Is there a “Disconnect” between Public Opinion and U.S. Immigrant Admissions Policy?
A large body of research suggests that immigration policy-making in liberal democracies overlooks most citizens’ preferences most of the time. To support this view, scholars often point to an apparent “disconnect” between the expansionary immigration policies prevailing in most of the West and the heavily exclusionary bent of public opinion. This paper argues that the “disconnect” thesis oversimplifies ordinary citizens’ preferences over immigrant admissions policies in ways that inflate the divergence of public policy from public opinion. It demonstrates that the U.S. public’s abstract preference for less immigration in general coexists with strong majority acceptance of the specific admissions policies that generate most immigration. This seeming inconsistency arises in part because concrete questions about admissions policies evoke stronger humanitarian and economic considerations than the standard, more abstract, gauge of immigration policy preferences does. Citizens by and large do not support rolling back the number of immigrants admitted through family reunification, provisions for refugees, and skills-based visas even when they are made aware that these three admissions categories combined account for nearly all foreigners admitted permanently into the country.
Responsiveness – how closely public policies track public opinion – is a critical gauge of democratic accountability. A high degree of responsiveness indicates that citizens are guiding policy, and this is both a normative benchmark and an empirical expectation in standard models of representative democracy (Downs 1957; Dahl 1971; Schattschneider 1960). Political scientists disagree about how often and in what domains and contexts procedurally democratic states enact popular policies and refrain from enacting unpopular ones (Barabas 2007; Wilson 1980; Achen 1977; Gilens 2005; Shapiro 2011). But there is virtual consensus that the principle of accountability is breached when status quo policy survives in the face of durable public opposition or when it changes in ways that run counter to mass preferences.
A large body of survey data suggest immigration unambiguously exemplifies a policy domain in which responsiveness is low. Scholars often point to a “disconnect” (Schuck 2007) between mass opinion and immigration policy in nearly every liberal democracy in the world, with governments admitting many more immigrants than the majority of citizens would prefer (Freeman 1995; Joppke 1998; Tichenor 2002; Citrin & Sides 2008). This “expansionary bias” (Freeman 1995, p. 882; see also Brubaker 1995) relative to the median voter’s preference for more restrictive policy is viewed as evidence for weak popular control over policy. Despite most citizens’ apparent preference for rolling back status quo policies that admit large and growing volumes of immigrants each year, these policies persist or evolve in ways that further increase immigration.
This apparent disconnect between opinion and policy is central to claims that immigration policy-making in liberal democracies is largely oligarchic. The power of pro-immigration elites and clienteles is seen as thwarting the desires of ordinary citizens. Cosmopolitan elites who decry populist anti-immigrant appeals as xenophobic or racist make grass-roots mobilization of overtly anti-immigration opinion unappealing to most mainstream politicians, and so public opposition to the status quo remains largely unorganized.1 Business sectors such as agriculture and high technology that employ immigrant labor and co-ethnic lobbies that support immigrants from a shared ancestral background pressure policy-makers to sustain the unpopular status quo (Freeman 1995; Joppke 1998; Schuck 2007).2 Organized minorities rule while diffuse majority opposition goes unheeded (cf. Wilson, Dilulio, and Bose 2012).
By pushing past the customary reliance on a single survey question most often invoked to support the “disconnect” thesis we argue that alleged unresponsiveness of immigration policy to public opinion is less convincingly established than often assumed. The evidence usually marshaled in support of this view oversimplifies the nature of public preferences on immigration policy and overstates public opposition to the policy status quo. The disconnect thesis implicitly holds that an abstract belief that there should be less immigration is tantamount to a preference for rolling back the specific and well-entrenched rules governing the admission of most legal immigrants. A simple survey experiment we conducted demonstrates that this assumption is unwarranted. We find that a generalized and abstract preference for less immigration in fact coexists with majority support for keeping or expanding the number of immigrants covered by the main admissions preference categories recognized in U.S. immigration policy. This remains true even when respondents are informed that the admissions categories they are considering – family reunification, skilled workers, and refugees – together constitute virtually all legal immigrant admissions.
The “disconnect” we observe instead is psychological and might be termed “internal” in that many individuals prefer limiting immigration in the abstract but nevertheless resist altering the admissions policies that sustain the expansionary status quo. We tie this apparent inconsistency to differences in the considerations evoked by each: the standard admissions categories listed above engage humanitarian or economic considerations (Zaller 1992) to a significantly greater degree than does the more abstractly-worded standard item that asks about the level of immigration in general. Different facets of immigration policy also evoke different sets of considerations from one another and from those that undergird general and abstract reactions to immigration. The upshot is that immigration policy attitudes are complex and multifaceted, and thus worthy of attention beyond the well-worn debates over the relative importance of cultural, economic, and ideological predispositions purported to broadly explain “opposition” to or “support” for immigration (e.g. Hainmueller and Hopkins 2014; Newman et al. 2014). Studies that reduce public preferences over immigration policy to a single dimension ranging from hostility toward immigrants or immigration to “pro-immigrant attitudes” can overlook important aspects of the foundations of Americans’ policy preferences and, as we demonstrate here, can generate misleading understandings of opinion-policy divergence and democratic responsiveness in this domain.
A more nuanced assessment of Americans’ immigration policy attitudes should re-open broader questions about whether the immigration policy status quo is solely a byproduct of clientelism, in James Q. Wilson’s terms, with concentrated benefits trumping diffused costs or the outcome of the dominance of elite values in the policy debate (cf. Freeman 1995). Clearly, pro-immigration elites and organized clients deserve a central place in any account of immigration policy-making in liberal democracies (Freeman 1995; Schuck 2007, 2008; Joppke 1998; Hollifield 2006; Andreas 2000; Tichenor 2002). But the results reported here show that these forces cannot simply be said to override mass preferences, as is often alleged.3 Whatever ordinary citizens’ generalized worry about or even antipathy toward current levels of immigration, when concrete features of admissions policy are queried there is broad public acceptance of retaining the main pillars of the prevailing immigration regime.
