April 16 th ) and some were not exposed. Our results point consistently towards the conclusion that individuals living in Catholic countries interviewed from April 16 th onward are more likely to report that the EU is not doing enough on the migration issue. Keywords: Perception, Migration, Papal visit JEL codes: D83, R23 PRELIMINARY DRAFT. We thank seminar participants at the European Commission seminar series (2017). Opinions expressed herein are those of the authors only. They do not necessarily reflect the views of, or involve any responsibility for, the institutions to which they are affiliated. Any errors are the fault of the authors. European Commission, Joint Research Centre, Directorate I - Competences, Competence Centre on Microeconomic Evaluation (CC-ME), Via E. Fermi 2749, Ispra (VA), 21027, Italy. Corresponding author: elena.meroni@ec.europa.eu;

2 1 Introduction In the last two years, hundreds of thousands of refugees, escaping war, violence and economic instability in their home countries (Middle East, Asia and Africa) arrived on the shores of the Greek Island of Lesbos after crossing the Aegean Sea from Turkey. Hundreds of people drowned while attempting this sea crossing. Rapidly, Greece has become a gateway to Europe for thousands of refugee, especially Syrians, fleeing the war in their homeland. To alleviate this tragedy, on 20 March 2016, the EU signed a deal with Turkey which shut the doors of Europe and decreed that refugees arriving in Greece would now be sent back to Turkey if their asylum claims were rejected. 1 The closure of the so-called Balkan route drastically decreased (by 80%) the influx of refugees entering Europe trough Greece. 2 Nevertheless, the situation has been worsened by the impact of the EU-Turkey deal that stated for every Syrian refugee on the Greek islands who is returned to Turkey, a Syrian asylum seeker in Turkey will be resettled in Europe. The Greek authorities have struggled to implement the deal on the ground though, meaning many migrants who arrived on the islands after 19 March have ended up being detained for months, rather than days. Thus, camps which had initially been designed for short-term use and set up haphazardly in often remote locations without electricity, sewage and water access, suddenly had long-term residents. Perhaps as important, the ongoing refugee crisis poses many critical questions on how the EU has managing this dramatic status quo and how citizens perceptions and their beliefs are changing accordingly. To do so, in the context of Pope Francis visit to Lesbos Island (Greece) on April 16 th 2016, we study the causal impact of His speech on the perception of EU citizens about the EU action in the field of migration. We take advantage of the Special Eurobarometer survey of the European Parliament that was conducted in the 28 Member States (MS) of the European Union, from 9th to 19th April 2016 that focuses on the perceptions and expectations Europeans have regarding the EU s action, the fight against terrorism and to the mutual defence clause. We exploit the unintentional timing of the interviews in a novel research design to show the causal effect of persuasion during the Pope Francis s visit on European citizens perception and beliefs of EU action about the recent refugee crisis. 1 The facts on the EU-Turkey agreement are available at this link MEMO _en.htm. 2 A total of 173,450 refugees and migrants arrived by sea in Greece between January and December 2016, which is 80% lower than in the same period in 2015 (856,723). The number of people arriving in Greece in December 2016 (1,665) decreased by 97% compared to January (67,415) and by 98% compared to December 2015 (108,742). Arrivals also decreased by 16% compared to November 2016 (UNHCR, 2016). 1

