Asian Journal of Political Science. ISSN: (Print) (Online) Journal homepage:

Transcription

1 Asian Journal of Political Science ISSN: (Print) (Online) Journal homepage: Attitudes toward immigrants and immigration policy in Asia and the Pacific: a quantitative assessment of current theoretical models beyond western countries Joan Barceló To cite this article: Joan Barceló (2016): Attitudes toward immigrants and immigration policy in Asia and the Pacific: a quantitative assessment of current theoretical models beyond western countries, Asian Journal of Political Science, DOI: / To link to this article: Published online: 29 Feb Submit your article to this journal View related articles View Crossmark data Full Terms & Conditions of access and use can be found at Download by: [Joan Barcelo] Date: 29 February 2016, At: 13:28

2 ASIAN JOURNAL OF POLITICAL SCIENCE, Attitudes toward immigrants and immigration policy in Asia and the Pacific: a quantitative assessment of current theoretical models beyond western countries Joan Barceló Department of Political Science, Washington University in St. Louis, St. Louis, MO, USA ABSTRACT This article empirically assesses the validity of current theoretical models of attitudes toward immigrants and immigration policy in the Asia-Pacific region. This paper takes representative data from the World Values Survey and implements a multilevel model to test five of the main theories in the literature: the human capital, the social capital, the political orientation, the contact/group threat and the economic competition. The results from the analysis lend credence to the important effects of human capital, social capital and politicalideological variables on respondents attitudes toward immigrants and immigration policy. However, the results provide mixed and weak evidence for the contact/group threat and the role of economic determinants. Importantly, economic determinants tend to be more powerful in shaping people s attitudes toward immigration policy than shaping attitudes toward immigrants as people. Altogether, this article sheds new light on the validity of current theoretical models based on western countries for other areas of the world. Finally, the results from the paper also support the usefulness of non-economic, as opposed to purely economic, models in the understanding of individuals attitudes toward immigrants and immigration policy in the Asia-Pacific region. Introduction KEYWORDS Immigration; Immigrants; Attitudes; Prejudice; Discrimination; Survey; Public Opinion By mid-2013, the number of international immigrants was estimated at approximately 214 million people (UN Population Division, 2013). Even though migratory flows occur both across and within national boundaries, international migrants face numerous obstacles because of restrictive immigration regimes and strong societal opposition by host societies (Messina, 2007). Studies on the sources of negative attitudes toward immigrants have largely focused on the European continent or the Organization for Cooperation and Development (OECD) countries (e.g. Gang et al., 2002; Lahav, 2004; Coenders et al., 2005; Kessler and Freeman, 2005; Semyonov et al., 2006; Martínez-Herrera and Moualhi, 2007; Ceobanu and Escandell, 2008; Malchow-Moller et al., 2008; Meuleman et al., 2009; Rustenbach, 2010; Martínez-Herrera et al., 2013). Heterogeneity within the continent (Zick et al., 2008) and availability of high quality data (e.g. Eurobarometer, European CONTACT Joan Barceló 2016 Taylor & Francis joanbarcelosoler@wustl.edu

3 2 J. BARCELÓ Social Survey (ESS)) are the main reasons for this geographical specialization (Ceobanu and Escandell, 2010). Current theoretical models are based on a limited number of western countries, even though they are generally offered as universally applicable. The purpose of this article is to expand this research line to countries at different stages of development and migration histories that we encounter in the Asia and the Pacific region. The stock of international migrants in the Asia-Pacific region constitutes 1.3 per cent of the total population in the region, which amounts to about 53 million people. Even though this percentage is substantially lower than the average number of immigrants hosted in some European countries, about one-quarter of the total number of international migrants in the world is located in Asia and the Pacific. The region is complex and diverse. Some countries have nearly no immigration, such as Vietnam with a population with 0.08 per cent non-citizens and international refugees and some other countries have a large stock of international immigrants, such as Singapore with 38.2 per cent foreignborn population (UN Population Division, 2013). 1 Altogether, the heterogeneity in the region presents an excellent setting to test the array of explanations that researchers have provided for OECD member states to explain the development of anti-immigrant attitudes and policies in the Asia-Pacific countries: human capital, political orientation, social capital, contact (or group threat ) and economic competition. I take the World Values Survey (WVS), conducted during the period between 2005 and 2008, to explore individual and country-level economic and non-economic determinants of attitudes toward immigrants and immigration policy. This article makes several contributions to the study of natives attitudes. First, this study provides an overview of the main theoretical models that are susceptible to be imported to the Asia-Pacific region. Second, it takes into account methodological complexities with the incorporation of different weighting procedures that may have biased the results. Finally, this paper also accounts for theoretical and empirical discrepancies between the two dependent variables that are generally used in the literature: attitudes toward immigrants (as people) and openness to immigration (as policy). The article proceeds as follows. The first section briefly describes the main theories that explain attitudes toward immigrants and immigration in other parts of the world, and which constitute the theoretical hypotheses of the analysis. The second section details the data, the research method, and the operationalization of the independent and dependent variables. The third section presents the main empirical results. The final section discusses the main findings, the theoretical and policy implications and avenues for further research. Theoretical background There are two types of attitudes on the issue: views toward immigrants and views toward immigration policy (Ceobanu and Escandell, 2010). Even though most researchers choose to study them together (e.g. Berg, 2010; Burns and Gimpel, 2000; Dovidio and Esses, 2001; Luedtke, 2005; Segovia and Defever, 2010), this practice may invalidate the measure given that they tap into different theoretical considerations. However, I do expect both attitudes to arise from a similar set of beliefs, values and opinions. This common source justifies the use of explanatory models based on the same theoretical structure for both dimensions. For the most part, the theoretical differences between the two constructs generally

