Abstract
How do the preferences of political elites shape humanitarian immigration to the United States? Focusing on the asylum and refugee systems, we trace the ways that the preferences of political elites affect the number and characteristics of migrants who receive relief. Our findings suggest that presidential preferences remain crucial in determining who is admitted in the U.S. refugee system and that congressional preferences are important in determining the number admitted. The preferences of both appear to matter considerably less on asylum decisions. These results highlight the difficulty of eliminating or reducing the role of the executive branch in American immigration policy, and suggest the importance of the design of the immigration enforcement bureaucracy in limiting the role of elite preferences in determining humanitarian immigration outcomes.
Historically, presidents have tended to view the refugee system as a useful tool of foreign policy. Cold War presidents from Dwight D. Eisenhower to Ronald Reagan favored those fleeing communist regimes (Loescher and Scanlan 1986; Gibney 1988). A clear example is the treatment Cuban and Haitians refugees received from the 1970s through the 1990s. Generally, Cubans were welcomed in an effort to embarrass and shame the communist Cuban government, even as Haitians were denied relief (Cox and Rodríguez 2009). Indeed, evidence suggests that these strategic uses of refugee relief were a primary motivation for the Refugee Act of 1980: Congress sought to restrict the discretion the president had enjoyed with respect to refugee admissions (Leibowitz 1983). As Senator Edward Kennedy noted at the time, “One of the principal arguments for the Act was that it would bring the admission of refugees under greater Congressional and statutory control” (Kennedy 1981, 146). Simultaneously, liberals have pushed to broaden the groups of people to whom the humanitarian system will provide relief, something also embodied in the 1980 Refugee Act (Tichenor 2002). As Norman Zucker and Naomi Zucker note, “humanitarian and nondiscriminatory aspects . . . were to be emphasized” (1992, 63). Although the system created by the 1980 act was both broad and generous, ideological battles have continued over who counts as a refugee or an asylee, battles that have remained the “primary issues of American refugee-asylum policy” (Zucker and Zucker 1992, 63).
For example, in September 2015, in response to the growing Syrian refugee crisis, Secretary of State John Kerry announced that the United States would increase its annual admissions of refugee from seventy thousand to eight-five thousand in 2016 and one hundred thousand in 2017. Kerry’s announcement came amid growing demands that United States do more in response to the mass forced migration of Syrians fleeing the civil war. Opposition played out in the halls of Congress as Republicans on the House Judiciary Committee attempted to give Congress the power to set admissions, threatening refugee resettlement funding, and culminating in a House bill that would require the FBI director “to certify . . . [that] each Syrian or Iraqi refugee admitted to the United States . . . were not security threats” (DeBonis 2015). Notably, the president was able to resettle the promised ten thousand refugees despite vociferous opposition from Republican members of Congress and Republican governors. Are presidential preferences paramount in determining U.S. refugee policy, as suggested by this episode? Can coordinated congressional action work to restrain the extent to which presidential preferences matter?
In this article, we use an innovative weighting approach to characterize the refugee and the asylum systems to understand how the preferences of political elites affect the humanitarian commitments of the United States. The refugee and asylum systems together are the primary way the United States admits migrants for humanitarian reasons, yet the two systems are distinct. First, refugees are admitted from outside U.S. borders, whereas asylees are already in the country when they seek asylum. Second, who within the executive branch determines whether one is admitted differs greatly between the two programs. For refugees, primary decision making is granted to a group of workers in the Department of State; for asylees, it is primarily in the Department of Homeland Security and the Department of Justice.
We argue that these differences condition the extent to which presidential and congressional preferences affect the number and characteristics of those admitted under the two programs. Using both descriptive and multivariate analyses, we find that presidential preferences are a key determinant of the characteristics of who is admitted as a refugee and that congressional preferences matter with respect to the number of refugees admitted; we find less support for the notion that the preferences of elites matter in the asylum system. Ideological influence on who is admitted is especially important given that refugees typically make up 80 percent of the total humanitarian intake in our period of study. These findings suggest the extent to which the administrative structure of an immigration bureaucracy can cabin or promote the role that preferences play in immigration admissions.
