Sharing citizenship: economic competition, cultural threat, and immigration preferences in the rentier state

This paper proposes a framework of immigrant acceptance that accounts for both group-level and individual-level characteristics and conducts a novel test of the cultural threat hypothesis. Immigrants ’ individual traits are conceptualized as secondary to their identity-based claims. The empirical strategy leverages a set of survey experiments conducted in the extreme rentier state of Qatar, where naturalization poses tangible negative financial consequences for citizens by expanding the pool of government welfare beneficiaries. Findings demonstrate that citizens are willing to share citizenship with a narrow ethnic in-group while individual cultural and economic attributes are lower-order determinants influencing economically vulnerable citizens. Importantly, answers to direct survey measures are at odds with these findings, demonstrating their susceptibility to social desirability bias.


Introduction
Why do citizens reject immigrants? Decades of scholarship has debated the relative importance of economic versus non-economic factors in explaining anti-immigrant attitudes (see Hainmueller and Hopkins, 2014b). Despite compelling theory suggesting that bias may be rooted in fear over economic competition, public opinion evidence from a variety of settings has not favored this hypothesis (e.g., Citrin et al., 1990;Sniderman et al., 2004;Hainmueller and Hopkins, 2014a). Instead, survey data have pointed to the social-psychological challenge posed by newcomers whose ethnic or cultural identities differ from those of the native population, as the most potent and consistent driver of anti-immigrant sentiment (e.g., Citrin et al., 1990;Schildkraut, 2011;Wright and Bloemraad, 2012).
Recent work has sought to explain the puzzling lack of empirical support for the economic threat hypothesis. Malhotra et al. (2013) propose that the effects detected in national surveys reflect the greater prevalence, rather than impact, of cultural concerns over immigration, because not all citizens face labor market competition from immigrants. Other research has argued that negative views of immigrants may reflect categorical rejection of immigration as a policy, or of certain legal categories of immigrants (e.g., undocumented immigrants), rather than preferences over immigrants' personal characteristics (Wright et al., 2016;Soroka et al., 2017). The two lines of argument thus imply a methodological as well as a theoretical answer. The first is that typical survey samples are likely to privilege cultural threat models of anti-immigrant sentiment, while the second observes that opposition to immigration may form logically prior to judgments about which economically-or culturally-relevant attributes make an individual immigrant acceptable.
Based on these insights, this paper suggests a new model of immigration preferences that demonstrates how economic and identity-based considerations work together at both the individual and the group-level. Extending the approach of Malhotra et al. (2013) and leveraging a sudden shift in immigration policy due to an exogenous shock, we investigate immigrant acceptance in an extreme welfare state-the Arab Gulf state of Qatar-where all citizens face a clear zero-sum financial cost from newly-eligible permanent residents and kinship-based tribes are important political actors. We build on the intuition of Wright et al. (2016) and Soroka et al. (2017) to consider a range of criteria, beyond legal status, that may serve as the basis for categorical rejection (or acceptance) of immigrants. Citizens may deem membership in a favored ascriptive category a necessary and sufficient qualification irrespective of an immigrant's other traits. We interrogate this counter-notion of "categorical acceptance" by examining citizens' prioritization of competing justifications for permanent residency in Qatar, including those based on economic skills/contribution, coethnicity ( jus sanguinis), and birthright ( jus soli).
A considerable body of literature has been devoted to examining the relationship between immigration attitudes and the welfare state in European and North American contexts (Senik et al., 2009;Burgoon et al., 2012;Macdonald, 2021). However, the problem of immigrant rejection occurs in many societies around the world, including developing and nondemocratic settings, which present challenges for researchers. Often theoretical models used in Western contexts are not directly applicable, the identities in question are politically sensitive, and data collection presents its own difficulties. As a consequence, researchers may talk past each other for want of a common framework. This research adapts immigration theory to rentier states, demonstrating that extreme cases should not be mistaken for theoretical anomalies, given a sufficiently flexible framework. In doing so, we suggest a way forward for researchers in under-researched contexts who can test and refine the proposed model using similar experimental analysis.
Similar to many other contexts, capturing naturalization preferences in Qatar presents challenges associated with social desirability bias. Previous research suggests that traditional survey questions about identity politics in Qatar are likely to be impacted by the respondent's desire to appear progressive and avoid disclosing the importance of tribal politics, particularly to nonnative survey enumerators . Using rare nationally-representative survey data from Qatar, we employ an innovative combination of analyses. The first is an original choice experiment designed by the authors to assess citizens' relative acceptance of different claims for permanent residency.
A second experiment replicates the conjoint design of Hainmueller et al. (2014aHainmueller et al. ( , 2014bHainmueller et al. ( , 2015, estimating the effects of individual-level economic and cultural attributes simultaneously. A final stage in the analysis examines determinants of overall support for the new permanent residency law in relation to the BIDR impression management scale. 1 Experimental results reveal a strong preference for individuals who possess a descent-based rather than a merit-based claim for permanent residency. Qataris are more likely to reject permanent residents who pose a greater economic threat, but this effect operates only for candidates who lack a descent connection to Qatar, and only among lower-income citizens. Meanwhile, answers to direct survey question reveal substantial social desirability bias, underscoring the importance using experimental methods to mitigate bias.