3 Using a difference-in-difference set up, we provide novel evidence on the emotional reaction to a persuasive message related to Catholic Church doctrine as embodied in Papal speech and European citizen change in perception on EU actions in response to the visit. It appears clear that individuals beliefs are relevant to explain individual actions: voters, consumers, donors, investors and workers that decide on political candidates, products, charitable projects, investments and jobs based on their beliefs. For instance, recent political campaigns have been dominated by the anti-refugee sentiment which has been recognized as an important factor that causes a dramatic increase in the share of votes for the far-right parties across Europe (Steinmayr, 2016; Dustmann et al., 2016; Sekeris and Vasilakis, 2016; Becker and Fetzer, 2017; Bratti et al., 2017, among the most recent). The choice to study these questions in the context of the Papal visit to Lesbos Island is guided by two factors. First, the perception on migration related issues were salient in Papal speeches in this visit, with recurring themes being: (i) the condemnation on how the refugees are treated by the EU institutions; (ii) the importance of the current refugee crisis that are acerbating xenophobic and nationalistic sentiments. In the most recent political elections, the right-wing populist parties won 12.6% of the votes and 24.0% in Germany and Austria, respectively. In both cases, these parties disagreed to welcome the asylum seekers while they raised anti-refugee slogans and held protests against what they deem the Islamization of Europe. In this respect, we focus on three research questions about salient migration matters: (i) does the exposure to Pope Francis s message affect the perceptions of Europeans on EU action about migration?; (ii) Does this message persuade the beliefs of particular type of individuals? (iii) If so, what are the driven characteristics of the subgroup of people more exposed to the persuasive message? To address the effect of persuasion on key relevant issues related to EU actions, we exploit the Special Eurobarometer 2016 survey that covers a wide range of issues focusing primarily on individual perceptions of the EU, its main challenges and the European integration in general. The issue salient in Papal speech was about the treatment of refugees and migrants and the five-hour visit was exactly meant to call attention to the crisis. Therefore, we define the treated group those individuals coming from a country whose official religion is Catholic. Taking advantage of the accidental timing of the Papal event in Lesbos Island and the Special Eurobarometer, we define the post period if the interviews are held from April 16 th onwards. Indeed, their interaction measures the short run causal impact of a persuasion message on beliefs shift. The results show that that being exposed to Papal message significantly increases the insufficient perceived extent on EU actions about migration. We show that this is a short term effect due to a 2

4 temporary setting of the issue at the top of people s minds. Our estimates are similar across numerous robustness checks including the definition of alternative specifications of the indicator of exposure to Papal message, the addition of various time fixed effects that are meant to control for the occurrence of aggregate shocks and the running of a different placebo exercise To the best of our knowledge, the closest related paper to our contribution is by Bassi and Rasul (2017) on the Papal influences on fertility related beliefs and behavior. In particular, the authors study the effect of Pope Jonh Paul II s visit to Brasil in 1991 on fertility beliefs. Similar to our work, they exploit the fact that part of population was surveyed before and part after the papal visit. His speech contained explicitly persuasive messages related to fertility choice so that they are able to measure how persuasion shifts both short-run beliefs such as intentions to contraceptive and long-term fertility outcomes, such as the timing and total number of births. Their results point in the direction of large immediate impact on beliefs, impact on timing of fertility but not in the total fertility rate. Differently, we study the shift in perceptions on EU action in managing the current refugees crisis during the casual Papal visit to Lesbos Island. We do not focus on a single area within country as in Bassi and Rasul (2017) since we provide empirical evidence on this phenomenon across Eu countries. Our study is related to a recent strand of the literature that investigates the influence of media exposure on individual beliefs and behaviours. Several recent works have documented an impact of the media on individuals beliefs and choices as voters (Kearney and Levine, 2015; DellaVigna and Kaplan, 2007; Gentzkow, 2006), on the fertility decisions (Kearney and Levine, 2015; Ferrara et al., 2012), on the divorces (Chong and Ferrara, 2009), on crime (Dahl and DellaVigna, 2009) or on the perception of corruption (Rizzica and Tonello, 2015). However, the only paper focusing on the impact of media exposure on beliefs and perception about the migration issue is the one by De Poli et al. (2017). They estimate the effects of intensified media reporting on refugees drowning in the Mediterranean on individual xenophobic attitudes, elicited via a randomized survey experiment employing a version of the trolley dilemma. Their strategy combines a randomized survey experiment with variation in interview timing. They find that increased issue salience of asylum-seekers drowning in the Mediterranean reduced negative attitudes towards foreigners by 2.2 percentage points. 3

