European Citizens’ Attitudes on the Return of Refugees to Their Home Country: Results from a Survey in 13 EU Member States

European law requires member states of the European Union (EU) to accept refugees. It also allows them to return refugees to their countries of origin if the reason for asylum no longer exists. Both the reception and return of refugees has become a widely debated and controversial issue in many member states of the EU. Based on a survey conducted in 13 EU member states, we analysed whether citizens support the return of refugees when the cause for their displacement has become obsolete and how differences in attitudes toward the return of refugees can be explained. A clear majority of Europeans (70%) support the return of refugees. This is also the case for those who believe that their country should accept refugees in the first place. These results mean that existing law—the acceptance of refugees in need and the return when the cause becomes obsolete—is being supported by a majority of Europeans. However, there are considerable differences in approval rates among the countries. The results of multivariate analysis show that ideational factors—both on the micro and the macro level—influence attitudes toward the return of refugees, whereas structural factors, recurring to economic interest, do not contribute substantially to the explanation of attitudes toward the return. The political implications of these findings are discussed in the last section of the article.


Introduction
The number of refugees who seek protection in Europe has increased significantly between 2014 and 2016. 1 The escalation of the civil war in Syria pushed more and more people to abandon their home countries and try to find refuge in Europe. This development unleashed a controversial public debate in many European Union (EU) member states, focusing on different topics: first, whether and to what extent the EU member states are prepared to grant asylum to people who have been persecuted; second, the debate on an equal allocation of refugees among European member states; and third, the discussion about the return of refugees to their countries of 1 We use the term refugees for people who are covered in the directive 2011/95/EU and are defined as vulnerable persons (https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2011:337:0009:0026: EN:PDF; accessed 10 February 2020). On the one hand, these are persons who, according to the Geneva Convention and the defined criteria, enjoy subsidiary protection and most of the time had to leave their country because of war or civil war. In contrast to refugees, we can differentiate people who leave their home countries for economic reasons and want to settle in another country. We will call this group of people migrants. origin if their asylum application is rejected or the situation in their home country has changed so that the reason for being granted asylum has become obsolete. In many EU member states, we have seen a rise of right-wing or right-wing populist parties, which differentiate themselves from other established parties by nativist and excluding perspectives on migration (Mudde 2007;Norris and Inglehart 2019;Rydgren 2017). Accordingly, they do not approve of taking in refugees, do not want to see an equal allocation of refugees among the EU member states, and are in favour of a strict return policy. 2 These parties often claim to represent the attitudes of the majority of citizens. They understand themselves as their true representatives who express the citizens' negative attitude toward refugees.
A lot of research has been done on the attitude of citizens toward accepting refugees and also on the allocation of refugees in European countries (e.g. Bansak et al. 2016Bansak et al. , 2017Gerhards et al. 2016Gerhards et al. , 2019Steele and Abdelaaty 2018). We have shown in previous publications that an overwhelming majority of European citizens believe that refugees should be taken in if their home country was stricken by civil war (Gerhards et al. 2019). Even in Hungary, where people are very sceptical when it comes to refugees, the right to be granted asylum as codified in the Geneva Convention is supported by a clear majority.
But the European right of asylum also states that refugees can be returned to their home countries if their reason for being granted asylum has become obsolete. Accordingly, we understand the return of refugees as the right of states to return refugees whose residence permits have expired and were not extended. The states then have the right to send them back to their countries of origin in the event that the individuals do not want to return voluntarily. 3 If and to what extent European citizens support the return of refugees and which factors explain differences in citizens' attitudes toward the return of refugees is so far an unanswered research question that we will address with this paper. Our research builds on a survey carried out in 13 EU member states.
The European law that applies to all member states and their citizens constitutes the conceptual framework of our study. This law consists of two elements: the obligation of European states to take in refugees and the right of states to return refugees to their home country if the reason for seeking refuge has become obsolete and if they do not want to return voluntarily. We ask whether and to what extent citizens support the law and if the legal framework is thus supported by the citizens' belief in its legitimacy. Max Weber (1985) outlines in his sociology of domination 2 Some of these parties are the Finns Party in Finland, UKIP in the UK, the Danish People's Party, the National Rally (formerly known as the National Front) in France, Geert Wilders Party for Freedom in the Netherlands, the Alternative for Germany, the Freedom Party of Austria, Attack in Bulgaria, Fidesz in Hungary, Lega in Italy, and Law and Justice in Poland. 3 Unfortunately, only limited reliable data on the return rate in the different European countries exist. Eurostat publishes the number of people who are asked to leave a country, as well as the number of people who leave voluntarily or are forced to leave. We cannot, however, determine whether people left the country in the same year they were asked to leave. Accordingly, the return rates on the basis of these numbers vary among the countries examined, between 7% (Portugal and Hungary) and 108% (Germany) in 2016. There are also strong fluctuations between the years (https://ec.europa.eu/eurostat/web/asylumand-managed-migration/data/database; accessed 4 March 2020). that the legitimacy of an order depends on the citizens' belief in its legitimacy. In the first section, we will outline the existing legal framework in Europe to return refugees. In the second section, we will describe the factors that can explain citizens' attitudes toward refugees by referring to the relevant literature. We will differentiate between structural factors that are linked to the interests of people and countries and ideational factors. Section three provides an overview of our dataset and the methods applied. In the fourth section, we will present the results of the descriptive and multivariate analyses. A clear majority of all Europeans (approximately 70%) are in favour of returning refugees. Interestingly, this also applies to those who support taking in refugees in need. European law-taking in refugees in need on the one side and their return if the reason for being granted asylum has become obsolete on the other side-is supported by the majority of Europeans. At the same time, significant differences between countries exist. The results of the multivariate analyses show that ideational factors influence attitudes toward return both on the individual level and on the country level. Structural factors that refer back to the economic situation of the respondents and the country they live in do not help to explain the results. In the last section, we will summarise our results and discuss their political implications.