4 ASIAN JOURNAL OF POLITICAL SCIENCE 3 attribute a stronger degree of ethno-racial prejudice when explaining generalized opposition to immigrants rather than opposition to the policy. By contrast, orientations toward the policy tend to be associated more strongly with perceptions about the economic and social consequences of an influx of immigrants for the entire country. The literature on attitudes toward immigrants and immigration policy preferences has provided a number of explanations for negative attitudes toward immigrants and immigration (see e.g. Ceobanu and Escandell, 2010; Zamora-Kapoor et al., 2013). The most prevalent of them are the human capital, the social capital, the political orientation, the contact/group threat and the economic competition. The Human Capital Hypothesis proposes that higher educational attainment reduces anti-immigrant attitudes and increases the ideal level of immigration in the country (e.g. Gang et al., 2002). Two sorts of theoretical argument for this effect can be differentiated: labour-related and non-labour-related arguments. The former suggests that education provides the skills to the members of the host society to avoid competing for jobs against immigrants in the labour market. Additionally, educated people benefit more from the low wages paid to unskilled immigrants. The latter asserts that education is understood as increasing reflexivity and open-mindedness, which is supposed to lead educated people to have higher levels of tolerance toward racially or culturally different people. Even though some authors have taken education as a proxy for worker s skills (Mayda, 2006), when both dimensions have been tested against each other, non-labour arguments account for most of the variance of the relationship between educational achievement and attitudes toward immigrants (Hainmueller and Hiscox, 2007). The Social Capital Hypothesis is based on the idea that those who possess high levels of social capital are more likely to respect the rights of others and be more tolerant toward difference (Putnam, 2001; Herreros 2008; Herreros and Criado, 2009). Social capital can be defined as the set of features of social organization such as networks, norms, and social trust that facilitate coordination and cooperation for mutual benefit (Putnam, 1995: 67). Therefore, individuals with high levels of social capital will tend to expect strangers to be more trustworthy, regardless of the cultural background of the others. And so it seems reasonable to expect these individuals to be more accepting of international immigrants (Herreros and Criado, 2009). Following this rationale, the empirical strategy focuses on the individual social capital rather than on the aggregate social capital. Another potential source of anti-immigrant and anti-immigration attitudes comes from individuals political-ideological orientation. Those people with low political interest may be willing to blame others for societal misfortunes, and immigrants are an easy target (Espenshade and Hempstead, 1996). Yet it could also be the case that those with relatively low political interest also lack political information and, therefore, the tools to assess the actual impact of immigration on society. Additionally, negative attitudes toward immigrants have been linked to rightist political movements, especially in Europe with the emergence of right-wing parties that push for stricter controls over immigration, and have become more politically powerful in the last decades (Sides and Citrin, 2007). Related to the ideological spectrum explanations, some authors argue that nationalist attitudes, defined as those who express feelings of national superiority over others, tend to discriminate against immigrants (De Figueiredo and Elkins, 2003; Raijman et al., 2008).