Understanding the cultural threat hypothesis
Political economists have long theorized that some form of material self-interest underpins antiimmigrant sentiment. Immigrants may threaten natives by competing for jobs, depressing wages, or dissipating welfare benefits (Harwood, 1986;Gang and Rivera-Batiz, 1994;Hanson and Spilimbergo, 1999;Mayda, 2006;Scheve and Slaughter, 2006). Yet public opinion evidence from a variety of settings-mainly the United States, Canada, and Europe-has favored a competing explanation of bias against immigrants rooted in political psychology: rejection of individuals seen as culturally or ethnically incompatible (Citrin et al., 1990;Schildkraut, 2011;Wright and Bloemraad, 2012;Hainmueller and Hangartner, 2013). Reviewing studies published over the previous 15 years, Malhotra et al. (2013, 394) report that all find some support for the cultural threat hypothesis, while results consistent with the economic threat hypothesis are less common and often disputed.
Scholars have explored the cultural threat hypothesis from several angles. Some contend that opposition to immigrants is rooted directly in prejudicial attitudes: racism (Schildkraut, 2011), xenophobia (Wright, 2011), or ethnocentrism (Citrin andSides, 2008). Others working in the tradition of social identity theory understand anti-outsider bias as a consequence of the social and cultural distance that results from dissimilarity in attributes such as language (Hopkins, 2015), religion (Adida et al., 2010), skin color (Fouka et al., 2021), nationality, and class (Shayo, 2009). An influential literature in cognitive psychology and economics assumes that in-group members rank cultural and ethnic out-groups according to their perceived social distance, and seek to avoid psychological costs of interaction with out-group members that increase with distance (Tajfel et al., 1979;Turner et al., 1987;Akerlof, 1997). This self-classification produces in-group favoritism that has been observed both under minimal laboratory conditions (Brewer, 1979;Huddy, 2001) and in real-world settings ranging from the United States (Sigelman and Sigelman, 1982), to Africa (Adida, 2015;Carlson, 2015), and the Middle East (Shockley and Gengler, 2020).
Another interpretation of the cultural threat hypothesis is that bias reflects sociotropic concerns about maintaining national values and identity. This perspective distinguishes more clearly between ascriptive and non-ascriptive out-group distance by observing that, although ethnic and racial differences between immigrants and the native population increase the cost of social integration for newcomers (Dustmann and Preston, 2007), bias against immigrants from ethnically distant groups can be reduced if they signal conformity to national norms (Hainmueller and Hopkins, 2014b;Hopkins, 2015), for instance by learning the local language (Schildkraut, 2005) or demonstrating valued work ethic (Hainmueller and Hopkins, 2014b). In this view, opinions about immigrants are primarily driven by their individual-level characteristics that signal the degree of conformity to the host country's cultural norms.

Locating the economic threat from immigration
One critique of existing tests of the economic threat hypothesis is that they fail to capture how ordinary people actually think and form opinions about immigration. This recent strand of literature has sought to move beyond the economic/cultural debate to show that negative views of immigrants may stem from principled, categorical rejection of immigration generally, or of particular (legal) types of immigrants. Wright et al. (2016) use a conjoint experiment to show that a substantial proportion of Americans reject illegal immigrants out of ontological concerns over justice and respect for the rule of law, irrespective of immigrants' identity-and economicrelated attributes. Similarly, Soroka et al. (2017) conduct a survey experiment with citizens in both the United States and Canada and find that willingness to extend welfare assistance to an individual is impacted more strongly by the candidate's legal status than ethnic category. Such results imply that opposition to immigration may logically precede attribute-based judgments about what makes individual immigrants acceptable. By this view, many extant survey-based studies focus on the wrong behavioral process.
Still, despite their theoretical divergence, the attribute-based and categorical frameworks of immigration attitudes agree in concluding that orientations are not primarily, if at all, driven by rational economic calculations about the private financial impact of immigration. Malhotra et al. (2013) suggest a straightforward reason for the relative lack of support for economic threat models: not all citizens face an economic threat from immigrants. When the authors focus on a vulnerable group-US high-tech workers-they do observe more negative attitudes toward immigrants who represent job competitors. Although they find that cultural concerns also remain strong determinants of attitudes among members of this sector, the key implication of their study is that the lack of public opinion evidence supporting the economic threat hypothesis may simply reflect a relative lack of prevalence of a labor market or other material threat across an entire population, whereas the emotions and often implicit biases that underlie cultural rejection of foreigners are more widespread. Indeed, recent evidence from another country in the MENA region, Morocco, demonstrates that job competition on a sub-national level can drive economic-related rejection of immigrants (Buehler et al., 2020). However, this research may overestimate the general importance of economic compared to cultural factors due to social desirability bias since it relies on non-experimental questions.
While not mentioned by Malhotra et al., we suggest that lower cognitive burden is a separate way that surveys may privilege cultural threat models of immigration attitudes. Individuals may find it difficult to infer the impact of immigration, or of specific immigrant profiles, on their personal financial situation based on information presented in a survey. Such cognitive burden can lead to satisficing response behavior-providing survey answers that are just "good enough"and, ultimately, measurement error (Krosnick, 1991(Krosnick, , 1999. By contrast, feelings of cultural or physical threat largely tap unconscious emotions and prejudices that are more readily primed via survey questions describing immigrants or immigration scenarios. We employ treatments with specific financial information in order to preclude this possible issue.

