Introduction

Canonical work on immigration has long sought to explain native citizens’ preferences toward restrictive and liberal immigration policies. However, few studies have examined an equally fundamental question: do immigration policies themselves shape attitudes toward immigrants? That is, would two otherwise identical immigrants be evaluated differently by natives if they are admitted through different immigration policies? Existing work tends to treat immigration policies as an outcome rather than a determinant of natives’ attitudes toward immigrants. Despite the lack of empirical evidence, policymakers are cautious about the potential for liberalizing immigration policies to cause public opinion backlashes. Even in the case of highly skilled immigration, states often swing between restrictive and liberal policies. Most existing talent schemes contain a labor market test to ensure that highly skilled immigrants will not undermine the occupational interest of natives. The more restrictive ones come with labor protection, fixed time limits, and points-based tests that account for applicants’ proficiency in the language of the host country.

Immigration policies attract intense public scrutiny because they determine immigrants’ profiles, rights, and responsibilities (Czaika & De Haas, 2013). The salience of these issues in contemporary politics justifies the question of whether restrictive policies strengthen or weaken natives’ support for immigrants. In this paper, I develop a theory of attitudinal policy feedback within the context of immigration. Furthermore, to investigate to what extent the attitudinal effects of immigration policies can be accounted for by existing theories, I examine how responses to policies vary across natives with different socioeconomic characteristics and psychological attributes.

The attitudinal effects of immigration policies are theoretically ambiguous. Scholarship on policy feedback provides a useful framework for assessing the resource and interpretative effects of public policies but is ambivalent regarding when and how those effects occur in specific policy domains. In the context of immigration, one might expect that restrictive policies would generate more support for immigrants because they protect natives from economic and cultural threats (Traunmüller & Helbling, 2022). Furthermore, restrictive policies may signal strong anti-immigration sentiments in society and thus drive moderate individuals to distance themselves from accusations of xenophobia and racism (Schwartz et al., 2021).

While the positive attitudinal effects of restrictive immigration policies might seem intuitive, the literature offers strong theoretical reasons to hypothesize otherwise. For instance, restrictions on labor market rights and access to welfare have been described as “heuristic devices” that amplify natives’ threat perceptions toward immigrants (Ryo, 2017). Policies based on ascriptive traits, such as language, may solidify “hierarchical constructions of citizenship,” entrenching negative stereotypes (Ellermann & Goenaga, 2019). Furthermore, certain classic models in social psychology suggest that group boundaries tend to increase intergroup bias (e.g., Gaertner et al., 1990). Based on these models, we could argue that immigration restrictions reinforce boundaries between immigrants and natives, thereby fueling anti-immigrant sentiment.

These conflicting theoretical predictions raise an important empirical question. However, most studies to date have not treated policies as an explanatory variable, but rather as an outcome measure. The limited research examining the feedback effects of immigration policies has largely focused on the impact of these policies on immigrants’ sense of belonging, identity, and potential for integration (e.g., Jiménez et al., 2021; Mohamed, 2017; Wright & Bloemraad, 2012). Efforts to explain natives’ immigration attitudes have yielded mixed results. Schlueter et al. (2013) observe more anti-immigrant sentiment in countries with stringent integration policies. By analyzing discussions on Twitter, Flores (2017) shows how a high-profile anti-immigrant law passed in Arizona triggered more negative sentiments against Mexican and Hispanic immigrants in the US. Contrarily, Schwartz et al. (2021) note that anti-immigrant attitudes softened after the Brexit Referendum. In Asia, Lee et al. (2022) report a positive association between perceived stringency of border restrictions and support for immigration.

One reason for these mixed findings is the multidimensional nature of immigration policies. To test the attitudinal effects of policies, many scholars rely on cross-national or longitudinal comparisons. The first strategy compares immigration-related attitudes across policy regimes, whereas the second examines attitude changes following the implementation of a new immigration policy. These empirical strategies enable scholars to broadly estimate how policies likely influence attitudes. However, they do not allow for comprehensive testing of the effects of competing hypotheses because, in practice, measures with potentially divergent impacts on attitudes are often combined within single-policy packages. For instance, a relaxation of language requirements may be enacted alongside tighter labor market restrictions for immigrants. Instead of disaggregating policies into smaller groups, existing studies have either subsumed them under a single generic rubric of policy restrictiveness or have focused exclusively on just one or a few restrictions, which risks eliding the individual effects of different policies.

A more serious challenge is confounding. Policy changes are often tied to powerful but context-specific political processes (Mettler et al., 2023). These processes may long precede the actual policy change and thus cannot be effectively controlled for by before-and-after comparisons. For example, restrictions formulated against the backdrop of anti-immigrant mobilization may galvanize negative predispositions against immigrants. Similarly, immigration liberalization may be accompanied by media campaigns or education reforms that socialize the values of multiculturalism. These processes can dilute the effects of policies, leading to potentially erroneous conclusions that immigration reforms are inconsequential or counterproductive.

