Good for the Common Good: Sociotropic Concern and Double Standards toward High- and Low-Skilled Immigrants in Six Wealthy Countries

Abstract

Immigration policy has conventionally implied a double standard, in which high-skilled immigrants are more acceptable due to their potential contribution to the national economy, little welfare burden, and better cultural adaption, while low-skilled ones are not favored, because of a belief in their limited contribution to the common good. In contrast to the egocentric interest explanation, we emphasize the importance of such sociotropic concerns and suggest that acceptance of immigrants with different skill levels is an outcome of perceived growth and distributional impacts or threatened cultural boundaries. Drawing data from the 2011 Transatlantic Trends: Immigration survey, we performed seemingly unrelated regression modelling to compare natives’ attitudinal responses in six wealthy countries. We found that in addition to the evidence that high-skilled immigrants are favored over low-skilled ones, the worry about welfare burden to the nation is one of the main factors causing locals to dislike low-skilled immigrants. The public who perceive immigrants’ threats to the national economy in terms of taking jobs away in general are also likely to disfavor high-skilled immigrants. Expectations of cultural assimilation are somewhat detached from acceptance of high-skilled immigrants. As the research results imply clear limitation of the double-standard perspective, we propose a new scheme for understanding both double- and single-standard views and incorporate these variations into the sociotropic theory and future research design.

1 Introduction

Over the last two decades, admission of skilled immigrants has become a policy preference over low-skilled immigrants (Mukhopadhyay and Zou 2020; Helbling and Kriesi 2014). This double standard in judging which sort of people from abroad are welcome seems to be prevalent across the globe. Skilled immigrants are considered to increase a nation’s competitiveness (Kerr 2019; Brücker et al. 2012), while lower-skilled ones are utilized to fill manual jobs avoided by natives¹ and to cater to temporary needs of the labour market. Many nations have attempted to attract “the best and the brightest” talent from around the globe by loosening immigration regulations, sometimes even enticing them with favourable policies. In contrast, the entry and continued stay of low-skilled immigrants within a contractual period are accepted only reluctantly (Tseng and Komiya 2011). Although since the 2008 economic crisis many governments have responded to protectionist calls by restricting immigration policies, even toward the highly skilled, others have taken advantage of the crisis for further liberalisation in attracting high skilled immigrants (Cerna 2016). With the high-skilled immigrants defined as people having a university degree (Iredale 2001), UK and Italy have shown a decline trend since 2010 while US, Germany, France and Spain have held a consistently increasing pattern (Table 1).² As of 2015, the high-skilled immigrants registered 36.3% and 44.8% for US and UK, respectively, and ranged between 23% and 27% for Germany, France and Spain. In contrast, Italy had a much lower proportion at 13%. These countries will be investigated more extensively regarding to natives’ attitudes toward high- and low- skilled immigrants.

Table 1 Proportion of immigrants^a who were high-skilled in US and major European countries, 2000–2015 (proportion of foreign born residents with tertiary education).

The differential treatment toward immigrants is an example of classism based on two sets of evaluative criteria: immigrants with higher economic profiles are welcomed because of their presumed economic contributions and “trouble-free” backgrounds (Tzeng 2006), while their counterparts are accepted only conditionally because of their presumed social and cultural incompatibility (Lan 2006; McLaren 2003; Pietsch and Marotta 2009). This “divide and judge” policy labels the high-skilled immigrants as “wanted” assets—colloquially, people would say that “they are good” for enhancing the nation’s common good (or, well-being), whereas the low-skilled are “unwanted” burdens. The double standard, given off often in popular immigration policy debates, tends to endorse a distinction between different categories of immigrants on the basis of rational calculation of what is best for the common good of one’s society. On the other hand, even if able to be attentive to the diversity of immigrants, people might place all of them in the same category of outsiders threatening to undermine local culture, compete for jobs with locals and use social services excessively. Thus, the locals are likely to look at immigrants through the same lens, which not only avoids cognitive dissonance but also helps advance clear policy positions toward outsiders. Or people might have mixed feelings, both good and bad, toward immigrants of different skill levels. In other words, low-skill workers are not completely unwanted and high-skill workers are not definitely wanted. Two important research questions are thus derived: (1) Are the mass public more likely to respond to immigration with the double-standard perspective, or do they perceive incoming migrants as one homogeneous group and react indiscriminately, or is their reaction a mixture of both? (2) What explains natives’ attitudes toward high- and low-skill immigrants, respectively?

