Refugees through the eyes of locals; a study on tolerance towards outer groups in Turkey and Germany

doi:10.21203/rs.3.rs-3090696/v1

In 2019, around 1% of the world’s population was displaced, among whom 26 million are expected to be refugees. The arrival of refugees has been related to prejudice and intolerance towards outer groups. This paper investigates the effect of refugee influx on tolerance towards outer groups, immigrants and foreigners, other races, and religions in Europe’s two largest refugee-receiving countries, Turkey and Germany. For this study, individual-level data from the World Value Survey (WVS) and European Value Survey (EVS) is used in a quasi-experimental setup, difference in difference (DID) strategy for the Turkish context, and pre-post intervention regression analysis for the latter. The results showed that the refugee influx positively influenced tolerance in both countries in the short run. However, the Turkish population became less tolerant over time. As the European Union (EU) has a definite role in influencing migration and refugee integration through its laws and funding, individuals with confidence in the EU became more tolerant toward outer groups. The results are robust to several falsification and placebo tests. These findings provide important implications for government bodies and policymakers to promote integration and understanding among diverse communities.

Development Economics

Behavioral Economics

refugees

tolerance

prejudice

immigrants or foreigners

other races

other religion

confidence in EU

Tolerance is acceptance of diversity in values, behavior, and appearance among human beings and respecting the right of everyone to live peacefully. In contrast, intolerance occurs through violence and marginalizing and discriminating against the minority group or excluding such members from social and political participation (UNESCO, 1995). Tolerance towards others promotes individuality and diversity, peace, and secular civil society. The current literature on tolerance varies across fields of study. For instance, social psychologists primarily studied tolerance in terms of negative attitudes toward social groups (i.e., prejudice). Intolerance or the existence of discrimination is linked to the mental health and well-being of medical scholars. The economics literature has linked the tolerance of individuals with social, economic, and political factors. Intolerance towards people of different races and the color is a present-day problem around the world, causing phenomena like xenophobia, racism, islamophobia, antisemitism, etc., leading to hate speech, physical injury, or even murder across the globe (Côté and Erickson, 2009; Gniewosz and Noack, 2008; Mazzoni et al., 2020). An increase in the popularity of right-wing political parties and protectionist leaders is observed across the European region due to the fear of concerns of globalization and immigration (Vieten and Poynting, 2016; Davis and Dole, 2017). This phenomenon has prompted studies linking tolerance with individuals’ voting behavior and elaborating on factors resulting in intolerance or prejudice. Such studies are insightful, however, limited, as they do not inform public policy on whether human behavior, such as tolerance, can change instantaneously. For instance, shocks, such as refugee influx, cause deterioration of tolerance towards outer groups according to Stephan and Stephan’s (2000) social threat theory, harming societal peace and harmony. Or does refugee influx make people more tolerant towards outer groups? supporting Allport’s (1958) social interaction theory driving governments to be more optimistic in accepting and blending refugees into the local community. Therefore, keeping these two possibilities, a systematic investigation is conducted to understand the changes in tolerance levels towards outer groups in the aftermath of the refugee influx across Turkey and Germany. This paper builds on social dominance theory and studies the effects of the refugee influx in Turkey and Germany on tolerance towards outer groups using a difference in differences (DID) study and simple pre-post regression analysis. In addition, as European Union (EU) plays a significant role in immigration laws for European nations, this paper tests whether having confidence in the EU moderates the relationship between refugee influx and tolerance using the two-stages least square (2SLS) method.

However, specific challenges exist while testing whether tolerance changes instantaneously. For instance, the first hurdle would be determining the type of shock that might affect tolerance among individuals. Second, ensuring no other shock influenced individuals' tolerance before and after the shock. This study overcomes such hurdles by using the 2011 refugee influx in Turkey and the 2015-16 refugee influx in Germany as unexpected events that can potentially influence tolerance among individuals. The refugee influx is measured by the number of asylum applications collected from the Ministry of Interior of Turkey and Statistisches Bundesamt (statistics from the German federal office). This data is merged with regional and individual-level data from the longitudinal World Value Survey (WVS) and European Value Survey (EVS). The details of the questionnaires and survey years used in this study are listed in Appendix A.

This paper has three significant contributions. First, this paper examines whether tolerance or prejudice is reshaped in the shorter term compared to the current literature’s approach to treating bigotry or discrimination in a longer-term window. The goal is to help policymakers design social interventions to promote societal tolerance. This is followed by proving a causal relationship between refugee influx and tolerance among individuals and systematically examining how having confidence in European Union can be a causal mechanism through which refugee influx may influence discrimination among individuals. Third, a major empirical challenge exists to address observable and unobservable endogeneities. Therefore, DID and pre-post regression analysis along with potential additional effects (such as fear of financial or physical threat and crime) and unobservable biases using selection on observables are addressed, which will likely provide causal inference on whether refugee influx changes tolerance levels among individuals.

The DID estimates for Turkey indicate that locals in Turkey become more tolerant in the short run but less tolerant in the long run. For instance, locals became 12% and 17% more tolerant towards people of other races and religions. However, in the long run, in 2018, when the refugee problem persisted, locals were 17% less tolerant towards immigrants and foreigners and 8% towards other religions. Germans become 6% more tolerant towards immigrants, 7% more tolerant towards other races and religions after the 2015-16 refugee influx. In addition, having confidence in the EU tends to positively influence tolerance towards outer groups in Turkey and Germany; however, it loses significance in the long run. For instance, individuals with more confidence in the EU than those without faith in the EU tend to be more tolerant towards immigrants and people from different races and religions.

In this paper, several consistency checks are conducted to evaluate the robustness of the findings. The robustness of the model is tested by using alternative measures of refugee influx (e.g., foreigner number for the Turkish context, whereas migrant inflow numbers for the German context). In addition, falsification tests are conducted. For instance, a false refugee influx year is used, followed by running the analysis using the wrong treatment dummies. The external validity tests are done by omitting variable bias. Lastly, due to simple pre-post regression analysis for the German context, additional controls, such as fear of losing jobs, the spread of disease, and the threat of crime are added to further test the robustness of the model.

The next section of the paper elaborates on the background of the study, expanding on Turkey and Germany’s refugee influx. Whereas section 3 discusses the related studies on tolerance, and addresses the gap and contribution of this paper. The data used for this study is discussed in section 4. Section 5 presents the examination of the Turkish context, and section 6 on Germany, with an overview of the methodologies employed, and discussions on the results and analysis. Lastly, section 7 provides the implications and conclusion of this study.

2.1 Refugees

The UN defines refugees as individuals forced to flee their country because of persecution, war, or violence and cannot return home (Black, 2001). They may fear persecution for race, religion, nationality, political opinion, or other factors (Shacknove, 1985). Refugees from Syria, Afghanistan, South Sudan, Myanmar, and Somalia contribute to two-thirds of the refugees worldwide. At least 79.5 million were forced to flee their home, the highest global displacement on record. According to UNHCR, 26 million are refugees, around half of whom are children (UNHCR, 2019). The 1951 Geneva Convention obliges the member countries to provide shelter, protection, and assistance to refugees (United Nations, 2019).

2.2 Syrian civil war and European Migration crisis

The civil war in Syria caused 9 million refugees to flee their homes. Since the conflict started in 2011, families have agonized under horrific violence, killing thousands of people, destroying the nation, and setting back the standard of living by decades. Refugees sought asylum in neighboring countries such as Turkey, Iraq, Jordan, Lebanon, and Egypt. The Syrian refugees relied on the Eastern Mediterranean and Balkan route to reach Europe arriving in Greece, Cyprus, Bulgaria, Germany, Sweden, and other EU nations. In addition, in Iraq, the decades of conflict in the region, followed by a surge in Islamic State (ISIS) attacks in Northern Iraq, created 3 million refugees, among which 260 000 were forced to flee Iraq in 2014. Lastly, hundreds of refugees from Afghanistan, Kosovo, and Albania still seek asylum in European countries. The arrival of the large number of refugees creates a significant economic, social, and political challenge for the host European countries such as Turkey and Germany, prompting the European migration crisis.

2.3 Turkey and refugees

Figure 1 illustrates that the number of individuals seeking international protection increased from 2010 onwards to 2018 and fell afterward.

In April 2011, Syrian refugees started to enter Turkey. A group of two hundred and fifty-two refugees was the first Syrian refugees to seek asylum in Turkey, weeks after President Bashar al-Assad’s response to anti-regime demonstrations throughout Syria turned violent. From the end of April, the borders between Turkey and Syria became unstable as thousands started fleeing Syria to seek refuge in the Turkish province of Hatay. The Turkish government was persistent and vocal about open borders and offering sanctuary to anyone seeking it. According to the Turkish government, by the end of October 2011, around twenty thousand refugees were found refugees in Turkish borders, the majority living in camps. In mid-2012, the surge in fighting and ceasefire in Syria activated a torrent of refugee arrivals that continued for a decade. Turkey's government announced to restrict on the acceptance of Syrian refugees to a hundred thousand in August 2012. However, it reached forty thousand in 2012 to over one million by the end of 2014. In 2016, the Turkish government took other steps to limit refugee numbers and generate control over the refugee influx in Turkey. It ended the visa-free policy for Syrians and started building security barriers along its Syrian border. They amended their law on foreigners to allow the deportation of individuals linked to terrorism or other threats. Turkish border guards were criticized for shooting toward refugees to prevent refugees from crossing borders. According to Human Rights Watch reports, nearly 1.7 million Syrian refugees were residing in camps on the Syrian side of the Syrian-Turkish border. The Turkish government's strict measure saw refugees decline to 198 000 in 2018, the smallest increase since the refugee influx took off. Most Syrian refugees being registered were newborns, not arrivals from the increasingly impenetrable border. Figure 2 shows how the refugees have spread across Turkish provinces by the year 2018.