5 4 J. BARCELÓ In a more complex way, the contact hypothesis proposes that under conditions of equal status, common objectives, authority sanction and the absence of competition, interaction between two groups should lead to more positive attitudes between them (Allport, 1954; Pettigrew and Tropp, 2006; Paluck and Green, 2009). As regards the attitudes toward immigrants, it is unlikely that the benign effects of contact ever happen since the four requirements are rarely met simultaneously. Therefore, in its negative version, this is sometimes presented as the group threat or group position (Blumer, 1958; Hainmueller and Hangartner, 2013; Quillian, 1995). In the group threat, members of the dominant group consolidate their superior position through prejudice toward the subordinate group and foster a clear in-group/outgroup differentiation. When people come into contact with immigrants in the presence of competition for scarce resources, intergroup hostility should be expected. As the requirement for the beneficial effects of intergroup contact is rarely realized, this paper takes the group threat side of the contact hypothesis to argue that as the size of the minority group increases, especially if it occurs quickly, the feeling of threat by the majority will increase and they will feel they may have to compete for cultural hegemony (Hopkins, 2010). Finally, economic competition is a commonly analysed source of negative attitudes toward immigrants and immigration. This postulates that the more an individual is exposed to direct competition for scarce material resources, the more discrimination toward the rival out-group it will foster (Mayda, 2006). While the contact hypothesis links the size of the immigrant population to group threat through the competition for cultural hegemony, the economic competition hypothesizes that the source of prejudice is the dispute over material resources. At the individual level, low-skilled native workers who enter into competition with low-skilled immigrants are expected to be more strongly against immigrants and immigration (Scheve and Slaughter, 2001). For similar reasons, individuals employment status might be a potential determinant since an unemployed respondent might tend to blame immigrants for their unemployment (Rustenbach, 2010). As a consequence, those low-skilled native workers, unemployed or with low family income, are expected to compete more directly for economic resources with immigrants and, consequently, to hold more negative attitudes toward immigrants and immigration. In conjunction with individual determinants, macro-economic factors may also play a role in shaping natives attitudes. Alterations in the state of the economy, unemployment rates, or the economic prosperity of the country might change the feeling of threat of the native population vis-à-vis immigrants (Coenders et al., 2008; O Rourke and Sinnott, 2006). At the same time, a country s high levels of inequality may lead to fiercer competition between low-skilled native workers and immigrants because the lower strata has a smaller portion of the national product to share and, therefore, more is at stake in the battle between native and immigrant population. Methodology: data and measurements Data To test the relative importance of the theories in the Asia-Pacific region, I use data from the World Values Survey ( period) (World Values Survey Association, 2013). The WVS is a global research project that explores people s beliefs and values and is

6 ASIAN JOURNAL OF POLITICAL SCIENCE 5 the most widely used worldwide survey centre for research in the social sciences. Certainly, WVS, ESS, International Social Survey Programme, and all other international survey agencies have been criticized. The most prevalent critique is the assumed comparability of cross-national survey measurements (Davidov et al., 2014). However, it is also true that these data sources, and also the WVS, remain the main source of survey data around the world and the source of a number of influential works in the discipline as large, and in research on attitudes toward immigrants in particular (e.g. Ceobanu and Escandell, 2010; Mayda, 2006; Strabac and Listhaug, 2008). In particular with the WVS, the breadth of countries surveyed has made this cross-national survey data a standard resource for studying variation of political attitudes across countries around the world. Wave 5 of the WVS included representative survey data for 11 of the Asian-Pacific countries as classified by the regional activities of the United Nations: Australia (1,421 interviews in 2005), China (1,991 interviews in 2007), Georgia (1,500 interviews in 2008), India (2,000 interviews in 2006), Indonesia (1,000 interviews in 2006), Japan (1,000 interviews in 2005), Malaysia (1,200 interviews in 2006), South Korea (1,200 interviews in 2005), Thailand (1,500 interviews in 2007), Turkey (1,346 interviews in 2007) and Vietnam (1,584 interviews in 2006), with a total of 15,742 valid respondents. The sample of countries available from the region is representative of the geographical area. Empirical specification The empirical strategy should take into account the hierarchical nature of the data where individual data are grouped within countries. Multilevel models are a commonly used method and an excellent way to address contextual determinants of individual behaviour (Gelman and Hill, 2007) when independent variables are located at both individual (e.g. respondent s age) and country level (e.g. stock of international immigrants at country level). Even though limitations of the minimum sample size of 50 groups (level 2) are a difficult hurdle to overcome with the current availability of data (Maas and Hox, 2005), the alternative of setting the group level variance to 0 or infinity seems worse (Gelman, 2007). In this case, the number of second-level groups is 10 and 11 for the dependent variables of attitudes toward immigrants and large-scale immigration respectively. Yet, some relevant control and independent variables are missing for some countries in the sample, so models are replicated several times to capture the benefits of the trade-off between countries in the sample with valid values, and so the number of observations at level 1 and 2, and the number of variables included in the models. The modelling structure chosen to test my empirical hypothesis is a varying-intercept model with individual-level and country-level covariates. Thus, the hierarchical structure of the data has two levels, with individuals nested within countries. More complex models (varying-intercept and varying-slopes models) were too unreliable to be implemented due to the small number of groups available for some models. More formally, this is the modelling specification implemented in the empirical section: Level 1: Pr(Immigrants or Immigration policy) ij = log it 1 (b 0j + b 1 X 1ij + b 2 X 2ij b k X kij ) Level 2: b 0j = g 00 + g 01 W 1 + g 02 W g 0M W M + u 0j

