What factors determine attitudes to immigration? A meta-analysis of political science research on immigration attitudes (2009-2019)

Attitudes toward immigration have attracted much scholarly interest and fuelled extensive empirical research in recent years. Many different hypotheses have been proposed to explain individual and contextual differences in attitudes towards immigration. However, it has become difficult to align all of the evidence that the literature has produce so far. The present article contributes to the systematization of political science empirical research on public attitudes toward immigration in the last decade. Using a simplified combined-tests technique, this paper identifies the microas well as macro-level factors that are consistently linked to attitudes toward immigration. It reports findings from a meta-analysis of the determinants of general attitudes toward immigration in published articles in thirty highly ranked peer-reviewed political science journals for the years 2009 – 2019. The results warrant a summary of factors affecting attitudes to immigration in a systematic, measurable and rigorous manner.


1
Introduction 1 Attitudes to immigration, immigrants and refugees have become a highly salient issue in many countries, particularly in the aftermath of the so-called "migration crisis". While increasing proportions of immigrants in Western societies are viewed positively by some, stressing immigration benefits, others view these demographic changes with suspicion. Consequently, social scientists and, in particular, political scientists, have dedicated considerable attention to the factors that might explain individual attitudes toward immigration in recent years. However, as many hypotheses and factors have been proposed, ranging from intergroup contact and residential context to the role of personal predispositions, it has become increasingly difficult to see the wood for the trees.
Disagreement over what drives people's attitudes to immigration persists. The literature on public opinion toward immigration predominantly focuses on two main types of factors affecting these attitudes. The first factor is individual-level indicators such as age, gender, education, left-right positioning, etc. The second approach is to look at macro-level indicators such as GDP per capita, the share of the population that is unemployed, or the share of immigrants in the country. This article aims at assessing recent empirical evidence on what individual and contextual level factors are consistently linked with general attitudes to immigration and which are not. I make a contribution to the literature by (a) providing a systematic overview of factors linked to individual-level attitudes toward immigration in the political science literature and (b) evaluating which of these factors were consistently found to explain individual-level attitudes toward immigration in empirical research.
Several publications (Ceobanu and Escandell 2010;Hainmueller and Hopkins 2014;Dennison and Dražanová 2018) have already provided comprehensive reviews regarding attitudes to immigration. While these reviews have many merits, this article intends to advance previous work in several ways. Firstly, the above-mentioned studies are mostly concerned with attitudes to immigration policy in North America and Western Europe (Hainmueller and Hopkins 2014) or they limit their review to studies that make use of multinational survey projects (Ceobanu and Escandell 2010). In contrast, this article assesses general attitudes regarding the effect of immigration, without limiting its scope geographically and also including single country studies. For instance, data from Central and Eastern Europe are also included, as well as those from New Zealand. Secondly, given recent developments in many countries around the world, epitomized by Donald Trump's victory in the American presidential election in 2016 and the Brexit vote, empirical research regarding attitudes to immigration has unprecedently flourished in recent years. Therefore, new data have been released (new waves of cross-country longitudinal datasets such as European Social Survey (ESS), European Value Study (EVS) and the World Value Survey (WVS)). New theories (see for example Pardos-Prado and Xena´s (2019) theory regarding individuals with low transferable skills in the labor market articulating a subjective sense of job insecurity and consequently higher hostility toward migrants) and new hypotheses (see for example Aarøe, Bang Petersen and Arceneaux (2017)´s proposition of individuals high in behavioral immune sensitivity being more opposed to immigration) have been also proposed. Thirdly, I propose a more advanced technique to assess the determinants of attitudes to immigration compared to the systematic reviews available so far (Ceobanu and Escandell 2010;Hainmueller and Hopkins 2014;Dennison and Dražanová 2018). I advance a quantitative (meta-analytical) procedure used to systematically and statistically combine the results of previous studies regarding factors that are consistently linked to attitudes to immigration. I ask two research questions: (1) What indicators are most frequently used in quantitative studies to explain individual attitudes to immigration? (2) Within the political science literature, what individual and contextual indicators are consistently found to influence individual attitudes to immigration? 1 The author would like to thank Catriona Harris and James Dennison for research assistance in the initial process of identifying relevant articles. Robert Schuman Centre for Advanced Studies Working Papers This article is structured in four parts. I begin by presenting the sample of quantitative studies and the research strategy. I then systematically analyze the individual factors which are most frequently used in these quantitative studies to explain attitudes to immigration. I then shift to contextual level factors at the regional and country level. Finally, I summarize my findings and discuss some possibilities for further research.