In 1980, U.S. humanitarian migration policy dramatically changed with the passage of the Refugee Act.1 The act was intended to restrict the president’s unconstrained use of the parole power to admit refugees (Rodríguez 2010) and to promote the use of humanitarian factors in decisions about who should receive relief in the United States in both the refugee and asylum systems (Gibney 1988). The law passed 328–47 in the House, strongly resisted by Republicans, who opposed the generosity of the law to refugees (Tichenor 2002, 247–48).2 The goal, in other words, was to reduce the extent to which the president could use humanitarian relief to further strategic goals and to reinforce the U.S. commitment to the international refugee resettlement system. In place of the old parole-based system, Congress substituted a regime in which the president has authority to set the number of refugees admitted in consultation with Congress, leaving asylum adjudication to be determined on a case-by-case basis. Summarizing the Refugee Act, Zucker and Zucker suggest that though the idea was to create a nondiscriminatory relief program, the broader humanitarian aspects of the law have instead been conditioned by “perceived foreign-policy needs and domestic politics” (1992, 63).
Together with the Refugee Act, the Immigration and Nationality Act (INA) establishes the legal requirements for admission under both humanitarian programs. Under the INA, to be eligible for refugee or asylum status, an applicant must be unwilling to return to his or her country because of a well-founded fear of persecution (101(a)(42) INA). It establishes that the number of refugees admitted each year shall be determined by the president before the beginning of the fiscal year and “after appropriate consultation with Congress.” It stipulates that the number be “justified by humanitarian concerns or is otherwise in the national interest” (8 U.S.C. 1157, Sec. 207).
The president is required to provide a significant level of analysis in the consultative documents prepared for Congress. Yet typically this consultation is no more than cosmetic, and the executive’s position is approved without much vetting (Zucker and Zucker 1992). Congress has no similar consultative role with respect to the number of asylees admitted yearly, nor is any numerical cap set on asylum admissions (Miller, Keith, and Holmes 2015), though occasionally Congress will intervene to change asylum policy.3
To date, we have little empirical evidence about whether this now four-decade-old statutory scheme works as intended, constraining presidential discretion in the admission of refugees and asylees. In the next section, we further distinguish the refugee and asylum systems.
COMPARING THE U.S. ASYLUM AND REFUGEE SYSTEMS
The most important distinction between the asylum and refugee systems is where an applicant seeking humanitarian admission to the United States is when applying. Asylees seek asylum from within the United States or on arriving at a point of entry. Asylum applications may be either affirmative or defensive. Affirmative applications are those filed by an applicant not under threat of removal from the country. Immigrants placed under removal proceedings by the Department of Homeland Security (DHS) may file defensive asylum applications. Those applying for relief affirmatively have their applications heard by asylum officers, who are part of the DHS bureaucracy. Immigration judges (IJs), who are similar to administrative law judges and appointed within the Department of Justice (DOJ), hear appeals from denials of affirmative decisions and all defensive claims of asylum.
Prior work shows that IJs tend to pursue their own policy preferences when adjudicating asylum claims (Miller, Keith, and Holmes 2015; Keith, Miller, and Holmes 2013), though they also remain susceptible to pressure from the attorney general (Kim 2018). A key point on individual decision making in the asylum system is that case-by-case adjudications frequently hinge on an assessment of the credibility of the applicant by the decision maker—indeed this is a major route by which preferences determine admissions (Miller, Keith, and Holmes 2015). Further, no ex ante (before arrival) discretion about who is allowed to apply for asylum is possible (Cox and Rodríguez 2009). In this way, asylum mirrors the larger U.S. immigration system, under which executive discretion influences who is allowed to stay through selective enforcement, not who is allowed to enter in the first place (Cox and Rodríguez 2009, 2014).