7 6 J. BARCELÓ In the above equations, the βs are individual-level regression parameter, where β 0j is the intercept of the country of the respondent. The γs are group-level regression parameter estimates. Weighting procedure A methodological issue that appears in the multilevel analysis refers to the weighting method of countries. If models are left unweighted, each country is weighted for the equivalent of its sample size. Since sample size is based on the specific survey design, and not on substantive criteria, weights based on sample size are inappropriate constructions. To solve this weighting issue, I complement the unweighted results with two different weighting strategies. First, I construct a model that represents the total population of the region under study. In this case, each country is weighted by its population size. Second, I construct a model that equally represents each country in the region. With the aim to provide the most complete analysis on the attitudes toward immigrants and immigration in the Asia-Pacific region, the three sorts of analyses are explored: unweighted models (supplementary information), weights by population size (supplementary information) and equal-sized models (reported in the paper). Since this paper s aim is to make inferences across all countries in the region, the latter weighting strategy is the strategy presented throughout the paper. Measurements and empirical hypotheses In this paper, I link each to a set of independent variables to test the validity of some of the main theories in the literature. My outcome of interest is generalized animosity toward immigrants and immigration policy preferences. Given that the WVS provides a measure for each of them, my empirical test analyses the factors that affect both separately. The first dependent variable refers to attitudes toward immigrants. This is a dummy that takes the value of 1 if, given a list of societal groups, immigrants/foreign workers are mentioned as a group that the respondent would not like to have as neighbours. Otherwise, it takes the value of 0. 2 The second dependent variable captures attitudes toward immigration policies. The original wording is the following: How about people from other countries coming here to work. Which one of the following do you think the government should do? a) Let anyone come who wants to? b) Let people come as long as there are jobs available? c) Place strict limits on the number of foreigners who can come here? d) Prohibit people coming here from other countries? e) Don t know. To make this variable comparable to the first dependent variable, I recode it to take the value of 1 if the respondent chooses either a or b, and 0 if c or d is chosen instead. The e choice is left as a missing value. The first hypothesis that comes to explain natives attitudes toward immigrants and immigration is the human capital. There are two potential variables that might capture this same underlying concept: total number of years of formal schooling of members in host societies and a set of categorical values of educational attainment. The former tends to perform better as a predictor of these attitudes toward immigrants

8 ASIAN JOURNAL OF POLITICAL SCIENCE 7 (Hainmueller and Hiscox, 2007). Therefore, education is operationalized with two dummy variables: secondary education, which takes the value of 1 if the highest educational attainment is secondary education, and 0 otherwise; and tertiary education, which takes the value of 1 if the highest educational attainment is at the tertiary level, and 0 otherwise. Primary education is the category of reference. The second hypothesis is the social capital. This is formed by two dimensions: social trust and involvement in social organizations. Social trust is measured with the following question: Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?, which takes as a dummy variable between trusters (1) and not trusters (2). The latter is measured by the number of organizations in which the respondent is involved. This is asked as follows: Now I am going to read off a list of voluntary organizations. For each one, could you tell me whether you are an active member, an inactive member or not a member of that type of organization?: membership of church or religion organization; sport or recreational organization; art, music, educational organization; labour union; political party; environmental organization; professional association; charitable/humanitarian organization; consumer organization or any other organization. An index that sums the value of 0 for not being a member of the organization, 1 for being an inactive member and 2 for being an active member is created. The index ranges from 0 (not a member of any type of organization) to 20 (active member of all types of organization). Even though both are usually understood to measure social capital, the interpersonal trust dimension captures better the theoretical meaning of trust with strangers. Thus, it is expected to have a stronger effect. The third hypothesis is the effect of political orientation. The first measurement refers to the potential effect of interest in politics. Two dummy variables for the position of somewhat interested and very interested have been created. The category of reference is generated by those who express themselves not to be very interested or not at all interested in politics. I use dummies to avoid since I do not necessarily expect a linear and equidistance relationship among categories. Secondly, ideology is measured with a standard ideological self-placement question This scale is a continuous variable that ranges from the far left (0) to the far right (9). However, this hypothesis will not be tested for China and Malaysia since it is not available. Finally, national pride is measured with the following question: How proud are you to be [nationality]? Very / Quite / Not very / Not at all proud, using a dummy, which takes the value of 1 if the respondent is very proud of their own nationality, and 0 otherwise. The categorization into a dummy of this variable is motivated by the lack of variability in the data across individuals. In particular, less than 9 per cent of the sample reports feeling not very proud or less of their nationality. Therefore, to make the coefficient stable, I decide to collapse the three lowest categories into a single category by creating a dummy that differentiates between those who report being very proud (58 per cent of the sample) and all other categories (42 per cent of the sample). The contact or group threat hypothesis is operationalized in two ways. First, I take into account the stock of international immigrants in a country as a percentage of the total population (in 2008). Second, I introduce a measure of the recent pace of immigration. There are reasons to believe that recent increases in the net migration rate, calculated as the difference between emigrants and immigrants divided by 1,000 units of population,