More broadly, this study of immigration attitudes illustrates the complexity of measuring public support for changing existing policy in any domain and the difficulty of assessing the extent of democratic accountability and responsiveness (cf. Gilens and Page 2014). Failure to solicit preferences about the actual details of policies and their implementation risks generating incomplete and potentially misleading conclusions about the degree to which policy and opinion are aligned. In the present case, existing research suggests that immigration policy is a relatively clear or “easy” case for demonstrating the presence of a “disconnect.” Thus reassessing this conclusion in the domain of immigrant admissions should warn us that a similar misinterpretation of whether public policies override public opinion may prevail in other domains too.
DOES THE PUBLIC OPPOSE THE LEGAL ADMISSIONS POLICY STATUS QUO
On the surface, the United States, a self-styled “nation of immigrants,” exhibits the same gap between permissive policy and exclusionary opinion as do publics in most other liberal democracies (Freeman 1995). The U.S. government grants more than one million foreign nationals a year the permanent right to live permanently in the country. More than two-thirds are relatives of U.S. citizens or permanent residents, and the rest are split between refugees, skilled workers (Batalova & Lee 2012) and a relatively small number of “diversity” visas (Schuck 2003). And policy actually has become ever more permissive in the fifty years since the Hart-Celler Act established the contours of the current regime. The 1990 Immigration Act, for example, increased annual visa provisions by several hundred thousand even though its sponsors’ initial intent was to prevent immigration levels from expanding (Tichenor 2002; Zolberg 2006).
This expansionist direction of policy seemingly overrides majority preferences. A half-century of polling indicates that far more Americans want to limit rather than to increase legal immigration. One recent review finds “continuing negativity and ambivalence” in public attitudes toward immigrants, with episodic spikes in anti-immigrant sentiment punctuating more moderate periods (Muste 2013). Peter Schuck remarks that Americans “do not oppose immigration in principle, in general, or unalterably, but they do want less of it (or at least no higher)” (2008: 351). Scholars continue to debate the role of moral norms and perceived cultural or economic threats to the individual and the nation in fostering durable mass opposition to immigration (for a recent review, see Hainmueller and Hopkins 2014), but the fact of opposition to increasing immigration seldom is disputed.
Inferences about the level of support or opposition to current immigration policy, and hence the validity of the “disconnect” thesis, rest on strong assumptions about the meanings of the responses to standard poll questions. The most commonly cited indicator of Americans’ immigration policy attitudes, first asked by Gallup just prior to the Hart-Cellar Act’s passage in 1965, asks whether the current level of immigration (probably unknown to most respondents) should be increased, decreased, or left the same and on its face seems to solicit opinions about whether and how to change status quo admissions policies.4 Indeed, poll questions that have response options “should be increased, decreased, or kept about the same” have even been described as “always ask[ing] about policy preferences” (Barabas 2007, p. 12).
We are not the first to criticize this particular question. For one, Schildkraut (2013) notes that most versions fail to differentiate legal from illegal immigration, an important caveat in principle given Americans’ routine conflation of the two (Ramakrishnan, Esterling, and Neblo n.d.). Yet specific references to “legal” immigration do not much change the patterns of response (see Figure 1, below).
Others – most recently Newman et al. (2014) – have pointed out that much rides on interpreting the middle “remain about the same” response. If this answer indicates indifference or poor understanding of the question, then respondents with genuine preferences tilt heavily toward restriction, consistent with the disconnect thesis. But the middle – leave things as they are – response could also indicate acceptance of either the prevailing stocks or flows of immigrants. Acceptance of prevailing stocks would still be consistent with the disconnect thesis since large-scale influxes continue under current policies. In fact, acceptance of current flows would not necessarily undermine the thesis, even if it would imply more receptivity to immigration than is usually assumed, because current policies permit flows to increase each year. Thus the disconnect thesis is still sustainable under these interpretations. However, if the middle category mostly represents a more general acceptance of “what is happening” in the realm of immigration, then the distribution of opinion would not be consistent with a sharp disconnect between opinion and policy since, taken together, those favoring an increase or sanctioning the status quo often – though not always – outnumber those advocating a decrease. Without knowing more about how often middle category responses reflect each of these distinct reactions to immigration, one cannot say how important a critique of the disconnect thesis these caveats entail.
We focus instead on what we see as two more fundamental problems with interpreting responses to the “levels” question as furnishing evidence for an opinion-policy disconnect over legal immigrant admissions. First, this item may be a better indicator of measuring generalized anxiety about immigration (cf. Brader, Valentino, and Suhay 2008) than of a concrete preference over whether the prevailing level of immigration should be reduced. Second, even if responses can be interpreted as indicating a true majority or plurality preference for lowering the level of immigration generally, it does not follow that most Americans would support rolling back specific status quo admissions policies in order to achieve this aim. We consider each of these problems in turn prior to developing an alternate approach that calls the evidence for the widely accepted disconnect thesis into question.
Preference for Reducing Admissions or General Anxiety? Temporal trends in responses to minor variants of the standard “levels” item suggest that it may gauge generalized anxiety people harbor about immigration and not an explicit preference over the actual volume of admissions. Legal admissions have risen steadily and substantially since the 1965 Hart-Celler Act, driven mainly by rises in family reunification, the legalization of nearly three million illegal immigrants under the 1986 Immigration Reform and Control Act (Rytina 2002), and the Immigration Act of 1990 that, among other provisions, created the Diversity Visa Program and increased quotas for skilled workers. Yet, as shown in Figure 1, rapidly rising admissions levels have produced no enduring movement in the percentage of Americans saying immigration should be decreased.5 Sparse data that generally coincides with immigration’s salience in national media during the earlier years notwithstanding, the most recent twenty years of more frequent polling show remarkable stability in the percentage of Americans backing a reduction in immigration.