Refugees apply for relief from within their country of origin or a safe third-party country. Refugee status is determined with the help of the UN High Commissioner for Refugees (UNHCR) and, since 2006, a group known as the Refugee Corps. President George W. Bush created the Refugee Corps to provide a highly trained and specialized set of adjudicators (and supervisors) of overseas refugee claims, rather than rely on circuit-riders borrowed from the Asylum Office and elsewhere. The goal was to provide more consistent outcomes across adjudicators and refugees (U.S. Congress 2006). Decisions about whether an individual is eligible for relief are typically made by officials from the U.S. Department of State (State), secondary screening being performed by elements of the Department of Health and Human Services and DHS. Decisions occur at nine regional resettlement support centers located abroad and staffed by the Department of State.
Under the 1980 Refugee Act, the president is to determine how many may receive relief as refugees with the stipulation that the number be “justified by humanitarian concerns or is otherwise in the national interest” (8 U.S.C. 1157, Sec. 207). The president is required to provide a significant level of analysis in the consultative documents prepared for Congress.4 Further, the president has wide latitude to determine admissions under the proposed cap. Usually this is a designation made based on the country of origin for a refugee, where country of origin is considered a proxy for “immigrant type” (Cox and Rodríguez 2009, 458). As Aristide Zolberg notes, the provision in the law for groups of “‘special humanitarian concern to the United States’ opened the selection process to bargaining by . . . ideological . . . interest groups” (2006, 349). Indeed, the UNHCR makes recommendations about groups of refugee for U.S. resettlement, not individuals (Martin 2005).
Scholars continue to argue that presidents perceive humanitarian migration policy as a foreign policy tool divorced from U.S. treaty commitments to protect the vulnerable (Tichenor 2002; Zolberg 2006; Bon Tempo 2008; Salehyan and Rosenblum 2008; Rottman, Fariss, and Poe 2009). Our argument is slightly more nuanced in that we suggest a finer distinction to be made on the notion that presidents are strategic, namely, that conservative administrations will tend to use the system to embarrass geopolitical opponents more than liberal presidents will. Another way to phrase this difference is that liberal presidents might place more emphasis on humanitarian concerns than conservative presidents do. We develop these expectations.
Table 1 summarizes the key differences between the asylum and refugee systems. Aside from displaying the discussed characteristics of where individuals are when they apply for relief and who makes relief determinations, table 1 also shows that over the period of study—FY 1982 through 2018—relief in the refugee system has been more important numerically. On average, about four times as many people gain relief through the U.S. refugee system as through the asylum system. The next two rows highlight characteristics of who gets relief through each system. Human rights repressions scores range from 1 to 5, higher scores representing a more repressive regime. The aggregate data in table 1 show that the asylum system tends to protect people fleeing slightly more repressive regimes than the refugee system does.
Similarly, in the penultimate row of table 1, we calculate the percentage of asylees and refugees fleeing countries that are military allies of the United States. Perhaps not surprisingly given the long noted strategic use of the refugee system, it appears that 39 percent of refugees and 54 percent of asylees are fleeing military allies. The final row shows that the average number of economic sanctions for countries fled by asylees is higher (1.66) than the average for refugees (0.94). All of these differences are statistically significant (p<.00).
IMMIGRATION PREFERENCES OF U.S. POLITICAL ELITES
Generally, scholars assert a consistent difference in the approach of conservative and liberal parties on immigration issues. As Idean Salehyan and Marc Rosenblum suggest, the political divide on humanitarian migration and immigration in general lines up in predictable ways, parties on the right “tend[ing] to oppose the expansion of asylum, refugee, and family reunification flows and the extension of rights to immigrants,” and parties on the left “tend[ing] to be stronger proponents of international human rights protections, including asylum” (2008, 107). For instance, Randall Hansen and Desmond King (2000) note that the conservative party led the increase in migrant deterrence policies in the United Kingdom (for Australia, see Mughan and Paxton 2006).
Salehyan and Rosenblum assert that such a pattern has “held in the United States, where Democrats have generally advocated more liberal humanitarian migration policies than have Republicans” (2008, 107; see also Schrag 2000). Strategic concerns arise because admitting refugees and asylees “acknowledges human rights problems in the sending country and may be seen as providing shelter to dissidents” (Salehyan and Rosenblum 2008, 105).5 In a comprehensive policy history of U.S. immigration, Daniel Tichenor notes that “the struggle among modern policymakers has reflected differences over who [emphasis in original] should benefit from generous refugee relief” (2002, 248).