9 8 J. BARCELÓ might be more relevant than the stock of already settled immigrant population. As the sample includes countries with a significant recent migration history as well as countries with a low rate of recent immigration, the rapidly increasing size of the immigrant population might increase natives perceived threat. To capture this, the annual average net migration rate for the period is included in the analysis. Even though the number of immigrants naturally increases the likelihood of contacts between natives and immigrants, this measure most likely captures the defensive reaction of natives who view immigrant out-groups as a threat to the natives dominant position. This measurement is quite standard and follows previous work from this theoretical framework (Blumer, 1958; Hainmueller and Hangartner, 2013; Quillian, 1995). Finally, economic competition is measured at two levels: at the individual and at the country level. From an individual level, the first variable to consider is the relative income scale. This divides the population in 10 steps of family income, from low (0) to high (9), standardized by country in order to make it comparable cross-nationally. The second dimension that might capture economic competition is whether the respondent is a skilled worker or not. This variable takes the value of 1 if the respondent is an employer/manager, professional worker, office worker, foreman or supervisor or skilled manual worker, and 0 otherwise. In these cases, competition with immigrants is expected to be relatively lower. And it takes the value of 0 if the respondent is a semi-skilled or unskilled manual worker, an agricultural worker, a farmer who owns his own farm, a member of the armed forces, employed in security, or has never had a job. In these cases, competition is expected to be relatively high. Other jobs are taken as missing values. 3 Lastly, employment status is measured with a dummy that takes the value of 1 if the respondent is employed, and 0 otherwise. Another dummy takes into account whether the individual has been employed in the last five years (1) or not (0). At a country level, the perceived economic competition is measured by using several static indicators: national unemployment rate (in 2008), gross domestic product per capita (in logarithmic scale, in 2008), and dynamic indicators of economic conditions: average annual unemployment rate in the last five years (between 2003 and 2008) and average annual change of gross domestic product in the last five years (between 2003 and 2008). Lastly, the measure of country inequality is the Gini coefficient (in 2008). Even though these measures are not ideal for the of economic competition, they follow the standard approach for operationalizing perceived economic competition (e.g. Coenders et al., 2008; Coenders and Scheppers, 2008; Strabac and Listhaug, 2008). To summarize the linkage between and measurements and my empirical expectations, Table 1 reports the variables that correspond to each and the hypothesized direction of the effects. Notice that the set of theories that are meant to explain one dependent variable is the same set of theories that are meant to explain the other dependent variable. Therefore, I jointly report the expectations for both, although the empirical results will be provided for each variable separately. Results Country-level surveys suggest an extreme cross-national variation in natives attitudes toward immigrants and immigration. Figure 1 shows that there are four types of countries. First, tolerant countries are those with tolerance levels above the mean of the sample

10 ASIAN JOURNAL OF POLITICAL SCIENCE 9 Table 1. Summary of theoretical hypotheses and their operationalization. Theories Indicators (expected sign on positive attitudes) Human Capital Hypothesis Educational attainment (+) Social Capital Hypothesis Social trust (+) Involvement in social organizations (+) Political Ideological Hypothesis Interest in politics (+) Left right ideological placement ( ) National pride ( ) Theory / Group Threat Stock of international immigrants, 2008 ( /+) Migration rate, 2008 ( /+) Competition Relative income scale (+) Skilled worker (+) Employment status ( ) National unemployment rate, 2008 ( ) Average annual unemployment rate, ( ) GDP per capita, log scale, 2008 (+) Average annual growth of the GDP per capita, (+) Gini coefficient, 2008 ( ) Italics indicate country-level variables. regarding both immigrants and immigration, such as Australia (19 and 27 percentage points, respectively) or China (11 and 12 percentage points, respectively). Second, intolerant countries are those countries, such as Malaysia ( 30 and 25 percentage points, respectively) or Thailand ( 18 and 11 percentage points, respectively), which have levels below the mean in both indicators. Third, reluctant acceptant countries are those that may not like immigrants but would rather have a more permissive immigration policy, such as Vietnam or South Korea. And, finally, friendly antagonist countries are those countries which are tolerant toward immigrants but hold a restrictive view toward immigration, as occurs in Georgia. Figure 1. Observed values of positive attitudes towards immigration by country.