2.4 Germany and refugees

In 2015, there was a major increase in individuals seeking asylum in Europe, becoming the highest annual flow of refugees since 1985. Among the EU members, Germany received the highest number of asylum applicants, peaking in 2015, as shown in Fig. 1. The arrival of refugees in Germany has been associated with the surge in xenophobic attacks and the popularity of right-wing party, Alternative für Deutschland (AFD) in Germany (Otto and Steinhardt, 2017). This relationship questions the tolerance of Germans towards outer groups.

Since 1993, asylum seekers have been distributed across regions using their national system, EASY. A quota-based distribution is carried out, called the Königsteiner key. 2/3 of it comes from tax revenues and 1/3 from the population. However, the state allocates refugees to municipalities randomly. At maximum, during the first six months of the asylum application, asylum seekers are obliged to live in initial reception centers, where they are assisted with necessities such as food and clothing, from where they must find collective housing spread across the federal region. The asylum seekers are provided medical care for acute illnesses and pain. Asylum seekers can work only after they move from initial reception centers but require a work permit from the Alien’s authority. However, once asylum seekers are granted asylum or ‘refugee’ status, with that residence permit, they can work without requiring any other permission from governmental bodies (UNHCR, 2019). Throughout the application process, depending on individual scenarios or district restrictions refugees might face movement restrictions. For instance, some might be obliged to remain in reception centers, while others are allowed private housing. Refugees are more likely to stay in their assigned location rather than move to a different state. For instance, first, free goods and social services are provided in their assigned locations for helping them integrate, and (ii) refugees are legally obliged to remain until a decision is made (Gehrsitz and Ungerer, 2022). In addition, in 2016, the Integration Act 2016 imposed geographical restrictions for a minimum of 2 years for refugees arriving in 2016, and onwards (UNHCR, 2019). The table below reports the number of asylum applications and migrant numbers across all the regions in Germany in 2015 during the refugee influx.

Table 1

The number of asylum applications and migrant numbers by regions in Germany in 2015.
Number	Regions	Asylum applications	Migrant numbers
1	Baden-Württemberg	57578	320942
2	Bayern	67639	328561
3	Berlin	33281	99867
4	Brandenburg	18661	37519
5	Bremen	4689	20465
6	Hamburg	12437	44425
7	Hessen	27239	173192
8	Mecklenburg-Vorpommern	18851	30173
9	Niedersachsen	34248	193408
10	Nordrhein-Westfalen	66758	463195
11	Rheinland-Pfalz	17625	91922
12	Saarland	10089	21728
13	Sachsen	27180	61126
14	Sachsen-Anhalt	27810	42250
15	Schleswig-Holstein	15572	45706
16	Thüringen	13455	41762

Note: The data is collected from Statistisches Bundesamt (federal office statistics of Germany) for 2015. Asylum applications are submitted to the Federal Office for Migration and Refugees (BAMF) seeking refugees; In contrast, migrant numbers are the inflow of migrants in Germany measured by the Federal statistics office of Germany.

The asylum applications are used to measure the refugee influx, whereas the migrant numbers are used as an alternative measure to test the robustness of our results in section 6.4.1.

3.1 Theoretical underpinning

This paper draws upon the principles of two underlying theories, Allport’s (1958) social interaction theory, and Stephan and Stephan’s (2020) social threat theory. Both theories address the relationship between interaction and acceptance or prejudice, however, have opposing views on the issue. Relating to Allport’s (1958) social interaction theory, Turks and Germans would have greater interaction with the refugee population after the refugee influx, initiating better mutual understanding and appreciation, improving their tolerance towards outer groups. In contrary, Stephan and Stephan‘s (2020) threat theory would imply that such interaction might raise misconception and conflict, harming tolerance. The findings would extend the understanding of the social dominance theory, which implies that group-based inequalities are maintained through institutional discrimination, individual discrimination, and behavioral asymmetry. This implies individuals with greater social dominance orientation are more likely to dehumanize individuals from lower orientation, outer groups such as refugees, resulting in less favorable attitudes, such as tolerance (Sidanius, Devereux and Pratto, 1992; Sidanius and Pratto, 2001; Pratto and Stewart, 2012).

3.2 Tolerance and prejudice

Several papers studied the factors or contexts developing prejudice and tolerance. Janmaat and Keating (2017) found that though prejudice exists among youth, they are more accepting of racial minorities or homosexuality than their older generations. Their finding was aided by Evans and Kelly’s (2019) paper which showed prejudice fell to its lowest points in European nations until just increasing recently. Hungsberger and Jackson (2005) found religion to make individuals both tolerant and intolerant. In Cowling and Anderson’s (2019) study, participants showed more significant prejudice towards refugees of a different religion from their own; similarly, religious primes also increased respondents’ donations supporting refugees (Lazarev and Sharma, 2017). Threat to mortality or health of individuals nurtures prejudice. For instance, news of terror attacks or threats of diseases harmed prejudice toward outer groups (Das et al., 2009; Hung et al., 2011). Lastly, central cultural values tend to influence the nature of prejudice, including prejudice toward immigrants and foreign workers, highlighting the importance of understanding the cultural dynamics of discrimination. For instance, compared to East Asian cultures, in North European heritage, a stronger association exists between economic competitiveness, and prejudice towards immigrants and foreign workers (Shin and Dovidio, 2017).

3.3 Tolerance and its effects

Tolerance is a measure of social capital. Crowley and Walsh (2018) used data on twenty-one thousand individuals from fourteen European countries from the Lathe life in Transition Survey (2016) to examine the relationship between social capital and life satisfaction. The three critical dimensions of social capital examined were structural (personal ties), cognitive (trust), and tolerance. The results from the probit analysis showed that solid structural ties, tolerance, and trusting individuals improved life satisfaction. Tolerance and threat perception influenced political preferences in Germany and Netherlands. Eriksen and Kentmen-Cin (2017) found that people are more cautious about the standing of democracy in their country than personal security. In Germany, Muslim cues increased political intolerance, whereas, in the Netherlands, people became more tolerant. Similarly, country-specific differences were found to have a role to play in the psychological dynamics of prejudice toward immigrants and foreigners, influencing the acceptance of such groups in the country. Shin and Dovidio (2018) studied the discrimination towards immigrants and foreign workers in US, Germany, and Australia. They examined the tendency to perceive social and biological-based differences; physical, economic, and symbolic threats; and cultural values such as emphasizing individual uniqueness and achievements as predictors of prejudice. In comparison, the study’s results showed that the effect of biological-based differences was more substantial in Australia. In contrast, the development of symbolic threat was more significant in Germany, and the impact of individual uniqueness was more significant in the US.

3.4 The outcome of refugee migration

The refugee migration in Europe encouraged several studies in the literature. In Germany’s context, Gehrsitz and Ungerer (2016) studied the influence of refugee influx on labor market, crime, and voting behavior using administrative county-level data. They found that local employment did not suffer; however, the crime rates increased slightly. and refugees suffered from unemployment. Interestingly, though anti-immigrant parties found an increase in support on the macro level, the effect was the opposite at the micro level. Similarly, Stahler (2017), using the New Keynesian DSGE model, found that if the migration qualification structure matches the native population, do not lead to significant changes in GDP or consumption. Certain prominent events can transform electoral competition as actors are forced to take definite positions on such issues and allow citizens to align their partisan preferences and political attitudes. Mader and Shoe (2019) used panel data to analyze how German political parties reacted to the influx of refugees at Europe’s borders in 2015/16. It could be understood that German citizens changed partisan preferences relying based on their immigration attitudes. This refugee crisis is thus considered a critical juncture in transforming party competition in Germany. Radical right voters tend to show stable partisans with entrenched anti-establishment views; however, such voters were seen to change allegiances when established parties accommodate their preferences. However, significant parties may lose their voters when they try to attract such radical right voters via restrictive immigration policies. Thus, a status quo can be understood in equilibrium (Chou et al., 2018). Other European countries, such as Italy, had one of the most significant influx of refugees among the European nations. The impact of immigration on the political preferences of the natives in Italy was explored by Chou et al. (2018). They found that in municipalities with greater arrivals of immigrants, electorates are more likely to vote for a coalition of center-right, which lobby for tighter policies against refugees. Papers also on the American context found that if prejudice were minimal, Obama would have won an electoral vote landslide back in the 2008 US elections (Highton, 2011)

3.5 Research gap, questions, and objectives

The current literature explores the origin and influence of intolerance and prejudice. The global refugee crisis has initiated significant attention on the refuge crisis. Therefore, numerous studies and discussions explore the implications of the refugee influx relating to economic, political, and social factors. Relating to Allport’s (1958) social interaction theory and Stephan and Stephan’s (2000) social threat theory, interaction can either improve tolerance through or harm tolerance towards outer groups. The assumption that tolerance or prejudice may change instantaneously is overlooked. This paper extends the understanding of social dominance theory, for instance, the development of social stratification, such as prejudice or intolerance, in the short run. For example, this study contributes by learning how the tolerance of locals transforms following the refugee influx. This paper is built on this gap in the literature and aims to answer;

Q1. Whether refugee influx makes individuals intolerant towards outer groups (e.g., tolerance towards immigrants or foreigners, other races, and religions?

Q2. Does having confidence in the European Union (EU) mediate between the refugee influx and tolerance of individuals towards outgroups?