Within the United States, and specific to the humanitarian relief system, the battle has not necessarily been over whether broad refugee protections should be in place, given that both conservatives and liberals historically found some common ground in expanded protections. Disagreement instead centers on whether, as conservatives tended to prefer, the system be used to shame strategic adversaries (historically communist adversaries) or whether, as liberals would prefer, the system be used to protect human rights more broadly (Tichenor 2002). Put slightly differently, it may be that liberal presidents simply put less emphasis on strategic concerns in the context of refugee admissions.
Further, the relative broadness of relief provided by the refugee system narrowed in the United States (as in most other affluent democracies) with the end of the Cold War (Zolberg 2006). Yet within this narrowed scope for relief, we believe that ideological differences over who should benefit will remain. Additionally, a trade-off is not always required between providing humanitarian relief and using the system strategically. It is possible, of course, to admit people fleeing a geopolitical adversary that is also highly repressive of human rights.6
Our main expectations flow from this general split on immigration policy as reflected in four measures of the asylum and refugee admissions flow. These are the overall number of asylees or refugees admitted, the level of physical repression faced by those admitted, the proportion fleeing U.S. military allies, and the average number of U.S.-imposed economic sanctions on countries that asylees or refugees are fleeing. These measures give broad coverage over time and across aspects of humanitarian and strategic concerns. Including these measures leads to four hypotheses:
Hypothesis 1a: Conservative presidential administrations tend to allow fewer refugees than liberal administrations.
Hypothesis 1b: Conservative presidential administrations tend to select fewer refugees threatened by physical repression than liberal administrations.
Hypothesis 1c: Conservative presidential administrations tend to select fewer refugees fleeing military allies than liberal administrations.
Hypothesis 1d: Conservative presidential administrations tend to select more refugees from countries under U.S. economic sanctions than liberal administrations.
We also expect that, given the histories of the two programs and the decision-making structure in each that these partisan and ideological differences will be more prevalent in the refugee system than in the asylum system. This expectation stems from the fact that the asylum bureaucracy is not as centralized as the refugee bureaucracy. Interventions into asylum policy by the courts (Zucker and Zucker 1992; Zolberg 2006) and by Congress have been frequent. Indeed, the model in the asylum system is one of adversarial legalism (Hamlin 2012, 942), under which decisions are made on a case-by-case basis with attorneys before a judge-like bureaucrat (Hamlin 2012; Miller, Keith, and Holmes 2015).
Congress has also occasionally shown interest in changing the policies at work in the asylum system as it did with both IIRIRA and the REAL ID Act. Indeed, the REAL ID Act of 2005 increased the difficulty of gaining relief through asylum by sharply raising the requirements for documenting proof of persecution (Hamlin 2012). Given that bureaucratic insulation is much higher for the asylum system than the refugee system, we expect the refugee system to be more revealing:
Hypothesis 2: Differences between presidential administrations will be greatest in the refugee system.
Finally, we expect that congressional preferences may matter, but that if they do it will be primarily in regard to the number of refugees admitted. This expectation arises for three reasons: first, because the law explicitly directs the president to consult with Congress on the number of refugees—but not asylees—to be admitted; second, because the relative insulation of the asylum bureaucracy makes congressional intervention less likely to be effective; and, third, because the much higher numbers in the refugee system imply that if legislative concerns center on costs, then the refugee system should be the focus of attention. Indeed, as noted, at least some members of Congress saw increased congressional control as the express purpose of the Refugee Act of 1980. Congressional preferences on the number of refuges admitted mirror the partisan breakdown of hypothesis 1a:
Hypothesis 3: Conservative Congresses will tend to allow fewer refugees than liberal Congresses.