This paper exploits an original paired profiles conjoint survey experiment conducted in Hong Kong to tackle the above challenges. The experiment features immigration schemes for highly skilled immigrants that randomly vary in restrictiveness along economic, cultural, and welfare dimensions. This method not only distinguishes among highly skilled immigrants admitted under policies of varying restrictiveness but also simultaneously tests for the attitudinal effects of policy items of importance to major theoretical explanations of attitudes toward immigrants. To examine the rationale behind the participants’ choices, I conducted in-depth qualitative interviews with 54 participants from diverse industries. I focus on highly skilled immigration because it is far less studied than refugee and low-skilled migration despite having outgrown them (Kerr et al., 2016). It also poses a hard test for the attitudinal effects of policy, as natives should favor or be indifferent to the granting of additional rights to foreign talent.

The main finding from the experiment is that even in the context of highly skilled immigration, policies still significantly and substantively influence how natives view immigrants. Specifically, an increase in policy restrictiveness always enhances the expected probability of an immigrant being preferred by the natives. More strikingly, contrary to predictions derived from major theories, the attitudinal feedback effects of policies vary little across groups. The scattered interaction effects are significant only for nationality requirements and, to a smaller extent, preferential welfare treatments. Furthermore, only social identities partially change the direction of policy feedback generated by nationality requirements, which suggests that policies exert distinct attitudinal effects not captured by conventional explanations.

The findings from qualitative interviews complement those of the experiment. They not only confirm that people understood the experimental policy profiles and developed clear policy-based preferences and sociotropic concerns but also suggest two novel mechanisms of attitudinal policy feedback: moralizing and quality assurance effects. The former triggers moral judgments regarding loyalty and fairness, whereas the latter shapes the perceived quality of immigrants. These effects are mostly non-deliberative but are so resilient that the resultant immigration preferences remain unchanged even when the merits of liberal policies are highlighted.

This paper fills a long-standing gap in immigration attitude scholarship. Although considerable work has been done on native citizens’ preferences for restrictive or liberal immigration policies, we have yet to tease out how those preferences are grounded in the heterogeneity of immigration policies. Closing this knowledge gap would undercut major explanations of attitudes based on individual differences. For example, sociotropic economic concerns have long been seen as a powerful predictor of natives’ support for immigrants. However, once the effects of immigration policies are accounted for, these concerns no longer play a leading role in explaining attitudes. Natives with different predisposed sociotropic economic concerns invariably prefer highly skilled immigrants admitted through restrictive policies. This shared preference is all the more remarkable given the well-documented perception of highly skilled immigrants as net contributors to the national economy. Sociotropic economic concerns are important, but my findings suggest that they are conditioned by immigration policies.

This paper also contributes to the growing policy feedback literature, which has overwhelmingly focused on US welfare policies (Béland & Schlager, 2019). Highly skilled immigration policies are distinctive in that their beneficiaries include non-citizens, which presents policymakers with a “control dilemma”: Generous incentives may attract more or better-qualified immigrants but may also fuel anti-immigrant sentiment (Kolbe, 2021). This paper reveals fresh evidence of the attitudinal policy feedback arising from these policies, extending policy feedback research to a new and theoretically interesting policy area.

Background

This study focuses on highly skilled immigration in Hong Kong. Since 2003, the government has expanded talent schemes for foreigners. Table A1 summarizes the city’s immigration policies for highly skilled workers in 2021. These policies have varying requirements, admission procedures, and incentives. For example, the General Employment Policy (GEP) admits non-mainland Chinese citizens with a job offer, while the Quality Migrant Admission Scheme (QMAS) is a points-based system without nationality or job requirements. Immigration policies in Hong Kong are subject to frequent changes, enhancing the experiment’s realism and reducing the risk of status quo bias.

Despite these policy efforts, Hong Kong continues to face a severe talent shortage. In a survey conducted in 2022, 83% of businesses in the city reported difficulties finding skilled workers (Manpower Group, 2022). This shortage is not unique to Hong Kong. Global talent competition has intensified over recent decades (Czaika & Parsons, 2017). Not only have governments raced to relax admission requirements, but some also offer generous financial incentives. Unlike low-skilled workers and refugees who may face prejudice from the public, highly skilled immigrants are widely viewed as net contributors to national economies. Therefore, if policy liberalization does reduce support for immigrants, as hypothesized in this study, such an effect should be less salient in the context of highly skilled immigration.

Of Hong Kong’s 7.5 million people, immigrants comprise 38% (Census and Statistics Department, 2021). Despite mounting concerns over mainland China encroaching upon Hong Kong’s autonomy, by 2021, the city remained the principal regulator of immigration, particularly for highly skilled immigrants. Non-Hong Kong citizens, including those from other parts of China, must apply for a visa issued by the Hong Kong Government to work and settle in the city. Also, as in many parts of the world, immigration in Hong Kong is a contentious political issue (Appendix I.1). Much of the hostility is directed toward mainland Chinese immigrants, who have been blamed for raising welfare costs and spreading China’s cultural and political influence (Jiang et al., 2023). This politically motivated antipathy may heighten the salience of identity-related concerns. However, Hong Kong is not exceptional in how political considerations and international relations fuel anti-immigrant sentiment. For example, attitudes toward Russian immigrants in some former Soviet countries are heavily shaped by memories of Soviet domination and suspicions of Russian influence (Klvaňová, 2019). Appendix I.3 presents data from previous studies and cross-national surveys to further illustrate why Hong Kong’s immigration politics is highly comparable to that of many industrialized societies.