11 10 J. BARCELÓ However, observed values in tolerance levels might conceal huge cross-national heterogeneity that may drive the results. Unequal levels of human capital, social capital, politicization of immigration, different features of immigration or the state of the economy, have been regarded as potential explanations of cross-national differential attitudes toward immigrants and immigration in the Asia-Pacific region. The validity of each argument is empirically tested and separately analysed in this section by using several model specifications. 4 Tables 2 and 3 report the results of attitudes toward immigrants (as people) and immigration (as policy) with equally weighted countries, respectively. Columns in these tables represent different model specifications (null model with no predictors, individual-level model with only individual-level predictors, model to test the contact, and model to test the economic competition ). In particular, since the number of countries in the sample is small, I need to test each hypothesis involving country-level predictors separately to ensure enough degrees of freedom per model. The empirical analysis is conducted in such a way that I test as many theories as possible to as many countries as possible. This is done by gradually removing those variables from the model that are missing for entire countries. In particular, the first four columns from the tables show the results of all countries with valid responses in all variables (altogether, six countries). The following three columns show the results of all countries with valid responses in all variables except the variable immigrant and size of the town of residence (eight countries for the model of immigrants from Table 2, and nine for the policy model from Table 3). And, finally, the last three columns show the results of all countries with valid responses for all variables except for the variable of immigrant, size of the town of residence, and left right ideological self-placement (10 countries for the immigrants model from Table 2, and 11 for the policy model from Table 3). Human capital hypothesis The human capital hypothesis is tested by looking at the impact of the highest educational level on individual s attitudes toward immigrants and immigration. Tables 2 and 3, and A1 A6 in the Appendix, indicate a consistent relationship between having received secondary or tertiary education versus holding a primary degree and having positive attitudes toward immigrants and the inflow of immigrants. Even though the reliability of the estimates varies by models, the direction is consistent throughout all the models (complete pooling, no pooling, and all models for the three weighting procedures) for both immigrants and immigration policy. Figure 2 plots the predicted probabilities of attitudes toward immigrants and immigration policy at different levels of the highest educational attainment for the selected countries when using the full models with countries equally weighted (Table 3). After controlling for all the first- and second-level variables of the model, an average respondent with primary studies, or less, holds positive attitudes toward immigrants with a probability of 62.5 per cent. This likelihood for an individual with the same personal characteristics and from a country with the same indicators would increase to 66.1 per cent and 67.8 per cent with secondary and tertiary education, respectively. These results point to the relevant effect of education, as it is effectively increasing open-mindedness and tolerance regarding racially and culturally different people, which in turn translates into greater support for more open policies for the entry of new

12 Table 2. Determinants of positive attitudes towards immigrants (countries equally weighted). DV: Immigrants Null Individual competition Full Individual competition Full Individual competition Full IV Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Sex Age 0.02* 0.02* 0.02* 0.02* Age square 0.00* 0.00* 0.00* 0.00* Size town 0.05*** 0.05*** 0.06*** 0.06*** Immigrant 0.24 (.14) 0.26 (.12) Education secondary * 0.12* 0.12 (.11) *** 0.15*** 0.15*** 0.15*** (.13) (.19) Education university *** 0.25*** 0.26*** 0.22** 0.23*** 0.23*** 0.23*** 0.23*** Truster 0.30*** 0.31*** 0.30*** 0.30*** 0.34*** 0.34*** 0.34*** 0.35*** 0.27*** 0.27*** 0.27*** 0.27*** Social organization 0.03*** 0.03*** 0.04*** 0.04*** 0.02*** 0.02*** 0.02*** 0.03*** 0.01** 0.01** 0.01** 0.01** Somewhat interested (.15) (.12) (.15) Very interested 0.14* 0.14* 0.15* 0.15* Left right scale 0.03** 0.03** 0.03** 0.03** 0.07*** 0.07*** 0.07*** 0.07*** Very proud of 0.19*** 0.20*** 0.20*** 0.20*** 0.12** 0.12* 0.11* *** 0.16*** 0.16*** 0.16*** nationality (.19) Income * 0.02 (.11) 0.02* 0.03* 0.02 (.11) 0.02 (.12) 0.02 (.11) 0.02 (.12) Skilled worker 0.10 (.12) 0.09 (.16) 0.09 (.13) 0.10 (.13) 0.17*** 0.17*** 0.18*** 0.19*** 0.14*** 0.13*** 0.14*** 0.13*** Employment status Constant 0.80** 0.97* ** 0.55* ** 0.72** Random-effects parameters Intergroup variance component Stock of international immigrants 2008 Net migration rate National unemployment 2008 Unemployment rate *** 0.62*** 0.01*** *** 0.03*** 0.13*** *** 0.36*** 0.29*** 0.11*** 0.14** 0.11*** 0.09*** 0.09* 0.07 (.16) 0.01 (.10) 0.01 (.15) (.19) 0.03* (Continued) ASIAN JOURNAL OF POLITICAL SCIENCE 11

13 12 J. BARCELÓ Table 2. Continued. DV: Immigrants Null Individual competition Full Individual competition Full Individual competition Full IV Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. GDP per capita, log 1.06*** (.13) 0.10 scale 2008 (.18) Growth of GDPpc, 0.24** Gini coefficient, ** 0.12* 0.02** 0.08* 0.04 Reliability estimate Log likelihood Deviance Number of observations Number of countries