Legal Framework
European law serves as the conceptual framework of our research. Its sociophilosophical justification has already been formulated in Immanuel Kant's text "Vom ewigen Frieden: Ein philosophischer Entwurf" (Perpetual Peace: A Philosophical Sketch; Kant 2008Kant [1795). Kant believes that peace between states is not a given but must be promoted by a comprehensive legal framework. Three legal systems are necessary for this: civil law, which coordinates the relation of people within a state; law of nations, which codifies intergovernmental relations; and the so-called law of world citizenship (ius cosmopoliticum), which applies to all people worldwide and is especially important for our research question. The latter includes the right to visit other states without being treated in a hostile way. According to Kant, a visitor can be turned away only if his or her life is not threatened. This implies that people who seek refuge in another country because of a war in their home country or because they are being personally persecuted by the regime may not be denied access. Kant points out that this is a right to visit and not a right to stay in the sense of a permanent right of residence: "The Law of World Citizenship Shall Be Limited to Conditions of Universal Hospitality" (Kant 2008(Kant [1795, p. 357). 4 Therefore, a country is not obliged to host a persecuted person once his or her reason for persecution has become obsolete and he or she is no longer in danger in the home country. The right of residence granted by the cosmopolitan law is of a temporary nature.
Kant's idea of a cosmopolitan law and the question under which conditions persons in need of protection have to be taken in, as well as when they will have to return, is codified in different European directives. The most important legal document concerning the right of refugees to be accepted in another country is the directive 2011/95/EU. 5 This directive includes the criteria specified in the Geneva Convention. People who are persecuted because of their race, religion, nationality, political opinion, or membership of a particular social group have the right to refuge. The Geneva Convention does not, however, include persons who had to leave their country because of war or civil war. The directive determines that these persons as well are eligible for subsidiary protection. This protection has to be granted if a person faces torture or a death sentence in their home country or if his or her life is in danger because of an international or civil war. War and civil war refugees as well as politically persecuted persons therefore have a guaranteed right to be taken in by the European Union. During the asylum procedure, which varies from one country to another, it is determined whether a person has been persecuted. It depends on the country's legislation how long people are allowed to stay after their status has been verified. 6 The right to stay, however, is a temporary right linked to the reason of persecution in the home country. States may return refugees to their home countries if the reason for being granted refuge has become obsolete. This basic principle is also set out in the directive from 2008 on common standards and procedures in member states for returning illegally staying third-country nationals. 7 This directive specifies that the member states have to take action if persons do not hold a residence permit (anymore). This applies, for example, to refugees whose reason for seeking refuge in Europe has become obsolete. The member states are allowed to return people to their home countries if they adhere to certain procedures.
In general, European law complies with Kant's idea of a temporary right to visit. Nation states can, of course, extend the temporary right to visit or even grant citizenship for pragmatic reasons or because it is in a state's interest. This is set out in this directive as well. Granting permanent residence can be useful to promote integration into the host society if it is clear that the reasons for persecution will continue to exist. However, this does not affect the principle of temporality or the right to stay.
European law constitutes the reference point for our empirical analysis. We asked European citizens whether they think that refugees should be returned if the situation in their home country has improved and, therefore, the reason for asylum has become obsolete. Hence, we examined whether the existing legal norms were supported by the citizens' belief in the legitimacy of these norms. As described above, the 5 https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32011L0095; accessed 11 February 2020. The directive has not been adopted by the United Kingdom, Ireland, or Denmark. 6 For example, in Germany asylum seekers (according to Article 16a of the German Basic Law) and recognised convention refugees (according to the Residence Act) have a temporary residence permit of 3 years. They receive an unlimited settlement permit if the reasons for their recognition have not changed. Subsidiary beneficiaries of protection generally receive a limited residence permit of 1 year that can be prolonged by another 2 years. It must be verified that the reason for their protection still exists, e.g. that an armed conflict in the home country has not been resolved. 7 See https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32008L0115; accessed 27 February 2020. nation states decide whether the return of refugees is pursued and implemented or whether they allow refugees to stay after their status as refugees has expired. This is the difference between returning and accepting refugees, as the member states are legally obliged to take in refugees in need.