[Figure 1 About Here]
People may not know the number of immigrants in the United States (Sides & Citrin 2007), but most are aware that immigration is increasing (Hopkins 2010). By implication, if the “levels” questions actually were indicating reactions to the policy status quo, we might have expected a gradual increase in those wanting less immigration (cf. Page & Shapiro 1992). If responses to these questions are taken literally, the absence of a steady increase in the share of Americans stating a preference for less immigration would be illogical unless people’s ideal level of immigration rose markedly over time. Although anti-immigrant sentiment has been tied to deep-seated political predispositions and normative views about national identity that may be difficult to change (e.g. Citrin et al. 1997), it is altogether possible that Americans have become more tolerant of large volumes of immigration over a long period of substantial exposure. Of course if Americans’ “ideal” level of immigration did increase or does continue to increase as immigration increases, it suggests that the disconnect fades over time. Public opinion may be always anxious about demographic change but evolves to accept it. This possibility would at least suggest a reinterpretation of the “disconnect” thesis. Instead of being exacerbated by ongoing influxes, the opinion-policy gap would be constant as people become, gradually, ever more comfortable with ever higher levels of immigration. They may not want more immigrants right now, but when the immigrants inevitably come, they gradually become habituated and do not continue to strongly push for reducing the level of immigration. Along these lines, research by Newman (2013) and Hopkins (2010) indicates that opposition to immigration spikes when there is a sudden influx of newcomers where relatively few immigrants live, but that the impact of change is greatly diminished in contexts where a large number of immigrants have already settled.
While this interpretation is possible, the trend in responses in Figure 1 seems to record the tenor and intensity of political debate and salient events rather than tracking the ongoing influx of immigrants. This episodic rise and fall is consistent with the idea that the “levels” question is a measure of context-driven anxiety (fear) rather than an enduring opposition to expansionist admissions policy (loathing). During a period of uproar about immigrants’ use of public services, from the late 1980s until the late 1990s, millions more Americans advocated lower levels of immigration than during most other periods. But as economic conditions improved and attention to Proposition 187, welfare reform, and immigrants’ use of public benefits waned in the late 1990s, exclusionary responses became much less prevalent even as the number of green cards issued rose steadily each year. The major exception was another spike in anti-immigrant sentiment following the September 11, 2001 terrorist attacks. These patterns do not seem to amount to a reaction to actual immigration volume, which increased steadily over the period, as much as the ebb and flow in the political salience of immigration, a cycle driven by events and political entrepreneurship.
Abstract Preference over Level of Immigration vs. Support for Specific Policy Change A second caveat about resting the case for the disconnect thesis on this survey item is that it does not probe beliefs about the concrete pillars of status quo policy. After all, visas are not handed out at random, yet the levels question is silent about the bases upon which immigrants are admitted. So even if the majority prefers in the abstract that the United States admit fewer immigrants, it is problematic to infer that this means they are willing to alter status quo policies in order to bring this about.
Public opinion research has identified many instances in which support for a general principle coexists with opposition to its implementation in specific cases. For example, people who endorse freedom of speech in the abstract will nonetheless often support censoring members of disfavored groups (Prothro and Grigg 1960; Sullivan, Piereson, and Marcus 1979). The perceived consequences of tolerating unpopular or disliked opinions presumably outweigh the impetus to be consistent with the general principle of free speech. Similarly, Americans who identify as ideological conservatives often are “operational liberals” who prefer an active federal government that taxes the rich and spends liberally on public goods and services (Free and Cantril 1967). Limited government and individual freedom are widely shared ideals, but when it comes to particular policy areas, the American public is willing to countenance the growth of government to promote desired social ends. In a third example, pervasive support for the principle of racial equality exists alongside strong opposition to policies such as affirmative action designed to implement this goal (Schuman et al. 1997). Here, the American public desires an outcome—closing racial divides in life chances-- but still rejects some commonly advanced means of achieving it. Though the particular source of such “principle-implementation” gaps (ibid.) are controversial, common explanations point to motives rooted in prejudice, ideology, or group interest that override general egalitarian values when it comes to policy choices (Sears, Sidanius, & Bobo 1999).
If there is any “disconnect” in the cases described above, it is internal to individuals, separating abstract political principles, values, or even preferred policy goals, from beliefs about the specific policies that have been proposed or implemented to realize them. One would not argue that California’s Proposition 209, which ended affirmative action in the public sector, reflected a disconnect between public opinion and public policy simply because people believe there should be more racial equality in the abstract. Nor would one conclude this if government imposed higher taxes on the rich (contrary to widespread abstract “conservatism”) or that the HUAC hearings were evidence of a policy-opinion disconnect because people believe in the abstract that government needs to respect civil liberties. Obviously these examples speak to cases that differ from one another and from immigration policy in many ways. As we discuss further below, however, the important connection between then is that, in each case, abstract and specific questions about policy generate distinct, if overlapping, sets of considerations and generate quite different response patterns.
By the same token, many Americans may believe in the abstract that there is too much legal immigration and say the level should be reduced but still support the policy status quo if they were asked about its implementation through the actual rules governing visa allocations. The standard levels question in fact has no specific policy content. It mentions no set of actions or rules that would realize the supposed popular goal of reducing immigration. Thus we cannot infer from such a question whether Americans would support changing the current admissions policy regime in order to reduce immigration or whether they would accept prevailing policies despite their abstract support for lower immigration.
Hypotheses Our purpose here is to ascertain whether and in what way abstract preferences for reduced immigration as expressed in responses to the “levels” question are related to dissatisfaction with specific expansionist admissions policies, to wit the number of immigrants admitted under the three largest status quo admissions preference categories: family reunification, high skilled workers, and refugees. We aim to measure the degree and direction of divergence between responses to the standard levels question and preferences vis-à-vis the level of immigration in these three categories.
We argue that the nature of the disjuncture, if any, between abstract predispositions and specific policy preferences depends on the particular considerations each evokes. Guided by Zaller’s (1992) “Receive-Accept-Sample” (R-A-S) model, we assume that people form responses to survey questions by sampling from among the accessible considerations they have absorbed from elite debates and not rejected as inconsistent with ideological predispositions and then average across the salient predispositions pulling in each direction (Zaller 1992; Zaller and Feldman 1992). If specific status quo policies evoke a wider set of positive considerations than the standard levels question, then relying on the latter alone would yield an overly negative assessment of public acceptance current immigration policy.
Indeed, speculating about the kinds of considerations that questions about concrete policy might evoke, there is good reason to anticipate greater support for preserving or augmenting expansionist status quo admissions policies than would be implied by a naïve interpretation of the levels question. Immigration policy ostensibly has humanitarian and economic goals, both of which are reflected in the preference categories U.S. law establishes. Both have a prominent place in elite discussions of immigration (Tichenor 2002), despite the frequent portrayal of culturally threatening images of immigrants in mainstream U.S. media (e.g. Brader, Valentino, and Suhay 2008). And there is evidence that both resonate in public opinion toward immigrants (Newman et al. 2014; Hainmueller and Hopkins 2013).