Existent research on attitudes toward immigrants mainly focuses on how natives’ demographic factors (especially education, income and occupation) explain their attitude. Such analyses are mainly based on the egocentric concern, that is, a native’s calculation of self-interest. In contrast, Hainmueller and Hiscox (2010) as well as Hainmueler and Hopkins (2014, 2015) point out that natives often form their attitudes toward immigrants based on sociotropic concerns,³ that is, what functions immigrants perform and therefore how immigrants can affect society in a favourable or negative manner. More specifically, it refers to how much they will contribute to the common good when they join the labor market and become a citizen of the nation. In accordance with Hainmueler and his colleagues, we will examine how various sociotropic concerns (individuals’ perceptions of immigrants’ impacts at the societal and national level) affect natives’ acceptance of high- and low-skilled immigrants differently.

We used the data set from the 2011 Transatlantic Trends: Immigration survey,⁴ which allows comparisons of public opinion patterns in six high-income destination countries, the US, UK, Germany, France, Italy and Spain. The results show that, in general, high-skilled immigrants indeed are better received, which seems to signal the popular practice of a double standard in response to immigration. Yet, highly skilled immigrants are not necessarily better received in every aspect of the impacts on national economy and culture. This research sheds light on individuals’ multifaceted and complicated attitudes toward immigrants, which are more sophisticated than a simple “divide and manage” policy. We show that sociotropic concerns are more important than egocentric concerns in the explanation of attitudes toward immigrants.

2 Sociotropic Hypotheses through the Lens of a Double Standard

Hainmueller and Hiscox (2010) argue that an individual’s sociotropic concern in reference to perceived national interests is a necessary consideration to overcome the obstacle of the conventional economic interest argument. A formal model of economic interest explanation advances hypotheses on the basis of the factor proportions (FP) theory that substitutability in the labor market determines support for certain types of immigrants (Borjas et al. 1996). As the inflow of lower-skilled immigrants will increase the overall market supply of this type of workers and lower the wages for low-skilled natives (due to labor replacement), the latter are likely to look unfavourably on such immigrants. High-skilled natives can be more supportive, because such inflow helps raise the wages of high-skilled natives (due to growth potential and their managerial functions). In contrast, the opposite effects on attitudes will happen with an increase of higher-skilled immigrants. The FP model, although theoretically parsimonious, has received little supportive evidence. The literature does not find refusal of immigrants by natives with similar education and skill levels (Hainmueller and Hopkins 2014; Naumann et al. 2018). For the sociotropic viewpoint, natives in fact have a major concern with the impact of immigrants on two fronts, which exerts high influence on how they perceive incoming foreign workers and develop a stance for or against. First, immigration can be seen as a fierce challenge to the native culture, such that group boundaries they have made efforts to maintain are tarnished, damaged or even overthrown. This symbolic threat is more influential in affecting attitudes toward immigration than is self interest, which can be highly difficult to identify in daily life. Second, national economic interest, compared to personal interest, is a surprisingly powerful conceptual frame to formulate attitudes toward incoming foreigners. We discuss the issue of contribution to the national economy first.

High-skilled immigrants, seen to be contributing to economic performance across various native groups, are consistently preferred in many survey findings (Hainmueller and Hopkins 2014). This sociotropic perspective can also explain why most people hold a double-standard perspective to evaluate immigrants possessing different skills. The local residents oppose low-skilled immigrants mainly because of perceived threats to the national economy by taking jobs away and lowering wages. Indeed, conventional research has provided evidence linking such threats to antipathy toward low-skilled immigrants (Ben-Nun Bloom et al. 2015; Hellwig and Sinno 2016), although there is no denying that low-skilled immigrants can be functional in filling unwanted lower-paid jobs (Helbling and Kriesi 2014). On the other hand, highly skilled immigrants are presumed to be strong in capital input, innovation ability, and entrepreneurship, which contribute to increased productivity, employment, and income to the whole society (Greenwood and McDowell 2011; Facchini and Mayda 2012; Orrenius and Zavodny 2014). Entry of high-skilled immigrants is most welcome because locals see their functional contribution to the economic needs of the host country and overlook negative impacts of their competition with members of the middle class and even the upper-middle class (Malhotra et al. 2013). Two sociotropic hypotheses through the lens of a double standard can be proposed accordingly:

H1

Perceived immigrants’ economic contributions to the nation increase support for highly skilled immigrants, but have no effects on support for low-skilled immigrants.