Factors that can Influence the Citizens' Attitudes Toward the Return of Refugees
To what extent do citizens support the idea that refugees are to be returned to their home countries once the reason for seeking refuge has become obsolete, and how can possible differences in attitudes be explained? These are the main questions we were trying to answer with our survey. We distinguish between structural and ideational factors that can have an influence on the individual and on the country level and are directly linked to attitudes toward the return of refugees. Two further factors exist that can influence people's attitude toward the return of refugees. Their stance toward taking in refugees is one of these factors: Respondents could believe that refugees should not be taken in in the first place even though they might legally be entitled to have access to a country. One can assume that people who hold this opinion also believe that refugees should be returned to their home countries. Furthermore, we assume that people who are in regular contact with refugees have a higher probability of opposing their return. This so-called contact hypothesis was developed by Robin M. Williams Jr. (1947) and Gordon W. Allport (1955). It states that intergroup contact under certain conditions can reduce prejudices. 8 While sporadic and superficial contacts tend to confirm prejudices, regular and close contacts reduce prejudices against other groups, according to Allport (1955) (see also Pettigrew 1998 andMcLaren 2003). 9 On the basis of the analysis of the German Socio-Economic Panel (SOEP), Eisenecker and Schupp (2016) have shown for Germany that around 10% of the respondents are in regular contact with refugees. In addition, there are people who are in contact with refugees in their social environment or at work. We expected that more contact with refugees would correlate with a higher rejection of returning them to their countries of origin.
When determining the structural and ideational variables that could explain citizens' attitudes toward the return of refugees, we relied on those factors that also influence the attitudes toward taking in refugees. This is logical because the consequences of not returning and taking in refugees may overlap. Refugees who remain in the host country or are taken in can, for example, be seen as a threat to the citizens' own sociostructural positions. We assume, however, that structural factors are 8 These conditions include equal group status within the situation, common goals, intergroup cooperation without competition, and support of authorities, law, or custom (Allport 1955). Whether the conditions are necessary or only helpful to reduce prejudices is widely discussed in the literature (Pettigrew 1998;Pettigrew and Tropp 2006). 9 According to Pettigrew (1998), a change in attitude through contact takes place through four different processes: (a) Negative expectations can be refuted by contact, (b) a change in attitude can lead to behaviour change, (c) affective relationships can lead to more empathy, and (d) contact with other groups can lead to a different evaluation of the individual's own group.
K more important when it comes to explaining attitudes toward the return. Attitudes toward taking in refugees are probably more influenced by moral motives that might cover reasons based on self-interest, while attitudes toward the return of refugees are probably more influenced by economic factors because the remaining refugees will sooner or later have access to the labour market and to social benefits.

Structural Factors
If one sees refugees as a threat to one's own status, it is more likely that one is in favour of their return. This hypothesis stems from the realistic group conflict theory (Campbell and Levine 1972) and the integrated threat theory (Stephan and Stephan 2000). The basic assumption of both theories is that the (perceived) competition with other groups leads the affected persons to a feeling of threat and to negative attitudes (Stephan and Stephan 2000, p. 25;Brief et al. 2005, p. 831). The economic situation of individuals and groups especially plays an important role in that context. According to Quillian (1995, p. 590), a poor economic position leads to the rejection of immigrants, as immigrants compete with natives for scarce resources. In addition, such natives often hold immigrants accountable for their own situation (scapegoating). A similar argument applies to the group a person belongs to. If people believe that their country is economically disadvantaged because of migration, they tend to oppose immigrants (McLaren 2003, p. 915).
Refugees who have come to Europe and stay longer will either enter the education system and, later, the labour market or will depend on social benefits. Brenzel et al. (2019, p. 42) show on the basis of the refugee sample of the SOEP that refugees in Germany have, on average, lower qualifications and are less often employed than the German population. Blossfeld et al. (2016, p. 79) analysed data of the German Federal Employment Agency (Bundesagentur für Arbeit) and found similar results. Assuming that refugees (especially in the first years) receive more social benefits and have lower qualifications than the average of the domestic population, one can expect that especially people who receive social benefits and have lower qualifications will feel threatened and therefore favour refugees' return as soon as the reason for seeking refuge has become obsolete. Several studies have examined the correlation between competition in the labour market and attitudes toward immigrants. The results are inconsistent, however. 10 O' Rourke and Sinnott (2006, p. 857) used data of the International Social Survey Programme 1995 and found that the attitude toward a specific group of refugees compared with the attitude toward immigration in general was not related to the qualification level of a person. One reason for this could be the moral obligation that comes with taking in refugees, as their life is threatened. However, if the reason for seeking refuge became obsolete, a feeling of moral obligation might decline as well.