Data and Methods
This section provides information about the literature search, research design and meta-analytical strategy. The aim of this meta-analysis is not the selection of all studies, or a representative selection of all studies, but rather the selection of 'best' studies published in the last decade. Published work in the top-ranked political science journals has gone through the process of rigorous peer review and is therefore supposed to be of high quality and report more reliable results. My goal is to include general political science journals as well as journals specializing in attitudinal research. I also aim to strike a balance between European and American journals. I have chosen the top-ranked political science journals as a combination of the ranking of Clarivate Analytics and Google Scholar rankings. Table 1 provides a list of political science journals that have been included as information sources. The data used for this study are collected as part of a larger project that seeks to systematize the literature on the factors affecting attitudes to immigration. The sample of articles for the meta-analysis below has been selected based on a search for all articles fulfilling the established criteria directly in peer-reviewed political science journals' own database, within the publication timeframe 2009-2019. Admittedly, the ten-year timeframe is not based on a rigorously defined pre-established criterion but has been selected as an ad hoc reflection of recent developments in the empirical political science research of attitudes toward immigration. The first pre-selection was done by at least two independent coders, who used the search terms "immigrant" and "immigration". In cases where they disagreed, a third coder assessed the inclusion of the article. The criteria for the final selection of the dependent variable used are described in detail below. The relevant studies and estimates were identified using a detailed protocol describing the inclusion criteria (e.g. how attitudes toward immigration are defined, the unit of analysis for the dependent variable etc).

Dependent variable
This meta-analysis focuses only on studies explaining attitudes regarding the consequences of immigration for the receiving societies. This may include attitudes regarding the general effect of immigration on society. It might also include other attitudes regarding the effects of immigration on culture, economics, crime rates, etc., but only when these items are aggregated into summative or averaging multi-item indices. Studies that look at only one factor on the effect of immigration (for instance, the effect of immigration on the national economy) are excluded from this analysis. Composite measures were chosen in order to avoid measurement error and to reveal issue preferences that are well structured and stable (Ansolabehere, Rodden and Snyder 2008). Similarly, focusing on one type of dependent variable and excluding, for instance, dependent variables that focus on the willingness to admit immigrants with questions such as "should the number of immigrants from foreign countries permitted to come to live here be increased a lot or decreased a lot?" is done for comparability reasons. It is plausible that factors affecting attitudes regarding the effects of immigration might affect attitudes to immigration policy differently.
As shown in Figure 1, 2692 potentially relevant studies were identified. From these, 152 articles have been identified as using attitudes to immigration, broadly defined, as their dependent variable. Further restrictions concerning the eligibility criteria have been applied to make the models explaining attitudes toward immigration as comparable as possible. These have been further reduced to 33 research articles. Nevertheless, as the overall number of variables used in the studies is extremely high (115 in total), describing each of the variables separately is well beyond the possibilities of this paper. I therefore present only a selected number of variables, which further reduces the number of studies de facto used for the principal analysis into 23. Table 2 provides a list of studies that have been used for the metaanalysis of variables presented in this article. Some of the articles include more than one independent samplefor instance, Aarøe, Bang Petersen and Arceneaux (2017) use four samples of respondents across the United States and Denmark to test their hypothesis. Therefore, although the meta-analysis is based on 23 published studies, it relies, in fact, on 37 independent survey samples. Robert Schuman Centre for Advanced Studies Working Papers

Figure 1. Flow diagram of article identification for meta-analysis
The items comprising the different indices are not necessarily the same across all studies. Furthermore, even when items used to measure attitudes to immigration are, in fact, the same and based on the same data source, many researchers seem to apply a different and unique terminology. For instance, while using the same set of items, the explanatory variable might be labelled as diversely as anti-immigration attitudes (Pardos-Prado and Xena 2019), anti-foreigner sentiment (Frølund  or perceived consequences of immigration . Thus, the outcome variables to be considered in this meta-analysis are selected according to the question(s)/indicator(s) used to measure them rather than according to their name. Table 3 provides an overview of the questions/indicators used to measure the outcome variable in each study, as well as the name for the dependent variable used by the authors of the study.   Immigrants take jobs away from people who were born in the U.S. [Denmark].
 Immigrants spur higher economic growth.
 Immigrants contribute to higher crime rates.
Berg (  Most people who come to live here work and pay taxes. They also use health and welfare services. On balance, do you think people who come here take out more than they put in or put in more than they take out?