Table 2 summarizes these expectations in terms of how we expect conservative control of government to correlate with a particular characteristic. Cells with tildes indicate that we do not expect to observe a statistically significant difference between administrations. Unfortunately, we cannot test the argument Tom Wong (2017) puts forward that ideological alignment on immigration occurred in Congress only following contentious events in 2005 because it overlaps with two key interventions—the Real ID Act in the asylum system (see Hamlin 2012) and the creation of the Refugee Corps in the refugee system.
Evidence suggests that members of Congress will be less concerned with the characteristics of the refugee and asylee flow—who is entering rather than how many. The general immigration literature holds that Congress has crosscutting influences that pit pro-migration interest groups against the “overall pattern of [popular] hostility to migration” (Rosenblum 2004, 32). As Marc Rosenblum (2004) reports, members of Congress are generally uninterested in immigration issues because they are of little value to constituents and this is likely to be reflected in little impact for congressional preferences on the characteristics of the refugee and asylee flows.
The one exception might be concerns with the costs of refugee resettlement. The Refugee Act of 1980 specifically requires a projection of the costs of any refugee resettlement program. As Zucker and Zucker note, “large numbers of refugees incur the fear of . . . heavy costs for long-term refugee support . . . thus galvanizing the legislature into making refugee or asylum policy” (1992, 64). Cost is related to the number of refugees admitted, so if Congress is concerned with costs it should be primarily concerned with the number of refugees admitted. Indeed, three of the seven consultation requirements in the Refugee Act concern the costs of resettling refugees and other provisions of the law repeatedly emphasize the importance of the “economic self-sufficiency” of refugees (8 U.S.C., Sec. 1522(a)). Given that typically four times as many people are admitted as refugees than as asylees, and limited congressional attention to immigration more generally, we expect that any concern with the number of migrants admitted is likely to focus on refugee admissions.
Figure 1 displays the overtime trends in refugee and asylum admissions across three historical eras. The Cold War era is the period (in our data) from FY 1982 through FY 1991 and the collapse of the Soviet Union; the post–Cold War era covers FY 1992 through FY 2001, ending with the 9/11 attack; the 9/11 attack defines the third era on the figure, through FY 2018. A few things are notable. First, refugee admissions always exceed—usually greatly—asylum admissions except immediately following the 9/11 attacks and in FY 2018 under the Donald J. Trump administration. Indeed, FY 2018 shows the lowest refugee admission in the period covered by our data (22,484).
Second, asylum admissions are quite low until the creation of the modern asylum system in 1991. The creation of a group of asylum officers to handle affirmative asylum cases to ensure that asylum rulings are “fair and sensitive” (Koehn 1991, 231) defines this change. Indeed, writing at the creation of this newer asylum system, Zucker and Zucker suggest that it would be fairer and more consistent and that “foreign-policy concerns will no longer be a major influence on asylum decisions, which instead will draw on human rights conditions” (1992, 68). With the implementation of the new adjudicatory system, the number of asylees admitted rose correspondingly from approximately two thousand to thirty-nine thousand by FY 2001.
However, since the 9/11 attacks, the number of asylees admitted to the United States has held steady around twenty-five thousand with minimal yearly fluctuations. The decrease in fluctuations is most likely because courts have intervened to stabilize definitions and practices and the bureaucracy itself relies to some extent on precedent in its decision making. Finally, it is clear from figure 1 that variability is greater in the number of refugees admitted than in the number of asylees. This is the first indication that the refugee system may be more responsive to the preferences of political elites than the asylum system is.
DATA AND METHODS
As noted, we take a unique approach to gauging the influence of political elites. A number of studies analyze the broad dynamics of how many refugees or asylees are admitted to the United States in a given year (Miller, Keith, and Holmes 2015; Keith, Miller, and Holmes 2013; Salehyan and Rosenblum 2008; Rosenblum 2004). Our approach differs because we are interested in not only the number of people admitted under each program, but also the characteristics of that population. In addition, our focus in this work is on how the preferences of political elites condition the characteristics of the asylum and refugee flows. These preferences are most likely to be manifest in broad patterns at the aggregate level. Therefore, we seek to create a data set that captures this aggregate picture for both systems.
