The determinants of individual attitudes towards immigration
Kevin H. O’Rourkea,b,c*, Richard Sinnottd
a Department of Economics and IIIS, Trinity College, Dublin 2, Ireland
b CEPR, London, U.K.
c NBER, Cambridge, Massachusetts, U.S.A.
d Department of Politics and ISSC, University College Dublin, Belfield, Dublin 4, Ireland
Received 14 January, 2005; received in revised form 17 October 2005; accepted
The paper formulates hypotheses and reports on individual attitudes towards immigration based on data for 24 countries on socio-economic position, socio-demographic characteristics and political attitudes. The results are consistent with the predictions of factor proportions trade theory but also suggest that a range of other economic and cultural factors influence attitudes towards immigration.
JEL classifications: F22, D72.
Keywords: Immigration; Political economy
* Corresponding author. Telephone: +353 1 608 3594. Email: email@example.com
Economic theory suggests substantial welfare gains from free migration: according to one general equilibrium estimate, freeing up world migration could double world income (Hamilton and Whalley, 1984), a gain that substantially outweighs the much-heralded estimated benefits of liberalization of world trade.1 Despite the welfare arguments in favour of migration, however, governments tend to restrict immigration in practice. In 2001, 21 out of 48 developed country governments had policies designed to reduce immigration, while only 2 had policies designed to increase immigration (UN, 2002, Table 3, p. 18). In democracies, policies presumably to a large extent reflect the individual preferences of voters. Reasons why voters might not want immigration can be classified into non-economic and economic. The non-economic reasons include racism, xenophobia, and milder forms of nationalist sentiment such as social norms or cultural preferences. The economic reasons reflect voters’ economic interests. This paper empirically explores both determinants of preferences with regard to immigration.
In a paper examining the growing restrictiveness of late 19th century immigration policy, Timmer and Williamson (1998) argued that economic factors were sufficient to explain the anti-immigration backlash that occurred in the major host countries of the New World at that time. This backlash was manifested in such legislation as head taxes, Chinese exclusion acts, the definition of various categories of persons as ‘excludable’, and so on. Timmer and Williamson constructed an index of immigration barriers in the US, Canada, Argentina, Australia and Brazil from 1850 to 1930, based on a careful reading of each country’s immigration legislation. They then regressed this policy measure on a number of explanatory variables, and found that, once economic variables had been controlled for, in particular inequality, there was no independent role for xenophobia of the sort frequently stressed by qualitative histories of the period.
Does this conclusion still hold? In this paper, we do not look at the determinants of government policies, as do Timmer and Williamson. In order to think systematically about what determines policies, we could appeal to a political economy model.2 An alternative is to look at the determinants of individual voters’ attitudes towards immigration, which we do using cross-country survey data. We consider differences in attitudes between skilled and unskilled voters.3 In a sense, we are testing one of the key assumptions underlying many political economy models, namely that individuals’ attitudes towards globalization vary systematically with their endowments. In so doing, we follow Scheve and Slaughter (2001), who used survey data to consider the question of who is in favour of immigration, and why. However, Scheve and Slaughter looked at survey data for just one country, the US. As will be emphasized later, cross-country data are required to properly test hypotheses regarding the determinants of attitudes towards immigration.
Our paper is closest in spirit to previous work that we, and also Mayda and Rodrik, have done on the determinants of individual attitudes towards trade (Mayda and Rodrik, 2005; O’Rourke and Sinnott, 2001). The independent work of Mayda (2005) raises many (but not all) of the issues addressed in this paper, and indeed uses the same dataset. Since Mayda’s study differs from ours in various respects, her results serve as a robustness check on several of our key conclusions, and we consider her conclusions in more detail later on.
The plan of the paper is as follows. Section 2 first reviews what standard trade theory has to say about the determinants of attitudes towards immigration, before introducing a range of other relevant influences that go beyond the admittedly narrow confines of such models. Section 3 introduces the survey data set that we use, and indicates how the data are applied to test the hypotheses. Section 4 reports the results of ordered probit regressions, asking what are the determinants of individual attitudes towards immigration. We also report a number of bivariate probit regressions, which allow us to simultaneously explore the determinants of attitudes towards trade, immigration and refugees. We find that allowing for the fact that these attitudes are all inter-related can be important for the results obtained. Section 5 compares our results with those of Mayda (2005). Section 6 concludes.