A crucial question is whether these considerations are more important influences on Americans’ preferences regarding the level of immigration admitted through status quo preference categories than on responses to more general questions about the preferred “level” of immigration in the abstract. It seems sensible to expect that this is the case. Newman et al. (2014), for example, find that priming humanitarian concerns evokes more pro-immigrant responses to the standard “levels” question, suggesting that even though they also find a positive correlation between humanitarian values and pro-immigrant sentiment the question itself does not fully bring to mind the types of considerations that specific references to family unity or fleeing from plight in one’s country of origin might. Hainmueller and Hopkins (2014) and Wright, Levy, & Citrin (2013) both find that Americans prefer immigrants whose characteristics suggest an ability to contribute to the nation’s economic well-being and fiscal balance (Hainmueller & Hopkins 2014). Even if some such sentiments are already tapped by the standard levels question (e.g. Citrin et al. 1997), it is reasonable to expect that they would be even more salient considerations when it comes to public preferences over the appropriate level of high skilled immigration.
We thus hypothesize that fewer Americans will favor reducing immigration when they are asked about the existing policies that allocate visas than when they answer about the level of immigration in the abstract. In other words, we expect that the standard “levels” question overstates the degree of “disconnect” between expansionist immigration policies and public opinion. Applying the R-A-S model also implies that which psychological predispositions will be linked to support for particular aspects of status quo admissions policy will vary. This leads to two additional hypotheses. We anticipate that humanitarian values will be a stronger predictor of support for retaining or expanding family and refugee immigration than of responses to the levels question. And we expect that the salience of the federal deficit, a gauge of concern about fiscal solvency, will be a stronger predictor of retaining or expanding the admission of immigrants with skills U.S. employers say they need than of preferences over the level of immigration generally. Here we predict that the structure of support for immigration will vary as a function of the specificity of the policy question and as a function of which particular admissions policy is queried.6
Our argument is that references to the actual parameters of status quo admissions policies themselves prompt respondents either to (1) “self-prime” with considerations that would have been less accessible when answering the standard levels question or (2) apply frames they might overlook when thinking about their preferred level of immigration in the abstract. In a conventional priming experiment (eg. Iyengar & Kinder 1987), the researcher would artificially make family values or humanitarian concerns accessible to respondents and then ask the standard item. By contrast, in a conventional framing experiment (Chong and Druckman 2007) the researcher would expose respondents to some explicit argument that immigration policy is “a question of family values” or “first and foremost an issue of costs and benefits to the U.S. economy” and then see if this influences their abstract belief about the appropriate level of immigration or their feeling toward immigrants. We expect that substituting specific questions about support for the immigration policy status will work similarly, causing respondents to draw on considerations and to think of the issue in terms they would not when confronted with the standard levels question. But generating this effect requires no priming or framing manipulation beyond simply referring to the specific policy in question.
We underline that in pointing to self-priming or framing as a causal mechanism, we are not simply advancing the well-accepted idea that priming and framing can influence policy attitudes. Instead, our argument is that if simply rephrasing a question to ask about specific policies in a manner that better reflects reality sets such cognitive processes in motion and these processes lead respondents to endorse or at least not oppose the status quo, it suggests that the immigration opinion-policy disconnect is less stark than research based on abstract beliefs about the level of immigration in general concludes.
We accept that overlaying some of the virtually limitless threatening or reassuring primes and frames one might apply to immigration policy questions could push public opinion in one direction or the other when it comes to support for retaining or expanding status quo admissions policies. And of course one cannot be certain which indicator would serve as a better measure of potentially mobilized opposition to immigration under different political scenarios. But this caveat applies to any survey question, including the standard levels question that underpins the “disconnect” thesis. It makes little sense, in other words, to reject our argument on these grounds but still to accept inferences based upon the standard “levels” item about opinion-policy congruence. In any case, the disconnect thesis alleges a sustained divergence of opinion and policy, not merely one that could be brought about by elite mobilization under some circumstances.7
It is also not clear that another standard approach to measurement – averaging multiple questions about immigration policy that seem to tap a common factor or latent construct – is an appropriate way of gauging opinion-policy congruence or of determining how disposed Americans would be to follow the lead of elites seeking a change in the legal admissions policy status quo. Scales composed of multiple items are the best way to measure a latent attitude tapped commonly by the items (e.g. Lord and Novick 1968; Ansolabehere, Rodden, & Stewart 2008). But this technique is only useful if the latent attitude the items have in common is in fact conceptually very similar to the construct one wishes to measure. If different immigration policy questions tap different considerations idiosyncratically, then there is no reason to assume that the latent attitude estimated by averaging across multiple survey items would have much at all to do with an actual preference over whether to roll back particular policy status quos. It might furnish some general gauge of “anti-immigrant sentiment,” but, as we have argued, such abstract and general attitudes may diverge substantially from one’s actual preference over whether to retain a particular policy that bears on the object of the general attitude even in some important way.
To be clear, our argument is not that researchers should replace polling based on the abstract levels question with specific questions about policy or that positive responses to specific questions establish a clear pro-immigrant bent in American public opinion. It is simply that scholars have been too quick to assume that abstract support for reducing immigration as measured by a traditionally used survey item is a sufficient gauge of ordinary citizens’ feelings about the immigration policy status quo and willingness to replace it with a more restrictive regime and to introduce this inference as a building-block in models of immigration policy-making. If people who support reductions in immigration generally are still not willing to alter the status quo policies that generate most immigration, then those assumptions are unwarranted and at a minimum deserve further scrutiny.
Robustness to Information about the Status Quo Even if people respond more favorably to specific policies than to abstract notions, some of this may be traceable merely to ignorance about the primary status quo admissions categories rather than a substantive policy preference. Studies indicate that Americans’ preferences over the level of immigration to be quite resistant to argumentation and information about the size of the immigrant population as well as its composition and characteristics (Citrin and Sides n.d.; see also Citrin and Wright n.d.). Yet given widespread public innumeracy on this and related matters (Wong 2010; Citrin and Sides n.d.; Sides and Citrin 2007), respondents made aware that nearly the entire volume of legal permanent resident immigration (95%) comes from the three sources they are being asked to weigh in on might bring their preferences over family, refugee, and skills-based admissions into line with their preferences for lower admissions levels overall. But if support for increasing or preserving the level of family, skills, and refugee immigration is robust to information about the admissions policy status quo, then we can be confident that the specific policy preferences we measure do not merely stem from a lack of awareness that almost all legal immigration comes from these three preference categories.8 To this end, we examined whether informing a random portion of our sample of the percentage of legal admissions attributable to each of these three visa preference categories would alter preferences over whether and how to change the status quo in each of these categories.