Insights from Theory and Hypotheses

Policies for highly skilled immigrants can be categorized into three groups (Czaika & Parsons, 2017): (1) labor market tests and work rights, which manage job competition between natives and immigrants; (2) language requirements, designed to address perceived cultural threats; and (3) financial incentives and permanency rules, which set tax and welfare obligations. Most countries implement some of these restrictions, the effects of which on attitudes toward immigrants are explored below.

Policy Feedback

The idea that public policies can reshape political processes and attitudes is not new but has recently attracted substantial attention. Building on the work of Pierson (1993), this scholarship has focused on two broad mechanisms of policy feedback: resource effects and interpretive (or cognitive) effects. Resource effects arise from the uneven distribution of policy benefits and costs across different groups of people. Beneficiaries of a public policy are incentivized to defend and expand the policy (Campbell, 2012). For example, in the US, the Affordable Care Act has gained more support among its most likely beneficiaries (i.e., those who are low-income and in their early 60s) (Sances & Clinton, 2021). We should observe similar effects in immigration policies because they too have significant distributional consequences (Beerli et al., 2021). If materialistic concerns do shape attitudes toward immigrants, as suggested by a large body of literature, we should expect significant resource effects from immigration policies.

Interpretive (or cognitive) effects describe how policies influence beliefs regarding what is “possible, desirable, and normal” (Soss & Schram, 2007). People may extrapolate normative judgments about broader political and social issues from the design of public policies and their direct experience with them. In the case of highly skilled immigration, public policies may convey interpretive effects primarily because they make group boundaries salient. They define, on legal grounds, the membership and content of the citizenry and the conditions under which immigrants can be included (Mettler & Soss, 2004). Group boundaries do not necessarily produce hostilities, but they provide a basis for intergroup comparisons (Brewer, 1999). Specifically, the policies governing immigrants’ rights and responsibilities can be subjects of intense equity assessment, in which natives evaluate immigrants’ inputs and obtained outcomes relative to their own (Cohen-Charash & Spector, 2001). Prior studies in social psychology (e.g., Doosje et al., 2013) have shown that unfavorable comparisons may be seen as a threat to one’s ingroup status and thus stoke defensive anger toward the outgroup of, in this case, highly skilled immigrants. Finally, policies offer “credible cues” for moral judgments (Ryo, 2017). Natives’ preferences concerning selection mechanisms and eligibility criteria can influence their perceptions of immigrants as deserving or undeserving of rights and citizenship.

Policies can exert negative and positive attitudinal effects simultaneously (Campbell, 2012). Explaining why specific effects prevail has been a focus of recent policy feedback research. Using longitudinal data collected in China, Zhang et al. (2023) propose a “positive–negative asymmetry” in policy feedback, according to which the public is more sensitive to losses from a policy than to gains of similar magnitude. Their argument is rooted in what psychologists call “negativity bias,” in which negativity tends to have a more substantial influence than positivity (Rozin & Royzman, 2001). Based on this body of work, we may assume that natives are more heavily influenced by policy cues about protection from immigrant threats (or the lack thereof) than by those regarding the benefits of immigration.

To further theorize the direction of attitudinal feedback, we must disentangle different types of highly skilled immigration policy. To begin, let us consider the effects of labor market regulations. Highly skilled immigration policies typically include specific laws governing immigrants’ rights in the labor market. For example, labor market tests require immigrants’ employers to demonstrate that they cannot find a suitable native worker. The aim of such regulations is to alleviate natives’ concerns regarding the potential harm that immigrants may bring to the country’s economy. By protecting natives from foreign job competition, these policies should encourage the public to see highly skilled immigrants as a complement rather than a threat to the local workforce. While it is true that stringent restrictions may limit immigrants’ potential economic contributions, given the insights of the negativity bias literature, we should expect the effect of immigration benefits to be less salient than that of threats in attitude formation. My first hypothesis is therefore as follows:

H1

Natives prefer highly skilled immigrants admitted under restrictive labor market policies to those under liberal policies.

In addition to labor market regulations, highly skilled immigration policies often give (implicit) preference to people from certain cultural or ethnic backgrounds (Ellermann & Goenaga, 2019). Immigrants from cultures believed to be compatible or similar to that of the host country often enjoy more rights than other immigrants, with language requirements as the most common assessment tool. These “hierarchical constructions of citizenship” are driven by concerns about national identity and social cohesion (Castles, 2005). When immigrants possess valuable skills or arrive in large numbers, policies that recognize native culture or erect barriers against certain immigrant groups may placate natives’ concerns about their declining status (Neureiter, 2022). In the case of Hong Kong, where widespread fears loom regarding the influx of immigrants from mainland China, people’s attitudes may be influenced by policies that offer protection against “mainlandization.” My second hypothesis is therefore as follows:

H2

Natives prefer highly skilled immigrants admitted under policies that favor local culture or limit mainland China’s influence to those who are not subject to such policies.