H2

Perceived immigrants’ economic threats to the nation decrease support for low-skilled immigrants, but have no effects on that for high-skilled ones.

Another often-stressed impact of immigrants pertains to perceived welfare threat. This consideration, which focuses on public social services, asserts that high-skilled immigrants are preferred over low-skilled ones because the latter’s use of public services is disproportionally high, that is, higher than their contributions to tax revenues (Bonin et al. 2000; Borjas 1995). Whereas high-skilled immigrants can quickly become net contributors in terms of public finance, low-skilled migrants are viewed as potential net burdens, resulting in either increased taxes or reduced welfare services and benefits for the rest (Facchini and Mayda 2009). Several studies have shown that natives disapprove of low-skilled immigrants due to the belief that they erode welfare benefits for locals (Hainmueller and Hiscox 2010; Helbling and Kriesi 2014). As an increasing number of countries are reducing welfare benefits because of budget constraints, low-skilled immigrants are being increasingly described as significant threats to limited social resources. This divergence also explains the existence of a double-standard viewpoint derived from the perceived fiscal burden of high- and low-skilled immigrants. A hypothesis on erosion of welfare spending is advanced:

H3

Aperceived threat to the nation’s welfare system from immigrants does not decrease support for high-skilled immigrants, but will hinder support for lower-skilled immigrants.

Immigrants are perceived as a threat to host societies in cultural arenas involving language, religion, and ethnicity. This perspective deserves some additional elaboration. In addition to threatening social cohesion, the increasing presence of “strangers” can be seen as challenging traditions, values, and vernacular identities, resulting in an “us versus them” social divide that serves as a foundation for anti-immigrant hostility (Pietsch and Marotta 2009). The underlying rationale of such an identitarian ideology is the fear that the presence of immigrants will alter national culture and disturb the ways of life which have been followed and cherished for a long time (Zincone 2006). These fears of negative impact are used to justify a policy of limiting and screening immigration flows and citizenship and favouring immigrants from similar ethnic backgrounds (Ben-Nun Bloom et al. 2015).

The double-standard perspective would hypothesize that perceptions of cultural threats tend to be especially strong against low-skilled immigrants, who are viewed as lacking in terms of education, cultural capital, and language skills, and who are therefore perceived as being less able to adapt to a host society’s mainstream culture, regardless of how it might be defined. In contrast, highly skilled immigrants are more likely to be viewed as culturally acceptable because of their “trouble-free” middle- or upper-middle-class backgrounds (Tzeng 2006). High-skilled immigrants are believed to be more liberal, less religious, and more open-minded (Helbling and Kriesi 2014). They tend to have better language skills and to quickly adapt to the norms and customs of their new societies (O’Connell 2011), thereby adjusting to local cultures faster and with lower assimilation costs (Facchini and Mayda 2012; O’Connell 2011). Accordingly, high-skilled immigrants are generally not viewed as threats to local cultures. Thus, the fourth hypothesis is proposed:

H4

Perceived immigrants’ threats to the national culture decrease support for lower-skilled immigrants but have no effects on high-skilled immigrants.

3 Data, Measures and Estimation Method

Data were drawn from the 2011 Transatlantic Trends: Immigration (TTI) survey (https://doi.org/10.3886/ICPSR34423.v1). TTI surveys are designed to measure public opinions about immigration and social integration based on nationally representative samples in large migrant-receiving countries in Western Europe and North America. This cross-national project has been conducted by the German Marshall Fund of the United States⁵ (GMFUS) in cooperation with its various partners since 2008 (Gustin and Ziebarth 2010). The participating countries and the contents of each TTI survey are not exactly the same every year. The 2011 TTI survey included separate questions of attitudes toward immigrants with different education levels and covered six high-income countries, the US, UK, Germany, France, Italy and Spain. The field method used multi-stage random sampling techniques with computer-assisted telephone interviews involving individuals aged 18 and older living in households with access to landline telephones. However, since 20% or more of the populations in Italy, Spain, and the US had access to cell phones only at the time of survey, 20% of the samples in those three countries were contacted via their mobile phones.