14 Table 3. Determinants of positive attitudes towards immigration policy (countries equally weighted). DV: Immigration Influx Null Individual competition Individual competition Full Individual (10) Model economic competition Full model IV Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Sex * 0.07* * Age (.13) 0.01 (.13) 0.01 (.12) 0.01 (.11) 0.02** 0.02** 0.02** 0.02** Age square Size town 0.02 (.13) 0.02 (.13) 0.02 (.11) Immigrant Education *** 0.18*** 0.17*** 0.18*** secondary Education 0.15 (.11) 0.15 (.11) 0.14 (.13) 0.17** 0.17** 0.16* 0.16* 0.29*** 0.29*** 0.29*** 0.29*** university Truster 0.33*** 0.33*** 0.33*** 0.35*** 0.35*** 0.34*** 0.35*** 0.36*** 0.36*** 0.36*** 0.37*** Social organization *** 0.02*** 0.02*** 0.02*** Somewhat *** 0.13*** 0.13*** 0.13*** interested Very interested 0.32*** 0.32*** 0.33*** 0.24*** 0.24*** 0.24*** 0.25*** 0.37*** 0.37*** 0.38*** 0.38*** Left right scale ** 0.03** 0.03** 0.03** Very proud of 0.22*** 0.22*** 0.23*** 0.33*** 0.33*** 0.34*** 0.33*** 0.22*** 0.22*** 0.22*** 0.22*** nationality Income 0.07*** 0.07*** 0.07*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** Skilled worker ** 0.12** 0.11** 0.11** Employment status Constant 0.51* 0.57* (.10) Random-effects parameters Intergroup variance component Stock of international immigrants 2008 Net migration rate National unemployment *** 0.70*** 0.65*** 0.02*** 0.52*** 0.49*** 0.13*** 0.05*** 0.84*** 0.81*** 0.51*** 0.19*** * ** * 0.68** ** (Continued) ASIAN JOURNAL OF POLITICAL SCIENCE 13

15 14 J. BARCELÓ Table 3. Continued. DV: Immigration Influx Null Individual competition Individual competition Full Individual (10) Model economic competition Full model IV Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Unemployment 0.88** ** rate GDP per capita, log 0.79* 1.23** * scale 2008 Growth of GDPpc, 0.41 (.10) 0.56** * Gini coefficient, *** 0.08* 0.11** 0.11* 0.17** Reliability estimate Log likelihood Deviance Number of observations Number of countries

16 ASIAN JOURNAL OF POLITICAL SCIENCE 15 Figure 2. Predicted probability of Attitudes towards Immigrants and Immigration Inflow by Educational attainment. immigrants. At the same time, this finding follows the same pattern as the rest of the literature with a sample of OECD or European countries (e.g. Gang et al., 2002; Mayda, 2006; Hainmueller and Hiscox, 2007). Social capital hypothesis Either participating in social organization or thinking that others are generally trustworthy is expected to have a positive impact on attitudes toward immigrants and immigration. Even though the results presented here lend further support to the social capital hypothesis, they also suggest that we need to approach the concept of social capital carefully. The two dimensions taken into consideration here propose that involvement in social organizations and trust in others facilitate cooperation among citizens and lead to a number of positive outcomes. However, empirical tests reveal that when put together that is, each dimension is controlled for the other interpersonal trust is the truly positive dimension that is behind the scenes of social capital and pulls individuals toward more tolerant attitudes vis-à-vis immigrants and a more open migratory policy. Interpersonal trust is, in fact, the only individual-level variable that is consistently significant for every single model and is unaffected by the different model specifications. Thus, analyses that only regard the role of involvement in social organizations overlook the powerful effect of interpersonal trust. Figure 3 displays the magnitude of the effect of interpersonal trust on attitudes toward immigrants and immigration inflows. A respondent who believes that most of the people can be trusted is expected to hold a positive predisposition toward immigrants with a probability of 71.9 per cent, while this figure would decrease to 66.1 per cent if the respondents feel that most of the people cannot be trusted. The same effect is shown when using opinion regarding the openness of the migratory policy, as being a social truster boosts respondents attitudes toward more open immigration policy by 8.2 per cent. Again,