Finally, immigration policies for highly skilled immigrants also determine their tax and welfare entitlements. Despite their economic contributions, foreign talents can still be subject to taxation and welfare criticisms (Magni, 2024). For example, they may be accused of exacerbating economic inequalities, particularly if low-skilled natives and industries are crowded out in the process of economic transformation. These criticisms are compounded by the proliferation of preferential policies such as fast tracks to citizenship, generous tax breaks, direct access to welfare benefits, and housing subsidies (e.g., The Scheme for Housing for Foreign Talents of Singapore in 1997). Such preferential treatments may engender public concerns about procedural inequity and provoke perceptions of favoritism. Therefore, my third hypothesis is:

H3

Natives prefer highly skilled immigrants admitted under restrictive welfare policies to those under liberal or preferential policies.

In summary, policy feedback theory provides a compelling rationale for anticipating attitudinal changes caused by immigration policies. While other theories may suggest similar hypotheses, a crucial distinction exists. Specifically, by providing a common basis for intergroup comparisons, immigration policies do not necessarily imply stark subgroup differences in attitudinal responses. This contrasts with traditional models centered on natives’ characteristics (Hainmueller & Hopkins, 2015). Below, I organize these models into three families with corresponding auxiliary hypotheses.

Labor Market Competition and Sociotropic Concerns

Economic explanations have long been central to the immigration literature, with one strand focusing on labor market competition and economic self-interest. The most basic models assume that immigrants increase competition for scarce jobs and that natives more likely reject immigrants with similar skills or occupational interests (Hainmueller & Hiscox, 2010). These economic self-interest models have direct implications for the attitudinal feedback effects of labor market regulations, as they not only explain why restrictions might enhance support for immigrants but also imply that this positive effect should vary among natives. Assuming that highly skilled respondents are more likely to compete with foreign talent (Hainmueller & Hiscox, 2010), we may hypothesize that:

H4 a

The positive attitudinal effects of labor market restrictions are concentrated among more-educated natives.

Not all economic explanations prioritize short-term material or tangible interests (Weeden & Kurzban, 2017). Many models posit that natives may reject immigrants on sociotropic economic grounds. That is, natives may fear that immigrants will intensify job competition and wage depression for other native workers. Whether these general economic concerns are essentially driven by individual self-interest is subject to debate (Gerber et al., 2017; Melcher, 2021). What is clear, however, is that they have been a major explanation for attitudes, even in the context of highly skilled immigration (e.g., Tzeng & Tsai, 2020). Thus, we may hypothesize that:

H4b

The positive effects of labor market restrictions are concentrated among natives who perceive highly skilled immigrants as posing an economic threat to Hong Kong.

Cultural and Identity Threats

A common objection to economic explanations is that attitudes toward immigrants are more a matter of emotion than materialistic calculation. In this view, hostilities are driven by psychological and cognitive factors, such as prejudices (Brewer, 1999), ethnocentrism, implicit beliefs (Lee & Chou, 2020), and, perhaps most fundamentally, social identities. According to social identity theory (Tajfel & Turner, 1986), people are motivated by their self-esteem needs to maintain a positive image of the groups to which they belong. When the status of their ingroup is threatened or insecure, they may develop negative attitudes toward outgroups. This body of work complements the hypothesis that cultural and nationality requirements enhance natives’ support for immigrants (H2) and suggests that attitudinal policy feedback effects vary significantly between high and low identifiers. In the context of Hong Kong, where a strong local “Hong Konger” identity has emerged, we should expect:

H5

The positive attitudinal effects of cultural and nationality requirements are more potent among natives who see themselves exclusively as “Hong Kongers.”

Welfare-Related Concerns

The final cluster of theories delves into welfare-related concerns. Focusing on low-skilled immigration, many accounts take welfare competition and tax burden as their premise (e.g., Dustmann & Preston, 2007). Others ground “welfare chauvinism” in group-based politics, in which immigrants are generally considered less deserving of social welfare than natives (e.g., Magni, 2024). With regard to highly skilled immigration, the literature on natives’ preferences for equality may be more pertinent since highly skilled immigrants are less likely to use welfare benefits. Highly skilled immigration can deepen inequality in the host country by altering its skill composition, relative factor supplies, and return to capital, which may disproportionately benefit high-wage natives (Lin & Weiss, 2019). It is reasonable to assume that preferential treatments for highly skilled immigrants may escalate public apprehensions about inequality and that their negative attitudinal effects are conditioned by natives’ tolerance for inequality:

H6

The negative effects of preferential treatments are concentrated among natives who have a stronger preference for egalitarianism.

This theoretical discussion is not intended to be exhaustive. Yet, drawing on insights from the policy feedback and immigration attitude literature, it explains why immigration policies may exert significant attitudinal feedback effects. It also articulates the rationale behind immigration policy decision-making and derives hypotheses to distinguish between major theoretical models. To the extent that immigration restrictions are developed to address theoretically distinct concerns among natives, reliable tests must be devised to extricate their effects and examine whether they vary across natives with different socioeconomic characteristics.

Methods

Design of the Conjoint Experiment

This study used an original paired conjoint experiment embedded in a two-wave online survey. The experiment directed respondents to consider the policies under which highly skilled immigrants are admitted to work and live in Hong Kong. Following a short introduction that asked respondents to evaluate different policy options to attract highly skilled immigrants to Hong Kong, a conjoint table presented two hypothetical immigration schemes, as displayed in the format shown in Fig. 1.