Therefore, this study aims to find the effect of refugee influx on tolerance towards outer groups in Turkey and Germany. And how individuals having confidence in European Union (EU) may moderate the relationship between refugee influx and tolerance toward outer groups;

The tolerance towards outer groups is measured using individual-level data capturing tolerance towards immigrants and foreigners, other races, and religions used from the World Value Survey (WVS) and European Value Survey (EVS). Binary variables are created to represent individuals who expressed aversion to having immigrants or foreigners, other races or religions as neighbors. These variables are used as predictors to measure the dependent variable. Both surveys are large-scale, cross-national surveys measuring cultural values, attitudes, and beliefs, providing valuable insights into how they have changed over time. The dataset includes survey years for both Turkey and Germany that spans years prior to and following the year of refugee influx in respective countries, enabling measuring the impact of the refugee influx. For instance, in Turkey, the refugee influx started in the year 2011 and continued until 2018, with data collected from both pre-influx 2001 and 2007 to post-influx 2011 and 2018. Similarly, for Germany, data was collected prior to influx in 2013, and post-influx in 2018, as the refugee influx occurred in 2015. Two different model specifications are adapted to study the change in tolerance towards outer groups across provinces and districts of Turkey and Germany and are discussed in Sections 4.1 and 5.1. The regional data is collected from the Turkish Statistics Institute and the German statistical office. The details of the questionnaires used for this study can be found in Table A1 of Appendix A.

5.1 Difference in differences (DID) strategy for Turkey

Difference-in-differences (DID) method is used to find the causal effect of refugee influx on tolerance of the Turkish population in the short run when refugees were only close to reception centers in 2011, and in the long run, when refugees are spread across Turkey by 2018. The DID approach allows to study the differential impact of treatment on a specific sample. It removes the biases in post-treatment period comparisons that could result from permanent differences between those groups and tendencies from comparisons over time formed by trends due to other causes of the outcome. The DID model takes the following form:

Tolerance_ipt = α_p + β₁Treatment Area_p + B₂Post_t + B₃(Treatment_p × Post_t) + X_ipt + ε_ipt (1)

In this model, Tolerance ipt is the tolerance of individual i towards outer groups in province p in WVS year t, measured using three dependent variables, tolerance towards foreigners and immigrants, tolerance towards other races, and tolerance towards other religions. All specifications include fixed effects for province and year. The province-specific fixed effects control for the time-invariant characteristics across provinces. Technically, it accounts for permanent differences across provinces, such as exposure to refugees and cultural traits toward tolerance, which are likely to remain constant during this analysis period. The treatment area is a dummy variable, with 1 for areas with refugees and 0 otherwise. For short-run analysis, in 2011, only areas close to reception centers were assigned a value of 1 (i.e., Hatay province), whereas for long-run analysis, in 2018, areas with comparatively high refugee numbers, approximately areas with greater than 10,000 Syrian registered refugees, were assigned 1 and 0 otherwise. The treatment and control provinces are treatment and control for this study are listed in Appendix B. Post_t is a dummy variable for the post-influx period that takes the value of 1 if observations are of the WVS year from 2011. For instance, we capture the short-term and long-term effects of refugee influx on tolerance by allocating 1 to WVS years 2011 and 2018. In our setting, the treatment variable is an interaction between our measure of the influx of refugees by province, the treatment area, and an indicator for the post-influx period. Specifically, the coefficient B₃ in Eq. (1) estimates the average treatment effect. We restrict our estimation model by controlling, for X_ipt —a vector of time-variant observable characteristics of individuals, such as age, marital status, number of children, and education. ɛ_ipt is the idiosyncratic error term that captures time-varying unobserved heterogeneity. The details of the questionnaires and survey years used for this analysis are listed in Appendix A.

5.2 Results and analysis

5.2.1 Descriptive statistics

To understand the nature and characteristics of the data used in this study, the summary, central tendency, and variability of the dataset are exhibited in Table 2, along with the correlation of the covariates with the outcome variables of this study in Table 3.

Table 2

Summary statistics of our sample.
Variable	Observations	Mean	Std. dev.	Min	Max
Tolerance (other races)	8,727	0.68	0.47	0	1
Tolerance (immigrants & foreigners)	8,702	0.63	0.48	0	1
Tolerance (other religions)	8,715	0.66	0.48	0	1
Confidence in EU	5,040	0.39	0.48	0	1
Age	8,762	37.69	13.63	17	95
Edu	8,767	1.07	0.99	0	3
Gender: Female	8,767	0.50	0.50	0	1
Marital status: married	8,762	0.68	0.47	0	1
No children	8,767	0.34	0.48	0	1
Employment rate	8,280	0.41	0.49	0	1
Note: The table shows the distribution of the WVS/EVS sample used in this study.

The table above indicates the total number of observations for the outcome variable, tolerance towards other races, or towards immigrants and foreigners, are 11,665 and 11,639. From this sample, on average, 67% are tolerant towards other races. Most of the variables studied are dummy variables, whereas age is continuous, and education is a categorical variable. Followed by Table 3, which indicates that the treatment area being investigated has a significant negative correlation with tolerance towards other religions. Individuals as they get older, have a greater negative relationship with tolerance towards outer groups. Similarly, being married also is negatively correlated with tolerance towards outer groups. Table 3 indicates the bivariate relationship; however, other factors might influence this relationship.

Table 3

The correlation of covariates used in this study.
	Correlation
	Tolerance (other races)	Tolerance (immigrants and foreigners)	Tolerance (other religions)
Treatment area	-0.07*	-0.09*	-0.03*
Confidence in EU	-0.03	0.01	0.004
Age	-0.08*	-0.04*	-0.07*
Edu	0.20*	0.19*	-0.34*
Marital status: married	-0.08*	-0.07*	-0.08*
No children	0.08*	0.07*	0.08*
Employment rate	0.02*	-0.02	0.01
Note: The table shows the correlation of covariates with three dependent variables; tolerance towards other races (column 1); tolerance towards immigrants and foreigners (column 2), and tolerance towards other religions (column 3). ∗Significant at 5% level

To conduct a difference in differences (DID) analysis, the treatment and control group need to be similar in terms of characteristics for our analysis to be valid. No significant difference would imply that differences in outcomes between treatment and control groups are due to the shock, refugee influx. Therefore, the treatment and control groups with means of various variables that might influence our outcome variables in the long run are compared in Table 4.

Table 4

The mean difference between treatment and control groups.
	Treatment (After influx) (1)	Control (Before influx) (2)		Standard error (4)
Independent variables
Age (years)	36.53	36.64	0.12	0.69
Married (1 = yes)	0.72	0.71	-0.01	0.02
Education	1.36	1.39	0.034	0.04
Dependent variables
Tolerance_immi & foreigner	0.55	0.66	0.11	0.01
Tolerance_ other races	0.61	0.70	0.008	0.02
Tolerance_ other religions	0.63	0.66	0.04	0.01
Confidence in EU	0.35	0.34	0.02	0.01

Note: The table shows the mean difference between the treatment and control groups (column 3), followed by the standard error (column 4). ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The table above shows there is no significant difference in the mean for most of these factors. The only significant difference observed was in the measure of economic threat, with the treatment group reporting a higher level of economic threat than the control group. Lastly, the local population, refugee numbers, and gross domestic product (GDP) between the provinces in treatment and control groups can be found in Appendix B.

5.2.2 Benchmark analysis

To investigate the effect of refugee influx on tolerance levels towards outer groups, the difference-in-differences (DID) method is used to compare changes in an outcome between treatment groups, provinces that had refugee influx, and control groups (no or minimum refugee influx). This method allows segregation of the effect of the refugee influx on the tolerance towards outer groups.

4.2.2.1 Short run

Table 5

The short-run effect of refugee influx on tolerance towards outer-groups
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions
Treatment area	0.0200	-0.0255	-0.114**
	(0.0392)	(0.0444)	(0.0458)
Post influx	-0.0108	-0.0863***	-0.0684***
	(0.0140)	(0.0173)	(0.0176)
Treatment*Post	0.0254	0.120**	0.173***
	(0.0582)	(0.0575)	(0.0595)
Age	0.00416	0.0148***	0.00926**
	(0.00262)	(0.00367)	(0.00367)
Age square	-3.03e-05	-0.000155***	-0.000110***
	(2.88e-05)	(4.07e-05)	(4.04e-05)
Marital status: married	-0.0277	-0.105***	-0.0178
	(0.0193)	(0.0250)	(0.0258)
No children	0.0536**	-0.0225	0.00792
	(0.0213)	(0.0276)	(0.0278)
Education	0.0664***	0.104***	0.110***
	(0.00679)	(0.0110)	(0.0113)
Constant	0.478***	0.341***	0.418***
	(0.0566)	(0.0813)	(0.0811)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes	Yes Yes
Observations	6,346	2,951	2,951
R-squared	0.021	0.074	0.079

Note. This table shows the results of the difference-in-differences method. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The positive significant coefficient for ‘Treatment * Post’ indicates individuals became more tolerant towards outer groups. For instance, individuals in areas with refugee influx became 10% more tolerant of people of other races after the influx. The finding supports Gordon Allport’s (1955) contact hypothesis, where positive intergroup contact reduces prejudice and improves tolerance and relation. In this scenario, local residents become more tolerant due to positive intergroup contact with their refugee neighbors, colleagues, or friends. Through interaction and shared experiences, local residents identify refugees as individual human beings with their own unique experiences and stories rather than a homogenous and threatening group, resulting in better tolerance and acceptance.