What factors determine attitudes to immigration? A meta-analysis of political science research on immigration attitudes (2009-2019)
European University Institute 9  People who come to live and work here generally harm the economic prospects of the poor more than the rich.
 If people who have come to live and work here are unemployed for a long period, they should be made to leave.
 People who have come to live here should be given the same rights as everyone else.
 If people who have come to live here commit a serious crime, they should be made to leave.
 If people who have come to live here commit any crime, they should be made to leave. (2018) 2 perceptions of threat related to immigration

Homola and Tavits
People have different views on the consequences of immigration. Thinking about the effects of immigration, please tell us how concerned (not at all concerned, not too concerned, somewhat concerned, very concerned) you are about each of the following? The effect of immigration on  violence and vandalism in my community.
 violence and vandalism in the country as a whole.
 the economic conditions in my household.
 the economic conditions in the country as a whole.
 the American national identity the American culture. Hopkins (2011) 2 levels of concern about local immigration as a problem  How big a problem is immigration in your local community?

Hooghe and Quintelier
Johnston, Newman, and Velez (2015) 2 threat from ethnic diversity  Ethnic change has had positive effects on the local culture (e.g., opportunities for new experiences) or negative effects (e.g., a threat to status quo)?
Just and Anderson (2015) 1 the perceived consequences of immigration  Immigration is bad or good for their country's economy.
 Immigrants undermine or enrich the country's cultural life.
 Immigrants make the country a worse or better place to live.

Langsaether and
Stubager (2019) 1 immigration attitudes The index is based on questions related to how the respondent feels about immigration and immigrants: are they seen as a cultural threat, should they maintain or abandon their customs and traditions, are they a strain on the welfare system or not etc.  2 concern about immigration  Are you very concerned or not concerned about immigration?
Pardos-Prado and Xena (2019) 2 anti-immigration attitudes  Immigration is bad or good for the country's economy.
 The country's cultural life is undermined or enriched by immigrants  Immigrants, make the country a worse or better place to live.

Meta-analytical strategy
This study aims to provide an overview of recent research evaluating factors affecting general attitudes to immigration. Thus, I analyze results from previous studies and summarize the findings via quantitative methods (Smets and van Ham 2013). By doing so, I conduct "an analysis of analyses" (Glass 1976, 3). The articles in my sample use different statistical techniques and therefore provide different test statistics. Following common practice within meta-analytical studies in such cases, I use the vote-counting procedure (Imbeau et al. 2001;Smets and van Ham 2013) which classifies each independent variable into one of three categories. A variable is considered a "success" if it is significantly related to the dependent variable and its effect is in the hypothesized direction. On the other hand, an independent variable is considered a "failure" when its effect on the dependent variable is nonsignificant. Finally, a variable is considered an "anomaly" when its effect is statistically significant, but in the opposite direction than hypothesized. I consider the two-tailed p < 0.05 level as the cut-off point for significant effects.
While the majority of studies frames the dependent variable in negative terms (with higher numbers meaning more anti-immigration attitudes), some studies focus on favorable attitudes towards immigration, thus using the scale of the dependent variable in the opposite direction. To make the effects comparable across studies, the classification of variables from studies where higher numbers of the dependent variable will reflect more positive attitudes to immigration has been reversed. Therefore, all hypotheses and classifications of the independent variables have been done in relation to antiimmigration attitudes.
In a second step, the success rate of each independent variable under consideration is calculated. The success rate is an estimate of the most frequent relationship between the independent variable and attitudes to immigration. It is calculated by dividing the number of "successes" by the total number of times that the variable has been tested in a model. The success rate gives an overview of a variable´s influence on attitudes toward immigration. The higher it is, the more confident one can be that the independent variable has the hypothesized effect in terms of direction as well as significance.
Some studies include more than one analysis regarding the factors affecting attitudes to immigration. For example, certain studies present more than one model per analysis, with each model including additional variables, or they conduct additional robustness tests on the same data. In such cases, there is a risk of artificially inflating the success rate, as these multiple tests might be based on the same data and thus not, in fact, independent from each other. Therefore, I have only included the most inclusive model (i.e the one where all variables are included) to maintain the independence of observations. 2 On the other hand, certain studies conduct several analyses based on different data. For example, they conduct a separate analysis for two countries based on two country-specific surveys. In these cases, both models have been included in the meta-analytical review as the independence of models is preserved.
Given that I am interested in the average effects on attitudes toward immigration, I would ideally only include variable estimates based on representative samples from the general population. However, this is not possible in practice given sampling bias and other issues. As a second-best option, I excluded estimates that refer to very specific or idiosyncratic populations. Such estimates may be the result of the sampling strategy in a given study. For instance, many researchers limit their analysis to only the native population in the sample, although their operationalization might differ. Some samples are even more specific, for instance, they include only males (Aarøe et al. 2017;van Assche et al. 2017) or only the young (van Assche et al. 2017). In some instances, the variable is part of an interaction model and thus 2 Sometimes, an independent variable is, however, only included in one model and not in another. In these cases, the single variable was recorded from the other model.