Structural features can influence the attitude toward the return of refugees not only on the individual level but also on the country level. If fewer refugees enter a country, and if they stay a shorter time, there will be less competition in the labour market. Accordingly, one can assume that the relative group size of the refugees has an impact on the question of to what extent they are seen as a threat (Quillian 1995, p. 589;McLaren 2003, p. 916;Meuleman et al. 2009, p. 354). We assumed that the higher the number of refugees in a country (in relation to the size of the population), the more likely citizens were to support the return of refugees to their country of origin.
A similar hypothesis can be formulated when looking at the economic situation of a country. A low unemployment rate in a country and a well-developed welfare state leads to less competition so that refugees are less seen as a threat (Meuleman et al. 2009, p. 354;Kehrberg 2007, p. 266). Accordingly, we expected that countries with a low unemployment rate and high social expenditure rates would have a lower approval of the return of refugees than countries with a high unemployment rate and a weak welfare state. On the basis of the analysis of the Eurobarometer 1997, Kehrberg (2007, p. 274) shows that countries reject immigration more after an increase in the unemployment rate than in countries where this is not the case.

Ideational Factors
The wish to return refugees to their home country can be influenced not only by structural factors related to economic interests but also by ideational motives. Some people see the different cultural attitudes and practices of migrants and refugees as an enrichment of their own culture. Others see them as a threat to their cultural identity. We suspected that this general attitude also influences the attitude toward the return of refugees. Accordingly, we assumed that people who see multiculturalism as a threat to their national culture tend to reject a long-term stay of refugees more than people with cosmopolitan attitudes do. Hainmüller and Hopkins (2014, p. 233) argue that especially groups that are culturally very different to the host society are rejected (Hainmüller and Hopkins 2014, p. 233). Since 2015 (in the course of the refugee crisis), the majority of refugees have come to Europe from Muslim countries. Muslim refugees experience the strongest rejection (Bansak et al. 2016, p. 218;Gerhards et al. 2019). Therefore, we expected that xenophobic attitudes play an important role when it comes to approving or rejecting the return of refugees.
Furthermore, we believed that people who strongly identify with the nation state tend to disapprove of refugees remaining in their country because they do not see refugees as part of their society (Nickerson and Louis 2008, pp. 798;Ivarsflaten 2005, p. 23). Consequently, an identification with supranational bodies such as the European Union could increase the willingness to grant refugees a long-term stay in the host country, as the idea of solidarity transcends national borders (Merkel 2017, p. 13; see also Zürn and De Wilde 2016).
Finally, we expected that political orientation influences attitude toward the return of refugees. Left-wing ideologies are more connected with ideas of equality, solidarity, and internationalism (Fuchs and Klingemann 1989). We assumed that people who identify as left-wing would be more strongly in favour of taking in refugees K than people from the political centre or from the right-wing camp. Nickerson and Louis (2008, p. 808) showed in an experimental study with Australian students that right-wing political orientation as well as a strong identification with the nation state increases the probability of rejecting asylum seekers. In their study based on the analysis of the Eurobarometer, Kessler and Freeman (2005) also found a clear link between a conservative political orientation and rejection of immigration to Europe (see also Sides and Citrin 2007; for the United States, see Chandler and Tsai 2001). Literature has also come up with the hypothesis that not only rightwing but also left-wing parties oppose migrants, as voters of left-wing parties tend to have a lower social status and feel threatened with losing this status because of immigration (Alonso and Claro da Fonseca 2011, p. 868). If this is the case, the relationship between the left-right scale and attitudes toward the return of refugees would therefore be U-shaped.
Ideational factors can also influence the attitude toward return of refugees not only on the individual but also on the country level. The so-called bandwagon effect says that people adapt their political attitude to what they perceive as the prevailing opinion (Schmitt-Beck 2015, p. 1). With the rise in the number of refugees since 2015, the question of taking in, integrating, and returning refugees has become one of the most debated topics in almost all countries of the EU. 11 Assuming that the majority opinion shaped by public debate also influences citizens' attitudes, one can expect that a negative social climate toward refugees leads to stronger support for their return. Such a climate can reveal itself through negative reporting on refugees in the media, asylum-critical discourse by politicians, or negative attitudes of people in one's personal environment or social media. We chose the share of people with a negative attitude toward taking in refugees in the country of the respondents as a proxy indicator, as we do not have reliable information on the above-mentioned factors. We assumed that a higher number of people who are critical of the right of asylum goes hand in hand with negative reporting on refugees. Accordingly, we also expected stronger support for the return of refugees in these countries.