5.2.2.2 Long run

There was an increasing trend of refugee applications until 2018 in Turkey (see Fig. 1). To investigate whether tolerance towards outer groups changed over time in the long run, the relationship between tolerance towards outer groups and refugee influx is investigated for the year 2018. As the severity of the refugee influx might influence the relationship of treatment areas with tolerance, which the dummy variables would not capture; along with the dummy treatment measures, two proxy severity measures of the refugee influx from the Turkish Statistical Institute are used. (i) The total number of foreigners in 2018 (see columns 4 to 6 of Table 6) (ii) The ratio of Syrians living in the city to the total number of people living in the city in 2022 (see columns 7 to 9 of Table 6). All three-regression analyses are illustrated separately in Appendix C.

From columns (4,6,7 and 9) of Table 6, it can be observed that the DID coefficients are significantly negative. The results portray that in 2018, eight years after the influx of refugees, individuals become less tolerant of immigrants or foreigners, and other religions. Relating to the contact hypothesis, the effect of intergroup contact on tolerance is found as non-linear. The findings support Lord, Ross, and Lepper’s (1979) study, where they discussed how people’s attitudes might be positive in the short run but become less positive or negative with time, depending on the intensity, duration, and context of the contact. For instance, lengthier contact among groups that are perceived as forced or undesired might create stress and conflictual situation, leading to harming intergroup relations. Lastly, from 2012 onwards, Turkey suffered from an increasing trend in unemployment (ILO, 2022), further augmenting the competition for limited resources in the economy between locals and refugees, increasing prejudice or intolerance. Related to realistic group-conflict theory, Sassenberg et al. (2007) found that negative interdependence between groups, like competition, results in prejudice towards the outer groups. The results indicate that though tolerance towards outer groups improved around the time of the refugee influx, in the long run, the tolerance towards outer groups diminished.

Table 6

The long-run effect of refugee influx on tolerance towards outer-groups
	Treatment: dummy						Treatment: continuous ‘22 refugee no.
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions
Treatment area	0.381***	0.357***	0.489***
	(0.0964)	(0.0888)	(0.0897)
Treatment cont.				-119.1	-265.2*	-40.68
				(176.4)	(151.8)	(147.2)
Treatment cont. 2022							-0.464	0.013134***	-0.159
							(0.682)	0.003653	(0.569)
Post influx	0.259***	0.309***	0.328***	-1.182	-0.0465***	-0.187	1.306	-0.04888***	0.702
	(0.0748)	(0.0638)	(0.0656)	(1.861)	(0.0110)	(1.553)	(1.440)	0.010578	(1.202)
DID (dummy)	-0.169***	-0.0533	-0.0828**
	(0.0421)	(0.0425)	(0.0418)
DID (cont.)				-13.91***	-4.393	-6.828**
				(3.475)	(3.502)	(3.449)
DID (cont.2022)							-0.0502***	-0.00589	-0.0246**
							(0.0125)	0.004217	(0.0124)
Age	0.00663**	0.00939***	0.00757***	0.00663**	0.00939***	0.00757***	0.00663**	0.007634***	0.00757***
	(0.00297)	(0.00294)	(0.00293)	(0.00297)	(0.00294)	(0.00293)	(0.00297)	0.001914	(0.00293)
Age square	-6.70e-05**	-0.000105***	-9.35e-05***	-6.70e-05**	-0.000105***	-9.35e-05***	-6.70e-05**	-9.6E-05***	-9.35e-05***
	(3.37e-05)	(3.34e-05)	(3.33e-05)	(3.37e-05)	(3.34e-05)	(3.33e-05)	(3.37e-05)	2.14E-05	(3.33e-05)
Married	-0.0492**	-0.0676***	-0.0310	-0.0492**	-0.0676***	-0.0310	-0.0492**	-0.08586***	-0.0310
	(0.0197)	(0.0192)	(0.0193)	(0.0197)	(0.0192)	(0.0193)	(0.0197)	0.013649	(0.0193)
No children	0.0192	0.00981	0.00892	0.0192	0.00981	0.00892	0.0192	0.020134	0.00892
	(0.0211)	(0.0207)	(0.0206)	(0.0211)	(0.0207)	(0.0206)	(0.0211)	0.014606	(0.0206)
Education	0.0581***	0.0955***	0.108***	0.0581***	0.0955***	0.108***	0.0581***	0.09367***	0.108***
	(0.0109)	(0.0105)	(0.0108)	(0.0109)	(0.0105)	(0.0108)	(0.0109)	0.004854	(0.0108)
Constant	0.205*	0.101	0.00315	2.030	3.674**	0.00757***	2.145	0.495724	1.026
	(0.113)	(0.105)	(0.106)	(2.139)	(1.842)	(0.00293)	(2.291)	0.040228	(1.912)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes
Observations	5,298	5,324	5,311	5,298	5,324	5,311	5,298	5,314	5,311
R-squared	0.161	0.161	0.172	0.161	0.161	0.172	0.161	0.147	0.172
Note. This table shows the results of the difference-in-differences method. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (columns 1,4 and 7), towards other races (columns 2,5 and 8), and tolerance towards other religions (columns 3,6 and 9). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

5.2.3 Confidence in European Union (EU)

The European Union (EU) plays a role in influencing migration in Turkey through formal negotiations and agreements. For instance, Turkey received €6 billion to improve the humanitarian situation, and Turkish nationals were granted visa-free travel to Europe in exchange for barring Syrian refugees from entering Europe through the Turkish border (International Rescue Committee, 2022). In addition, as the EU’s influence on migration policy and commitment to upholding principles of human rights and dignity, it is important to examine the role of the EU in shaping tolerance towards outer groups. Therefore, the hypothesis of whether confidence in the EU moderates the relationship between refugee influx and tolerance has been tested. The figures below show a bivariate relationship between tolerance towards outer groups and confidence in the EU.

Table 7

The role of having confidence in the EU between refugee influx and tolerance
	(1)	(2)
VARIABLES	confidence EU (Short run)	Confidence EU (Long run)
Treatment area	0.292***	-0.235**
	(0.0906)	(0.103)
Post influx	0.141***	-0.162*
	(0.0188)	(0.0845)
Age	-0.338***	0.182***
	(0.0665)	(0.0445)
Age square	0.00243	0.00261
	(0.00306)	(0.00306)
Marital status: married	-2.66e-05	-2.66e-05
	(3.44e-05)	(3.44e-05)
No children	0.0298	0.0301
	(0.0211)	(0.0212)
Education	0.0318	0.0348
	(0.0228)	(0.0228)
Constant	0.221***	0.397***
	(0.0698)	(0.122)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes
Observations	5,036	5,036
R-squared	0.079	0.076

Note. This table shows the results of first-stage regression in the short run. The observations are at the individual-level unit. The dependent variable measures whether the individual is confident in the EU. Dummy variables are created of individuals who have confidence in the EU. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The first stage regression results indicate that in the short run, in 2011, the refugee influx had a positive impact on having confidence in the EU, however, in the long run, in 2018, the impact became negative, and individuals lost their confidence in the EU. The initial improvement in confidence in the EU might be attributed to the perception of the EU as a competent institution due to its humanitarian response to accepting refugees. Turkey sought to join the EU, and accepting refugees was to demonstrate commitment to EU values and integration measures. However, the EU has not yet granted the extended commitment to Turkey’s membership in the EU and failed to deliver technical and economic support, harming the trust of Turks in the EU (Adam, 2017; Seufert, 2016).

Table 8

The role of having confidence in EU between refugee influx and tolerance in long run
	Short run			Long run
	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religions
Confidence in EU	0.158	0.338**	0.511***	-0.194***	-0.233***	-0.283***
	(0.176)	(0.170)	(0.176)	(0.0515)	(0.0486)	(0.0480)
Age	-0.00657**	-0.0103***	-0.00898***	-0.00279	-0.00886***	-0.00832***
	(0.00304)	(0.00299)	(0.00299)	(0.00306)	(0.00302)	(0.00305)
Age square	6.50e-05*	0.000115***	0.000108***	1.51e-05	9.91e-05***	0.000102***
	(3.43e-05)	(3.39e-05)	(3.37e-05)	(3.42e-05)	(3.40e-05)	(3.43e-05)
Marital status: married	0.0431**	0.0576***	0.0158	0.0537**	0.0869***	0.0570***
	(0.0202)	(0.0197)	(0.0199)	(0.0209)	(0.0203)	(0.0205)
No children	-0.0245	-0.0216	-0.0267	-0.0191	-0.0210	-0.0271
	(0.0219)	(0.0215)	(0.0215)	(0.0221)	(0.0217)	(0.0218)
Education	-0.0532***	-0.0851***	-0.0928***	-0.100***	-0.0756***	-0.0749***
	(0.0120)	(0.0119)	(0.0122)	(0.00634)	(0.00609)	(0.00620)
Constant	0.441***	0.486***	0.424***	0.598***	0.637***	0.663***
	(0.0768)	(0.0760)	(0.0759)	(0.0679)	(0.0670)	(0.0674)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes	Yes Yes
Observations	5,298	5,324	5,311	5,298	5,324	5,311
R-squared	0.158	0.161	0.172	0.049	0.039	0.039

Note. This table shows results of second stage regression. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (columns 1 and 4), towards other races (columns 2 and 5), and tolerance towards other religions (columns 3 and 6). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The results from the two-stage least squares (2SLS) showed that individuals with confidence in the EU tend to have a more tolerant attitude towards outer groups in the short run (see columns 1,2 and 3 of Table 8). For instance, individuals who are confident in the EU are 33.8% and 51.1% more likely to be tolerant towards individuals of other races and religions than individuals who are not confident in the EU. A probable reason is being confident that locals view refugees as part of the shared European community and associate integration with collective responsibility. Therefore, improving unity and willingness to cater to others’ needs, rather than their own self-interest. Whereas individuals with no confidence in the EU might be skeptical about the EU’s jurisdiction and ability to manage or help the refugee crisis, thus perceiving refugees as a threat and burden to society. Therefore, being confident about the EU is important as initiating confidence in the EU might promote greater tolerance and acceptance of refugees in Europe.