2.1. Trade theory
Empirically, it is the case that labour demand curves slope downwards, and that immigration lowers wages (Borjas, 2003; Hatton and Williamson, forthcoming): this is the basic fact that leads people to oppose immigration on economic grounds. Whose wages are lowered, however, depends on the composition of immigration: we expect low and high-skilled immigration to be opposed by the equivalent skill class.
What determines whether it is the high-skilled or the low-skilled who are more opposed to immigration? In order to answer this question, we need to begin with a theoretical framework where immigration affects factor prices, as in the Heckscher-Ohlin family of factor proportions models of trade.4
For example, consider a model in which three factors (capital, skilled labour and unskilled labour) produce two commodities, and in which all three factors are mobile across sectors. Assume that capital and skilled labour are the ‘extreme’ factors, and that unskilled labour is the ‘middle factor’ (Ruffin, 1981; Thompson and Clark, 1983; Davies and Wooton, 1992).5 Take two countries that are initially identical, and increase the endowment of skilled labour in one country. This will raise unskilled wages and lower skilled wages in that country. Lower the endowment of unskilled labour in the same country, and the result will again be to raise unskilled wages and lower skilled wages. Thus if we consider two countries identical in all respects but one, namely the proportion of the total workforce that is skilled, then in the country with the more skilled workforce, skilled wages will be lower, and unskilled wages will be higher. Skilled workers will migrate from skill-abundant (which we denote as rich) to unskilled-labour-abundant (which we denote as poor) countries, and unskilled workers will migrate from poor to rich countries. Immigration will hurt skilled workers in poor countries, but benefit the unskilled there; in rich countries, immigration will hurt the unskilled, but benefit skilled workers. Thus, the prediction is that the impact of skills on anti-immigrant sentiment should be related to a country’s GDP per capita.6 In the richest countries, being high-skilled should have a negative impact on anti-immigrant sentiment. In the poorest countries, being high-skilled should have a positive impact on anti-immigrant sentiment. An interaction term between skills and GDP per capita should therefore enter with a negative sign in a regression explaining anti-immigrant sentiment.
In a factor proportions trade model, it is also possible to predict through the Stolper-Samuelson Theorem who will favour free trade, and who will favour protection. 7 When countries are distinguished solely by their relative factor endowments, agents are consistent in their attitudes towards globalization. That is, in rich countries skilled workers favour both liberal trade and liberal immigration, while unskilled workers are protectionist and anti-immigration. In poor countries, it is the unskilled who are liberal in their attitudes towards both trade and immigration, while the skilled favour both protection and immigration restrictions. We thus predict that ceteris paribus being protectionist should increase the likelihood that an individual is anti-immigrant; while ceteris paribus being anti-immigrant should increase the likelihood that an individual is protectionist.8
Matters are more complicated if technology or capital endowments differ across countries: it is then possible that rich countries, with superior technology and/or capital/labour ratios, will see inflows of skilled as well as unskilled labour, despite being relatively skill-abundant. Indeed, this is what we observe. In such circumstances, can we still say anything about whether a country will experience inflows of predominantly skilled or unskilled labour? Presumably, the composition of inflows will depend largely on the ratio of skilled to unskilled wages; the higher this ratio, the higher the proportion of skilled immigrants, other things being equal.9 If skill differentials and inequality more generally are positively correlated, it follows that the impact of skills on anti-immigrant sentiment should be related to a country’s level of inequality.10 In the most income-unequal countries, being high-skilled should have a positive impact on anti-immigrant sentiment. In the most income-equal countries, being high-skilled should have a negative impact on anti-immigrant sentiment. An interaction term between skills and inequality should therefore enter with a positive sign in a regression explaining anti-immigrant sentiment.11
2.2. Beyond trade theory: further economic considerations
The sorts of trade models considered above are extremely restrictive in their assumptions. All people in the models are workers; there are no children, pensioners, or other people outside the labour force, and neither is there unemployment by assumption. The models are static, so there are no life-cycle issues; nor do the models allow for externalities, public goods, cultural preferences, taxes, welfare benefits, or other complications that almost certainly influence peoples’ attitudes about immigration (Hillman and Weiss, 1999). We now consider some of these complications, and indicate how they relate to the empirical analysis that follows.