We have categorised the different factors that can influence the attitude of people toward refugees as structural and ideational factors. However, such a classification is not as distinctive as one might assume. For example, individual education cannot be clearly assigned to one of the two categories. The influence of individual education can either be ascribed to the structural position in the labour market or to a higher cognitive mobilisation that correlates with more tolerance toward cultural diversity (Bobo and Licari 1989, p. 303;Hainmüller and Hiscox 2007, p. 405).

Data and Methods
The basis for our study is a general population survey, the Transnational European Solidarity Survey (TESS), which was carried out in different European countries between May 2016 and November 2016. 12 As we did not have enough resources to conduct the survey in all 28 member states, we had to limit ourselves to 13 countries (Austria, Cyprus, France, Germany, Greece, Hungary, Ireland, the Netherlands, Poland, Portugal, Spain, Slovakia, and Sweden). We are aware that it is not possible to reconstruct a complete picture of the attitude of citizens toward refugees in all 28 EU member states based on data from only 13 countries. However, in order to sample a broad spectrum of countries, we systematically selected the 13 countries according to different features that could have an impact on the dependent variable. Accordingly, we considered wealthy and less wealthy countries; western, middle, and eastern European countries; and countries with different welfare state regimes. 13 In 12 of the 13 countries, 1000 telephone interviews (both landline and mobile) were conducted. In Cyprus, the sample was reduced to 500. Respondents were exclusively registered national citizens aged 18 years or older at the time of the survey. For the analyses, the sample was limited to those cases that did not show any missing values in the relevant variables, yielding to an analysis sample of 9579 cases. Further information on the sampling, the survey, and the weighting, as well as the questions and recoding of all used variables, can be found in Tables A.1-A.3 of the Electronic Supplementary Material.
We operationalised our central dependent variable-the respondents' attitude toward the return of refugees-by asking whether the respondents agreed (yes/no) with the statement that refugees who have been granted the right to stay should be sent back to their home country when the situation has improved. 14 As explained, we assumed that the attitude toward the return of refugees was influenced by the general attitudes toward taking in or rejecting refugees. Accordingly, we asked whether the respondents agreed with the statement (using a 4-point approval scale) that people who are threatened by war in their home country should be granted the right to stay in the respondent's country. 15 To examine the contact hypothesis, we surveyed whether the respondents had regular contact with foreigners. 16 The respondents' structural position in the labour market was measured with a scale similar to the 12 TESS is a joint venture between two research groups from (1)  14 We did not expect that respondents would relate the question to the return of rejected asylum seekers, as the formulation of the question clearly relates to refugees with the right of residence. 15 In the survey, we asked for other reasons why refugees fled their home countries, such as persecution because of their religion, because they are homosexual, or because they campaigned for human rights. Fleeing from war was chosen for the analyses as the item does not contain information on the refugees' characteristics, which could distort the approval rate because of triggering certain stereotypes. Fleeing from (civil) war is also in the focus of the public debate, especially after the stark increase in the number of refugees from Syria, where most refugees have come from since 2013 (http://appsso.eurostat.ec.europa. eu/nui/show.do?dataset=migr_asyappctza&lang=en; accessed 1 July 2019). 16 The question does not survey the contact with refugees; there are two reasons, however, to use the item as an indicator of the operationalisation of the contact hypothesis. First, the group of those who have regular contact with foreigners also includes those who have regular contact with refugees. Second, studies have shown that contact with members of a specific group of immigrants and the respective change in attitude is transferred to other groups and not just to the specific group one has contact with (Pettigrew 1997).
Erikson-Goldthorpe-Portocarero class scheme (EGP), which is also used in the Eurobarometer. We recoded the different class positions in five categories: 1) upper and lower service class, 2) routine non-manual, 3) technicians and skilled workers, 4) self-employed, and 5) agricultural/unskilled manual workers. We used the International Standard Classification of Education (ISCED) to survey the highest schoolleaving certification and used three different categories in our analyses: 1) high (tertiary education), 2) medium (higher and middle secondary education), and 3) low (no formal qualification/lower secondary education). We used three variables in our analyses to operationalise the ideational factors: first, the respondents' belief or denial that cultural life in their country was enriched by immigration (4-point approval scale); second, identification only with the nation state compared to identification with a transnational identity (dichotomous), which includes people who at least partially identify themselves as Europeans or citizens of the world; and third, political self-placement on a left-right scale divided into five groups-1) left, 2) moderate left, 3) centre, 4) moderate right, and 5) right. Control variables were the age (in 10 years) and gender (male/female) of the respondents. The univariate distribution of relevant variables can be found in Table A.4 of the Electronic Supplementary Material.