European University Institute 13
its effect is estimated across a very specificand in some cases non-existentsubgroups. These might include estimates for individuals with "low right-wing orientation" (van Assche et al. 2017), or skilled workers with offshorable jobs . The coefficients of the independent variables across these narrowly defined subgroups have been excluded. I include, however, estimates for more common samples such as for individuals living in countries considered young democracies (Frølund . To compare the effects of each variable across several studies and classify whether they fall into the category "success", "failure" or "anomaly", it is crucial to keep the hypothesized direction of the variables´ effect constant across studies. For instance, certain studies might hypothesize that being religious may affect attitudes to immigration negatively, while other studies may expect that religiosity affects attitudes to immigration positively. For reasons of comparability, it is important to maintain a single expectation (hypothesis) for an independent variable across all studies. In some cases, the theoretical expectations regarding the direction of an independent variable´s effect are apparent (for instance, education is almost exclusively theorized to affect attitudes to immigration positively), while for certain variables a priori theoretical expectations are much less clear and differ from study to study. For instance, following ethnic competition theory, a high number of immigrants in one´s own country can be expected to have negative effects on attitudes to immigration. On the other hand, following contact theory, one could expect higher numbers of immigrants to affect individual attitudes to immigration positively. Table 4 shows the hypothesized relationship with anti-immigration attitudes for each independent variable that is held constant across all studies independent of their own theoretical expectations.

Predictors of anti-immigration attitudes
In this section, I report the most commonly included variables thought to explain anti-immigration attitudes. 3 For ease of comparison, I group my findings into several categories, such as sociodemographic or economic factors, and present individual results under each of these categories. However, these categories are arbitrary and not necessarily exclusive. For instance, income can be considered a sociodemographic variable, however, here income is categorised under the "economic factors" category.

Sociodemographic characteristicsage, education, gender and place of residence
Often, a person's age tends to be included in regression models explaining attitudes toward immigration as a routine demographic control variable. Older respondents are often hypothesized to hold more antiimmigration attitudes. However, recent studies show that when isolating the effect of birth cohorts, a person's biological age is no longer significant (Gorodzeisky and Semyonov 2018) and that older individuals are more averse towards immigrants not because they become more critical towards immigration policies over the life-cycle, but because of cohort or generational effect (Schotte and Winkler 2018, Jeannet and Dražanová 2019). The unclear relationship between age and antiimmigration attitudes is supported by the meta-analytical results. Older respondents are found to be more anti-immigration roughly half of the time, which is reflected in the success rate of 54 percent.
Educational attainment is also one of the most commonly used predictors of anti-immigration attitudes. Theoretically, there is a rather strong consensus within the literature that higher educated individuals hold less anti-immigration attitudes compared to those with lower education. However, 3 I do not report on independent variables that have been used only in few studies as results with only very few observations are not as reliable.
several studies have shown that the strengths of the relationship depends on contextual factors (Borgonovi and Pokropek 2019) and does not always hold outside the established Western realm (Dražanová 2017). The expected educational impact is confirmed by the meta-analysis. Higher educated individuals hold, indeed, more pro-immigration attitudes than the lower educated. The meta-analysis demonstrated an 87 percent success rate, with 27 tests out of 31 were classified as successful. However, it should be noted that the countries covered by the meta-analytical sample are mostly the United States and Western Europe. Therefore, more analysis is needed to assess whether this relationship holds also in countries with a different political, cultural, economic and historical background.
Research on anti-immigrant attitudes does not typically systematically analyze gender patterns. Theoretically, it is not entirely clear whether women or men have more anti-immigration attitudes. Studies generally assume men should hold more anti-immigration attitudes, due to their more authoritarian personalities (Adorno et al. 1950) and conservatism (Harteveld et al. 2015: 107). However, with the recent politicization of gender in immigration debates (Farris 2017), native women might view certain immigrants as a threat to gender equality (Ponce 2017). The meta-analysis shows that gender in most studies is a non-significant factor in explaining individual differences in attitudes to immigration. When significant, women are roughly equally likely, or even slightly more likely, to hold antiimmigration attitudes than men.
Individuals living in urban areas are often predicted to hold more positive immigration attitudes. This is theoretically based on the contact hypothesis or compositional effects . The empirical link between living in a rural area and being anti-immigration is as strong as expected, with a success rate of 80 percent.