Gulf immigration and citizenship
Attitudes toward immigrants in North America and Western Europe have drawn considerable scholarly attention, but comparatively little is known about public opinion on immigration in the Arab Gulf states. This is despite the region having both the world's largest foreign-born populations relative to local citizens, and also the most diverse non-citizen populations (Czaika and De Haas, 2014). These countries-Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates (UAE)-are collectively home to millions of foreign workers who are needed to build and maintain their hydrocarbon-based economies (Winckler, 1997). Although they represent an estimated 88 percent of the resident population in Qatar and constitute a majority or significant plurality in every Gulf country, immigrants are employed under a temporary labor "sponsorship" (kafāla) system that affords no political rights or path to citizenship (Okruhlik, 2011).
This practical impossibility of obtaining citizenship has invited scrutiny of the structural economic and political exclusion of vast segments of Gulf populations (e.g., Dito, 2015;Lori, 2019). But for scholars of the rentier state, it is a predictable consequence of the Gulf's enduring political-economic model, wherein autocratic rulers distribute a sufficient portion of resource wealth to citizens to buy political autonomy (Luciani and Beblawi, 1987;Hertog, 2010). The rentier monarchies share a material interest in minimizing the pool of citizens-cum-welfare recipients, so as not to dissipate the finite rent income that sustains them. By this view, as Longva (2006) famously argues of Kuwait, what might appear as xenophobia or identity politics is actually a case of economic protectionism.
Gulf citizens have much to protect. With some variation according to the wealth of individual states, Gulf citizenship confers an array of direct and indirect financial benefits. The government employs around 85 percent of working citizens in Qatar (Halaoui et al., 2017). Gulf states also subsidize nationals via free health care and education as well as below-market prices for water, electricity, food, and other commodities (Gengler and Lambert, 2016). Gulf nationality is governed by strict descent-based criteria and is only passed patrilineally. Naturalization is subject to annual limits and opaque exercise of royal prerogative, and is extremely rare (Babar, 2014). Representative surveys of Qatari nationals conducted in 2013 and 2014 found that 90 percent of citizens supported extending citizenship to the children of Qatari mothers married to non-Qataris (Mitchell et al., 2013). Although these non-experimental survey questions may present inflated levels of acceptance due to social desirability bias, the public is clearly sympathetic to this group and desires the government to reform nationalization laws (Al-Malki, 2016).