However, the positive impact was not endured in the long term, rather created a decline in tolerance towards outer groups. The EU failed to provide enough support for the crisis. For instance, the EU-Turkey agreement, aimed towards restricting Syrian refugees to migrate to Europe through Turkey, in exchange for allowing visa-free travel of Turks to Europe failed. The agreement failed with legal challenges due to human rights accusations of human rights violations. In addition, Turkey is yet to be accepted as a member of the EU. Lastly, the popularity of right-wing parties in Turkey and the relationship between refugees and the EU may have contributed to further stigmatizing outer groups (Cagaptay, 2020; Moldovan, 2017; Niemann and Zaun, 2018; Kaya, 2020).

5.2.4 Robustness checks

5.2.4.1 Placebo

To test whether the model is robust, a placebo test was conducted. The treatment and control groups were randomly chosen from the sample, and the benchmark analysis is run using the randomly allocated groups. The purpose of the test was to test whether our model produces significant results when it should not, as the treatment group is randomly assigned and not related to refugee allocation.

Table 9

Placebo test using the random allocation of treatment and control groups in the short run
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religion
Placebo area	0.0574**	0.0356	0.0390
	(0.0234)	(0.0233)	(0.0240)
Post influx	-0.00185	-0.0548**	-0.0397*
	(0.0234)	(0.0231)	(0.0236)
Placebo*Post	-0.0606	-0.0436	-0.0283
	(0.0326)	(0.0329)	(0.0335)
Age	0.00574	0.0144***	0.00859**
	(0.00369)	(0.00365)	(0.00366)
Age square	-4.75e-05	-0.000151***	-0.000104**
	(4.10e-05)	(4.06e-05)	(4.04e-05)
Marital status: married	-0.0351	-0.105***	-0.0178
	(0.0254)	(0.0250)	(0.0259)
No children	0.0158	-0.0242	0.00429
	(0.0276)	(0.0278)	(0.0280)
Education	0.0667***	0.106***	0.113***
	(0.0115)	(0.0110)	(0.0113)
Constant	0.467***	0.323***	0.403***
	(0.0822)	(0.0825)	(0.0824)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes	Yes Yes
Observations	2,951	2,951	2,951
R-squared	0.068	0.074	0.077

Note. This table shows the results of the difference-in-differences method using randomly allocated treatment and control groups. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

Table 10

Placebo test using random allocation of treatment and control groups in long run
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races	Tolerance Other religion
Placebo area	0.0578**	0.0343	0.0395
	(0.0235)	(0.0233)	(0.0240)
Post influx	0.185***	0.297***	0.289***
	(0.0713)	(0.0602)	(0.0617)
Placebo*Post	-0.0469	-0.0371	-0.0187
	(0.0276)	(0.0273)	(0.0279)
Age	0.00613**	0.00926***	0.00734**
	(0.00297)	(0.00293)	(0.00293)
Age square	-6.03e-05*	-0.000104***	-9.04e-05***
	(3.36e-05)	(3.33e-05)	(3.32e-05)
Marital status: married	-0.0489**	-0.0678***	-0.0312
	(0.0197)	(0.0192)	(0.0193)
No children	0.0183	0.00974	0.00790
	(0.0211)	(0.0207)	(0.0206)
Education	0.0580***	0.0955***	0.108***
	(0.0109)	(0.0105)	(0.0108)
Constant	0.467***	0.323***	0.403***
	(0.0822)	(0.0825)	(0.0824)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes	Yes Yes
Observations	5,298	5,324	5,311
R-squared	0.159	0.161	0.172

Note. This table shows the results of the difference-in-differences method using randomly allocated treatment and control groups. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The results from the table above showed that the DID coefficients for all three dependent variables measuring tolerance towards outer groups show no significant effect, which provides evidence that the model is not sensitive to random variation in the data and is robust.

5.2.4.1 Falsification tests

In addition, to further validate the results. the robustness of our model is evaluated by testing our model on the dataset that falls outside the assumption on which the benchmark analysis is built. The year 2007 was assigned as the false influx year rather than the expected year 2011. Whereas the prior influx years set as 1990 and 1996.

Table 11

Falsification test changing influx year
	(1)	(2)
VARIABLES	Tolerance Immigrants and foreigners	Tolerance Other races
Treatment area	-0.0414	0.0620
	(0.0889)	(0.104)
False post	-0.151*	-0.00957
	(0.0864)	(0.102)
T area * f post	0.0985	-0.0381
	(0.100)	(0.114)
Age	0.00390	0.00672**
	(0.00308)	(0.00307)
Age square	-4.45e-05	-9.05e-05***
	(3.50e-05)	(3.50e-05)
Married	-0.0728***	-0.0958***
	(0.0213)	(0.0205)
No children	-0.0176	-0.00202
	(0.0224)	(0.0213)
Education	0.110***	0.141***
	(0.00850)	(0.00804)
Constant	0.690***	0.468***
	(0.0988)	(0.111)
Province fixed effect Year fixed effects	Yes Yes	Yes Yes
Observations	4,245	4,245
R-squared	0.103	0.124

Note. This table shows the results of the difference-in-differences method using false influx year. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The falsification test result showed no significant coefficient for DID (difference-in-differences). As the incorrect year of 2007 is used for the falsification test, when there was no refugee influx, indicates the model is robust and capable of identifying the lack of effect of the false treatment, strengthening the validity of the results.

6.1 Simple Pre-Post Intervention Regression Analysis

A simple pre-post intervention regression analysis to evaluate the effect of refugee influx on the tolerance of individuals across sixteen German states. As refugees are allocated across each state following a quota-based system as discussed in section 2.0, pre- and post-regression allow capturing the effect on tolerance. In this setting, outcomes are measured at different time points before and after the refugee influx, allowing to capture changes in tolerance levels of individuals to evaluate intervention effects. The design allows testing of potential biases such as autocorrelation, seasonality, heteroskedasticity, and random fluctuations. In addition, a control series that has not been exposed to the intervention can be added to increase the study’s internal validity. The assumption for the large refugee influx to be completely exogenous from the view of the locals would allow for estimating the causal impact of the refugee influx on the tolerance of individuals. The intuition is that due to the exogeneity of the refugee influx, the population of locals before and after the refugee influx have similar characteristics. This relates to the assumption in the model, where the distributions of X_i, and θ_i do not change across time. Thus, any discontinuity in the conditional distribution of outcomes, such as a change in tolerance levels towards immigrants and foreigners, other races, and religions, could be interpreted as the effect of the refugee influx. Thus, evaluating differences in outcomes pre, and post-refugee shock periods. For instance, let r* be the year in which the refugee influx occurred; we can define the treatment indicator Di = 1{t ≥ r ∗}. Therefore, our estimation would be,

Tolerance_ist = α + β₁ Asylum applications_s + B₂Post_t + B₃(Asylum applications_s * Post) + X_ist + ε_ist

In this model, Tolerance _ist is the tolerance of individual i towards outer groups in state s in WVS year t, measured using three dependent variables, tolerance towards foreigners and immigrants, tolerance towards other races, and tolerance towards other religions. The asylum applications are the number of asylum applicants by state, whereas Post is a dummy variable 1 if the year is 2018 (after the 2015-16 refugee influx), 0, otherwise. β₃ would be the parameter (i.e., the average causal effect of the refugee influx on tolerance). According to current literature, additional factors might influence the tolerance of individuals in this pre-post regression setup, therefore additional individual and macro level controls are added. For instance, X_ist is a vector of individual covariates, including the participant’s age, relationship status, number of children, and education. ɛ_ist is the idiosyncratic error term that captures time-varying unobserved heterogeneity. The details of the questionnaires and survey years used for this analysis are listed in Appendix A.

6.2 Results and analysis

6.2.1 Descriptive statistics

Tables 12,13 and 14 provide descriptive statistics of the sample used for this analysis, illustrating an overview of the nature and the characteristics of the data. Table 12 shows the number of observations, mean, standard deviations, and range of the variables used in this study.

Table 12

Summary statistics
	Observations	Mean	Std deviations	Min	Max
Independent variables
Age (years)	3,574	50.04	17.88	17	96
Married (1 = yes)	3,548	0.59	0.49	0	1
No children	3,551	0.30	0.46	0	1
Education	3,574	1.71	0.77	0	3
Dependent variables
tolerance_ immigrants	3,565	0.91	0.29	0	1
tolerance_ other races	3,563	0.86	0.34	0	1
tolerance_other religions	3,568	0.91	0.28	0	1

The table above illustrates that the average age of the sample used in this study is 50.04, whereas around 59% are married. The variables measuring tolerance towards outer groups show high mean values, indicating strong tolerance towards outer groups. In Table 13 the correlation of dependent variables with individual-level characteristics is presented.