Being native, immigrant background, ethnicity and race
Citizenship and race are included in models of attitudes to immigration based on the idea that immigrants and ethnic and racial minorities are more favorable to immigration because they can identify more strongly with other immigrants due to their own migration history (Becker 2019) or due to their similar outgroup status.
Being non-White or a member of an ethnic minority such as Maori  was significantly related to attitudes to immigration in about half of the models (50 percent success rate). Interestingly, in two studies being Black (compared to Whites and others) has resulted in significantly more antiimmigrant attitudes (Bramlett et al. 2011;. Different studies operationalized being part of the native population in different wayssome include a question specifically asking whether a respondent is a citizen of the country  while some focus on asking whether the individual has been born in a country  or whether at least one parent has been born outside of the country (Hooghe and Quintelier 2013). I combine all these possible operationalizations in a variable comparing the immigration attitudes of respondents who are citizens of the country they reside in and where the respondent and both of their parents have been born, and compared this to individuals who do not satisfy all of these three conditions at the same time. The expectation is that native respondents will be more anti-immigration than individuals who are not citizens; have been born outside the country; or have at least one parent born outside the country. The meta-analytical results overall confirm these expectations, although the success rate is only around 55 percent.

Religiosity and church attendance
The relationship between religiosity, church attendance and ethnic prejudice, in general, has been classified as complicated ever since Allport (1954, 413) puzzled over the fact that people who endorsed religious teachings of egalitarian and humanitarian values also showed high levels of (racial) prejudice.
It was described as a paradox in which religion both "makes and unmakes" prejudice. The paradox likely reflects basic group dynamics in which identification with a religious in-group promotes out-group derogation and that religious ingroups are often divided along ethnic or racial lines.
In line with this, previous research has found religiosity to play a positive effect on attitudes to immigration (Bohman and Hjerm 2014). On the contrary, others find that greater religiosity leads to more prejudice (Scheepers, Gijsberts, and Hello 2002). Moreover, some research observes that religion has no effect on attitudes toward immigrants (Creighton and Jamal 2015).
Despite expecting that religiosity would have negative effects on attitudes to immigration, results from the meta-analysis show that religiosity (0 %) and church attendance (20 %), in fact, do not contribute to more negative attitudes toward immigration. Had the hypothesis been framed in the reverse and if we had expected that religiosity and church attendance would lead to more positive attitudes to immigration, their success rate would be around 50% for religiosity and 40% for church attendance. However, it should be noted that the number of tests is arguably low to jump to any definitive conclusions.