The case of Qatar
Our empirical strategy leverages the extreme rentier state of Qatar, where newly-eligible permanent residents pose direct and obvious negative financial consequences for citizens by expanding the pool of welfare beneficiaries. Enabled by the world's highest per capita resource rents, Qatar transfers a generous portion of state wealth to its citizenry of only around 350,000 individuals. Furthermore, studies have shown that citizens of rentier societies judge their economic conditions not in absolute terms but relative to those of their peers, and that perceptions of unfairness in allocation can be a powerful force for mobilization and opposition (Okruhlik, 1999). Survey evidence from Qatar, notably, has demonstrated that perceptions of relative deprivation generate negative political attitudes irrespective of a citizen's objective level of wealth (Mitchell and Gengler, 2019). Qatar's citizenry is highly homogeneous along the primary descent-based categories of race (Arab), religion (Islam), and confessional school/denomination (Hanbali Sunni). This atypical combination of economic and cultural protectionism permits a unique test of the relative importance of economic and non-economic factors in shaping citizen attitudes toward immigrants and immigration policy.
Previous public opinion work in Qatar has established that citizens possess highly ambivalent views of immigrants, recognizing their necessary function in the economy while also perceiving them to be financial opportunists and a corrupting influence on conservative local culture and national identity (Diop et al., 2017). The vast majority of Qataris report that they would prefer a reduction in the number of foreign residents and a further tightening of the labor sponsorship system (Diop et al., 2017). Notably, survey experiments have revealed that Qataris express more negative attitudes toward immigrants when they are prompted to think about blue-collar laborers originating mainly from South Asian countries, compared to white-collar professionals from the Arab world and the West (Diop et al., 2012(Diop et al., , 2015. In 2017, neighboring Arab states initiated a surprise economic blockade of Qatar over a diplomatic dispute. In place until January 2021, the crisis created a need to retain and attract foreigners able to bolster economic output and external security. A law issued in 2018 made non-Qataris eligible for the first time for permanent residency, a quasi-citizenship status that confers many of the economic entitlements enjoyed by citizens. Permanent residents are granted priority access to government jobs, free health care and education, and the ability to start businesses, own property, and make investments without a local partner (Law no. 10 of 2018 regarding permanent residence, 2018). Thus, in the context of Qatar, the immigration debate centered around permanent residency. Already competitors for scarce resources, permanent residence could now access a larger and longer-term share of Qatar's economic pie, even as the embargo threatened citizens with the loss of jobs, business income, and investment returns. Interestingly, the law does not specifically mention the children of Qatari mothers as being eligible for this new status. However, due to the political relevance of this group in Qatar and throughout the Gulf, it is included in the analysis.

Theorizing immigrant acceptance
Our conceptualization, illustrated in Figure 1, builds on the intuition of Wright et al. (2016) and Soroka et al. (2017) to treat citizen evaluation of immigrants as a process-but one of acceptance rather than rejection. This choice is consistent with theory from social psychology in which individuals form groups in order to differentiate themselves positively from others, and bias in favor of in-group members is observed more readily than out-group prejudice (Brewer, 1979). In other models, citizens screen out candidates, and form general opinions about immigration, on the basis of immigrants' legal status. We argue that ascriptive group-level criteria such as tribal descent may serve the same discriminating function: principled acceptance of immigration may stem from its very association with ethnically or culturally proximate groups that citizens universally prefer.
In our model, citizens judge candidates for immigration on the basis of their claim for admittance into society-coethnic descent ( jus sanguinis), birthright ( jus soli), economic contribution, or security contribution-with claims inhering different degrees of social distance. We hypothesize that Qataris, and likely citizens elsewhere, favor in-group members and give priority to ascriptive over non-ascriptive claims. Those judged to possess the requisite descent-based qualifications are accepted categorically. To be categorically accepted, is to be accepted because of ones group membership, without reference to ones individual characteristics. Thus, membership in the favored ascriptive category (children of Qatari mothers) may be deemed a necessary and sufficient qualification irrespective of an immigrant's other, economically-relevant traits, with affinity for in-group members overriding financial disincentives for acceptance.
Meanwhile, immigrants lacking blood relation with citizens are evaluated based on their individual-level characteristics, including cultural similarity and their economically-relevant traits. For example, those from other ethnic groups who possess skills desirable to the state will be scrutinized more carefully on the basis of their individual attributes. In the case of Qatar, we predict that while cultural criteria are more widely used by most of the population to determine the acceptability of a migrant, a smaller segment of the national population perceives themselves as economically vulnerable to competition from foreign workers and will only accept migrants whose individual traits signal that they are not an economic threat.
The remainder of this paper tests a number of predictions following from the conceptual model summarized in Figure 1. A first set of analyses, based on an original choice experiment, examines the propositions (Hypothesis 1a) that citizens in Qatar accept or reject categories of permanent residency candidates based on their claim to naturalization (Hypothesis 1b), that citizens prioritize the descent-based claims of in-group members, and that (Hypothesis 1c) the likelihood of acceptance decreases when descent-based claims are absent. A second, conjoint experiment explores the hypotheses (Hypothesis 2a) that out-group candidates for immigration are judged individually rather than categorically, on the basis of their economically-and culturally-relevant traits, and that (Hypothesis 2b) the latter factors are more influential in determining citizen acceptance (Hypothesis 2c) except among the most economically vulnerable citizens, who will not accept competitors.