Table 13

Correlation with tolerance towards immigrants or foreigners, other races, and religions
	Tolerance Immigrants			Tolerance Other race			Tolerance Other religion
	All (1)	Post-influx (2)	Pre-influx (3)	All (4)	Post-influx (5)	Pre-influx (6)	All (7)	Post-influx (8)	Pre-influx (9)
Male (1 = yes)	-0.02	-0.02	-0.02	-0.04*	-0.06*	-0.03	-0.02	0.01	-0.028
Age (years)	0.004	-0.01	-0.003	-0.02	-0.06*	-0.02	-0.02	-0.11*	-0.004
Married(1 = yes)	-0.02	-0.01	0.0004	-0.01	-0.01	-0.003	-0.02	-0.03	0.002
No children (1 = yes)
Education	0.03*	0.05*	0.08*	0.02	0.05*	0.07*	0.01*	0.03	0.04*

Note: The table shows the correlation between individual controls with tolerance towards foreigners and immigrants, other races, and religions. Columns (1,4,7) show correlation using data from the full sample, whereas columns (2,5,8) show correlation after influx and columns (3,6,9) show before influx.

A positive correlation indicates that both the variables move in tandem, whereas a negative correlation would mean movement in the opposite direction. For instance, education tends to show a positive significant correlation with tolerance towards outer groups. This would mean that individuals with better education are likely to be more tolerant towards outer groups. Followed by Table 14, which exhibits the mean of variables of interest, where columns (1 & 2) show the average of the variables prior to the refugee influx, column (3) reports the mean difference between the treatment and control, whereas column (4) contains the standard errors.

Table 14

Mean and difference between the years prior to refugee influx
	Year 2006 (1)	Year 2013 (2)	Mean Difference (3)	Standard error (4)
Independent variables
Age (years)	49.48	50.44	0.96	0.55
Married (1 = yes)	0.61	0.64	0.03	0.02
No children	0.26	0.29	-0.03	0.01
Education	1.81	1.70	-0.01*	0.02
Dependent variables
tolerance_ immigrants	0.79	0.87	0.08	0.01
tolerance_race	0.86	0.92	0.07	0.01
tolerance_religion	0.87	0.96	0.09	0.01

Note: This table gives the average characteristics of the treatment and control households and the differences before the influx in 2015. Standard errors (S.E.) of the differences are presented in columns (4). The statistical differences are marked as stars. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.

The table above illustrates that there have been no significant differences prior to the refugee influx for tolerance towards outer groups. For instance, the mean difference between tolerance towards immigrants and foreigners were not significant at 0.08 between the year 2013 and 2006.

Table 15

The impact of influx of refugees on tolerance.
	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Tolerance Immigrants Foreigners	+controls	Tolerance Other Races	+controls	Tolerance Other Religions	+controls
Asylum applications	0.510**	0.501**	0.181	0.208	0.205	0.204
	(0.205)	(0.206)	(0.159)	(0.158)	(0.176)	(0.175)
Post influx	-0.547***	-0.504***	-0.621***	-0.627***	-0.608***	-0.582***
	(0.179)	(0.184)	(0.163)	(0.165)	(0.164)	(0.168)
Asylum * Post	0.0669***	0.0635***	0.0695***	0.0705***	0.0675***	0.0654***
	(0.0171)	(0.0176)	(0.0156)	(0.0158)	(0.0157)	(0.0161)
Age		-0.000992		-0.000811		4.35e-05
		(0.00190)		(0.00176)		(0.00164)
Agesquare		1.18e-05		3.71e-06		-4.64e-06
		(1.79e-05)		(1.69e-05)		(1.58e-05)
Married		-0.00521		-0.00934		-0.00790
		(0.0132)		(0.0111)		(0.0108)
No children		0.000743		-0.0219		-0.00435
		(0.0152)		(0.0140)		(0.0133)
Education		0.0311***		0.0198***		0.0129**
		(0.00710)		(0.00618)		(0.00613)
Constant	-4.003**	-3.955**	-0.794	-1.042	-1.027	-1.026
	(1.965)	(1.973)	(1.521)	(1.519)	(1.683)	(1.678)
State fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3,543	3,511	3,545	3,513	3,548	3,516
R-squared	0.089	0.093	0.071	0.075	0.056	0.058

Note. This table shows the results of the regression analysis. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (columns 1 and 2), towards other races columnsn 3 and 4), and tolerance towards other religions (columns 5 and 6). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

6.2.2 Benchmark results

Table 15 shows Eq. (1) estimates illustrated in section 4.2. The regression results of the effect of refugee influx dummy on tolerance of individuals towards immigrants (column 1), other races (column 2), and other religions (column 3). All the regressions are controlled for state-fixed effects.

6.2.1 The effect of refugee influx on tolerance levels of individuals

The positive coefficient of the interaction terms indicates that the influx of refugees makes people more tolerant towards outer groups, immigrants or foreigners, other races, and religions. For instance, individuals after the refugee influx were 6% more tolerant towards immigrants or foreigners, and 7% more tolerant towards people of other races and religions. These results are similar to findings from the Turkish context of this study. The finding can be related to Allport’s (1985) social interaction theory, as interaction with outer groups improves acceptance and tolerance of outer groups. Governments worldwide strive to socially immerse immigrants or refugees in the local community as part of their integration into the new society. The benefits lie both to the overall community and migrants or refugees themselves. Studies have shown how locals are more accepting of the migrant or refugee population when they know, bond with, and trust them. Therefore, the refugee influx, allowed such interaction making them more tolerant towards them (Eaton, 2015; Morrice, 2007; Larsen, 2011). Such development in tolerance indicates that the claim of an overall increase in hate crime is not due to the 2015 refugee influx or is not statistically significant.

6.2.3 The effect of having confidence in the European Union (EU).

The European Union (EU) has a strong role to play in the migration laws of European countries. According to Dublin Regulation, a European Union law, refugees must claim their asylum claim examined in the first EU country they enter (Hurwitz, 1999). Germany, a member of the European Union (EU), currently receives the highest number of asylum applications among other EU countries (Byrne, Noll and Vedsted-Hansen, 2004). Thus, it can be predicted that confidence in the body making decisions regarding migration, the European Union (EU), moderates the impact of refugee influx on intolerance of individuals.

The mean of having confidence in the EU is plotted against the mean of tolerance towards (a) foreigners or immigrants (b) other races sand (c) other religions.

Figure 4 indicates an existent positive trend among variable confidence in the EU with tolerance towards (i) immigrants or foreigners (ii) other races and, (iii) religions. For instance, when confidence in the EU increases, tolerance of individuals towards immigrants or foreigners increases. To investigate, whether individuals with confidence in the EU moderate the impact of refugee influx on intolerance levels; two stages of least square regression is run.

Table 16

The first stage of regression
	(1)
VARIABLES	Confidence EU
Asylum applications	0.471
	(0.416)
Post influx	0.127
	(0.310)
Asylum * Post	-0.00358
	(0.0295)
Age	-0.0147***
	(0.00293)
Age square	0.000136***
	(2.78e-05)
Married	0.0133
	(0.0192)
No children	0.0237
	(0.0231)
Education	0.0613***
	(0.0114)
Constant	-3.911
	(3.975)
State fixed effects	Yes
Observations	3,352
R-squared	0.067

Note. This table shows the results of first stage regression The observations are at the individual-level unit. The dependent variable measures whether the individual has confidence in the EU. Dummy variables are created for individuals who have confidence in the EU. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The first stage regression results above indicate that refugee influx had no influence on confidence in the EU. The individuals have not changed their confidence or support of the EU even after the influx. First, there might be other factors that have stronger effects on the confidence of German nationals towards having confidence in the EU, second, the relationship might be more complex with other factors moderating the impact of refugee influx on confidence in the EU.

Table 17

The second stage regression
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrants Foreigners	Tolerance Other Races	Tolerance Other Religions
Confidence EU	1.790***	1.239***	1.144***
	(0.127)	(0.110)	(0.110)
Age	0.0251***	0.0172***	0.0167***
	(0.00263)	(0.00227)	(0.00214)
Agesquare	-0.000230***	-0.000163***	-0.000159***
	(2.47e-05)	(2.15e-05)	(2.02e-05)
Married	-0.0285**	-0.0253**	-0.0227**
	(0.0132)	(0.0112)	(0.0108)
No children	-0.0438***	-0.0535***	-0.0336**
	(0.0157)	(0.0144)	(0.0136)
Education	-0.0780***	-0.0554***	-0.0566***
	(0.00957)	(0.00825)	(0.00827)
Constant	-0.286***	0.147*	0.188**
	(0.103)	(0.0871)	(0.0864)
State fixed effects	Yes	Yes	Yes
Observations	3,511	3,513	3,516
R-squared	0.090	0.070	0.053

This table shows the results of second- stage regression. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (columns 1 and 4), towards other races (columns 2 and 5), and tolerance towards other religions (columns 3 and 6). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The second stage regression results indicate that having confidence in the EU improves tolerance towards outer groups, tolerance towards immigrants and foreigners, other races, and other religions. As mentioned before, Germany is a member country of the EU, and has to abide by the European Union’s legal framework of migration and refugee laws. In addition, the EU has resource allocation and expertise in managing the refugee crisis, and confidence in the ability of the EU to effectively manage the crisis can be understood to improve tolerance towards outer groups (Carrera, 2016). Lastly, both Durbin (score) along with Wu-Hausman test shows a very low p-value of 0.0001, thus we reject H₀ and conclude that confidence in the EU is endogenous.

6.2.4 Robustness checks

6.2.4.1 Different measure for refugee influx

To test the robustness of the model, substitute measures for refugee influx have been used. The table above shows the regression analysis results using the number of migrants as a measure of refugee influx rather than asylum applications.