The two dependent variables are the respondent’s attitudes toward high-skilled and low-skilled immigrants, respectively. The original questions were expressed as “How much do you agree or disagree that (COUNTRY) should allow more immigrants with a high level of education to come and live here?” and “How much do you agree or disagree that (COUNTRY) should allow more immigrants with a low level of education to come and live here?” Responses were recorded along a scale of 1 (“strongly disagree”) to 4 (“strongly agree”) and responses of “don’t know” or refused to answer were recoded into the midpoint of the scale, 2.5, to fully utilize the dataset.⁶ Past studies have similarly referred to an immigrant’s education as a suitable proxy for skill level (Hainmueller and Hiscox 2007; O’Connell 2011). Higher scores indicate more favourable attitudes toward high- or low-skilled immigrants. To prevent the potential order effect of the two dependent variables, the two questions were rotated randomly by way of computer-aided interviewing. The Pearson’s correlation coefficients between the two dependent variables are moderate for the US, UK, Germany and Spain (ranging between 0.35 and 0.45). The French and Italian samples showed a high association (0.53 and 0.61). This indicates existence of an underlying common factor in the attitude formation, although the two measures address distinct issues. Other surveys also collected opinions about immigrants in rich democracies. However, their design did not solicit attitudes toward high- and low-skilled immigrants separately from the same respondent.⁷

The field work of the 2011 TTI survey was conducted in each country for 2 to 3 weeks between August and September. The response rates for the US, UK, Germany, France, Italy and Spain were 25%, 9%, 17%, 5%, 4% and 17%, respectively. The short span of field work seemed to be responsible for many noncontacts and thus low response rates. Indeed, the response rate of household telephone surveys has been diminishing because of the popularity of cell phones and other reasons (O’Toole et al. 2008; Beullens et al. 2018). In general, response rates for landline telephone polls dropped dramatically from 36% in 1997 to 9% in 2012 (Keeter et al. 2017).

Since nonresponse rates have traditionally been viewed as one key indicator of survey quality, lower response rates are a concern for unobserved bias in survey results. However, Leslie (1972), after reviewing several survey research results, demonstrates that a low response rate is unlikely to significantly bias the result when surveys target homogeneous populations (persons sharing similar backgrounds or opinions). Wright (2015) also demonstrates that probability-based samples can achieve a high level of accuracy with a surprisingly low response rate, while Krosnick (1999) and Yeager et al. (2011) agree that there is no need to presume that lower response rates signal lower representativeness when probability sampling methods are used (Hendra and Hill 2018; Yeager et al. 2011; Groves et al. 2009; Groves 2006). In other words, nonresponse may but does not necessarily induce bias in survey estimates (Groves et al. 2009). Because people who are more socially engaged are more likely to respond to survey requests (Keeter et al. 2017), a low response rate leads to bias on topics of civic engagement or volunteering but not necessarily on others like attitudes toward immigration. Recent research such as Rindfuss et al. (2015) and Amaya and Presser (2017) has proved that nonresponse bias may affect univariate estimates but often is less substantial in multivariate models and rarely alters the direction of association in the model. Yan and Curtin (2010) propose that respondents who have high missing-data rates are similar in some respects to those who refuse to participate in surveys. We think this is a key point for data checking. We found that those who provided answers are no different from those who did not on basic demographic characteristics.⁸ Thus, we decided that nonresponse bias is trivial in the dataset used.

Comparing two similar questions on attitudes toward high- and low-skilled immigrants between TTI and ESS7, we found several similar patterns (Table 2) in the two data sets. First of all, respondents favoured high-skilled immigrants over the low-skilled immigrants in each data set. Second, compared to any other countries, Spain contained a greater percentage of missing information (i.e., respondents refused to answer, had no answer, or answered “don’t know”) on respondents’ attitudes toward both types of immigrants. This tendency was also reported by Zapata-Barrero (2009) in his study of public opinions toward immigrants in Spain. Third, the amount of missing information was small, although TTI had slightly higher percentages. Due to these consistent patterns, we have additional confidence in using the TTI data set, which contained lower response rates.⁹

Table 2 Respondents’ opinions on allowing more immigrants of different skills (%)—TTI and ESS7 compared

Andreß et al. (2013, p. 58) propose that weights be used when making representative statements about a population, while using weights in statistical modelling should be avoided if control in sampling selectivity is not performed. With these concerns in mind, we remain cautious on potential bias derived from low response rates, especially in France and Italy. We report the outcomes for the weighted samples for cross-validation (more explanations follow). As for statistical modelling, we also use suitable techniques for preventing potential influences owing to low response rates. The TTI sample size was 1000 participants from each country, except the US with a sample size of 1001. We selected respondents with age between 18 and 69 for analysis to represent the majority of the adult populations.¹⁰ The proportion of foreign-born respondents varied from highest in Germany (10.3%) to lowest in Italy (4.5%) (see Table 3).