We considered four possible explanatory factors on the country level. The unemployment rate and government social spending from 2015 were used as independent macro-level variables to measure the influence of a country's economic situation on the citizens' attitude toward the return of refugees. We used the number of asylum applications per 100 national residents to determine how different countries were affected by refugee migration. Finally, the share of approval for taking in refugees was calculated for every country and included as a macro variable factor in the models. This variable tries to measure the general mood in a society toward taking in refugees. 17 To test our hypotheses, we calculated multivariate logistic regression models with country fixed effects to control for the clustered sample structure. The control variables, attitudes toward the acceptance of refugees, contact variables, sociostructural variables, and ideational factors were added stepwise to the model. The presented effect plots are based on the average marginal effects (AMEs), not including the country fixed effects. 18 The partial models and coefficients of determination are displayed in Table A.5 of the Electronic Supplementary Material.
We use the two-step regression approach to measure the influence of macrolevel factors on citizens' attitudes. In the first step, a regression model with all independent micro variables and dummy variables for all countries was calculated. In the second step, a macro regression model was calculated, using the country fixed effects (AMEs) from the first step as the dependent variable and the country-17 All descriptive calculations are weighted according to the drawing probability on the basis of age, gender, labour market status, region (NUTS 2), and employment status, as well as the number of inhabitants of the countries in the transnational descriptive analyses. 18 Average marginal effects in a binary logistic regression model can be interpreted as the average change of the probability over all cases. An AME of 0.1 means that the probability of supporting the return of refugees increases on average by 10 percentage points if the independent variable increases by one unit. level factors as independent variables. Due to the low number of only 13 countries included in this analysis, one regression model was calculated for each independent variable. With this approach, we could test whether the macro variables could explain some of the country differences, while controlling for composition effects due to different distributions of the micro variables in the countries. The results of the macro regression are presented in the form of scatterplots with regression lines. Two of the models include additional regression lines, not taking outliers into account.

Results
As shown in Fig. 1, a majority of around 70% of all respondents were in favour of returning refugees. With the exception of Ireland (49%), the return of refugees had a majority approval in all countries. The results underline that the legal provision concerning the return of refugees, as explained in Sect. 2, is supported by the majority of citizens of almost all surveyed countries. Figure 2 presents the distribution of approval concerning the return of refugees according to the four approval groups for taking in refugees. In contrast to the return, an overwhelming majority of the respondents totally agree (60%) or tend to agree (30%) with the acceptance of war refugees (for more detail, see Gerhards et al. 2019). The outlined results show that the people who approve of taking in refugees I r e l a n d S p a i n S w e d e n F r a n c e G r e e c e N e t h e r l a n d s P o l a n d G e r m a n y A u s t r i a H u n g a r y S l o v a k i a C y p r u s P o r t u g a l are less often in favour of their return than the two groups who oppose taking in refugees. Interestingly enough, 78% of the people who tend to agree with taking in refugees, and 61% of those who totally agree to take in refugees, are in favour of their return if the reason they were taken in has become obsolete. This result shows that the attitudes toward taking in and returning refugees are correlated, but that citizens clearly differentiate between the legitimacy of accepting and returning refugees. With regard to both dimensions, the citizens' attitudes comply with European law. But let us come back to the interpretation of Fig. 1. While only around half of all respondents in Ireland, Spain, and Sweden approve of the return of refugees, this number lies at over 80% in Hungary, Slovakia, Cyprus, and Portugal. The other countries show average approval rates between 65% and 75%. This high variance leads us to ask how these country differences can be explained. Therefore, we analysed whether and to what extent these differences in attitude toward the return of refugees could be explained using the hypotheses formulated in Sect. 3. The coefficient plot in Fig. 3 shows the result of the logistic fixed-effect regression. Of the two control variables, only age shows a weak positive and significant effect.
The attitude toward taking in war refugees has, as expected, a significant negative influence on the attitude toward the return of refugees: People who are in favour of taking in refugees have on average a 20-percentage-points-lower probability of agreeing to the return of refugees than people who reject the taking in of refugees. If the question of whether a person is in regular contact with foreigners is answered with a yes, the probability of being in favour of a return of refugees is on average 5 percentage points lower than for persons with no contact with foreigners. This result confirms the so-called contact hypothesis. Furthermore, we assumed that the structural position in the labour market has an influence on attitude toward the return of refugees. The data does not support this hypothesis with regard to the service classes. Unskilled and agricultural workers and skilled workers do not differ significantly from the service classes. Routine nonmanual workers show a small effect of 2 percentage points, which is just below the significance threshold of 5%. The self-employed have a 7-percentage-points-higher probability of being in favour of returning refugees. This result cannot be easily interpreted with regard to the interests of the self-employed.
The results are different when we look at the educational level. Having a low educational level leads to an increase in the probability of supporting the return of refugees by 8 percentage points compared with a high educational level. A medium educational level increases this probability by 5 percentage points. As mentioned above, education can only partially be interpreted as a structural factor that measures a person's interests. In total, the results indicate that the structural and the labour market position of a person have only a weak influence on their attitude toward the return of refugees.