5 2 Pope visits to Lesbos and On April 16 th 2016 Pope Francis visited Lesbos Island with the main purpose of meeting the refugees living in the Moria camp. 3 The trip lasted barely five hours, began on the asphalt of the island airport (10:00 am), moved to a prison-like camp, stopped at a harbour and ended on the runway again (3:15 pm). In particular, the visit consisted in a welcome ceremony, a visit to the Moria refugee camp together with His Holiness Bartholomew and His Beatitude Ieronymos (the ecumenical patriarch of Costantinopolis and the archbishop of Athens and of all Greece), where they held a speech and had lunch with the refugees. The Pope s visit comes at a pivotal time for Greece and sends a clear message on the Vatican s stance: Refugees are welcome in Europe. Francis Pope s travel has the official aim to draw the world s attention to the current grave humanitarian crisis in the Mediterranean Sea. In this respect, His Holiness states that immigrants are not numbers, but people, faces, names and stories and He strongly encourage the political leaders to employ every means to ensure that individuals enjoy the fundamental right to live in peace and security. 4 Figure A.1 show the words used by the Pope in the speeches held during that day. What it is important to notice is that during the day Pope Francis, His Holiness Bartholomew and His Beatitude Ieronymos mentioned several time that they wanted to call attention to the current humanitarian crisis, and that political leaders and the public opinion should stop ignoring this crisis and that concrete actions should be taken to solve it. This is somehow different form the usual speeches held by the Pope, which most of the time are directed to the Catholic population and have a more religious content. 5 Importantly, the Pope s visit evoked a lot of mediatic attention: in most of the newspapers across EU his visit to Lesbos was reported as an important news. If we look at the google trend for searches of the words Pope Francis or Lesbos Island around the date of the visit, we notice a pick on the searches of these words on the Saturday 16 th and the following day. (See figure A.2) 3 Data Data on individuals perception about EU action in the field of migration are taken from the Special Eurobarometer survey Europeans in 2016: Perceptions and expectations, the fight against terrorism 3 Moria reception centre on Lesbos Island, which is one of the place that Pope Francis visited, has a capacity of 1,500 people but at that time there were more than 4,000 refugees because of huge bottlenecks in the asylum process. 4 More details on https : //press.vatican.va/content/salastampa/en/bollettino/pubblico/2016/04/16.html. 5 In the Appendix C we report parts of the speeches where these concepts were mostly expressed. 4

6 and radicalisation which was held in Europe between April the 9 th and the 19 th. This survey was the first Eurobarometer asking directly questions about Europeans perception about such topics. The item we focus on is the question about what citizens think about the EU action related to migration, and they can reply stating that the EU action is excessive, right or insufficient. Our main specification we build a dummy which takes value 1 if the reply is The EU action on migration issue is insufficient. The interpretation of this reply is twofold: EU action could be considered insufficient either because citizens think that the EU is not doing enough to help the refugees once they get in Europe, or because they think the EU is not doing enough to prevent that migrants arrive in the first place. Thus, the question doesn t reflect a positive or negative perception on migration, it doesn t reveal whether Europeans are in favour or against migration, it just represents a sort of non-satisfaction of the respondent towards what the EU is doing. Our idea is that the reply to this question can change if individuals are exposed to massive news related to the issue of migration, and therefore more individuals should be more prone to reply that the action is not sufficient, no matter what is their opinion about migration in general. 4 Empirical strategy We exploit the fact that the Pope s visit to Lesbos Island is fortuitously in the week of the Eurobarometer survey, which implies that some individuals are randomly interviewed in the survey before the Pope s visit and others after Francis Pope s message. We expect that the individuals replying after the visit are more likely to modify their reply on the issue regarding migration, and we expect the effect to come from countries which are Catholic, and thus should be more likely to listen to what the Pope says and anyway are more exposed in the media to the actions of the Pope. We address some potential composition and time bias using a difference-in-difference strategy, where the individuals that come from non-catholic country are used as a control group. Our main equation is the following: y ict = α 0 + α 1 After t + βafter t Cath c + γx ic + µ c + κ c t + ɛ ict, (1) where y ijt is a dummy taking value 1 if respondent i, in country c interviewed in time t says that EU action on the issue of migration is insufficient; After is a dummy taking value 1 if the interview is held from April 16 th onwards; Cath is a dummy for Catholic countries 6 ; X ic is a vector of pre- 6 We considered as Catholic countries whose main religion is Catholicism: Belgium, Austria, Hungary, France, Ireland, Italy, Luxembourg, Poland, Portugal, Czech Republic, Slovak Republic, Slovenia, Spain, Croatia and Malta. 5