Table 3 Summary statistics—percentages or means

The main independent variables were perceived national impact on economy, culture and welfare burden associated with immigrants. The TTI survey contains two items regarding perceived economic functions: (1) “Immigrants generally help to fill jobs where there are shortages of workers”, and (2) “Immigrants help create jobs as they set up new businesses”. Additionally, two items represent perceived economic threats: (1) “Immigrants take jobs away from native born workers”, and (2) “Immigrants bring down the wages of (NATIONALITY) citizens”. Competition for social services by immigrants has been a controversial issue in public policy debates. Ideally, this measure should cover a variety of welfare use such as medical programs, public schools, food, housing, etc. (Camarota 2015). However, the TTI survey does not provide detailed measurement based on these items; thus, our measure is not optimal in this aspect. One single measure is used to tap perceived welfare burden: “Immigrants are a burden on social services”. These indicators above are measured on a four-point scale, with 4 = strongly agree and 1 = strongly disagree.

Perceived cultural threat was measured by two items: (1) “Some people think that immigration enriches (NATIONALITY) culture with new customs and ideas. Others think that these new customs and ideas negatively affect (NATIONALITY) culture”. Agreement with “Immigration negatively affects (NATIONALITY) culture” is coded as one (0 = immigration enriches culture). (2) “The people who come to (COUNTRY) should try to act like people from (COUNTRY)”. Answers were recorded from 1 (“strongly disagree”) to 4 (“strongly agree”). Higher scores indicate strong expectation that immigrants should assimilate into the native culture rather than celebrating hospitable multiculturalism, which in a way hints at a cultural threat.

The control variables in this study include the respondents’ demographics, socio-economic backgrounds and a number of important factors that have been considered in previous research. In addition to gender, age and birthplace (native- or foreign- born dummy), we used education and household financial situation to measure socio-economic background. Our preliminary analysis of occupational status did not reach significance in model estimation; therefore, it is not included in the final models. Education is a set of dummies capturing post-graduate, college, and secondary or lower degree. A self-rated improvement in financial situation (ranging from 1 (“a lot worse”) to 5 (“a lot better”)) is also used as a proxy for the respondents’ economic conditions, because individual income is not available from the TTI survey. As more cross-cultural networks increase support for immigrants (Tsai and Tzeng 2014), a measure of the number of friends who had been born in another country, but who were now living in the same country as the respondent, was used to indicate a cross-cultural network. A summary of all statistics for the independent and control variables is presented in Table 3.

Because we estimate attitudes toward both high- and low-skilled immigrants simultaneously rather than using a single equation for each, a specific type of generalized least squares estimators known as seemingly unrelated regression (SUR) modelling is employed (Fiebig 2003). It is highly likely in our case that the error terms across the regression equations are correlated, that is, there are unobservable variables operating such that an individual who favours high-skilled immigrants also tends to tolerate low-skilled immigrants. Sample information across equations as such can be employed to improve the precision of estimation of parameters in a given regression equation. The SUR model gains better unbiased estimation by joint analysis of the set of regression equations, especially when the contemporaneous correlations are large. However, the gained efficiency will reduce to single-equation least-squares estimation when (1) error terms across different equations are mutually uncorrelated, that is, the equations are empirically unrelated; and (2) equations use identical predictors (Baum 2006). The high correlations found between our dependent variables in respective country samples justify choice of the SUR method (more in the following section).

Recent analysis of cross-national research has increasingly adopted techniques of multilevel modelling with a number of survey datasets pooled together and estimated simultaneously (Luke 2004). In this study we decided not to follow this approach. Rather, we perform analysis of individual populations. The reasons are twofold. First, because there are data of only six countries, it is not efficient to estimate societal structural factors, in which the small N (total observations) problem renders their evaluation unreliable if not plausible. The literature suggests a threshold number of 25 populations for pooling (Bryan and Jenkins 2016). Second, the demographic compositions of the six analyzed societies differ, despite being relatively wealthy countries. Multivariate relationships can be unique and thus meaningful under particular contexts, but tend to become indistinct when pooled with other populations (Babones 2014). In light of these considerations, the section that follows presents empirical findings derived from analyses based on individual countries in reference to a pooled analysis. But for easy comparison, the reporting of the outcomes is structured to show comparable patterns across societies.