Providing this kind of information sets a high bar for our main thesis because it alerts people to a potential inconsistency between an abstract preference for less immigration and a specific preference not to lower or even to increase the level of immigration in the preference categories that generate the great majority of legal immigration. This invites respondents to support a reduction in one or more categories in order to bring about their abstract preference. The dominance of family reunification (66% of all Green Cards) over employment-based and refugee admissions (14% and 15% respectively) also furnishes respondents with a clear means of bringing abstract and specific preferences into line, should they feel motivated to do so: support a greater reduction in family immigration than you would have absent the information and make no change to your position on the other two categories irrespective of the information. If negative effects of the information on support for family immigration are instead paired with offsetting increases in support for refugee and skilled immigration, then this would point to a desire to achieve greater balance between categories but not to effect a net reduction in the level of legal immigration admitted under status quo policies in order to bring about an abstract preference for reducing immigration overall.
DATA AND METHOD
Our data come from a national Internet survey of 1,597 Americans fielded in December 2013. The survey was administered by SSI, which uses a matching-based methodology to approximate a representative sample of the public. Our sample under-represents blacks (7%) and probably also Latinos (5%; the survey was administered only in English), though since a single multiple-choice question was used to gauge race or ethnicity, it is possible that some Latinos identified as white or other, the first response option, rather than as Latino.9 College graduates (48%) are also heavily over-represented. Reassuringly, however, the distribution of opinion on our baseline immigration level question does not differ a great deal from recent ANES, Gallup, and New York Times / CBS polling data (see Figure 1). The experimental effects we show are also consistent across education categories and among Democrats and Republicans (see the Online Appendix). Ideally we would want to compare baseline distributions and treatment effects across racial and ethnic groups (cf. Masuoka and Junn 2013), but our minority group samples are not large enough to make such inferences. Results are fully robust, however, to the restriction of our sample to whites and to the exclusion of foreign-born respondents from the analysis.
The dependent variable in our analysis is respondents’ preferred level of immigration relative to the status quo. This question was asked four different ways (details of our randomization procedure follow), but consistently re-scaled to run from 0-1 where 1 indicated the greatest opposition to immigration. In the baseline or “control” condition, respondents were asked, “Do you think the number of immigrants from foreign countries who are permitted to come to the U.S. to live should be increased a lot, increased a little, left the same as it is now, decreased a little, or decreased a lot.” This question is very close to that asked in the ANES/GSS about the preferred level of immigration and modified slightly to allow for closer and less awkward standardization across treatment conditions. Variations inserted either the phrase “with skills U.S. employers say are needed” (“Skills”), “with family members living legally in the U.S.” (“Family”), or “fleeing violence or government oppression in their home countries” (“Refugees”) after the reference to the number of immigrants.
Additionally, in all three treatment versions of the dependent variable, the word “legally” was inserted after “come to the U.S.” in order to make it clear that respondents were to consider just legal immigration policy. We did this to help ensure that respondents knew we were asking them about legal admissions rather than policies concerning what to do with the millions of immigrants living without permission in the United States that often raise similar concerns about family unity and economic contribution. Still, given that prior polling suggests that using the word “legal” in and of itself does not substantially alter responses to the levels question (see footnote 4), we did not attempt to experimentally disentangle the impact of this word from the impact of the specific stipulation of preference category. Obviously even if, contrary to prior polling, this word did have some impact this would still suggest that the “disconnect” thesis overstates public opposition to legal status quo policies. In addition, as we show below, the different predispositions the three specific policy questions tap suggest that specifying a particular basis for admission we specify is in fact driving the results.
In addition to age, sex, educational attainment and standard measures of party identification, liberal-conservative ideology, and interest in politics, we measured respondents’ humanitarian values, measured using an additive scale composed of responses to four agree strongly – disagree strongly statements: “One should always find ways to help others less fortunate than oneself,” “A person should always be concerned about the well-being of others,” “It is best not to get too involved in taking care of other people’s needs,” “People tend to pay more attention to the well-being of others than they should.” Responses were coded so that higher values indicated greater humanitarian disposition, summed, and re-scaled to run from zero to 1 (Cronbach’s Alpha = .65). We also asked respondents to rank the importance of the following six problems facing the nation: the economy, the federal deficit, immigration, terrorism, Middle East policy, and health care. We then re-scaled each issue’s rank to run from 0 to 1, where 0 indicated it was given lowest priority and 1 indicated highest. Prejudice was measured using feeling thermometers for whites, blacks, Hispanics, and Asian-Americans (cf. Kinder and Kam 2009). For whites, we subtracted the mean ratings for the three minority groups from whites’ rating, and for minorities we subtracted whites’ rating from the in-group rating. We re-scaled the measure to run from -1 to 1 where 1 indicated the highest level of prejudice.
The several stages of randomization are summarized in Figure 2. First, two-thirds of respondents were randomly assigned to participate in a survey experiment. Of these, 60% received no information about either how many legal immigrants to the U.S. are admitted each year or the basis of these admissions. The other 40% initially viewed a pie chart, each slice of which represented the share of admissions in each of the three major preference categories, with a fourth slice representing the share admitted through “other” channels. The pie chart indicated that two-thirds of all legal admissions in 2012 were of immigrants who had family members living legally in the U.S., that approximately 14% of immigrants were brought in under employment-based preferences, and that the remaining 15% came in as refugees. We opted for a pie chart to illustrate the distribution of admissions to minimize confusion arising from misunderstandings of the meanings of percentages or populations shares. At a minimum, respondents should have been able to infer from the chart that virtually the entire volume of legal immigration comes as a result of those three categories (the “Other” slice was shown to account for only 5% of all permanent resident admissions), with family reunification the dominant basis for admissions.