7 determined individuals characteristics; µ c are countries fixed effects and ɛ ict is the error term. We also include country specific linear time trend, κ c t (day). 7 Standard errors are clustered at the day of interview level, following (Bassi and Rasul, 2017), as this clustering reflects that identification in our is based on time variation in the interview. The coefficient of interest is β, capturing the effect of being interviewed after the Pope s visit to Lesbos Island in Catholic countries. Thus variation come from overtime comparison among Catholic and non-catholic countries. The main identification assumption is that: the effect of the Pope is present only in Catholic countries, Catholic and non/catholic countries would have behaved the same before and after the Pope s visit in the absence of the visit. This can be tested by running placebo tests interacting the treatment effect with the days dummy variables to provide an estimate of the effect by day. We test for the existence of trends in the pre-adoption period using the following specification: m q y ict = α 0 + β τ After t τ Cath c + β +τ After t+τ Cath c + γx ic + µ c + ξ t + ɛ ict, (2) τ=0 τ=1 where ξ t are day of the interview fixed effect. The coefficients β ±τ estimates the impact of being in a Catholic country on a given day of the interview, including both days before the papal visit to Lesbo (from April 9 th to 14 th, τ = 1,..., 6) and after the visit (from April 16 th to 19 th, τ = 1,..., 4). If the common trend assumption holds, we should expect the coefficients from β 1 to β 6 to be not different from Results In Table 1 we report the main results of Equation 1. We start with a simple model where we only include the main variables of interest: being interview after the Pope visit, being in a Catholic country and their interaction. The coefficient associated to the interaction term is the main coefficient of interest and it is positive and significantly different from 0. This means that being interview in a Catholic country, after the Pope visit to Lesbo Island increases the probability of declaring that The countries in the control group are: Denmark, Estonia, Finland, Germany, Great Britain, The Netherland, Latvia and Sweden (Protestant) and Bulgaria, Greece, Romania, Cyprus (Orthodox). 7 In the results we show different versions of this main specification: one with the full set of controls; one with only countries fixed effects; one with only the main variables of interest: After, Cath and their interaction. The main control variables included in X ijt are: age, gender, type of community, level of education, marital status, number of children, and type of occupation. 8 Excluded category is the day before the visit, April 15 th. 6

8 the EU action on migration is insufficient. The results is stable to the inclusion of country fixed effects (column 2); control variables at the individual level (column 3) and country specific linear time trend (column 4). Table 1: Baseline results: the effect of exposure to Papal visit on European migration perceptions EU action on migration Insufficient (1) (2) (3) (4) After Pope visit*catholic countries ** ** ** *** (0.0131) (0.0126) (0.0122) (0.0086) After Pope visit ** ** ** (0.0156) (0.0082) (0.0080) (0.0114) Catholic countries *** (0.0075) Observations 23,344 23,344 23,344 23,344 R-squared Country FE Control variables Linear Trend Note. In the Table we report the effect of being interviewed after the Pope s visit to Lesbo on the perception about the EU action towards the migration issue. In column (1) we report the regression with no controls; in column (2) we add country fixed effects; in column (3) we add control variables; in column (4) we add country specific linear time trends. Robust standard errors in parentheses clustered for the day of the interview. * p<.10 ** p<.05 *** p<.01. As mentioned in Section 4 the identifying assumption is that trends in answer to the migration questions would be the same in Catholic and non Catholic countries in absence of the treatment (i.e. the Pope s visit to Lesbos Island). In order to check the common trend assumption we interact each day with the dummy for Catholic country as explained in Equation 2. The results are reported in Figure 1. We clearly see that before the visit all the coefficients are not different from 0, supporting our assumption. We also notice that the effect is positive and significant in the day of the visit and the day after (April 15 t h and 16 th ), suggesting the very short run effects of this messages, as it is found in this literature most of the time (seerizzica and Tonello (2015), for instance). We also try different specification including additional or different fixed effect. All the results are presented in Table 2. In column (1) we report the baseline results, as presented in column (4) of Table 1. In the first specification we include NUTS2 level fixed effects rather than country fixed (column 2), to better control for geographical difference within countries, and we see that results are still positive and significant, even thought the magnitude of the coefficient halves. In the following specifications we leave the country fixed effects, but we include other fixed effects. In particular in column (3) we add day of the interview FE, to better control for potential differences, common to all the sample, in answering to the survey according to the day of their 7

9 Figure 1: Placebo regression April 9 April 10 April 11 April 12 April 13 April 14 April 16 April 17 April 18 April 19 Table 2: Alternative specifications: different fixed effects EU action on migration Insufficient (1) (2) (3) (4) (5) (6) After Pope visit*catholic countries *** * *** *** *** *** (0.0086) (0.0086) (0.0087) (0.0089) (0.0089) (0.0098) Observations 23,344 23,344 23,344 23,344 23,344 23,344 R-squared Individual characteristics Linear Trend Country FE Nuts2 FE Interview day FE Day of the week FE Interview hour FE Interview day x hour FE Note. Robust standard errors in parentheses clustered for the day of the interview. In column (1) we report the baseline specification; the specification in column (2) uses NUTS2 level fixed effect rather than country fixed effect; the specification in column (3) adds to the baseline day of the interview fixed effect; the specification in column (4) adds to the baseline day of the week fixed effect; the specification in column (5) adds to the baseline hour of the interview fixed effect; the specification in column (6) adds to the baseline hour of the interview, day of the interview fixed fixed effect and their interaction; * p<.10 ** p<.05 *** p<.01. interview. In column (4) we add day of the week FE to capture differences in answering according to which day the interview was conducted. The main source of differences comes, in our opinion, from individuals answering on weekends rather than week days, which should also capture for example 8