4 Analysis and Results

In Table 2, we arranged our data into “agree” and “disagree” groupings to facilitate clearer presentation of our results for respondent attitudes toward the high- and low-skilled immigrants, respectively. The results from unweighted and weighted samples are displayed for comparison. The weighting scores are computed by considering use of landlines and cellphones in the first stage, and then incorporating basic socio-demographics (age, gender, region and education) to compute a weighting score for each respondent. This weighting scheme provided by the TTI survey is well rounded and reliable for statistical estimation. In general, more than half of the respondents in each of the six countries agreed with the idea of allowing more high-skilled immigrants to stay, with the highest level of acceptance noted among the German respondents (73.2%), followed by Spain (68.5%), Italy (66.0%), the US (61.7%), France (56.2%), and the UK (55.4%) (column 1). The weighted results (column 2) indicate that Spain registers the highest percentage (69.2%), while Germany ranks second (67.4%). The last two countries also switched their positions, but the original difference between the two was very trivial.

Support was considerably less for low-skilled immigrants, with Italian respondents being the most positive (44.2% accepting), followed by French (36.3%), Spaniards (33.7%), Americans (33.5%), Germans (24.3%), and Britons (15.3%) (unweighted scores). The ranking order is about the same for the weighted scores except that the US and France switch positions. Again, the percentage changes between the two scores for the two countries are small. According to these results, the German respondents are very enthusiastic about welcoming high-skilled immigrants, while the Britons are the least welcoming, regardless of immigrant skill level. The nearly least welcoming attitude toward both types of immigrants in the UK (based on either weighting or unweighting techniques) had foreshadowed Brexit in 2016, as many studies have suggested that the Brexit referendum had been perceived as a vote against migration in general (Walter 2019). The unweighted percentage difference between accepting the two types of immigrants was lowest in France (19.9%), followed by Italy (21.8%), the US (28.2%), Spain (34.8%), the UK (40.1%), and Germany (48.9%). The weighted difference shows exactly the same ranking. The disparity of acceptance for the two types of immigrants appears substantial, with a wide range from 20% to nearly 50%, indicating a much warmer welcome for skilled immigrants in these high-income countries.

A relevant, interesting question is: do respondents hesitate more to express their opinions about low-skilled workers? This question is raised because social desirability may be at work such that a choice of “don’t know” or not giving any answer is more socially acceptable. However, the differences of percentage of both “don’t know” and “no answer” for the two types of immigrants are approximately the same [see columns A and B, (2)–(4) of Table 2]. This indicates that respondents are equally willing to provide opinions regarding high- and low-skilled immigrants.

The SUR results are presented in Table 4. Model 1 only contains our main independent variables (reflecting the sociotropic concerns), and Model 2 includes the demographics and other control variables which are not reported in order to save space. Note that only control variables which reached significance were included in the analysis. This practice is a necessity to gain more efficiency from SUR estimation; otherwise the regression outcome would reduce to that of the conventional least squares method (Fiebig 2003). The correlations of residuals in the seven SUR models with control variables have an average of .29 (highest in Italy and lowest in the UK), and all reach significance. The explained variance (R²) is substantial for both immigrant groups across the six samples, all together having an average of .25. We also note that education, being foreign-born and cross-cultural networking are especially influential in estimation.

Table 4 Attitudes toward immigrants in six countries: SUR estimates

Additionally, the multicollinearity problem was checked to avoid potential estimation instability (that is, inflated standard error) from including conceptually correlated predictors in a model, particularly from the perceived threat variables. The diagnostic estimation based on the variance inflation factor indicates that multicollinearity is not a concern because they all stay below 5, a substantial distance from a rigorous threshold value of 10 (O’Brien 2007).

Comparing models 1 and 2, we found only a little difference in R squares and coefficients. In other words, addition of demographic and other control variables does not increase explained variance to a substantial degree. This finding not only is in line with previous research findings that self-interest does not effectively account for individual attitudes toward immigrants but also supports evidence that individuals’ sociotropic concerns have more influence. The following sections show how different aspects of the sociotropic concern relate to the individual attitudes.