An important economic issue facing western societies is the difficulty of funding public pensions systems in the context of aging populations. A frequently suggested solution to the problem of rising dependency burdens is immigration: for example, Storesletten (2000) shows that the fiscal problems associated with the aging of the baby boomer generation in the United States could be solved through the immigration of working age high- and medium-skill foreigners. Immigrants into western societies however often tend to be young but unskilled, and can end up as the recipients of welfare payments, thus potentially exacerbating rather than remedying their hosts’ fiscal problems. In the US immigrants aged between 20 and 30, and with less than high school education, have been associated with a net fiscal burden of $100,000 in present value; while an immigrant family with three children arriving in Germany in 1997 and staying for 10 years received a net benefit of €120,000 (Razin and Sadka, 2004, pp. 3-4). Nannestad (2004) notes the fiscal burden of immigrants in the Danish case. The political economy implications of the interaction between immigration and the modern welfare state are important for our study. The possibilities depend on whether the pension scheme is one of defined benefits or defined contributions, and whether the scheme follows Beveridgean or Bismarkian principles (Scholten and Thum, 1996; Haupt and Wolfgang, 1998; Krieger, 2003). Anti-immigrant sentiment can therefore in principle be positively or negatively correlated with age.12 In principle we could discriminate between the alternative possibilities by regressing our measure of anti-immigrant sentiment on age, and indeed this is what we do later in the paper. However, such an exercise has to take account of the fact that these two alternative possibilities are purely ceteris paribus predictions. Hillman (2002) for example provides a model in which intergenerational transfer issues are just one influence on voters’ preferences regarding immigration, and natives’ utility can be affected by a sense of diminished identity or a sense that the population’s social norms are being replaced by other norms. If older people place a higher value on traditional social norms than the young, this might lead them to be more anti-immigrant than the young, even if (as in Hillman’s model) they stand to gain financially from immigration. We attempt to deal with this by introducing separate explanatory variables measuring nationalist attitudes (see below), but, unless these completely control for all cultural factors influencing attitudes, the coefficient on age in our regression will be a reduced form coefficient picking up both inter-generational conflicts regarding pension and welfare systems and age-specific differences in cultural attitudes, and has to be interpreted with this in mind.
A key economic variable missing from the analysis up to now is unemployment, and we include in our empirical analysis a variable indicating whether the respondent is unemployed or not. One might think that the unemployed would be more anti-immigrant than the employed, for example because they view labour market competition from immigrants as the reason for their being unemployed or because immigrants provide an additional drain on the welfare system that may eventually leave them less well off.13 Yet fear of unemployment might lead those with jobs to be just as hostile to immigration as those already out of work, in which case one would not see the unemployed being more anti-immigrant
The Heckscher-Ohlin model assumes that all factors of production are perfectly mobile across sectors. This assumption clearly does not correspond with reality, especially insofar as workers are concerned. We therefore include a ‘national mobility’ variable in our analysis, since arguably, those willing to relocate within the country should be more sanguine about the dislocation implied by immigration than those who are immobile. This will be particularly true if immigrants tend to concentrate in particular regions or cities. The prediction is that those who self-identify as being willing to move elsewhere within the country should be less anti-immigrant than those who view themselves as being tied to a particular industry or location.
2.3. Beyond economics: cultural considerations
Hillman and Weiss (1999) point out argue strongly that immigration policies may reflect cultural preferences, ‘and perhaps likes and dislikes that are contained in the collective memories of different peoples’ (p. 76). There are different ways in which such ‘non-economic’ factors can matter (Hillman 2002). First, natives may harbour an irrational hatred for foreigners. Second, natives may derive utility from living in a society with a well-defined sense of national identity and well-understood and accepted social norms; in this case, natives may oppose ‘excessive’ immigration on the grounds that it undermines these norms, without disliking hating foreigners per se. Alternatively, others may approve of immigration on the grounds that it provides diversity through , which these individuals view as a good rather than as a bad, with for example ethnic restaurants being the most visible but not the only benefit which they derive from the presence of foreigners. As mentioned earlier, it is possible to introduce such preferences into formal economic analyses of immigration policy. 14 (Hillman and Weiss 1999;, Hillman 2002
In this paper, we take seriously the potential roles of such non-economic factors in determining attitudes towards immigration. In particular, we explore the possibility that anti-immigrant preferences may in part be a function of strong feelings of national identity and an associated set of patriotic and nationalist attitudes that include pride in country, sense of national superiority and, at the extreme, antagonistic attitudes towards those who are not part of the nation. Of course, nationalist ideology may have its origins in a conjuncture between identity and group interests, and particularly in a conjuncture between identity and perceptions of inequality (Gellner, 1983); the point here, however, is that, whatever their origins, nationalist attitudes are likely to have a certain autonomy and may exercise an independent influence on the way in which individuals react to immigration and to other globalization issues. Our prediction is that nationalism will be positively correlated with anti-immigrant sentiment, other things being equal.