Direct measures
Data for our study come from a rare representative face-to-face survey of Qatari households fielded in May 2018 by the Social and Economic Survey Research Institute (SESRI) at Qatar University. A total of 733 citizens completed the survey, with a response rate of 52.3 percent and maximum sampling error of ±4.4 percent. Our key findings come from two embedded experiments-one choice task and one conjoint task-and related non-experimental questions. Respondents were introduced to the topic of immigration using language from state press releases describing different categories of potential beneficiaries of the new quasi-citizenship status: the children of Qatari mothers married to non-Qataris, immigrants with special professional skills, and immigrants performing unique service to the country (i.e., military service).
Respondents were first asked to indicate their overall level of support for the new legislation on a scale from 0 to 10, where 0 signified "it is a very bad idea" and 10 meant "it is a very good idea." They were next asked to rate the importance of several factors that could influence their opinion about who should be granted permanent residency, including a candidate being Muslim, being Arab, speaking Arabic, having lived a long time in Qatar, having a blood relation to Qatari families and tribes, contributing to economic development, and contributing to Qatar's security and stability (see Appendix notes).
Previous studies suggest that they are likely to be affected by social desirability bias for several reasons. First, respondents may overstate general support for the permanent residency law, or understate the importance of ascriptive compared to economic factors (Shockley and Gengler, 2020), so as not to offend enumerators, all of whom are non-Qataris whose nationality status is apparent from outward cues such as accent and dress (Benstead, 2013;Carlson, 2014;Gengler et al., 2019). Respondents may also downplay the importance of ascriptive factors so as not to appear biased in favor of certain ethnic categories in general (Malhotra et al., 2013). Thus, using such answers to direct survey questions to make inferences about Qatari attitudes and decision-making is potentially misleading.

Choice task
To generate a more accurate ranking of citizen preferences for different categories of immigrants, respondents completed a randomized choice task in which they selected between two competing sets-or baskets-of non-citizen groups. Each basket contained two groups, and the task was completed three times. The task was self-administered on a touch-screen laptop computer such that the survey enumerator observed neither the experimental characteristics nor the respondent's choice. This design element was important to reduce the likelihood of social desirability bias. The choice task included five categories of candidates for permanent residency, corresponding to different claims to the new status: (1) children of Qatari mothers; (2) those born in Qatar; (3) those belonging to a tribe not native to Qatar; (4) those having important professional skills; and (5) those who give military service.
In the analysis to follow, we use each of the five previously-listed groups of immigrants as predictors of the probability that a set is selected. A positive (negative) change in the probability of selection when a category is included in a basket indicates that respondents give it greater (lesser) priority for permanent residency. We report Average Marginal Component Effects for each treatment (Hainmueller et al., 2014) which yields a measure of each item's importance relative to the other items included in the experiment, thus establishing a reliable preference ordering.

Conjoint task
The choice experiment was designed to elicit a reliable ranking of immigrant types based on their claim for permanent residency; yet previous studies have shown that specific immigrant traits also play a role in determining acceptance. We expect individual-level attributes to be important in contexts where the blood connection to Qatari nationals is ambiguous or missing. Following Hainmueller and Hopkins (2014a) and Wright et al. (2016), participants completed a conjoint task in which they rated six different profiles and selected between three pairs of hypothetical candidates for permanent residency. The experiment included only immigrant profiles lacking a descent-based claim to permanent residency, and therefore likely to elicit individual-rather than categorical-based assessment. Thus, it does not include a treatment for the children of Qatar mothers. Based on prior research on the Qatari case and the findings of the choice task, it seems that their group membership entitles them to acceptance irrespective of their personal traits, whereas those from other potentially privileged groups, such as those with high skills, would likely be evaluated according to their individual cultural, linguistics, experiential, and economic characteristics. The conjoint task included treatments for a candidate's religion, nationality, language, salary, occupation, and duration of residence in Qatar.
Most treatments are familiar and adapted from Hainmueller and Hopkins (2014a), yet two merit special introduction. First, because religious and confessional (sectarian) identity is a sensitive issue in Qatar, first names were used as proxies for candidate religion, cuing for Christian, Sunni Muslim, and Shi'i Muslim identities; otherwise, a pure control conveyed no information about the respondent's religious identity (Shockley and Gengler, 2020). By omitting the first name for some (random) profiles, we test the relative acceptability of an immigrant whose religious identity is unknown, compared to those whose names provide a clear cue about religious affiliation and thus social proximity (Shockley and Gengler, 2020). Three different first names were used in each identity treatment group to avoid repetition of names in a way that might reveal the intent of the treatment and undermine the experiment. 2 Candidate nationalities were selected from among the most common white-collar expatriate nationalities in Qatari in order to have a variety of socio-cultural distances from nationals. Yemenis, for instance, would be the closest, while British nationals would be the furthest. The personal pocketbook implications of particular immigrant characteristics may not be readily discernible to an individual either in conjoint survey experiments or in real life. It may be difficult, for example, for a respondent to calculate how extending benefits to a doctor rather than a teacher, or a Palestinian versus a Yemeni, might impact her own economic prospects. For this reason, an explicit treatment specified the salary of the candidate to offer the clearest possible information about the salary of the potential resident.
A full summary of treatments is presented in Appendix Table 1. Average Marginal Component Effects (AMCEs) reported in the following section give the impact of the treatment variables relative to their respective baselines, modeling the change in probability of selection at the level of the immigrant profile. AMCEs provide an estimate of which permanent resident characteristics have a greater impact on acceptance compared to the other factors. Because citizens' economic threat perceptions may depend on their own financial situation, we also compute a second set of AMCEs conditional on a respondent's monthly household income, coded to fall into categories above and below the middle salary treatment of 40,000 Qatari riyals (approximately $11,000 USD) per month. Finally, following Wright et al. (2016), respondents were allowed to reject both candidates rather than being forced to choose one, as in some experiments. But respondents could not select both permanent residents. This design choice was made to discourage respondents from satisficing, or accepting all the candidates without paying attention to details. Figure 2 presents the results of the choice task designed to generate a reliable and unbiased ranking of different claims to permanent residency. Overall, findings demonstrate the importance of categorical, in contrast with individual-level, judgment about the acceptability of potential immigrants, in line with Hypothesis 1a. They also provide strong support for Hypothesis 1b, that citizens will prioritize ascriptive-based over merit-or utility-based claims for permanent immigration to Qatar; as well as Hypothesis 1c, that the likelihood of categorical acceptance declines with perceived social distance.