We first report the results when all six samples are pooled with a set of country dummies (the coefficients for country dummies are not shown to save space), to show the overall patterns of association across countries. Perceived economic function is positively correlated with acceptance of immigrants, while perceived economic threat, social service burden and cultural threat carry negative influences. Two predictors, however, do not reach significance: “should act like locals” for the high-skilled immigrants, and “bring wages down” for low-skilled immigrants. While this pooled estimation seems to provide a quick summary, it ignores country differences owing to what are called “compositional effects” (Babones 2014). That is, the specific influences owing to the structural characteristics of the population can be substantial across countries but are easily overlooked in pooled regression analysis. Thus, to further investigate the patterns of multivariate correlation, we decided to analyse each country sample for detailed comparison.

As was found in the pooled analysis, we note a strong influence of the perception of job creation on accepting high-skilled immigrants across all six countries. However, when looking at individual country samples, we find favourable effects of the filling jobs factor on accepting high-skilled immigrants only in the US, UK and Germany. This is probably because these advanced countries are experiencing a shortage of talents to fill high-skilled jobs. As for the low-skilled immigrants, job filling increases their acceptance in all countries except Spain, and job creation consistently increases their acceptance except in Germany. While low-skilled immigrants are more likely to be welcome when perceived as filling jobs, in the case of high-skilled immigrants, it is mostly due to creating jobs. These results thus lend only partial support to H1 because the level of acceptance for low-skilled immigrants also increases when people have a positive evaluation of the economic function of immigrants as a whole.

While we did not expect an association between perceived economic threat and acceptance of high-skilled immigrants, the results indicate a visible association: the possibility of immigrants taking jobs away from locals was an important concern among respondents in all countries except France. This finding is similar to one recent study (Malhotra, Margalit, and Mo 2013). Furthermore, the concern of wages being brought down was strongly significant in France and slightly so in Germany. For the low-skilled immigrants, job replacement is a concern strongly significant in the US and Spain, and slightly significant in the UK.¹¹ However, the wage reduction factor exerts no effects at all across the six countries (as found from the pooled model). In sum, perceived economic threats, especially job replacement, decrease support for high-skilled immigrants more than for low-skilled ones. Thus, H2 is not supported.

Perceived burden on social services exerts negative impacts for accepting high-skilled immigrants with medium significance in Italy and slight significance in the UK and France. With regard to low-skilled immigrants, this negative association appears strongly significant across all six countries. These findings provide only limited evidence for H3, because a correlation between perceived welfare burden and lower acceptance of high-skilled immigrants was also found in some studied countries.

We predicted that potential cultural threat was more likely to lower acceptance of low-skilled immigrants. The findings show that in all six countries, respondents who perceived immigrants as posing a negative effect on local culture were strongly opposed to accepting high- as well as low-skilled immigrants, except that there was insignificant influence for low-skilled immigrants in the US. This fails to support H4. However, we also found that the expectation that immigrants “should act like locals” operates weakly for high-skilled immigrants: it exerted a slightly negative effect only in Germany and Italy. In contrast, for low-skilled immigrants this predictor is quite significant in Germany, France, Italy and Spain, but fails to reach statistical significance in the English-speaking countries, the US and UK. In other words, for high-skilled immigrants, requiring assimilation into the native culture might seem unnecessary, perhaps due to their “trouble-free” image. In this sense, H4 receives some supportive evidence. The analytical outcomes obtained herein also alert our attention to the need for fuller specification of various aspects of cultural threat in measurement.

In sum, not all independent variables show a double standard, which the sociotropic perspective would probably expect to happen. Of the 42 pairs of effects (7 independent variables x 6 countries) as summarized in Table 5, only six pairs support the differentiating double standards that treat immigrants differently according to their skill levels. Another seven pairs show reversed double standards, in the sense that low-skilled immigrants are not particularly disfavoured over the high-skilled immigrants when certain negative perceptions appear. (The example is that in both Germany and Italy, perception of the immigrants’ taking jobs away leads to an unfavorable attitude toward high-skilled immigrants but no such association for the low-skilled immigrants). However, 29 pairs show a consistent influence across both groups of immigrants, which indicates a single standard judgement. In the latter case, there are 13 positive impacts, and eight negative ones. Finally, eight pairs show no impact at all. On the basis of these outcomes, we conclude that the double standard is not the predominant viewpoint in the mass public’s attitude toward immigration. As the above results imply clear limitation of the double standard perspective, we propose a revision and a reformulation of the sociotropic viewpoint for future research design.