[Figure 2 about Here]
The respondents not shown the pie chart were then randomly assigned to two different conditions. One third of respondents were asked the baseline or “Control” standard levels question. The remaining two thirds (“No Info - Treatment”) instead were asked the three modified versions of this question (the order in which respondents received these questions was randomized) referring to family, skills, and refugee immigration. Respondents assigned to see the pie chart all answered the questions referring to skills, family, or refugee status. None were asked the standard baseline levels question. Respondents viewing the pie chart would already have been primed to think of legal immigration as originating from these three preference categories, so asking the control question again of those receiving the information would not have yielded any clear interpretation.
To assess baseline knowledge about the admissions categories we were asking about, we assigned a final third of respondents to guess at the share of all legal immigrants admitted to the U.S. each year as a result of having family members in the U.S., possessing skills employers say they need, or because they were escaping persecution in their home countries. These guesses were not constrained to sum to 100%. Instead, a residual “other” category automatically tallied on the screen the residual percentage of immigrants after the three guesses were summed. The purpose of these questions was to gauge respondents’ awareness of the parameters of immigration policy in the U.S. These respondents were not asked the category-focused levels questions since their responses probably would be contaminated by their prior guesses.
Not surprisingly, these guesses are consistent with widespread public ignorance about the basis upon which immigrants are actually admitted legally to the U.S. Most respondents who were asked to guess did surmise that 100% or close to 100% of legal immigrants came in through one of the three admissions categories specified. 73% made guesses about the shares of family, skills, and refugees that added to 100%.10 However, this is most likely a result of our question’s having provided only these three specific categories to allocate admissions shares to, automatically relegating the remainder to “other.” On average, respondents also correctly guessed that the largest group of legal immigrants fell into the family reunification category. But as illustrated in the histogram displayed in Figure 2, 90% of respondents underestimated the share of legal admissions coming from family reunification. Again on average, they guessed that only 35% of admissions (Standard Error = 1%) are from family reunification, 30% (SE = 1%) are from skills-based preferences, and 25% (SE = 1%) are refugees. The wide range of guesses and sharp departure from the true distribution suggests strongly that many or most Americans were guessing more or less at random and have little or no awareness of the true parameters of current legal admissions policy. Whereas this implies that there is considerable latitude for opinion change in response to corrective information, it also requires that we make sure that our results are not just a function of mass ignorance about the nature of current admissions policy. If people’s support for family-based immigration, for example, is decimated once people become aware of the true status quo, and if reduced support for immigration in this category is not offset by increased support for immigration through the other two categories, then our findings would not be a terribly meaningful or robust gauge of opinions about the policy status quo.
Our analysis proceeds in three steps. First, we compare the distribution of responses of those assigned to the baseline levels question to those who did not see the pie chart but were asked the variants of the baseline question referring to family, skills, and refugee immigration. Random assignment permits us to do this via simple difference in means and proportions tests. Second, we use multivariate regression in each experimental condition to assess whether the predictors of immigration policy attitudes vary depending on whether one asks abstract or specific policy questions and which facet of admissions policy one asks about (family, skills, or refugee admissions). Given our hypothesis that public opinion about admissions through specific policy status quo preference categories would be more tightly linked than responses to the standard “levels” question to humanitarian values or concerns about fiscal solvency, the predictive strength in each condition of humanitarianism and the importance the subject accorded the problem of the national deficit are of particular interest. Third, we compare the response distributions to questions about family, skills, and refugee immigration among respondents who were randomly selected to receive the information and those who were not, again using simple difference in means and proportions tests.
Do Americans really oppose the expansionist legal admissions status quo as the “disconnect” thesis suggests? The top panel of Table 2 shows the results from our survey experiment. Consistent with much prior research, a plurality of respondents (45%) who were given no information and asked a question similar to that included in the ANES and GSS said the number of immigrants admitted to the U.S. should be reduced (see Panel A, Column 1: “Control”). These respondents also formed a bare majority (51%) of those who provided a valid response to the question. Twice as many favored decreasing the level of immigration as increasing it (23%). Recoding the five-point variable to run from zero to one, with one indicating a desire to decrease immigration by a lot yields a mean of .62 (SE = .03), again indicating a strong tilt toward an abstract preference for restriction and seeming opposition to current policy.
However, this portrait of public opinion changes dramatically among respondents asked specifically about legal immigration for people with family living in the U.S., skills employers say they need, or refugee status. Only 21% of our respondents favored decreasing the level of family-based immigration, 24% favored decreasing the level of skills-based immigration, and 24% wanted to lower the level of admissions for refugees. These figures are about half the percentage of Americans in the control condition professing to want immigration reduced. In each case, the percentage of Americans who support increasing the level of immigration from each category now substantially exceeded the percentage favoring reductions (by margins of 12 points, 9 points, and 11 points for family, skills, and refugees respectively, compared to the opposite pattern (-22 points) in the control group asked about immigrants in general). The mean scores on these variables are .47 (family), .47 (skills), and .48 (refugees), all significantly different (t-test p<.01) than the mean in the control condition and all indicating that the balance of opinion is toward leaving access for family members, skilled workers and refugees, the bulk of all immigrants, the same or higher.
Further indications of how strongly the specific questions indicate the American public to tilt against exclusionary alterations to the status quo emerge when responses to these questions are combined. Only 40% of subjects who answered all three questions supported reducing immigration in one or more of the three categories (that is, 60% preferred not to reduce immigration in even one of the three categories), compared to the 51% who supported reducing immigration a little or a lot in the abstract question. So even if we set a very low bar for classifying a respondent as preferring to change the status quo in a way that reduces immigration, this is clearly a minority preference. A less conservative but perhaps more sensible way of looking at the data is to average the three responses for each individual and observe the prevalence of subjects whose responses are on balance exclusionary, neutral, and expansionary. By this measure, 46% of the public prefers to increase immigration in these three categories more than decrease it, 24% prefers either no change or perfectly offsetting increases and decreases, and 24% prefers to decrease it on balance. And while there could be some concern that “decrease a lot” responses frequently imply a preference for dramatic reductions, only 27% of the public prefers to reduce even one of these categories “a lot.”