The ideational factors, in contrast, seem to be better predictors. All of the considered explanatory factors show significant effects in the expected direction: Those who believe that immigration enriches the culture of their country have, per point on the 4-point scale, a 9-percentage-points-lower probability of being in favour of K return. People who identify themselves only with the nation state have on average a 7-percentage-points-higher probability of approving of return than do people with a transnational identity. The political self-placement of the respondents also influences their attitude toward the return of refugees: People with a right-wing (8 percentage points) or moderately right-wing (6 percentage points) self-placement tend to be more in favour of the return than people in the political centre. People who see themselves as left-wing or moderately left-wing have a probability of favouring return of 7 or 8 percentage points lower, respectively, than people in the political centre. This result contradicts the hypothesis of a U-shaped link between the political orientation of the respondents and their attitude toward the return of refugees.
In conclusion, contact with foreigners, attitude toward refugees, and, especially, individual educational and ideational factors influence the approval of returning refugees, while lower occupational classes do not favour the return of refugees above average as we had expected.
Looking at the changes of the country coefficients in Table A.5 of the Electronic Supplementary Material, one can see that adding explanatory variables on the individual level reduces the strength of the country effects; however, all country coefficients still show a significant difference compared to France, the country of reference in the last model. We draw the conclusion that the described country differences can only be partly attributed to a composition effect with regard to the examined individual characteristics. To further look into this unexplained variance, we conducted a two-step regression analysis that uses the country coefficients of the individual model as the dependent variable in macro regressions. Because of the low number of cases, we did not interpret the statistical numbers and significances. Instead, we interpreted the scatterplots depicted in Fig. 4 as a first indication of possible correlations between the macro factors and attitude toward the return of refugees, which must be verified in future studies on the basis of a higher number of cases.
We assumed that the unemployment rate, the strength of the welfare state, and the relative number of refugees in a country have an effect on the attitude toward the return of refugees. The unemployment rate shows a slight negative effect for all countries. This effect turns around, however, if one takes the two countries with an extremely high unemployment rate (Spain and Greece) out of the analysis. Neither the social expenditure rate nor the number of applications for asylum shows a clear, systematic connection to the approval rate of the return of refugees.
To further examine whether the relative change in the structural macro variables over time (unemployment rate, number of applications for asylum) has an influence on the individual attitude toward the return of refugees (see Hiers et al. 2017;Kehrberg 2007), we repeated the same analysis including the change in unemployment rate between 2008 and 2015 and the changes in the number of applications for asylum between 2011 and 2015 (results are available on request). The results do not differ from the results presented here. Structural factors play a substantial role neither on the individual nor the macro level in the explanation of attitudes toward the return of refugees. The result is different, though, when it comes to the ideational explanation. We assumed that the social climate toward refugees has an independent effect on individual opinion about their return. Indeed, the share of approval for taking in war refugees is negatively linked to the share of approval for returning refugees. This relation remains and even slightly increases if the outlier Hungary is excluded. 19 Despite the low number of cases, this finding can be seen as a first indication showing that a positive social climate of refugee acceptance correlates with a higher approval of not sending refugees back to their countries of origin even though the threat in their home countries might have disappeared.

Summary and Outlook
The EU legislation guarantees persecuted people access to Europe. But the right to stay in the host country is temporary and expires if the reason for asylum has become obsolete. We conducted a survey in 13 EU member states, with which we determined to what extent EU citizens believe that refugees should be returned to their home countries once the reason for seeking refuge has become obsolete. A clear majority of the respondents (approximately 70%) are in favour of the return of refugees. At the same time, an overwhelming majority of the citizens believe that refugees should be taken in if they are in need. The European respondents apparently differentiate between the legitimacy of taking in refugees on the one side and their return on the other side. With regard to both dimensions, the citizens' attitudes are mostly in line with European law. This, however, is not true for all countries examined, as the legal norm of the return of refugees in Ireland is not supported by the majority of the population.
At the same time, the empirical analyses demonstrate clear country differences when it comes to supporting the return of refugees. The results of the multivariate analyses show that ideational factors have an influence on the attitude on the individual as well as on the country level, while structural factors that refer back to economic interests do not contribute substantially to the explanation of attitudes toward the return of refugees. Even though some of the country differences are due to different composition of the countries and differences in the social climate directed at refugees, one has to state that we cannot explain the country differences in attitude toward the return of refugees very well. Other factors that we did not consider in our data collection might exist. But the analysis might also indicate that a systematic comparative analysis that tries to explain country differences with relatively broad macro indices has its limitations, as it does not do justice to the historical developments and characteristics of single countries. Social scientists using historically comparative methods have mentioned that for quite some time (e.g. Mahoney 2004).
We would now like to discuss some political implications of our results. The strong wish of the population to return refugees to their home countries if their reason for being granted asylum has become obsolete is for two reasons a dilemma for politicians. First, rejected asylum applicants often cannot be returned for different reasons. This holds true, for example, for people charged with a criminal offense whose criminal proceedings have not ended yet. Members of a terrorist group cannot be returned to their home country if they risk torture in their home country. The same applies to people who do not have any identity documents or have destroyed them deliberately; they cannot be returned because their identity has to be determined first, which is often not possible.