Economic factors: Income, occupational status, occupational type, and personal and national economic difficulties
The personal economic situation is often considered as one of the driving forces behind individual differences in attitudes to immigration. This is based on realistic group conflict theory which argues that groups are competing for scarce resources (Quillian 1995). Studies assume that the most economically vulnerable are the most likely to be convinced by propositions of protectionism and the elimination of competition from immigrants for jobs, welfare benefits and even housing (Hainmuller and Hiscox 2010). However, the meta-analysis presented here shows that the empirical link between a personal economic outlook and anti-immigration attitudes is less strong than expected. In most instances, income is not a significant predictor of individual anti-immigration attitudes. The success rate of income is arguably low (33 percent) since in most studies income does not reach statistical significance. Even economic satisfaction, which is sometimes considered a better predictor as a personal economic outlook on attitudes to immigration than income, has a low success rate and is found insignificant in most studies. 4 Based on a similar theoretical framework, known as labour market competition theory, I also expected that those who are unemployed or those working in unskilled or manual jobs were more likely to be against immigration. However, occupational status does not appear to have a significant effect on attitudes to immigration in most models. Experience with unemployment contributed to more antiimmigration attitudes in only around 20 percent of studies. The second variable related to occupation is the type of work individuals have. It proved slightly difficult to combine the employment categories, as different studies use different classifications. For instance, "manual worker" is the only employment category in Pardos-Prado and Xena (2019), which signifies that manual workers are significantly more or less likely to hold anti-immigration attitudes compared to all other employment categories. On the other hand,  distinguish between several categories of occupations such as higher and lower service class, routine non-manual employees, petit-bourgeois, skilled and unskilled manual workers. In the meta-analysis, I combined categories so that the results reflect the effects of being an unskilled manual worker compared to all other employment categories. Manual workers are, indeed, found to be more-anti-immigration than individuals from other occupations in a majority of studies with a success rate of about 60 percent. 4 Social class, on the other hand, seems to be a better predictor of anti-immigration attitudes. However, it has been used only in two tests and thus it is not included among the main results. Robert Schuman Centre for Advanced Studies Working Papers What should be emphasized is that anti-immigration attitudes are not necessarily based on personal economic outlooks, but also on how immigration is perceived to affect society as a whole. Immigrants may be perceived as a risk to the national economy rather than to the personal economic situation, for instance in the form of a tax burden. The meta-analysis indeed suggests that satisfaction with the national economy is a more successful predictor (success rate 75 percent) for anti-immigration attitudes than one's economic situation. These findings are in line with Citrin et al. (1997) who found that beliefs about the national economy and anxiety over taxes were a much stronger predictor of anti-immigration attitudes when compared to personal economic circumstances that played a little role.
Results of the present meta-analysis are in line with recent research, that fails to find empirical support for labour market competition theory (Hainmueller and Hiscox 2010;Hainmueller, Hiscox and Margalit 2015;Jeannet 2018). It appears that non-economic factors grounded at a deeper level (such as values and beliefs regarding social and ethnic identity, preferences for homogeneity, cognitive ability and socialization experience) might influence both attitudes to immigration directly, as well as individual economic factors.