Fig. 1
figure 1

Experimental design. Note: Key information is emphasized in bold. For a screenshot of the original experiment in Chinese, see Fig. A4

Full size image

Compared with designs that feature only one profile per screen, this “paired profiles conjoint” design has a higher level of external validity because it induces more engagement among respondents (Hainmueller et al., 2014). Following suggestions by Bansak et al. (2018), I asked each respondent to evaluate five pairs of schemes, each displayed on a new screen. Below each pairing, a panel presented questions to measure respondents’ attitudes toward immigrants admitted through different policies. To minimize satisficing, the respondents were not allowed to proceed to the next pair of schemes until each had been visible on the screen for at least 30 s (Hainmueller & Hopkins, 2015).

Each immigration scheme includes six types of policies: labor market test (“application eligibility”), work rights (“change of employment”), country of origin (“nationality”), language requirements, welfare entitlement, and tax responsibilities (“tax”). By varying their levels of restrictiveness, I can directly test hypotheses derived from the aforementioned theoretical perspectives. Finally, policy order was randomized across respondents to address potential primacy and recency effects.

Each policy had multiple levels of restrictiveness, which were randomly assigned to the immigration schemes. The contents of the policies replicated those currently implemented in Hong Kong. I varied their restrictiveness using Cerna’s (2016) index of immigration openness. This method ensured that the policy profiles were realistic and theoretically grounded. For example, Hong Kong’s General Employment Policy (GEP) includes a labor market test to limit employers’ ability to hire immigrants. This is regarded as a more restrictive policy in Cerna’s index (2016). In a liberal regime, immigrants only need to meet certain pre-determined skill requirements. Therefore, I created an alternative version that omitted the labor market test. Similarly, the Quality Migrant Admission Scheme (QMAS) requires applicants to be proficient in Putonghua, Cantonese, or English. To create a more restrictive version, I limited the options to Cantonese or English, the two languages preferred by natives (Lai, 2011). Finally, I randomly assigned quotas (10000, 25000, or 40000 per year) to each immigration scheme to address the potential effects of outgroup size on natives’ evaluations. To avoid unrealistic profiles, the largest quota (40,000) never co-occurred with the “mainland Chinese only” nationality requirement. Table 1 contains the complete list of policies. Ultimately, 865 unique profiles were reviewed by the respondents. Appendix II.3 further discusses considerations regarding construct and external validity.

Table 1 Immigration policies featured in the experiment
Full size table

The core outcome variable is the respondents’ attitudes toward immigrants admitted under a given scheme. I asked the respondents to choose between prospective immigrants applying to one of the two schemes. Additionally, respondents were prompted to separately rate their support for each immigrant using a scale from 0 to 10 for the respective immigration scheme presented.

Sample

To ensure that questions measuring general attitudes toward immigrants and threat perceptions would not influence respondents’ choices in the experiment, I conducted the online survey in two waves (Hainmueller & Hopkins, 2015). The first wave gathered data on respondents’ socio-demographic characteristics, threat perceptions, and general immigration attitudes. After a 3-week washout period, the same respondents were invited to take the second survey, in which they evaluated immigrants admitted by different policy proposals (the experiment) and then responded to questions about egalitarianism.

Working with a professional online survey agency in Hong Kong that maintains a sizeable and diverse online opt-in panel, I recruited 5463 Hong Kong natives aged 18 or above to participate in the first survey (December 2020), of whom 1432 completed the second survey (January 2021). Descriptive statistics for the sample are presented in the Appendix. Compared with the broader population, the respondents were more educated and younger. However, this should not cause strong bias since the target group affected by highly skilled immigration is typically in the labor market and reasonably educated.

Empirical Strategy

To test for the existence and direction of attitudinal feedback effects (H1, H2, and H3), I examined the average marginal component effects (AMCEs) of immigration policy liberalization on attitudes (Hainmueller et al., 2014). Specifically, a series of linear regression models were estimated in which the unit of analysis is an immigration scheme (a combination of different policies). The dependent variable is an indicator of whether the respondents chose the immigrant admitted by the scheme. The independent variables include the indicator for each policy item, excluding the baseline category.Footnote 1 The AMCE thus describes the average change in the probability of immigrants winning support when they are admitted under a liberal rather than restrictive (the baseline) policy, with standard errors clustered by respondents. As the contents of the policies were randomly assigned, variations in AMCEs for a given policy can be convincingly attributed to differences in policy restrictiveness.

To test the remaining hypotheses, I examined whether the attitudinal effects of selected immigration policies varied across subgroups with different characteristics in two steps. First, I conducted nested model comparisons via a series of F tests to evaluate whether specifying the interactions between a subgroup identifier and policy items of interest would improve the fit of a regression model that only uses those policy items as predictors of support (Leeper et al., 2020). Second, I compared pairwise differences in conditional marginal means (CMMs), which represent the probabilities for respondents in different subgroups to select immigrants who were admitted through a particular policy, averaging across all other policies.