Table 5 Summary of testing double-standard hypothesis

5 Conclusion

This study asserts that natives prefer high-skilled immigrants over the low skilled, as found in previous research (Hainmueller and Hiscox 2010; Helbling and Kriesi 2014; Hainmueller et al. 2015; Valentino et al. 2019; Ford and Mellon 2019). Unlike most previous research focusing on egocentric concerns to explain acceptance of immigration, we tested the sociotropic hypotheses (Hainmueller and Hiscox 2010; Hainmueler and Hopkins 2014, 2015) with which to explain whether double standard evaluations are often applied to immigration among the mass public. The empirical findings reveal that sociotropic thinking is more effective than egocentric concern or the FP prediction in explaining variation among the mass public. Focusing on six wealthy destination countries, the obtained outcomes challenge a conventional belief that emphasizes positive labour market impacts with no cultural threat or welfare burden in welcoming high-skilled immigrants, and negative labour market impacts, high cultural threat and welfare burden in rejecting low-skilled immigrants. The evidence shows that both low- and high-skilled immigrants carry wanted and unwanted assets simultaneously in the mass public’s perception. Thus, sociotropic thinking does not necessarily embrace a double-standard judgement; rather, it goes quite nicely with a single-standard attitude toward immigration. The reasons why a single standard emerges can be various. It probably is a result of avoidance of cognitive dissonance (Aronson 1969). Other suggested factors (or communicators) include prejudice, mass media, political partisanship and even elite rhetoric (Fusell 2014; Hainmueller and Hopkins 2014; Pereira et al. 2010), which may also contribute to a generalized and overall evaluative thinking, rather than practicing a finer categorization of immigrants according to their differential education, skills or abilities.

This one-standard thinking also is reflected in the finding that job replacement is a salient predictor for disfavouring high-skilled immigrants. The context of our study is high-income countries, where high competition between the locals and high-skilled immigrants causes graver concern than that with incoming low-skilled workers. It is observed that job creation by immigration is an important factor for locals to accept low-skilled immigrants, which is often neglected in immigration policy debates. While welfare burden is a strong predictor of the locals’ unfavourable attitude toward low-skilled immigrants in all six countries, it also occurs for the high-skilled ones in Italy, France and the UK, although in a moderated manner. A recent study of European countries indicated little evidence that immigrants were excessive receivers of the welfare system compared to locals, however (Barrett and Maître 2013). This hints that stereotypical rather than actual social service usage lies behind many cases of attitudinal responses to immigration.

Finally, belief in immigrants’ negative impact on local culture and way of life leads to restrictivism, regardless of their skill levels in all six countries except the US, where only high-skilled immigrants exert such cultural threat. In the UK and US, perceived cultural threat lowers acceptance for high-skilled immigrants, but it has only a mild influence on that for low-skilled immigrants. The opposite occurs in Spain, in which there is less demand for high-skilled immigrants in comparison. But when it comes to the issue of assimilation into host cultures, our findings support the double-standard argument: expectation that immigrants should “act like locals” is not linked to acceptance of high-skilled immigrants except in Germany and Italy. However, this factor is strongly associated with antipathy toward low-skilled immigrants, except in the US and UK. Since the US and UK felt little or only mild cultural threat from the low-skilled immigrants, there is no need to require them to act like locals. In contrast, Germans perceived incoming cultural threats from the low-skilled immigrants; they therefore require them to act like locals. Specific cultural legacies might serve as a factor for such differences across countries.

We note three limitations. First, the measures of household income are not optimal due to the lack of detailed work contents and precise personal (or household) income data. Second, measures of welfare burden and cultural threat were based on only one or two indicators whose coverage was constrained. Future studies can improve upon this by mobilizing more nuanced and composite measures to test these factors for further validation. Third, the low response rates in France and Italy might somewhat discount the reliability of the findings from the two countries, despite the fact that we have shown this issue might not be as serious and provided supportive evidence from resampling techniques. Nevertheless, this study is one of the few endeavours to apply a sociotropic approach in a cross-country comparative study on the public’s respective attitudes toward high- and low-skilled immigrants. In addition, this study also is the first one to demonstrate natives’ predominantly single standard concpetion toward immigrants regardless of their skill level. This research design and measurement of perceived national interests can be replicated in other popular destination countries. Higher income countries in other regions (e.g. East Asia or oil-rich countries) seem like a natural extension for such research interest to determine whether the single-, rather than, double-standard perspective is the leading tenet among the mass public in deciding which type of immigrant is more welcome, if either.