Table 18

The impact of the influx of refugees on tolerance by using the number of migrant inflows rather than asylum applications.
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrant Foreigner	Tolerance Other Races	Tolerance Other Religion
Migrant numbers	0.778**	0.323	0.333
	(0.334)	(0.256)	(0.282)
Post influx	-0.259*	-0.414***	-0.427***
	(0.140)	(0.126)	(0.121)
Migrant * Post	0.0353***	0.0442***	0.0446***
	(0.0117)	(0.0106)	(0.0102)
Age	-0.00103	-0.000834	3.76e-05
	(0.00190)	(0.00176)	(0.00164)
Age square	1.22e-05	3.96e-06	-4.52e-06
	(1.79e-05)	(1.68e-05)	(1.57e-05)
Married	-0.00552	-0.00979	-0.00839
	(0.0132)	(0.0111)	(0.0108)
No children	0.000305	-0.0221	-0.00445
	(0.0152)	(0.0140)	(0.0132)
Education	0.0313***	0.0199***	0.0130**
	(0.00710)	(0.00618)	(0.00613)
Constant	-7.478**	-2.502	-2.636
	(3.568)	(2.737)	(3.012)
State fixed effects	Yes	Yes	Yes
Observations	3,511	3,513	3,516
R-squared	0.093	0.075	0.059

Note. This table shows the results of regression analysis using different measures of refugee numbers. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The coefficients of the interaction terms show similar results with the alternate measure of the refugee influx. The coefficients remain negatively significant, indicating that the influx of outer groups makes individuals more tolerant towards them.

6.2.4.2 Falsification test

A falsification test was conducted by imposing a false refugee influx year 2013 rather than 2015. Table 19 reports the results of the manipulated results.

Table 19

The impact of influx of refugees on tolerance of locals using false influx year.
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrant Foreigner	Tolerance Other Races	Tolerance Other Religions
	-0.819***	-0.432***	-0.335***
False influx year	(0.146)	(0.110)	(0.113)
	0.0423	0.180	0.289**
Post	(0.174)	(0.145)	(0.116)
	-0.00231	-0.0151	-0.0235
False * Post	(0.0167)	(0.0140)	(0.0113)
	-0.000111	0.000378	4.49e-05
Age	(0.00160)	(0.00140)	(0.00123)
	-2.56e-06	-1.01e-06	3.44e-06
Age square	(1.52e-05)	(1.34e-05)	(1.19e-05)
	-0.00465	0.00313	0.0117
Married	(0.0111)	(0.00904)	(0.00817)
	-0.0152	0.0149	0.00991
No children	(0.0127)	(0.0111)	(0.0102)
	-0.0291***	-0.0186***	-0.0104**
Education	(0.00569)	(0.00489)	(0.00463)
Constant	7.978***	4.164***	3.201***
	(1.399)	(1.052)	(1.084)
State fixed effects	Yes	Yes	Yes
Observations	5,557	5,559	5,562
R-squared	0.029	0.023	0.024

Note. This table shows the results of regression analysis using false influx year. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The results show DID coefficient is statistically insignificant, suggesting that the assumption in this study that the surge in refugee numbers in 2015 resulted in an improvement in tolerance towards outer groups is robust.

6.2.4.3 Placebo test

A placebo test is run to test whether the estimates capture the real effect of the refugee influx. The placebo dummy is constructed by drawing random individuals from the full sample, irrespective of whether belonging the pre and post-influx periods. Then the benchmark model is run.

Table 20

The impact of the influx of refugees on the tolerance of locals using a random sample.
	(1)	(2)	(3)
VARIABLES	Tolerance Immigrant Foreigner	Tolerance Other Races	Tolerance Other Religions
Random	0.0192	0.0244	0.0220
	(0.0179)	(0.0153)	(0.0149)
Post	0.176***	0.126***	0.115***
	(0.0152)	(0.0134)	(0.0133)
Random * Post	-0.0295	-0.0294	-0.0196
	(0.0206)	(0.0177)	(0.0173)
Age	-0.00126	-0.00103	-0.000266
	(0.00189)	(0.00175)	(0.00164)
Age square	1.39e-05	5.33e-06	-1.92e-06
	(1.77e-05)	(1.67e-05)	(1.58e-05)
Married	-0.00389	-0.00900	-0.00762
	(0.0132)	(0.0111)	(0.0109)
No children	-0.000670	-0.0240*	-0.00692
	(0.0152)	(0.0140)	(0.0134)
Education	0.0324***	0.0205***	0.0143**
	(0.00711)	(0.00617)	(0.00614)
Constant	0.846***	0.924***	0.905***
	(0.0523)	(0.0459)	(0.0444)
State fixed effects	Yes	Yes	Yes
Observations	3,531	3,533	3,536
R-squared	0.091	0.072	0.056

Note. This table shows the results of regression analysis using a random sample. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The interaction term shows insignificant coefficients for all three dependent variables capturing tolerance towards outer groups (see columns 1 to 3) in Table 20. Therefore, further proving the estimates represents the true causal relationship, rather than spurious statistical correlation.

6.2.4.4 The effect of fear on tolerance

The current literature discusses how fear of diseases, loss of jobs, and increased crime create prejudice among individuals, as discussed in section 2.2. Thus, these threat variables are initially plotted against the mean tolerance of individuals towards immigrants, other races, and religions (See Appendix E). From the scatter plots above, it can be noticed that tolerance towards immigrants, other races, and religions is negatively related to physical threat, economic threat, and crime. For instance, figure E1 (a) of Appendix E shows that the greater the physical threat to individuals, the lower the tolerance of individuals toward immigrants or foreigners. Therefore, the influence of such existent fear is added as controls (1) physical threat (2) economic threat, and (3) family victim of crime.

Table 21

The effect of refugee influx on tolerance with additional controls
		(2)	(3)
VARIABLES	Tolerance Immigrant Foreigner	Tolerance Other Races	Tolerance Other Religions
Asylum applications	0.455*	0.307	0.155
	(0.258)	(0.187)	(0.207)
Post influx	-0.485**	-0.682***	-0.635***
	(0.201)	(0.179)	(0.173)
Asylum * Post	0.0522***	0.0686***	0.0660***
	(0.0192)	(0.0170)	(0.0167)
Age	-0.00235	-0.000763	-4.63e-06
	(0.00209)	(0.00189)	(0.00182)
Agesquare	3.22e-05	7.69e-06	-4.73e-07
	(1.99e-05)	(1.85e-05)	(1.76e-05)
Married	-0.0130	-0.00383	0.000826
	(0.0146)	(0.0117)	(0.0122)
No children	-0.00288	-0.0258*	-0.00328
	(0.0164)	(0.0145)	(0.0144)
Education	0.0138*	0.00361	0.00127
	(0.00764)	(0.00638)	(0.00672)
Physical threat	-0.291***	-0.349***	-0.239***
	(0.0248)	(0.0245)	(0.0235)
Economical threat	-0.150***	-0.0486***	-0.0376**
	(0.0180)	(0.0152)	(0.0160)
Family victim crime	-0.0799***	-0.0332*	-0.0378*
	(0.0256)	(0.0201)	(0.0216)
Constant	-3.360	-1.911	-0.496
	(2.475)	(1.796)	(1.987)
State fixed effects	Yes	Yes	Yes
Observations	2,741	2,743	2,743
R-squared	0.225	0.250	0.156

Note. This table shows results of regression analysis using additional controls. The observations are at the individual-level unit. The dependent variable measures whether the individual is tolerant towards immigrants and foreigners (column 1), towards other races (column 2), and tolerance towards other religions (column 3). Measured using WVS/EVS questionnaire, could you mention any that you would not like as neighbors? Dummy variables are created of individuals who chose (i) immigrants/foreigners (ii) people of other races, and (iii) other religions. In parentheses are the robust standard errors. ∗Significant at 10% level; ∗∗significant at 5% level; ∗∗∗significant at 1% level.

The DID regression is run with additional controls in the table above. All three variables physical threat, economic threat, and family victim of crime show a negative significant coefficient for tolerance towards outer groups (e.g., tolerance towards immigrants or foreigners, other races and religions). The results explain that individuals who tend to have physical (fear of diseases) or economic (loss of jobs) threats, or their family has been the victim of crime are less tolerant towards outer groups. These findings shed light on the understanding of the development of intolerance towards immigrants, other races, and religions. For instance, individuals are understood to associate fear of diseases, loss of jobs, or crime with immigrants, they are feared to bring diseases, create job scarcity, and indulge in crime. Nonetheless, the interaction terms (asylum*post) remain significant with the additional controls, further justifying our model is robust.

7.1 Implications

From the finding that interaction with refugees improves tolerance toward outer groups, policymakers should immerse the refugees in the local community through social events, employment opportunities, and training. Therefore, allowing better relationships to be formed, and overcoming fear or existing stereotypical beliefs. The EU along with the government bodies should strive to be more efficient and effective in resolving the refugee issue, as both Turks and Germans with greater confidence in the EU, showed better tolerance towards outer groups. Lastly, better information should be shared on outer groups to overcome physical or economic fears, and misconceptions of relating crime to outer groups.