10 potential differences in the characteristics of the samples interviewed in the weekend compared to the one interviewed during the weekday. In column (5) we add interview hour FE and in column (6) we add interview day, interview hour and their interaction fe. In all the specifications the results are equivalent to the main ones, both in terms of magnitude and in term of significance. Table 3: Heterogeneity effects Groups coeff st. err. (1) (2) Gender Male (0.033) Female *** (0.0192) Age (0.0567) *** (0.019) (0.0373) (0.0318) Household structure No child ** (0.0084) A Child (0.0272) More than 1 child *** (0.0454) Education Up to 15 years (0.0262) years ** (0.0192) 20 years and older (0.0277) Still Studying (0.0748) No full-time education (0.2083) Employment Not working ** (0.016) Self-employment (0.0748) Employed high skills ** (0.0637) Employed low skills (0.0199) Marital Married * (0.0211) Single without children (0.0282) Single with children (0.0486) Divorced (0.0549) Widowed (0.0772) Others (0.5337) Country FE Control variables Linear Trend Note. In the Table we report the effect of being interviewed after the Pope s visit to Lesbo on the perception about the EU action towards the migration issue. In column (1) we report the coefficient of each different groups and in column (2) the relative standard error. Robust standard errors in parentheses clustered for the day of the interview. * p<.10 ** p<.05 *** p<.01. Finally, we investigate the possible channels that shed some lights on the subgroups of individuals more affected by Papal message. We observe that the effect is concentrated for the female and age groups. The people in the working age population seems more affected which is confirmed by the the analysis for the different employment status although we find a strong statistical correlation for people not-working. This is possibly due to the fact that this category encompasses the inactive that are mainly female driven. Finally, we observe larger effect for household with 9

11 more than one child and married. 6 Robustness checks In this section we present some placebo we run to assess the robustness of our results. First, we change the definition of time and of group of countries; then we use other Eurobarometer Surveys to assess whether there are natural changed in responses happening a given number of days after the surveys start. In table 4 we first change the variable After to the two days preceding the Pope s visit, April 14 th and April 15 th. In columns (2) and (3) we see the absence of any significant results if we just pick a random date before to the Pope visit. Then we change the treated group of countries to Catholic and Ortodox, using as comparison group only the Protestant. Again we see absence of significant results, suggesting that the effect is driven by Catholic countries only. Table 4: Placebo on the definition of groups and timing EU action on migration Insufficient (1) (2) (3) After April 15th (0.0185) Catholic countries * After April 15th (0.0210) After April 14th (0.0177) Catholic countries * After April 14th (0.0138) After Pope visit (0.0157) Catholic and Ortodox countries * After Pope visit (0.0176) Observations 23,344 23,344 23,344 R-squared Country FE Individual characteristics Linear Trend Note. In the Table we report estimates changing the definition of after and grouping of countries. In column (1) after is define as after April 15 th, in column 2 after April 14 th. In column (3) we group Catholic and Orthodox countries. Robust standard errors in parentheses clustered for the day of the interview. * p<.10 ** p<.05 *** p<.01. As further a placebo check, we take four Eurobameter surveys held between 2014 and In each survey we define a placebo dummy (After) that replicates the number of days between 9 In details we use: 1) Eurobarometer 82.3 (2014), ; 2) Eurobarometer 84.3 (2015), ; 3) Eurobarometer 86.2 (2016), ; 4) Eurobarometer 87.3 (2017),