Categorical acceptance and ranking of immigrant types
More specifically, inclusion of blood-related candidates (children of Qatari mothers) in a basket of groups increases its probability of selection by more than 30 percent, whereas immigrants explicitly described as lacking a descent-based connection to Qatar (those from non-Qatari tribes) decreases a basket's likelihood of selection by 27 percent. A qualitative gap separates blood-relatives from the second-most favored category of permanent resident among citizens: individuals born in Qatar. This group is associated with a 5 percent increase in the probability of selection. Meanwhile, categories capturing immigrants with special skills and service elicit slight negative preferences, each being associated with around a 5 percent decrease in the likelihood that a basket of candidates is selected. 3 As the most socially proximate group, the children of Qatari mothers serve as a clear baseline against which the proximity of other groups can be judged by respondents.
We interpret the comparatively very weak preferences for groups whose ethnic and cultural attributes are unknown-i.e., those born in Qatar, those with special skills, and those who provide special service to the state-as additional evidence that citizens base their acceptance of immigrants principally on ascriptive criteria. Lacking such information, citizens cannot form a clear opinion about the suitability of a given immigrant type. Thus, the experiment shows that it is identity concerns that evoke categorical judgments about who deserves to share the privileges of (quasi-)citizenship in the rentier state, not economic ones. Qatari citizens do not fail to accept candidates for permanent residency because they view them as unable to make significant contributions to the country's economy or security (or alternatively as a drain on government resources and jobs); rather, citizens do not accept them because their ethnic and cultural identities are not sufficiently similar to their own. Stated another way, special skills or the ability to make a contribution to society is not a sufficient claim to gain access to permanent residency absent ascriptive qualifications.