Similarly, we predict that those who have lived abroad in the past (including those who were born abroad), and thus have had greater exposure to other cultures, would be less hostile towards immigration; and that those who had a foreign parent or parents would similarly be less anti-immigrant than those with two native parents. In terms of the previous discussion, it is more likely that such individuals view diversity as a good, rather than as a ‘bad.’ We include a variable indicating whether the respondent is Roman Catholic or not, on the grounds that previous research has found that religious beliefs can play an important role in determining individuals’ attitudes towards globalization (von der Ruhr and Daniels, 2003). In previous work, we found that Catholicism was positively correlated with protectionist sentiment regarding trade policy (O’Rourke and Sinnott, 2001); here, we explore if this anti-market sentiment applies to international factor markets as well as to international commodity markets.
Finally, we ask whether women differ systematically in their attitudes to men. Previous research has found that women tend to be less pro-market in their attitudes than are men, and this might lead them to oppose immigration (ibid.).
To accomplish our objectives, we need a data set that provides information on individuals’ attitudes towards immigration, socio-economic position, socio-demographic characteristics and political attitudes. Since trade theory predicts that skill levels will have different implications for immigration policy preferences in different countries, the data should be cross-national in scope.
What we have are data provided by the 1995 International Social Survey Programme (ISSP) module on national identity. The ISSP national identity survey was conducted in twenty-four countries in 1995-96. The countries were: Australia, West Germany, East Germany, Great Britain, the USA, Austria, Hungary, Italy, Ireland, the Netherlands, Norway, Sweden, the Czech Republic, Slovenia, Poland, Bulgaria, Russia, New Zealand, Canada, the Philippines, Japan, Spain, Latvia and Slovakia.16
The ISSP survey asked respondents two questions that bear on their attitude towards immigration. The first asked if the number of immigrants to their economy should be increased a lot (1), a little (2), remain the same (3), be reduced a little (4) or reduced a lot (5). The second asked if refugees should be allowed to stay in the country; responses ran from agree strongly (1) to disagree strongly (5). Table 1 reports the mean response to these questions in each country, where countries are ordered according to the mean value of their response to the question on immigration. A separate column reports the ranking of countries according to their mean response to the question on refugees. Scores greater than 3 indicate that on average respondents were leaning towards greater restriction. As can be seen, individuals tended to be more strongly opposed to immigration in general than to refugees, suggesting that the interviewees were making a distinction between forced migration due to political repression and migration more generally. Sample respondents in every country on average favoured lowering the number of immigrants; by contrast, the mean response to the refugee question only exceeded 3 in five countries (Slovenia, the Phillippines, Japan, Latvia and Slovakia). What is interesting here is that in most countries, asylum seekers (i.e. persons who are seeking but who have not been granted refugee status, but who are typically confused with ‘refugees’ in everyday discourse) are prohibited from working, and are thus a drain on the welfare system; and yet respondents are more favourably disposed towards ‘refugees’ than towards immigrants in general. This is a first indication, in our view, that non-economic factors might be important in determining attitudes towards immigration.