12 Table 5: Oster method Outcome Beta s.e. R 2 Delta Bound (1) (2) (3) (4) (5) After Pope visit*catholic countries [0.0375, ] Note. In the Table we report estimates using the method proposed by Oster (2017). the first day of the survey and the papal visit to Lesbos in the original data (7 days), we then create the variable Catholic countries comprising all the countries we use in the main analysis, and we run four regressions similar to one presented in table As in the four considered surveys there isn t the same variable used in the main regression (opinion on the EU action in the field of migration) we use another variable, common the these surveys: individuals are asked about the two most important issues facing the EU at the moment, and one of the possible replies is Immigration. So the main dependent variable used in the placebo regression is a dummy taking value one if the respondent mentioned Immigration as one of the two most important issues faced by EU at the moment. 11 In none of these specifications using other Eurobarometer surveys the coefficient of interest is statistically significant. 12 Finally, we adopt the method proposed by Oster (2017) to produce bias-adjusted estimates assuming the presence of unobservables and producing bounds to the coefficient of interest, in the presence of such omitted variables bias. We chose a Rmax = 1.3R 2 of the baseline regression, and calculate the bounds of the coefficient of interest, for δ = 0 - corresponding to the original estimates - and δ = 1 as an upper bound which correspond to the assumption of equal selection between observed and unobserved variables. We also calculate the value of δ such that the β would be 0. Estimations are done in Stata with psacacl. Results are reported in Table 5. We see that as the 0 is not included in the bound and as δ is greater than 0, we conclude that the estimates are stable to the inclusion of other unobservables which could be omitted in our main estimates. 10 So in the first dataset considered the dummy after takes value 1 after November 14 th. 11 The other possible answers were: crime, economic situation, rising price and cost of living, taxation, unemployment, terrorism,eu s influence in the world, the state of Member States public finance, pensions, the environment, energy supply, or climate change. 12 Results available upon request. 11

13 7 Conclusion We analyzed the perceptions of European individuals about the ongoing refugee crisis that poses many critical questions on how the EU has managing this dramatic status quo and how citizens perceptions and their beliefs are changing accordingly. Knowing how perceptions are shaped it is particularly important because they affect individual actions in many sensitive contexts, spannig from voting choices to work and investment decisions. In this paper, we exploited the furtuitous Pope Francis visit to Lesbos Island (Greece) on April 16 th 2016 to understand whether the EU individuals changed their perceptions about the EU action in the field of migration. This study extends the frontier of empirical evidence on persuasion because we address: whether the exposure to Pope Francis s message affect the perceptions of Europeans on EU action about migration; whether the message persuades the beliefs of particular type of individuals and finally what are the driven characteristics of the subgroup of people more exposed to the persuasive message. 12

14 References Bassi, V. and I. Rasul (2017). Persuasion: A Case Study of Papal Influences on Fertility-Related Beliefs and Behavior. American Economic Journal: Applied Economics 9 (4), Becker, S. O. and T. Fetzer (2017). Does Migration Cause Extreme Voting? Technical report. Bratti, M., C. Deiana, E. Havari, G. Mazzarella, and E. C. Meroni (2017, September). What Are You Voting For? Proximity to Refugee Reception Centres and Voting in the 2016 Italian Constitutional Referendum. IZA Discussion Papers 11060, Institute for the Study of Labor (IZA). Chong, A. and E. L. Ferrara (2009). Television and Divorce: Evidence from Brazilian Novelas. Journal of the European Economic Association 7 (2-3), Dahl, G. and S. DellaVigna (2009). Does Movie Violence Increase Violent Crime? The Quarterly Journal of Economics 124 (2), De Poli, S., N. Jakobsson, and S. Schüller (2017). The drowning-refugee effect: media salience and xenophobic attitudes. Applied Economics Letters 24 (16), DellaVigna, S. and E. Kaplan (2007). The Fox News Effect: Media Bias and Voting. The Quarterly Journal of Economics 122 (3), Dustmann, C., K. Vasiljeva, and A. P. Damm (2016). Refugee migration and electoral outcomes. CReAM DP 19, 16. Ferrara, E. L., A. Chong, and S. Duryea (2012). Soap Operas and Fertility: Evidence from Brazil. American Economic Journal: Applied Economics 4 (4), Gentzkow, M. (2006). Television and Voter Turnout. The Quarterly Journal of Economics 121 (3), Kearney, M. S. and P. B. Levine (2015). Media Influences on Social Outcomes: The Impact of MTV s 16 and Pregnant on Teen Childbearing. American Economic Review 105 (12), Oster, E. (2017). Unobservable selection and coefficient stability: Theory and evidence. Journal of Business & Economic Statistics,