Individual acceptance and rating of immigrant profiles
The choice experiment results show that citizens in Qatar tend to accept permanent residency candidates who are known coethnics and reject known non-coethnics, while withholding firm judgment on candidate types whose ascriptive attributes are ambiguous. The conjoint analysis complements these findings by zooming in on the latter group of candidates, to examine which individual-level characteristics might induce acceptance of permanent residents who lack a descent-based qualification for immigration (see Figure 3).
The left-most column, labeled "Unconditional," reports AMCEs for the treatments across the entire sample. These results show that identity factors retain some significance in shaping candidate acceptance at the individual level. Immigrants who lack a first name are penalized relative to those whose first name signals an in-group (Sunni) confessional religious affiliation. We infer that candidates whose first name was not given were less likely to be selected due to uncertainty surrounding their cultural identity, as observed in the choice experiment. Indeed, even profiles with first names that imply out-group (Shi'i or Christian) identities were preferred to those assigned no first name. It could be that profiles lacking this information were assumed to have non-proximate identities withheld due to undesirability. Yet, rather than suggest that one religious affiliation is necessarily preferred to the others, the results instead evidence a primary desire among citizens to know the name-and thus verify the social proximity-of the immigrant.
Findings for the other culturally-relevant treatments are also consistent with the idea that, once an immigrant's descent-based qualifications have been established, other traits only impact the chances of acceptance on the margins. Qataris make no significant distinction, for instance, between candidates of a diverse set of nationalities relative to the Yemeni baseline, which is the most proximate identity category.
With regard to language, respondents give clear preference to monolingual Arabic speakers, relative to both monolingual English speakers and immigrants who speak Arabic and English. The estimated coefficient is almost indistinguishable for the latter two categories, suggesting that English is not valued for its economic utility, but rather serves as an out-group marker applying even to fellow Arabs. Citizens prefer to extend the benefits of permanent residency to persons who are linguistically similar to themselves, with any English proficiency taken as an indicator of cultural distance. As in the case of nationality, however, in substantive terms the negative effect associated with English language ability is small.
The remaining attributes included in immigrant profiles, which did not provide ethnic or cultural data, failed to generate statistically significant treatment effects. First, the length of time a non-citizen has spent in Qatar does not increase, or at all affect, acceptance as a permanent resident. As discussed, many immigrants are born and live their entire lives in Qatar and other Arab Gulf rentier states with no pathway to permanent residency or citizenship. In light of the slight positive prioritization of individuals born in Qatar witnessed in the choice experiment, we interpret the null conjoint finding as reflecting the fact that some respondents in the former task likely conflated the "born in Qatar" category with preferred groups, especially children of Qatari mothers, who are also native-born.
Two suggestive findings correspond to the economically-relevant attributes of potential immigrants. Citizens appear to disfavor the highest-salaried category of candidates, who are perhaps viewed as taking more than their fair share of the country's wealth. Meanwhile, respondents slightly favor candidates in the military profession, who might contribute to the country's stability and security at a time when Qatar faced a traumatic regional embargo.
A final set of results investigate the possibility of differential treatment effects according to the respondent's degree of economic vulnerability, in line with the proposition of Malhotra et al. (2013). The middle and right-hand columns of Figure 3 present AMCEs conditional on a citizen's household income, divided above and below a 40,000 Qatari riyal per month cut-off that is slightly below the medium household income among citizens, and mirrors the middle salary treatment category in the experiment. As expected, Qataris in the more economically vulnerable subgroup are less likely to select profiles in which the immigrant would earn an above-average salary that also exceeds their own household income. Citizens in the higher-income group are not influenced by a candidate's salary, and this between-group difference based on respondent income level is highly statistically significant (p = 0.006). Yet the conditional results also reveal the clear influence of identity-related factors among the same lower-income citizens, including a strong aversion to immigrants whose name signals a Christian or ambiguous religious identity, and permanent residents of the only Western nationality (British) included in the experiment. 4 Meanwhile, it is the financially-advantaged subgroup that drives the overall negative treatment effect of English language ability and the slight positive preference for security contributors in the form of military professionals.
On balance, then, the conjoint findings provide but relatively weak support for our Hypothesis 2a, that acceptance of out-group candidates for permanent residency is judged at the individual rather than categorical level. In fact, we observe marginal impacts of individual-level attributes of candidates who lack a descent-based qualification, with no treatment estimated to alter the probability of selection by more than 5 percentage points in the overall sample. Even in the subgroup analysis, the largest estimated coefficient does not exceed 0.10. The results lend somewhat more support to Hypothesis 2b, that the ascriptive attributes of out-group members more strongly shape citizen acceptance than their economically-relevant traits. The hypothesis is largely upheld in the overall sample, where two of four factors included to capture social distance (religion, nationality, language, and length of residency) were seen to influence citizens' choices, compared to neither of the two economically-related treatments (salary and occupation), based on the standard level of statistical confidence. Finally, we see evidence in support of Hypothesis 2c, inasmuch as lower-income individuals are less likely to accept economic competitors in the form of high-salary immigrants.

Immigration-related preferences and social desirability bias
To illustrate the value of our methodological approach, here we analyze responses to nonexperimental questions included in the survey, which directly asked respondents to rate the importance they attached to the different immigrant attributes considered in the experiments, as well as their overall level of support for Qatar's new permanent residency law. Figure 4 presents the relative difference measures described above, which indicate the importance attached to a factor compared to all others. Overall, it is difficult to distinguish between the factors, as confidence intervals are large and overlapping, indicating that respondents did not give dramatically different responses for any of the attributes. This illustrates the utility of our experimental design, which forced a choice between competing categories. Still, having a blood relationship to Qatar was clearly rated as least important relative to the other factors, in direct contradiction with the results of the choice experiment. Likewise, an immigrant's contribution to security was rated as somewhat more important than the other factors, while this category of permanent resident was disfavored in our experiment. Results give the impression that Qataris care less about broad identity groups and more about the societal contribution of a prospective immigrant, suggesting that responses to these traditional, straightforward survey items are substantially impacted by social desirability bias.
The same conclusion obtains in the case of reported support for the permanent residency law itself. Mean support across the entire sample is 8.0 on a 0-10 scale, suggesting a surprisingly high degree of public enthusiasm for a measure whose clear and inevitable effect is to widen eligibility for state welfare and thus dissipate the benefits presently enjoyed by citizens. To assess the extent to which susceptibility to socially desirable reporting helps explain nominal support for the controversial legislation, we use an Arabic-language implementation of the BIDR Impression Management subscale, a well-established measure of susceptibility to socially desirable responding (Paulhus, 1984). We use ordered logistic regression to predict the reported level of support as a function of impression management. 5 4 The income-based disparity in selection of British nationals is highly robust (p = 0.006). 5 Support for the law is coded into terciles, labeled "high," "medium," and "low.." The middle category is excluded from Figure 5 as it does not vary with impression management and likely captures satisficing. Figure 5, reported support for the permanent residency policy is determined to a very large degree by personal susceptibility to social desirability bias. The estimated probability of expressing high support for the law increases dramatically with a respondent's engagement in impression management, measured by the BIDR scale in standard deviation terms. At a "high" value on the impression management scale (+2 standard deviations), for instance, the likelihood of expressing high support for the law is 2.3 times that of voicing low support, whereas the reverse is true among those not prone to engage in socially desirable reporting. Among this latter group at the "low" end of the BIDR scale (-2 standard deviations), the predicted probability of high support for the permanent residency policy is 29 percent, compared to a 46 percent likelihood of indicating low support. In short, after accounting for the effects of social desirability bias, one arrives at the opposite conclusion about the overall level of public support in Qatar for the permanent residency law, from the one suggested by answers to the direct survey question.