Second, the temporality of the residence status is in a very complicated way linked to the question of integration of refugees into the host society. The rights of refugees to be integrated into the host society is codified in the directive laying down standards for the reception of applicants for international protection. 20 This directive leaves quite some room to manoeuvre to limit the rights of refugees and to complicate their integration. The directive specifies that member states should enable asylum applicants to access the labour market after 9 months. But at the same time, countries have the possibility to limit this right to prioritise their own citizens-a right that is often used. A policy of non-integration of refugees is often justified with the argument that integration of refugees would obscure the difference between refugees and other migrants. This again can lead to the fact that people who support taking in refugees but at the same time do not wish them to stay longer start to oppose taking them in and therefore question the basic principle of the right of asylum.
Even though the outlined argumentation is correct on an abstract and legal level, a policy of weak integration of refugees will entail considerable consequences and high "costs". The important question is how probable it is that the conditions in the refugees' home country improve in a way that the reason for asylum becomes obsolete. According to the yearbook 2018 of the Stockholm International Peace Research Institute, the global security situation has deteriorated in the past century, the number and complexity of armed conflicts has risen, and many of these conflicts do not seem to be easily overcome (Smith 2018, p. 3). The situation in the home country does not usually improve in the short term, which means that people remain longer in the countries that have taken them in (Crisp and Long 2016, p. 145). This is probably true for those refugees who have come to Europe since 2012 and who have mostly come from Syria or Afghanistan. One can assume that people from these two countries will probably not be able to return to their home countries in the next 10 years or even longer. The Assad regime seems to have stabilised thanks to the Russian war intervention. Even if the situation in Syria is stable, the refugees returning to Syria might face persecution by the regime. 21 Also with regard to Afghanistan, one cannot expect that the situation will improve in the near future and that Afghanistan will become a safe country where refugees can be returned to. 22 If one can expect that a majority of refugees cannot return to their home countries in the near future, it is, from a humanitarian point of view, necessary but also in the interest of the host society to integrate refugees as soon as possible. Otherwise, they will depend on social expenditures. A policy of non-integration can also lead to the establishment of a parallel society and to political radicalisation leading to high social costs. Therefore, economists recommend providing refugees with access to the educational system as early as possible and enabling them to access the labour market so they can become economically independent and do not have to depend on social expenditures Wößmann et al. 2016;Wößmann 21 The Syrian president Bashar al-Assad has announced a general amnesty for deserters, but his words cannot be trusted. There are reports of cases where returned people disappeared or were arrested. (https://de. reuters.com/article/syrien-amnestie-idDEKCN1MJ0YT, https://www.asyl.net/view/detail/News/berichteueber-verschwinden-syrischer-rueckkehrer/; https://foreignpolicy.com/2019/02/06/a-deadly-welcomeawaits-syrias-returning-refugees/, http://www.taz.de/!5542925/; accessed 29 May 2019. 22 The United Nations Assistance Mission in Afghanistan has documented a consistently high level of injuries and deaths of civilians in the past years (UNAMA 2018), and the European Council on Refugees and Exiles recommends in a current paper to stop returning people to Afghanistan because of the deteriorating safety situation (ECRE 2019). 2016). Early integration measures are linked to high costs, but they are worth it as a failed integration will later on cost even more. 23 If those refugees whose reason for displacement continues to exist in the long run and for whom return is not possible were to receive a permanent instead of a temporary right of abode, this would collide with the basic principle of temporality of the right of asylum. We have seen that citizens support this very much. If the right to stay were to be changed, one would have to expect resistance by the population. This resistance would be fuelled by right-wing populist parties that are sceptical toward migration.
This dilemma can, from a political point of view, not be easily solved. If one does not want to further strengthen the prevalent resentments in the populist camps, the only option is to clearly communicate with the population that the return of many refugees is unlikely and that early and intensive integration is necessary to avoid long-term costs for the host society. Furthermore, it is necessary to strengthen the fact that workers are needed in different sectors of the labour market. A survey study that was conducted by the Sozialwissenschaftlichen Institut der evanglischen Kirche (own translation: Social Science Institute of the Evangelical Church in Germany) in 2017 showed that such information can convince citizens. This study showed that the majority of all those who were in favour of deporting, in any case, rejected asylum seekers revised their opinions retrospectively: In a second step, they said that rejected asylum seekers may stay in Germany if they have already built an existence in Germany (72%) or if they have lived in Germany for many years and are well integrated (75%; Ahrens 2017). The results imply that the approval or rejection of refugees remaining in the host country depends on their integration into the host society.
Funding Open Access funding provided by Projekt DEAL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4. 0/.