The subgroups were defined based on the theoretical discussion in the previous section. I assessed the level of labor market competition by the education level of the respondents (with or without a college degree). Perceived threat to the Hong Kong economy was measured by asking the respondents to express whether they agreed with the following statements: (1) “highly skilled immigrants occupy the resources of Hong Kong, such as jobs, housing, and healthcare services, and therefore are a burden to Hong Kong” and (2) “highly skilled immigrants reduce average wages” (7 = strongly agree; Spearman–Brown coefficient = 0.75). Responses to these questions were summed, with values above the median coded as “high threat.”

For the hypothesis concerning cultural and identity threats, I asked the respondents to choose an identity that best described them: Hong Konger, Hong Konger in China, Hong Kong Chinese, and Chinese. Egalitarianism was measured by two statements (5 = strongly agree; Spearman–Brown coefficient = 0.75): (1) “Large income disparities are unjust because people are equal in principle” and (2) “The government should take strong measures to reduce income disparities” (Houtman, 2001). Summed values below the median were coded as highly egalitarian. Table A3 displays the descriptive statistics related to the main variables.

Results

Figure 2 shows the AMCE for each policy feature. Notably, all of the immigration policies significantly influenced how the respondents evaluated highly skilled immigrants. Liberal policies nearly always reduced support for highly skilled immigrants. For example, consistent with H1, removal of the labor market test requirement reduced the probability of being chosen by ten percentage points. Preferential treatments and relaxation in language requirements had similar effects (H3). Although highly skilled immigrants are unlikely to rely on social welfare, immediate entitlement to such benefits penalized support by 27 percentage points. Immigrants were more welcome if admitted through schemes listing Cantonese as an alternative language (H2). These findings consistently demonstrate the adverse attitudinal effects of policy liberalization.

Fig. 2
figure 2

Effects of policy liberalization on attitudes toward highly skilled immigrants. Note: This plot shows the change in the probability of an immigrant being preferred for admission to Hong Kong under a liberal policy, compared to the most restrictive policy scenario (the reference category), averaged across all possible policy combinations. The coefficients are AMCE estimates with clustered standard errors. The horizontal bars denote the 95% confidence intervals. Numerical results and marginal means (MMs) are detailed in Appendix II.6

Full size image

The only exception to this pattern may be nationality requirements. Compared with immigration schemes that only accept mainland Chinese citizens, those with no nationality restrictions raised the immigrant’s probability of being supported by 33 percentage points. This effect is consistent with H2, which reflects the anti-mainland Chinese sentiments in Hong Kong, but it may also reveal a more general disapproval caused by preferential treatments. Unlike other policies, the restrictiveness of nationality requirements was contingent upon the immigrants’ original nationalities. For mainland Chinese citizens, immigration schemes that are exclusive to them ease their admission to Hong Kong. Given the close relationship between mainland China and Hong Kong, natives may see the requirement as preferential treatment to mainland Chinese. I will explore this explanation in the next section.

Table 2 summarizes the results of the nested model comparisons. The F statistics describe the extent to which the patterns of attitudinal policy feedback effects differed by the subgroups of interest. The hypotheses derived from the three major theoretical perspectives (H4a, H4b, H5, and H6) suggest that immigration policies should have heterogeneous impacts on natives with different socioeconomic characteristics. Strikingly, the resultant F tests yielded little supportive evidence. Specifically, the regression models that included the interactions between labor market restrictions (e.g., labor market test) and the relevant subgroup identifiers (i.e., college education and perceived economic threat) did not exhibit a better fit than those that only used the restrictions as predictors. This was also true for policies concerning welfare entitlement and tax responsibility. The only exceptions were language and nationality requirements, which interacted with social identities as hypothesized in H5 [F (5, 13721) = 8.05, p < 0.001].

Table 2 Formal test of effect heterogeneity
Full size table

To detect effect heterogeneity not covered by the original hypotheses, I conducted pairwise comparisons of CMMs for all policies (Fig. A5). The attitudinal policy feedback effects remained remarkably similar across diverse groups of respondents, with only a few exceptions. For example, schemes excluding mainland Chinese and mandating a year’s residency for welfare access attracted greater support for immigrants from college-educated individuals. Respondents who perceived a more severe sociotropic threat from highly skilled immigrants exhibited weaker support when an immigration scheme recognized Putonghua as an accepted language. Additionally, consistent with H6, inequity-averse respondents were less favorable than non-averse respondents toward granting immediate welfare entitlement to highly skilled immigrants.Footnote 2 Overall, however, the interactions between policies and respondents’ characteristics were scattered, inconsistent in direction, and modest in size, which implies that the attitudinal feedback effects of policies cannot be simply accounted for by traditional explanations grounded in individual differences.

The only exception was the interaction between nationality requirements and social identities. As shown in Fig. 3 respondents of different identities exhibited significantly different attitudes toward all of the nationality requirements. Specifically, immigration schemes that excluded citizens of mainland China elicited strongly positive attitudes among respondents who identified as “Hong Kongers” while producing a negative, though non-significant, response among other respondents. This heterogeneous treatment effect confirms H5, suggesting that social identities play a critical role in attitudinal policy feedback regarding nationality and cultural requirements.