The data set also provides individual-level measures of a range of demographic, socio-economic and political variables. Among the socio-economic variables, the most valuable from the point of view of testing the implications of the trade theories we surveyed earlier is the respondent’s skill level. This is arrived at by coding the answers to questions on respondents’ occupation using the International Labour Organisation’s ISCO88 (International Standard Classification of Occupations) coding scheme. ISCO88 is a radical revision of the ILO’s previous occupational coding scheme (ISCO68). The main thrust of the revision makes ISCO88 particularly relevant for our purposes. As Ganzeboom and Treiman put it, ‘… the logic of the classification is mostly derived from skill requirements at the expense of industry distinctions’ and the overall effort may ‘be seen as an attempt to introduce more clear-cut skill distinctions into ISCO88’ (Ganzeboom and Treiman, 1996, p. 206). While a complex coding scheme of this sort allows for very fine distinctions between different occupations, we are interested in the four main skill categories provided by ISCO88. In brief, these are: (1) ‘elementary occupations’ (i.e. ‘manual labour and simple and routine tasks, involving…with few exceptions, only limited personal initiative’ (ILO, 1990, p.7)); (2) ‘plant and machine operators and assemblers; craft and related trades workers; skilled agricultural and fishery workers; service workers and shop and market sales workers; clerks;’ (3) ‘technicians and associate professionals;’ and (4) ‘professionals.’ A fifth group, ‘legislators, senior officials and managers,’ do not have a skill coding under this four-step skill classification and were included as a separate, fifth, skill category. Finally, we excluded members of the armed forces, since it was unclear what their skill levels were.
Unfortunately, application of the ISCO coding schemes in the 1995 ISSP was somewhat uneven: the survey coded occupation in three different ways, depending on the country in question. The ISCO88 coding scheme was used in 12 cases, the earlier ISCO68 scheme was used in 6 cases and a further 6 countries used a variety of national coding schemes. However, we were able to construct an approximation to the ISCO88 skill classification either by recoding the ISCO68 data or, in three cases (Britain, the Netherlands, and the Philippines) by recoding the country-specific occupational codes. This provided us with skill data for 21 of our 24 countries.17
In order to test the predictions regarding inter-generational transfers outlined in section 2.2, we include age and age squared in all regressions. We know whether the respondent is unemployed or not, and include this in the analysis. We also make use of a subjective economic variable, namely the stated willingness of people to move from one location to another in order to improve their standard of living or their work environment. Respondents were asked: “If you could improve your work or living conditions, how willing or unwilling would you be to move to another neighbourhood or village; another town or city within this county or region; another county or region; outside [named country]; outside [named continent]?” Based on the responses to these questions, we derived two binary variables, indicating whether or not individuals were nationally mobile, and internationally mobile.18 As mentioned earlier, we expect that respondents viewing themselves as nationally mobile will be less hostile to immigration than those who are mobile.
The ISSP national identity data set includes a wide range of indicators of nationalist attitudes. Rather than focussing on just one or two of these as indicators of what is, after all, a complex phenomenon, the approach taken here is to seek to identify an underlying dimension (or dimensions) of nationalism that would be measured by a subset (or subsets) of the items. We focus on the following seven questions (versions implemented in Ireland, other country/nationality labels substituted as appropriate):
“Generally speaking, Ireland is a better country than most other countries”
“The world would be a better place if people from other countries were more like the Irish”
“I would rather be a citizen of Ireland than of any other country in the world”
“It is impossible for people who do not share Irish customs and traditions to become fully Irish”
“People should support their country even if the country is in the wrong”
“Ireland should follow its own interests, even if this leads to conflicts with other nations”
“How important do you think each of the following is for being truly Irish?”... ... ...“to have been born in Ireland”
In each case, respondents were asked to rank their responses along a scale, in the case of the first six items, from 1 (strongly disagree) to 5 (strongly agree) and, in the case of the seventh item, from 1 (very important) to 4 (not at all important). The seventh item was reordered to make it consistent with the other six. Principal components analysis of these responses yielded two underlying dimensions of nationalist attitudes. As can be seen from the rotated factor loadings in Table 2, the first dimension is a straightforward preference for and sense of the superiority of one’s own country (here labelled patriotism). The second dimension identifies a narrow or exclusive sense of nationality combined with a degree of chauvinism of the “my country right or wrong” variety (here labelled chauvinism). On this basis, patriotism and chauvinism scores have been calculated by averaging responses across the relevant subsets of items identified in the analysis.19
In addition, we include the international mobility variable described above, since being willing to live overseas may signal an openness to other cultures, and hence a greater tolerance for immigrants. By the same token, we also make use of a question which asks whether the respondent had ever lived abroad, on the basis that previous experience of living abroad may provide a signal regarding familiarity with foreigners. We also have information on whether respondents or their parents are native born or not (if so, we expect them to be more anti-immigrant); on whether they are Roman Catholic or not; and on their gender.