7.2 Conclusion

This study investigates the influence of refugee influx on tolerance towards outer groups in Turkey and Germany. I have used quasi-experimental study designs, the difference in differences (DID), and simple pre-post regression analysis for Turkey and Germany. To capture the tolerance levels and other demographic characteristics across both countries, World Value Survey (WVS) and European Value Survey (EVS) were used, and the macro indicators were taken from their government statistical departments. The results show that in the short run, the refugee influx made individuals in Turkey and Germany more tolerant towards outer groups. The finding supports Allport’s (1958) social interaction theory, which implies interaction improves tolerance due by fostering mutual understanding and appreciation of mutual perspective. However, as individuals got less tolerant with time, the effect did not sustain in the long run in Turkey. The humanitarian requirement to help or provide support and aid might have improved tolerance shortly after the refugee influx; however, the competition towards limited resources between the refugee and local population created the vice versa effect in the longer run. Two-stage least square (2SLS) test was used to investigate whether having confidence in the EU had a moderating role in this relationship. As the EU has a definite role in influencing migration and refugee integration through its legal framework and funding, individuals from both countries became more tolerant towards outer groups with confidence in the EU. These findings highlight the importance of effective policies and strategies for promoting integration and tolerance towards refugees and other outer groups. Overall, the study provides valuable insights into the factors that influence tolerance towards outer groups in the context of the refugee influx. These results have important implications for policymakers and others working to promote integration and understanding among diverse communities. However, there are certain limitations of this study. First, the study uses WVS and EVS data sources, which tend to have data points across provinces and states of Turkey and Germany; a greater micro-level study might provide a more holistic understanding. For instance, distinguishable differences in refugee numbers as certain districts have a negligible number of refugees, while others have significant, district-level comparison would allow better study of tolerance with their interaction and exposure to refugees. Second, the answers to questions that captured the outcome variables, aversion to neighbors who are immigrants or foreigners, or belong to other races, and religions can get influenced by social desirability bias. As respondents might answer what they might believe as socially desirable or expected rather than their real opinion.

Competing interests: The authors declare no competing interests.

Batalla Adam, L., 2017. The EU-Turkey deal one year on: A delicate balancing act. The International Spectator, 52(4), pp.44-58.
Byrne, R., Noll, G. and Vedsted-Hansen, J., 2004. Understanding refugee law in an enlarged European Union. European Journal of International Law, 15(2), pp.355-379.
Cagaptay, S., 2020. The new sultan: Erdogan and the crisis of modern Turkey. Bloomsbury Publishing.
Chou, W., Dancygier, R., Egami, N. and Jamal, A., 2018. The Illusion of Far-Right Partisan Stability: How Party Positioning Affects Far-Right Voting in Germany.
Côté, R.R. and Erickson, B.H., 2009. Untangling the roots of tolerance: How forms of social capital shape attitudes toward ethnic minorities and immigrants. American Behavioral Scientist, 52(12), pp.1664-1689.
Carrera, S., Blockmans, S., Gros, D. and Guild, E., 2015. The EU's response to the refugee crisis: Taking stock and setting policy priorities. CEPS essay, (20/16).
Cowling, M.M., Anderson, J.R. and Ferguson, R., 2019. Prejudice-relevant correlates of attitudes towards refugees: A meta-analysis. Journal of Refugee Studies, 32(3), pp.502-524.
Cowling, M.M. and Anderson, J.R., 2019. The role of Christianity and Islam in explaining prejudice against asylum seekers: Evidence from Malaysia. The International Journal for the Psychology of Religion, 29(2), pp.108-127.
Crowley, F. and Walsh, E., 2018. How important are Personal Ties, Trust and Tolerance for Life Satisfaction in Europe? (No. SRERCWP2018-1). SRERC Working Paper Series.
Das, E., Bushman, B.J., Bezemer, M.D., Kerkhof, P. and Vermeulen, I.E., 2009. How terrorism news reports increase prejudice against outgroups: A terror management account. Journal of Experimental Social Psychology, 45(3), pp.453-459.
Davis, L. and Deole, S.S., 2017. Immigration and the Rise of Far-right Parties in Europe. ifo DICE Report, 15(4), pp.10-15.
Crawley, H., Drinkwater, S. and Kauser, R., 2013. Regional variations in attitudes towards refugees: evidence from Great Britain.
Eagly, A.H. and Chaiken, S., 1998. Attitude structure and function. The handbook of social psychology, 1(2), pp.269-322.
Eaton, S.E., 2015. Integration nation: Immigrants, refugees, and America at its best. New Press, The.
Erisen, C. and Kentmen-Cin, C., 2017. Tolerance and perceived threat toward Muslim immigrants in Germany and the Netherlands. European Union Politics, 18(1), pp.73-97.
Eurostat 2019. Asylum and first time asylum applications [online] Available at: https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=migr_asyappctza&lang=en [Accessed 05 October 2020]
Fiske, S.T., 1998. Stereotyping, prejudice, and discrimination. The handbook of social psychology, 2(4), pp.357-411.
Foster, C. and Frieden, J., 2017. Crisis of trust: Socio-economic determinants of Europeans’ confidence in government. European Union Politics, 18(4), pp.511-535.
Gehrsitz, M. and Ungerer, M., 2017. Jobs, crime, and votes: A short-run evaluation of the refugee crisis in Germany.
Gniewosz, B. and Noack, P., 2008. Classroom climate indicators and attitudes towards foreigners. Journal of adolescence, 31(5), pp.609-624.
Huang, J.Y., Sedlovskaya, A., Ackerman, J.M. and Bargh, J.A., 2011. Immunizing against prejudice: Effects of disease protection on attitudes toward out-groups. Psychological Science, 22(12), pp.1550-1556.
Hurwitz, A., 1999. The 1990 Dublin Convention: a comprehensive assessment.
International Labour Organization. (2022). iLOSTAT. Available at: https://ilostat.ilo.org/data/ [Accessed: 07 February 2023].
International Rescue Committee. (2022). What is the EU-Turkey deal? Available at: <https://eu.rescue.org/article/what-eu-turkey-deal>[Accessed 6 February 2022].
Jackman, M.R., 1978. General and applied tolerance: Does education increase commitment to racial integration?. American Journal of Political Science, pp.302-324.
Janmaat, J.G. and Keating, A., 2019. Are today’s youth more tolerant? Trends in tolerance among young people in Britain. Ethnicities, 19(1), pp.44-65.
Kaya, A., 2020. Right-wing populism and Islamophobism in Europe and their impact on Turkey–EU relations. Turkish Studies, 21(1), pp.1-28.
Larsen, B.R., 2011. Becoming part of welfare Scandinavia: Integration through the spatial dispersal of newly arrived refugees in Denmark. Journal of Ethnic and Migration Studies, 37(2), pp.333-350.
Lazarev, E. and Sharma, K., 2017. Brother or burden: An experiment on reducing prejudice toward Syrian refugees in Turkey. Political Science Research and Methods, 5(2), pp.201-219.
Mazzoni, D., Cicognani, E., Tzankova, I., Guarino, A., Albanesi, C. and Zani, B., 2020. Civic Participation and Other Interventions That Promote Children’s Tolerance of Migrants. In Children and Peace (pp. 89-102). Springer, Cham.
Moldovan, Carmen. "Is the EU-Turkey Action Plan an effective or just an apparent solution to the refugee crisis?." CES Working Papers 9, no. 3 (2017): 195-212.
Morrice, L., 2007. Lifelong learning and the social integration of refugees in the UK: The significance of social capital. International Journal of Lifelong Education, 26(2), pp.155-172.
Niemann, A. and Zaun, N., 2018. EU refugee policies and politics in times of crisis: Theoretical and empirical perspectives. JCMS: Journal of Common Market Studies, 56(1), pp.3-22.
Petty, R.E. and Wegener, D.T., 1998. Matching versus mismatching attitude functions: Implications for scrutiny of persuasive messages. Personality and Social Psychology Bulletin, 24(3), pp.227-240.
Pratto, F. and Stewart, A.L., 2012. Group dominance and the half‐blindness of privilege. Journal of Social Issues, 68(1), pp.28-45.
Sassenberg, K., Moskowitz, G.B., Jacoby, J. and Hansen, N., 2007. The carry-over effect of competition: The impact of competition on prejudice towards uninvolved outgroups. Journal of Experimental Social Psychology, 43(4), pp.529-538.
Seufert, G., 2016. Turkey as partner of the EU in the refugee crisis: Ankara's problems and interests.
Shin, H. and Dovidio, J.F., 2018. Differences, threats, values, and country‐specific prejudice toward immigrants and foreign workers in three major receiving countries: The United States, Germany, and Australia. Journal of Social Issues, 74(4), pp.737-755.
Sidanius, J., Devereux, E. and Pratto, F., 1992. A comparison of symbolic racism theory and social dominance theory as explanations for racial policy attitudes. The Journal of Social Psychology, 132(3), pp.377-395.
Sidanius, J. and Pratto, F., 2001. Social dominance: An intergroup theory of social hierarchy and oppression. Cambridge University Press.
Stahler, N., 2017. A model-based analysis of the macroeconomic impact of the refugee migration to Germany.
Tajfel, H., Turner, J.C., Austin, W.G. and Worchel, S., 1979. An integrative theory of intergroup conflict. Organizational identity: A reader, 56, p.65.
UNESCO, 1995. Declaration of principles on tolerance. Culture of Peace Programme.
UNHCR. 2020. Figures At A Glance. [online] Available at: <https://www.unhcr.org/en-my/figures-at-a-glance.html> [Accessed 7 July 2020].
UNHCR 2020, Germany. [online] Available at: <https://help.unhcr.org/germany/rights-and-duties/rights-and-duties-during-the-asylum-procedure/> [Accessed 05 October 2020]
Vieten, U.M. and Poynting, S., 2016. Contemporary far-right racist populism in Europe.
Vogt, W.P., 1986. Education and Tolerance in Comparative Perspective. The Review of Education, 12(1), pp.9-18.
Vogt, W.P., 1997. Tolerance & education: Learning to live with diversity and difference. Sage Publications, Inc.

Appendices.docx