As depicted in
How then can we accurately gauge citizen support for the policy of offering permanent residency to immigrants, the vast majority of whom lack a blood connection to Qatar? We reason that the BIDR analysis presented in Figure 5 provides a more valid estimate of overall support. More precisely, the predicted probability of high support for the permanent residency law is 40.7 percent in the absence of impression management, or social desirability bias (i.e., when the BIDR scale equals 0). This percentage also closely approximates the proportion of immigrant profiles (40 percent) that were selected in the conjoint task. 6 Together, the two corresponding estimates paint a very different picture from that given by the non-experimental measure, suggesting public support at approximately one-half the level implied by the answer to the direct survey question (8 out of 10, or 80 percent).

Discussion
In light of these findings, how can we understand immigration attitudes? Integrating existing frameworks, we proposed that attitudes toward immigrants are best understood as a set preferences, involving both categorical judgments about candidate groups (Wright et al., 2016) and trait-based judgments about individual candidates (e.g., Hainmueller and Hopkins, 2014a). The model expands existing conceptualization of the factors that can motivate categorical judgments by demonstrating the relevance of ascriptive criteria (in our case, kinship). The framework also integrates economic and cultural considerations at the immigrant level, but subordinates such calculations to group-based preferences across competing immigration justifications for admittance into society.
Qatar presents a most likely case for detecting national-level evidence of perceived economic threat from permanent immigration because it combines extreme rentier wealth with an atypically homogeneous citizenry, alongside the largest proportion of foreigners in the world. Nonetheless, findings provided little evidence that Qatari attitudes are primarily shaped by immigrants' economically-relevant attributes. Instead, Qataris prioritize descent-based claims to permanent residency, categorically accepting individuals with a blood connection to the country. Candidates who lack a blood connection to the country cannot easily earn acceptance by living in the country an extended period of time or even by being born there, and it matters very little whether they are nationals of a Western, South Asian, or Arab country.
We believe these findings can be attributed to how respondents use information. Identity-group information is the clearest and strong signal of belonging, whereas nationality Recall that respondents were permitted to reject both candidates. The 40 percent estimate is likely somewhat deflated because the experiment excluded candidates with descent-based claims to permanent residency. and length of stay do not provide as clear signals. Expatriates can reside in the Gulf all their life, live within their own ethnic communities, and have little knowledge of or interaction with citizens and their culture. British nationals may be ethnically Arab and those holding a Yemeni passport may have a local mother and a Yemeni father (Al-Malki, 2016). Respondents are aware that nationality categories are fluid, thus do not rely on this information to determine with whom to share citizenship. Methodologically, this highlights the advantage to having both the choice experiment and the conjoint experiment working together.
This more nuanced picture of public opinion calls into question some assumptions of rentier state theory, which holds that state-society relations in resource-exporting states revolve around the distribution of oil rents. We push back against existing interpretations of Gulf public reluctance to expand the citizenship franchise, which view cultural rejection as epiphenomenal to the true issue of economic protectionism (Longva, 2006;Al-Nakib, 2014) and align ourselves with emerging research that takes seriously the role of ascriptive identity in modern Gulf politics (Freer and Al-Sharekh, 2021).
While our study presents data from one crucial case, the framework both helps make sense of past findings from Western settings and is flexible enough to be adapted and tested in a wide variety of new contexts. The results reiterate the dual-usefulness of survey experiments as a tool for studying complex evaluations and as a corrective for widespread social desirability bias influencing sensitive opinions and choices. The vast majority of experimental work on immigration preferences has been conducted in developed Western democracies. We believe that researchers in non-Western contexts can benefit from applying our theoretical framework and experimental methodology to their studies of immigration. As future research continues to explore other social and political contexts, conceptualization of attitudes toward immigrants and immigration will no doubt be further refined and enriched.