Political attitudes, political/societal participation, and contact
Past and recent scholarship seldom places political orientations and ideology at the centre of attention when explaining individual differences in attitudes to immigration. Rather, left-right orientation and/or political ideology are used as control variables. Scholarship studying motivated reasoning associates political ideology with attitudes formation (Brooks, Manza and Cohen 2016). Seeing oneself as liberal or conservative has consequences in differential settings (Lodge and Taber 2013) to avoid cognitive dissonance. This theory stresses one's internal need for consistency in attitudes and actions in order to avoid negative feelings and discomfort (Festinger 1962). Similarly, theories of voting behaviour expect voters to vote in a way to minimize the distance between their own ideological correspondence and that of their selected party (Harteveld, Kokkonen and Dahlberg 2017). Those who identify ideologically as conservatives or with fiscally and socially conservative parties are therefore expected to show greater hostility toward out-groups. Whereas self-identified liberals shall be more tolerant of ambiguity or differences in lifestyle or identity.
Most of the studies reviewed here are not primarily interested in explaining the association between political attitudes and attitudes to immigration (Pardos-Prado and Xena 2029; Aarøe, Bang Petersen and Arceneaux, 2017;). Nevertheless, the results are suggestive. The success rate lies between 65 and 100% for political ideology, party identification and left-right positioning. More identification with social and fiscal conservatism is connected with more antiimmigration attitudes.
The literature also observes a large impact of past attitudes to immigration on current attitudes to immigration (100 % success rate). This is not surprising as a series of studies shows that attitudes are relatively stable over the life span. While individuals hold other attitudes that also affect attitudes to immigration, explaining attitudes to immigration by other attitudes is often considered problematic, as it is not entirely clear which type of attitude comes first and then affects the other. Thus, it is a chicken or egg paradox. Does racial prejudice lead to hostility towards immigrants? Or do those who dislike immigrants become racist in the process? My meta-analysis shows that these attitudes are indeed positively relatedthe success rate of nationalism, ethnocentrism and racial prejudice lies between 66 and 75 percent. Nevertheless, as mentioned above, the direction of the effect is not clear.
Political and societal participation in the form of membership in political parties, social associations, sports clubs, and so on is linked to attitudes to immigration through socialization and resources. Union membership is thought to reduce bias through contact with a diverse set of people, as well as intra-class solidarity. On the other hand, membership in associations such as sports clubs often serves as a proxy for (usually higher) social classas explained above, individuals with higher social class are expected to hold more pro-immigration attitudes based on the fact that they do not feel threatened by immigrants. The meta-analysis presented here finds that no membership in voluntary organizations or unions leads to more anti-immigration attitudes roughly about half of the time (44 percent). The majority of tests (10 out of 18) report no relationship between societal participation and anti-immigration attitudes.
The meta-analysis also looked at overall general contact with ethnic minorities. The classical contact theory states that contact with the outgroup reduces prejudice towards its members, as individuals become familiar with and less threatened by this group (Pettigrew and Tropp 2006). On the other hand, intergroup threat theory and ethnic competition theory hypothesizes that increasing numbers of minorities in one´s everyday life will lead to the native's hostile attitudes as a result of increased competition for political, social and economic resources (Olzak 1992). Overall, the meta-analysis confirms the contact theory assumptions, with a 71.5 percent success rate. Therefore, more contact with members of other groups leads to more positive attitudes to immigration. However, it should also be noted that reverse causality may also play a role heredoes contact lead to more prejudice, or are those with less prejudice more prone to contact with other minorities? This question can hardly be answered with observational studies.

Contextual level variables
People's attitudes to immigration are affected not only by individual characteristics described above but also by the political, cultural and economic systems that they lived in. Although most of the literature on attitudes to immigration still focuses on the United States, cross-national studies have shown the importance of contextual characteristics in explaining variations in attitudes to immigration. In this section, I review the influence of regional and country variables on attitudes to immigration. Nevertheless, the number of studies researching contextual effects on attitudes toward immigration is quite limited. Only a handful of variables have been used in more than two independent samples. The results below should therefore be taken with caution, as the total number of tests for these variables is generally low.

Regional minority share and density
Studies controlling for the effects of a minority share in the area (neighbourhood, region or country) on attitudes to immigration are prevalently based on two theories -intergroup contact theory (Pettigrew and Tropp 2006) and intergroup threat theory (Olzak 1992). Although measured at the contextual level, the theories´ assumptions are inherently based at the individual level. Controlling for shares of minorities in one´s area implicitly assumes that the proximal (physical) presence of immigrants leads to contact with minorities, and this may influence ones' attitudes. Rarely do researchers control for actual (daily, weekly, monthly) frequency of contact with minorities and thus the foundation of contact theory. On the other hand, the underlying assumption of threat theory is also questionable -to feel threatened by a group, one must first know of its existence and its relative size. Yet, many natives overestimate the size of the non-native population and even when explicitly corrected, their attitudes to immigration do not change (Hopkins, Sides and Citrin 2019). This leads some scholars to conclude that misperceptions regarding the relative size of minorities might be a consequence, rather than the cause of attitudes to immigration (Hopkins, Sides and Citrin 2019).
The results of the meta-analysis show that lower minority proportion in a small spatial unit 5 such as a neighbourhood, county, or region leads to more anti-immigration attitudes (90% success rate). Robert Schuman Centre for Advanced Studies Working Papers Nevertheless, less regional population densitywhich should capture the implications of contact theory has only about 25 percent success rate in explaining anti-immigration attitudes.