15 Rizzica, L. and M. Tonello (2015). Exposure to media and corruption perceptions. Temi di discussione (Economic working papers) 1043, Bank of Italy, Economic Research and International Relations Area. Sekeris, P. and C. Vasilakis (2016, June). The Mediterranean Refugees Crisis and Extreme Right Parties: Evidence from Greece. MPRA Paper 72222, University Library of Munich, Germany. Steinmayr, A. (2016, March). Exposure to Refugees and Voting for the Far-Right: (Unexpected) Results from Austria. IZA Discussion Papers 9790, Institute for the Study of Labor (IZA). 14

16 Appendix A: Figures international good make wish build including care greek love dear assist godforced needs thank human way politicalhope violence for bartholomew solidarity generosity efforts alsoasylum first ieronymos may service among made lesvos the offer meansmeeting island since rather situation tragiccome patriarch this many difficulties together refugees christcan peace end need others even must greece religious dignity help you europe crisis bring great know humanity migrants muchcommunities archbishop wartoday one sisters humanitarian countries world flee christians conflict work people brothers express fundamental suffering rights Figure A.1: Words used in the Pope s speech Figure A.2: Google trend 15

17 Appendix B: Additional tables Table B.1: Differences in the sample before and after April 16 th - to be complete with starts in the 3 column + countries (1) after before ttesto/ttesto se b/se Age (0.264) (0.001) Female (0.007) (0.006) Rural area or village (0.007) - Small or middle town (0.007) (0.011) Large town (0.007) (0.021) Years of education: up to (0.005) (0.030) Years of education: (0.007) (0.028) Years of education: more than (0.007) (0.041) Years of education: still studying (0.004) (0.036) Years of education: no full-time education (0.001) - Marital status: Married (0.007) (0.062) Marital status: Single with partner (0.005) (0.047) Marital status: Single (0.005) (0.044) Marital status: Divorced (0.004) (0.059) Marital status: Widow (0.005) (0.071) Marital status: Missing / refusal (0.001) - Children in the household: (0.006) (0.008) Children in the household: (0.005) - Children in the household: 2 or more (0.004) (0.010) Occupation: Not working (0.007) (0.017) Occupation: Self-employed (0.004) (0.014) Occupation: Employed - higher skills (0.005) - Occupation: Employed - lower skills (0.007) (0.006) Constant (0.177) Observations

18 Appendix C: Parts of the Pope Francis speeches We have come to call the attention of the world to this grave humanitarian crisis and to plead for its resolution. As people of faith, we wish to join our voices to speak out on your behalf. We hope that the world will heed these scenes of tragic and indeed desperate need, and respond in a way worthy of our common humanity. World opinion cannot ignore the colossal humanitarian crisis created by the spread of violence and armed conflict, the persecution and displacement of religious and ethnic minorities, and the uprooting of families from their homes, in violation of their human dignity and their fundamental human rights and freedoms. From Lesvos, we appeal to the international community to respond with courage in facing this massive humanitarian crisis and its underlying causes, through diplomatic, political and charitable initiatives, and through cooperative efforts, both in the Middle East and in Europe. [...]we call upon all political leaders to employ every means to ensure that individuals and communities, including Christians, remain in their homelands and enjoy the fundamental right to live in peace and security. A broader international consensus and an assistance programme are urgently needed to uphold the rule of law, to defend fundamental human rights in this unsustainable situation, to protect minorities, to combat human trafficking and smuggling, to eliminate unsafe routes, such as those through the Aegean and the entire Mediterranean, and to develop safe resettlement procedures. [...]we urge all countries to extend temporary asylum, to offer refugee status to those who are eligible, to expand their relief efforts and to work with all men and women of good will for a prompt end to the conflicts in course. We urge the international community to make the protection of human lives a priority and, at every level, to support inclusive policies which extend to all religious communities. We must never forget, however, that migrants, rather than simply being a statistic, are first of all persons who have faces, names and individual stories. Europe is the homeland of human rights, and whoever sets foot on European soil ought to sense this, and thus become more aware of the duty to respect and defend those rights. [...]we need to encourage political efforts that are broader in scope and multilateral. It is necessary, above all, to build peace where war has brought destruction and death, and to stop this scourge from spreading. To do this, resolute efforts must be made to counter the arms trade and arms trafficking, and the often hidden machinations associated with them; those who carry 17

19 out acts of hatred and violence must be denied all means of support. Cooperation among nations, international organizations and humanitarian agencies must be tirelessly promoted, and those on the frontlines must be assisted, not kept at a distance. 18