Fig. 3
figure 3

Attitudinal effects of policy liberalization by social identity. Note: The left plot reports selected CMMs, and the right shows their estimated differences. The horizontal bars are 95% confidence intervals

Full size image

I conducted a battery of robustness checks (Appendix III). A potential concern is whether the experiment simply reflects people’s policy preferences (i.e., people who prefer restrictive policies prefer immigrants admitted through restrictive policies). However, as Fig. 4 shows, preexisting immigration attitudes measured in Wave 1 did not change the pattern of policy feedback effects reported here [F (32, 12466) = 1.27, p = 0.14]. Liberal policies reduced support for immigrants, even among those who preferred liberal policies.Footnote 3 My findings are also robust to alternative measures of attitudes (rating instead of forced-choice), economic self-interest (individual and household incomes, self-reported social class), social identity (ratings instead of categories, ethnocentricism), and welfare attitudes (generalized trust). Figure A15 displays results based on a probability sample (N = 204), which are substantively similar to the findings here. Fear about COVID-19, perceived percentage of immigrants in the respondent’s industry, and survey fatigue did not condition the effects of policies on attitudes. Finally, to address concerns about attrition, I examined the standardized mean differences of the key variables between respondents who completed both waves of the survey and those who did not. The respondents were similar in all characteristics. Weighting the data does not meaningfully affect the main results.

Fig. 4
figure 4

Effects of policy liberalization by pre-existing immigration policy preferences. Note: CMMs based on Wave 1 responses to “Do you think that the number of highly skilled immigrants in Hong Kong should increase, remain unchanged, or decrease?” Responses (on a scale from 1 to 5) were recoded as 1 = negative (decrease); 2 = neutral (unchanged); 3 = positive (increase). The horizontal bars are 95% confidence intervals. See Fig. A15 for additional checks

Full size image

Qualitative Evidence of Policy Feedback

The experimental findings highlight the importance of attitudinal policy feedback beyond traditional accounts of immigration attitudes. To understand how highly skilled immigration policies shape attitudes, I interviewed 54 participants of the survey experiment through convenience sampling. Tables A9 and A10 summarize their characteristics. Many worked in professional industries such as aviation, healthcare, engineering, and accounting. The interviews were semi-structured and began with a short online questionnaire. Similar to the experiment above, the questionnaire required the respondents to compare two immigration schemes in a pair-profile conjoint table (Table A8), including one restrictive and one liberal scheme. However, unlike in the experiment, I reminded a randomly chosen half of the participants that liberal policies might be more attractive to foreign talent. Despite this change, all 54 participants selected immigrants admitted through the more restrictive scheme. Key interview questions concerned why respondents preferred one immigrant over another and their impression of immigrants admitted through different policies. Appendix IV details the design and implementation of the study.

The interviews confirmed the validity of the experiment. The idea that “policy matters” was aptly illustrated in the metaphor utilized by participant 40, who likened immigrants admitted through liberal policies to scholars moving to low-ranking universities:

There will be a difference. This is because although they are the same person, the screening process is different. It is like you are from X university. You will have a different feeling about yourself if you move to another [lower-ranked] university, such as Y and Z. You will also give a different impression to others.Footnote 4

Furthermore, the participants’ preferences remained the same even when they were explicitly told that liberal policies might more effectively attract foreign talent. These findings add to the construct validity of the survey experiment.

How do immigration policies shape attitudes? Policy feedback theory articulates two pathways: resource and interpretative effects. I found limited evidence for the former. Regardless of their degrees of exposure to foreign competition, the participants invariably favored immigrants admitted under restrictive labor market rules. Many believed that immigrants who pass a labor market test and cannot change jobs would better serve Hong Kong’s economic interest. As one participant stated, “If the immigrants keep changing jobs, they will affect the development of other industries.”Footnote 5 This was a typical sociotropic economic concern but was formulated based on policy contents.

Regarding interpretative effects, the interviews suggested that immigration policies trigger at least two types of moral evaluation. First, liberal policies that grant immigrants greater flexibility in the labor market and easier access to local welfare invoked the moral judgment that immigrants would see Hong Kong as a “springboard” in their career.Footnote 6 In contrast, long residency requirements for welfare access and labor market restrictions elicited perceptions of immigrants’ loyalty. One participant remarked, “It’s admirable when they [the immigrants] overcome these barriers and settle here.” Policy requirements thus serve as cues for natives to gauge immigrants’ commitment to the host society.

The second type of moral evaluation centered on fairness and was most often triggered by nationality and job-changing requirements. As expected, the participants saw immigration schemes exclusive to mainland Chinese citizens as preferential and consequently unfair.Footnote 7 Such perceptions lessened their support for immigrants entering under these schemes. Conversely, those admitted through an open system, including mainland Chinese, were seen as “legitimate.”Footnote 8 The participants were also sensitive to granting immigrants labor market mobility and accused beneficiaries of cheating. One participant commented, “The ease with which immigrants can switch jobs suggests they are improperly exploiting the system.”Footnote 9 These remarks underscore the importance of fairness considerations, consistent with Levy and Wright’s (2020) discovery that perceptions of fairness are pivotal in shaping American opinions on immigration issues.