Unemployment, household income and GDP per capita
During an economic recession and high unemployment rates, natives may view immigration as economically threatening to themselves (Dancygier and Donnelly 2014). Therefore, it is hypothesized that high levels of unemployment at both the regional as well as country level will lead to more negative attitudes to immigration. This is due to increased competition from immigrants on the job market, as well as increased competition for a fixed supply of welfare benefits. Another hypothesis regarding the negative effect of unemployment on attitudes toward immigration is the "fiscal burden argument", which assumes that either taxes will be raised, or cuts will be made to the welfare budget (Facchini and Mayda 2009).
However, the meta-analysis does not point in this direction, with a success rate of only about 33% for both levels. Like unemployment, overall economic prosperity, in the form of regional household income and GDP per capita, appears to affect individual attitudes toward immigration to only a small extent.
It is worth mentioning that studies controlling for these macro characteristics are not primarily interested in the effect of national economic prosperity on attitudes toward immigration.

Discussion
In recent years and especially during and after the so-called "migration crises", an unprecedented number of variables have been investigated as affecting attitudes to immigration by political scientists. In this research context, it has become difficult to grasp what factors matter the most for general attitudes to immigration. This article assesses in a comprehensive manner where current political science research stands, reviewing 23 articles and 37 independent samples in the top 30 political science journals published between the years 2009-2019.
The vote-counting procedure used in this research does not allow us to take into account the size of the effect of each independent variable (Glass, 1976). With a still increasing amount of research regarding factors affecting attitudes to immigration, future studies may call upon directly comparable test statistics. By being able to use a more sophisticated meta-analytical approach, they may provide a meta-estimate of the relationship between attitudes to immigration and a given independent variable of interest.
This research reveals that only a handful of independent variables are systematically included in studies addressing attitudes to immigration. Namely, out of the 115 independent variables included in models throughout all the studies, only age and gender were included in more than half of the studies. Even more importantly, most of the time these variables are included without any motivation or expectations regarding their effects. For instance, recent research shows that although certain variables such as gender, age and income are routinely included in models explaining attitudes to immigration. What we observe however are not the effects of aging per se, but rather the effects of common experiences shared by cohorts during the impressionable years which persist even later in life (Jeannet and Dražanová 2019). Future research might therefore explicitly address the question of why we would expect certain variables to affect attitudes to immigration rather than just perpetually controlling for them.
Another critical question future research shall answer is how do the relationships between attitudes to immigration and their explanatory factors hold in different contexts. As Hoxhaj and Zuccotti (2020) rightly point out, many factors are often considered as additive explanations of attitudes to immigration and conditionality is rarely studied. However, it has been highlighted by some preliminary studies that the effect of even factors considered to be strongly associated with attitudes to immigration and tolerance, such as education (Dražanová 2017;Borgonovi and Pokropek 2019), are conditional upon levels of GDP per capita and lengths of democracy. Similarly, Hoxhaj and Zuccotti (2020) show that higher minority share in neighbourhoods is associated with more positive attitudes toward immigrants, but that this effect decreases as the area's socioeconomic conditions worsen. All these findings suggest that analysing them in isolation could seriously hinder our understanding of the relationship between factors thought to explain attitudes to immigration.
While a lot of scholarly attention has been aimed at explaining the individual-level factors affecting attitudes to immigration, it appears much less focus has been given to cross-national and cross-regional differences in attitudes to immigration in current political science literature. This is rather surprising given the availability of cross-sectional surveys such as the European Social Survey and World Value Survey. Macro-level conditions affecting individual attitudes to immigration may shed light on increasing differences between countries in their approaches to the increasing numbers of immigrants in recent years. For instance, the literature has paid very little attention to systematically analysing why the new EU Member States have been in sharp opposition to migration, although the number of immigrants in their countries is effectively very low. Moreover, there might not only be variations between countries but also within countries over time. Robert Schuman Centre for Advanced Studies Working Papers A quick overview of the literature reviewed in this meta-analysis points to a problem often found in other areas of political sciencean overt emphasis on Western countries. The analyses presented here has been largely employed in the USA and Western Europe, with a small amount of analysis on (thanks to the European Social Survey) Central and Eastern Europe and only one study on New Zealand. However, this western-focused approach leads to sample bias as these developed countries have relatively fewer immigrants and more capacity to absorb them (Alrababa'h et al. 2020;Gonnot et al. 2020). A critical step in future research is addressing the real causal factors affecting attitudes to immigration. This can be done only by also including countries from other parts of the world in the analysis.
Finally, as this meta-analysis focuses only on one type of attitudes to immigrationnamely, the perceived effects of immigration on the societyfuture research should assess whether the same factors also affect other types of immigration attitudes, such as attitudes to immigration policy.