We hypothesized that the specific status quo policy questions about family and refugee immigration would evoke humanitarian values in a way that neither the baseline question nor the specific question about skills-based immigration would. We also hypothesized that asking about the level of skills-based immigration in particular would evoke considerations related to fiscal impacts and that therefore people who viewed the national deficit as an important issue would be more likely than others to support increasing this particular category of immigration but not more likely to support increasing the others.
The results of multivariate OLS analyses in each condition bear out these expectations. In the baseline condition, humanitarian values are not associated with more support for immigration.11 Yet when we ask about family-based and refugee immigration, consistent with our predictions, humanitarianism is a significant and potent predictor of expansionist preferences. And the intuition that there is no relationship between a preference for more skills-based immigration and humanitarianism also is confirmed. More to the point, Chow Tests confirm that the impact of humanitarianism on support for increasing immigration is significantly greater (p<.05) in the family and refugee conditions than in the baseline and skills conditions. Thus we argue that the nature of these status quo policies evokes different sets of considerations that in turn effect changes in the distribution of support relative to the baseline.
Also consistent with our predictions, concern about the federal deficit is associated only with greater support for skills-based immigration and not with preferences over the level of immigration in the abstract or with preferences over family and refugee immigration levels. This suggests that respondents particularly concerned about the nation’s fiscal circumstances will be more supportive of immigration streams well positioned to contribute economically.
More broadly, humanitarian values and concern about fiscal solvency are more potent influences on opinions about specific admissions policies than on abstract beliefs about the overall level of immigration. Consistent with the “Sample” component of the Receive-Accept-Sample model that conceives of survey responses as constructed by sampling from salient considerations that poll questions evoke (Zaller 1992; Zaller and Feldman 1992), questions that are more likely to bring to mind positive considerations, such as those referring to family, skills, and refugee status should be expected to generate a more favorable attitude toward immigration. The particular considerations evoked vary in sensible ways depending on which facet of legal admissions policy one asks about. This variation helps explain why there is a great deal of acceptance of particular expansionary status quo admissions policies despite widespread preference for reducing immigration in the abstract.
Finally, we ask whether these results are merely an artifact of public ignorance. As we have seen, Americans appear to know little about the basis upon which most legal immigrants are admitted. Thus we must ask whether widespread support for preserving or increasing the level of immigration admitted through status quo channels merely reflects ignorance of the status quo and consequent willingness to admit more immigrants in these domains while assuming that reductions could still be made in other admissions preference categories. It is one thing to endorse more immigration in each of the major status quo policies, each of which is intrinsically imbued with a rationale for admitting more or at least no fewer immigrants, and quite another to retain this point of view even knowing that one is in effect going against one’s abstract conviction that there should be fewer immigrants on the whole. Perhaps Americans would bring their attitudes toward specific admissions policies into line with their general preference for lower immigration if they knew that the vast majority of legal immigrant inflows are a result of the three specific admissions categories we asked about.
However, our data show clearly that this is not the case. Informing respondents of the actual distribution of admissions by preference category does alter support for each of the three admissions types. When respondents are told the distribution of legal immigrant admissions in 2012 by category, support for limiting family reunification immigration increases. The percentage advocating reduction (30%) is still significantly (p<.01) lower than in the control condition but is now higher than the percentage supporting increases (23%). However, the differential between the percentage supporting reductions and the percentage supporting increases is still much lower (p<.01) than in the control condition, and 61% of respondents still advocate increasing family-based immigration or leaving it the same. The mean score is .53, hardly indicating that mass opposition to status quo policy is pervasive. Moreover, in the light of information about the annual admissions status quo, it is even more likely here than in the standard levels question that the middle (leave things the same) response indicates endorsement or at least acceptance of expansionary policy. After all, these respondents have just been shown what the status quo policy is and still choose an answer that expresses satisfaction with it.
As further evidence that there is no majority motive to reduce the level of immigration admitted through status quo preference categories overall, we find offsetting rises in support for skills-based and refugee-based immigration among respondents who viewed the pie chart. If the disconnect thesis accurately described the state of public opinion toward the immigration policy status quo, then at least after being confronted with the information that these three categories encompass nearly the entirety of legal immigration we should have seen more respondents shifting to an exclusionary position. Respondents could easily have endorsed a reduction in the level of legal immigration overall by advocating the limitation of family-based admissions and leaving their other two responses unchanged. Instead, the mean response to the three policy questions remains virtually identical (.47; SE = .01) among informed and non-informed respondents. The increase-decrease differentials for skills- and refugee-based immigration rise to 19 and 17 points respectively, and the mean scores decline (less exclusionary) in both cases to .44. Support for the principal categories of legal immigration to the U.S. is not diminished when respondents are made aware that their responses in fact combine to determine whether legal immigration is on the whole increased or decreased.
In short, our informed respondents do, on average seek to redress the balance between family-based admissions and admissions of skilled workers and refugees. But the information about the policy status quo does not induce respondents to endorse reductions in the number of immigrants admitted through these three categories. A handful support greater parity across preference categories, but confronted with a clear opportunity to simply cut the level of immigration in the largest category without compensating for this cut by raising the level of immigration in the other categories, Americans decline to do so.
Taken together, these results strongly corroborate our expectation that the American public’s abstract belief that current policy allows for too much immigration need not translate into a desire to roll back the particulars of status quo legal admissions policy. In fact, there is at a minimum no greater support for scaling back major current admissions policies than for expanding them. These data at least weaken the frequent allegation that that democratic responsiveness is low when it comes to immigration policy. Instead, the type of disconnect our experiment reveals is the combination of concern over the volume of immigration in the abstract and generally supportive attitudes toward liberal legal admissions policies. There is nothing inherently irrational about this juxtaposition of feelings. One might be anxious about ongoing influx of immigrants but still, partly out of countervailing humanitarian and economic considerations, reject changes to status quo policies even if those changes would stem the tide of newcomers.
Decades of research leave little doubt that business lobbies and ethnic advocacy organizations play an important role in promoting liberal immigrant admissions policies. It is widely assumed that these interests overwhelm public preferences for greater limitations on immigration. A “disconnect” therefore emerges between exclusionary public opinion and expansionist immigration policy. We have argued, however, that while the public clearly has important reservations about the level of immigration in the abstract, most nevertheless resist the implication that status quo admissions policies that generate large-scale legal immigration should be rolled back. This suggests that public opinion toward legal immigration policies is more complex, conflicted, and nuanced than the disconnect thesis implies and that much of the public would not change the policy status quo if it could. Instead, the public evinces a high degree of permissiveness when it comes to immigration policy, with large majorities accepting the status quo or even believing it should be extended and few believing strongly enough in reductions to demand making specific status quo policies more restrictive.