In addition to moralization, immigration policies wielded a quality assurance effect. Participants inferred immigrant quality from restrictions beyond skill requirements, especially those pertaining to welfare entitlement. As one participant stated, “I don’t think they are highly skilled immigrants if they are attracted by welfare.”Footnote 10 Although restrictions on welfare entitlement are unrelated to skills, they gave a (false) sense of assurance to the participants about the quality of admitted immigrants. In a similar vein, immigrants admitted under policies open to all foreigners and requiring a labor market test were seen as being more competitive and economically important.Footnote 11

The moralizing and quality assurance effects described above echo a core tenet of moral foundations theory in social psychology: that moral judgments and affective valence (e.g., good–bad) are triggered by situational factors and are non-deliberative (Graham et al., 2013). This supposition contrasts sharply with the widely held assumption in the immigration literature that natives form attitudes that align with their predisposed values and interest (e.g., sociotropic concerns and egalitarianism). As my findings demonstrate, whether those predispositions are activated and how they shape attitudes depend—at least partially—on the policy context. This may explain why hypotheses derived from the three major theoretical perspectives failed to account for the lack of subgroup differences in relation to specific policies.

Conclusion

Using an original conjoint experiment and in-depth interviews, this paper establishes for the first time the attitudinal feedback effects of immigration policies for highly skilled workers. The finding is that restrictive policies enhanced natives’ support for immigrants. Most strikingly, these effects were similar across respondents with different degrees of exposure to labor market competition with immigrants, predisposed sociotropic economic concerns, and preferences for equality. Follow-up interviews confirmed the experiment’s construct validity and suggested two novel causal mechanisms: moralizing and quality assurance effects. The former triggers moral judgments regarding loyalty and fairness, whereas the latter shapes the perceived quality of immigrants.

The finding that policies have similar effects across subgroups should not be taken lightly. Since Hainmueller and Hopkins’ (2015) discovery of a “hidden consensus” on immigration, there have been growing, but still scattered, efforts to explain shared immigration preferences between otherwise sharply divided groups. For instance, in their effort to challenge the attribute-based explanations championed by Hainmueller and Hopkins (2015), Wright et al. (2016) show that policy issues may motivate potent moralistic convictions among natives that render the attributes of immigrants irrelevant. Similarly, Margalit and Solodoch (2022) illustrate how a sense of moral obligation to accommodate current residents in the country has prompted Americans with vastly different socioeconomic characteristics to lend more support to the stock of immigrants than to immigrant flows. These findings, along those of this study, undercut explanations of attitudes grounded in natives’ differences or immigrants’ attributes. Those factors are important, but whether they are catalyzed to shape attitudes depends on a complex set of institutional factors.

Closing the knowledge gap on the attitudinal feedback effects of immigration policies also problematizes some taken-for-granted phenomena in the literature. The preference for highly skilled over low-skill immigrants is one such case. With a few exceptions (e.g., Newman & Malhotra, 2019), this preference has come to be seen as a social fact that sociotropic concerns can readily explain. The findings of this paper offer a cautionary note: highly skilled and low-skill immigrants differ not only in their skill levels but also in the policies that admit and govern them. Highly skilled immigration policies tend to be restrictive due to the influence of professional organizations and the cost of skill acquisition. In contrast, policies for low-skill immigration may appear liberal because of the state’s obligations to human rights norms such as family unification or the protection of vulnerable groups (Lin & Weiss, 2019). This distinction is vital if policies indeed shape attitudes toward immigrants. Should highly skilled immigrants be admitted by overly liberal policies or granted preferential treatments, they too may become targets of hostility. This was the case in Singapore in the late 2000s, when native Singaporeans protested against their government’s “favoritism toward foreign talent” (Gomes, 2014).

Three caveats are in order. First, this paper does not suggest that restrictive policies always enhance support for immigrants. Context-specific political processes and the stringency of preexisting immigration policies may bias the direction of policy feedback in unknown directions. While this justifies the use of experiments, future research should theorize and test how people in different political settings process information about immigration policies. Second, while the resource effects proposed by policy feedback theory receive little support from this study, such effects may become more salient as policy dynamics evolve. The long-term effects of highly skilled immigration policies are beyond the scope of this paper. Still, future research may identify industries of strategic importance and examine how reforms that ease the admission of skilled foreign workers activate natives’ self-interest in the long run. Third, although my findings have implications beyond highly skilled immigration, other types of immigration may evoke alternative moral reasonings. Instead of quality and loyalty concerns, natives may focus on humanitarianism with respect to refugee or low-skilled migration. This research must not be construed as justifying discriminatory practices against vulnerable groups.

Caveats aside, the results of this paper nonetheless underscore the role of immigration policies in attitude formation. Although policy changes in the real world are typically tied to complex political processes, understanding the attitudinal effects of different immigration policies remains critical to formulate appropriate framing strategies. My findings will become even more important as technology smooths the way for talents’ global mobility. Increasingly, governments and corporations are being compelled to experiment with new policy ideas to compete for skilled workers. Future policy changes may well take place before mass political movements. In situations of social and political uncertainty, the empirical findings of this paper can provide policymakers with valuable insights for the design of immigration schemes for highly skilled foreign workers.