17 16 J. BARCELÓ Figure 3. Predicted probability of Attitudes towards Immigrants and Immigration Inflow by Levels of Social Capital. with varying reliabilities for some coefficients, the direction of the coefficient of truster is consistent throughout all the models (complete pooling, no pooling, and all models for the three weighting procedures) for both immigrants and immigration policy. By contrast, the effect of the second dimension of social capital, involvement in social organizations, appears rather contradictory, since it has a positive effect on attitudes toward immigration policy but it does not stem from a more positive view of immigrants. In fact, the positive effect of social organization found in the full model of the sample when countries are equally weighted is not robust across different specifications and country weights. The lack of consistency of this coefficient suggests the reader should view the variable of social organization with caution. Political-ideological hypothesis The interaction between individuals and the political sphere is believed to be another source of positive/negative attitudes toward immigrants and immigration accounting for the relevant literature in Europe and America. Results indicate a certain variability in the predictive strength across the three indicators analysed: interest in politics, ideological orientation and national pride. Models presented in Tables 2 and 3 reveal that ideological orientation (self-placement on the left right scale) as well as pride in one s own nationality consistently predict the relevant attitudes in the expected direction. In this, Figure 4 shows that, all else being equal, a respondent who is not very proud of their own nationality is more likely to show a positive attitude toward immigrants of 68.3 per cent, while for the same individual in the same country with the opposite view the probability decreases to 66.5 per cent. Even though this change is already statistically significant for people s attitudes toward immigrants, the magnitude effect increases substantially when we regard individuals opinions toward immigration inflow policy. Also, if a respondents states they are very proud as opposed to only being quite proud, not very proud or not at all proud, this leads to a decline of 8.1 percentage points in their likelihood to view a more open migratory policy positively. In particular, this means a substantial reduction from 49.3 to 41.2 per cent in the expected probability to hold a positive orientation toward a more open migratory policy. The sign of the effects for all countries in the no-pooling mode, as well as the complete pooling, shows a high level of consistency across all countries in the sample.

18 ASIAN JOURNAL OF POLITICAL SCIENCE 17 Figure 4. Predicted probability of Attitudes towards Immigrants and Immigration Inflow by Political Variables. Ideological self-placement follows a similar pattern to national pride. It considerably depresses attitudes toward the phenomenon at both the people and policy levels. For example, a respondent with the same personal conditions and country of residence would reduce their positive view toward immigrants by 12.9 per cent if located in the extreme-right instead of the extreme-left position. This means a very significant fall from an expected value of 73.4 to 60.5 per cent acceptance of immigrants. This effect on negative prejudice is partially translated into policy preferences. The same ideological shift would mean a change from 44.4 to 38.7 per cent, and so a drop of 5.7 percentage points in the expected ideal of openness of the migratory policy. While being proud of one s nationality increases people s negative attitudes toward immigrants and immigration policy in a very consistent manner throughout all the models, the effect of ideology does not seem to hold so much consistency. However, the comparability of ideologies across nations might undermine these coefficients. By contrast, interest in politics only becomes significant as a predictor of respondents opinion on the ideal level of openness of the migratory policy, but not on attitudes toward immigrants. Those who are very interested in politics have attitudes that do not differ substantially from those who are only somewhat or not interested at all in politics. Since the relationship between interest in politics and an individual s level of political information is assumed to be high, increased political information does not affect people s opinion on the matter when the rest of the potential intervening variables are controlled for, especially education,. However, it does seem to affect people s preferences regarding migratory policy, since those who are more interested in politics, and presumably are able to gather more political information, tend to accept higher levels of immigrant flows with a significant increase of 6.1 per cent.

19 18 J. BARCELÓ Figure 5. Predicted probability of Attitudes towards Immigrants and Immigration Inflow by Immigrants stock and migration rate. /group threat Large numbers of immigrants can potentially lead to two diametrically opposed paths. The contact between natives and immigrants attenuates intergroup tensions which would lead to a more tolerant view among the former and a better accommodation by the latter in the host community. By contrast, the negative side of the model suggests that, with a large number of immigrants settled in a country, natives feel threatened not only economically but also culturally by immigrants. This would lead to higher levels of animadversion toward them. Nevertheless, the process seems to be even more complex than thought. Figure 5 demonstrates the relevance of the stock of international immigration in the country to understanding people s attitudes toward immigrants and immigration, although it does so in a complex way. Both sides of the seem to be playing a simultaneous but divergent role in the process. Whereas a high number of immigrants in the country leads to more contact between natives and immigrants, and increases the favourable opinion toward immigrants (contact ), it also leads to a more restrictive policy toward the entry of new immigrants (group threat). Similarly, the recent net migration rate has the group threat effect on migratory policy although this is not accompanied with the effect on immigration that the contact would expect (predictions of attitudes toward immigrants are not shown in Figure 5 due to an insignificant coefficient). These results suggest that a positive effect of the number of immigrants in a given country on the view of the natives toward immigrants is more likely to be caused by a country s history as a receiver of immigrants. This implies that tolerance is not increased by a short-term increase in immigrants, but, instead, by the stock of international immigrants that is, the reception of immigrants over a long period of time. competition The agreement shown in the literature on the economic competition hypothesis is based on the varying perceived threat of immigrants in the competition for scarce resources. This implies different expectations for some individual-level variables. Most obviously, individuals with higher income jobs, those who are more skilled and are employed, are expected to view immigrants as less threatening since they are likely to be more confident in their