This revised portrayal of public opinion is in part attributable to the greater power of status quo admissions policy questions (versus the standard abstractly-worded level of immigration question) to evoke considerations leading to acceptance of large-scale immigration. These considerations vary sensibly across different facets of the current admissions regime. Humanitarian values are salient when people assess family reunification and refugee immigration, while concerns about the federal deficit are a significant predictor only of skills-based immigration preferences. These differences help explain why there may be many more people who wish the level of immigration were lower in the abstract but still resist changes to the admissions status quo that would realize this outcome. They also point to the complexity and multidimensionality of Americans’ immigration policy attitudes. To the extent there is some latent attitude ranging from hostility to support for immigration, researchers have greatly illuminated the causes of variation along this spectrum (Hainmueller and Hopkins 2014). We would argue that an important new frontier in research on public reactions to immigration is to look more closely at the nuances that this single dimension may obscure. For example, recent research points to differences in the nature and foundations of opinion toward legal and illegal immigration policies (Wright, Levy, & Citrin 2013) and in support for different types of policies aimed at regulating each.
Obviously identifying the impact of public opinion on policy is beyond the scope of this paper. But we argue that it is also premature to infer that public opinion plays virtually no role in the process, or that policy was made in spite of a generally hostile public, conclusions that arise naturally from research proclaiming clearer opposition to the policy status quo than we uncover here. Much remains to be learned about how and how effectively support for and opposition to particular immigration policies are mobilized, whose voices among the public elected officials heed, and how the attitudes we have measured here evolve during a sustained campaign focused on this issue. There are intriguing cases in which elite and organized opinion would appear to favor policies that nonetheless repeatedly fail to muster sufficient political support. For example, the number of visas for highly skilled immigrants remains far below what most elites and certainly the high tech lobby would prefer. Similarly, the number of temporary visas for agricultural workers is capped far below the preference of agribusiness, always a key player in the immigration policy-making process (Tichenor 2002). These outcomes seem at odds with clientelistic models, and better understanding of the contours of mass opinion might shed further light on the reasons for these apparent deviations.
But for all this uncertainty, our results suggest that it is not at all obvious that attempts to scale back existing legal immigration policies would easily attract strong majority support. Populist anti-immigrant mobilization is sometimes cast as politically low-hanging fruit that is constrained in large measure by prevailing discursive norms (e.g. Freeman 1995). These norms prevent most politicians from embarking on such a strategy even if the public would be receptive to it. But a practical implication of our research is that expansionary policies may not be as vulnerable to populist backlash as some scholars have worried (e.g. Schuck 2008). This conclusion may well be policy-specific. Certain types of immigration policy (e.g. the Schengen Agreement in the European Union or perhaps an explicitly multicultural admissions category such as “Diversity Visas”) do indeed exist despite palpable mass opposition and are especially vulnerable to public backlash. It is well established that there are pockets of intense and rigid opposition to some expansionary policies (Wright, Levy, & Citrin 2013; Schain et al. 2002). Still, the major pillars of a decidedly expansionist status quo U.S. legal admissions policy are at present accepted by most of the U.S. public.
We have focused on the disconnect thesis as it relates to legal immigration policy. An important avenue for future research is to look more closely at whether a clearer disconnect between public opinion and public policy emerges when Americans are asked about illegal immigration. This question is in some ways trickier to address than the one we have undertaken here because it is not straightforward what American policy toward illegal immigrants actually is. The strict letter of the law diverges markedly from its lax implementation, and yet recent surges in deportation and state-federal enforcement cooperation suggest a more serious stance. Nor is there much of an established orthodoxy over what the public favors: one understanding is that it supports a moderate and mixed approach. Though Americans widely believe in robust border security and favor sanctions on employers who hire illegal immigrants, the great majority of Americans rejects mass deportation, and most are willing to countenance salutary neglect (once illegal immigrants are in the country) or some limited form of amnesty (Citrin, Levy, & Lenz 2013; Wright, Levy, & Citrin 2013). Thus there is some reason to believe that arguing for a disconnect between lax de facto policy toward illegal immigrants and public preferences for rigid enforcement would also oversimplify the state of public opinion.
Ascertaining when majority opinion clashes with status quo policy is critically important to our understanding of how democracy functions, whose preferences are represented, and what role the mass public plays in setting or vetoing policies. A recent ambitious study (Gilens & Page 2014) concluded that, where mass opinion conflicts with the preferences of economic elites, the latter always get their way. In most cases, mass and elite opinions converge, so this conclusion leans heavily on our ability to discern when majority opinion truly opposes the status quo. The disconnect thesis casts immigration policy as just such a case, and the outcome is congenial to the oligarchy thesis. By demonstrating that mass concern about immigration does not straightforwardly translate into opposition to the policy status quo, this study indicates that great caution is required in using survey evidence to arrive at these sorts of inferences about the frequency with which elite opinion overrides the preferences of the majority. From a normative perspective, the immigration policy-making process may not be as undemocratic, in the majoritarian sense, as is often assumed.
Future research would benefit from the use of multiple survey measures that address not only abstract beliefs but specific questions about support for various facets of the status quo. Before we can say that a policy overrides majority will, or that there is a disconnect between opinion and policy, we must recognize in what ways the meanings of majority preference are often ambiguous and multiple.
Achen, C. (1978). “Measuring Representation,” American Journal of Political Science (22):
Andreas, P. (2000). Border Games: Policing the U.S.-Mexico Divide. Ithaca, NY: Cornell
Ansolabehere, S., Rodden, J., and Snyder, J.M. (2008). “The Strength of Issues: Using Multiple
Measures to Gauge Preference Stability, Ideological Constraint, and Issue Voting,” American Political Science Review (102): 215-232.
Barabas, J. (2007). “Measuring Democratic Responsiveness.” Paper Presented at the Annual
Meeting of the Midwest Political Science Association, Chicago, IL.
Batalova, J. and Lee, A. (2012). “Frequently Requested Statistics on Immigrants and