To do so, we begin with a time-series cross-sectional (TSCS) data set of every country that sent an asylee or a refugee to the United States in our period of study. We are able to characterize these humanitarian flows according to the countries from which they flee. For instance, refugees from Somalia in FY 2009 are coded as fleeing a U.S. military ally. Then, for each fiscal year, we weight the contribution of a country to the total flow. For example, in FY 2010, Iraq is the single largest contributor in the refugee flow, at 25 percent of the total. In FY 2010, China contributes 32 percent to the asylum flow. The appendix provides additional information on weighting for both systems.
Using this approach, we can summarize how much of the flow in either system comes from a military ally or highly repressive regimes, for instance. Aside from the number in either system, we focus on three characteristics of the flow: whether the country of origin is a military ally, whether the country of origin is under economic sanctions by the United States, and the level of human rights repression in the country of origin. Each is selected to help us understand whether any focus is on strategic or humanitarian uses of the systems—the key difference highlighted in the literature on the politics of humanitarian migration (Miller, Keith, and Holmes 2015; Keith, Miller, and Holmes 2013; Salehyan and Rosenblum 2008; Katzenstein, Keohane, and Krasner 1999; Loescher 1993; Loescher and Scanlan 1986).
Military allies are defined by whether a country receives military aid in a given year from the United States, coded as 0 or 1 (more information on each variable is presented in the appendix). We code the number of U.S. imposed economic sanctions a country is under in a given year, which ranges from 0 to 11. Our measure of human rights repression comes from the political terror scale (PTS), which ranges from 1 to 5, higher scores indicating greater repression.7 Two versions of the PTS scale are used, one based on reports from the Department of State and one based on reports from Amnesty International (Gibney et al. 2015). We average these two to create our repression scores. To give an intuitive sense of these scores, the average for Somalia in 2008 is 4.5 (high repression); Cuba in 2008 is a three (moderate repression).
Figure 2 presents the change in weighted averages of these variables over time. The dashed lines divide each panel into the eras noted in figure 1. A few trends are noteworthy. Correlation is tight in the proportion of refugees and asylees receiving relief who are fleeing military allies until the post-9/11 era. In regard to economic sanctions, divergence is considerable in the two series starting in the mid-1990s, as those admitted in the asylum process are considerably more likely to be fleeing states under a higher number of economic sanctions from the United States; the opposite is true in the refugee system. Finally, with respect to human rights repression, the asylum system reaches something of an equilibrium in the mid-1990s (stabilizing around a mean repression score just below 4), but in the refugee system fluctuates considerably from a low of 2.67 in FY 1989 to a high of 4.48 in FY 2016.
Our analytical approach is straightforward. First we present descriptive differences across partisan control of the presidency and Congress for the number of asylees and refugees, as well as information on each of the three characteristics we have chosen to analyze. Then we support the descriptive analysis with a series of multivariate regressions that control for a host of potential confounders. We think it is important to look closely at the descriptive statistics, given potential concerns about the power of the analysis in a multivariate regression framework. Although a relatively small N does raise some concerns about how small changes might alter our results, we undertake a number of robustness checks to probe our regression results. We used all of the available data in constructing this analysis—the entire population of interest rather than only a sample.
In each analysis (both descriptive and multivariate), we lag all explanatory variables by one period to allow for a causal interpretation of the results, reducing our N from 37 to 36.8 Differences are generally few between the descriptive analysis and the multivariate analysis with a full suite of controls. Where differences are, the multivariate results tend to contradict theoretically inconsistent results from the descriptive analysis.
To account for partisan differences among political elites, we code for partisanship in the descriptive analysis and ideology in the regression analyses. We use ideology scores in the regression analyses because they allow for more flexibility in the particularistic policies of both presidents and the legislature at any given time. Coding partisanship is simple with respect to the president, but for Congress we are careful to note periods of mixed control (where one party controls each House). In the descriptive results we present, mixed control is excluded from the analysis: for the thirty-six years in the analysis, we have fourteen years of Republican control, eleven years of Democrat control, and eleven years of mixed control. In the regression analyses, we use ideology scores rather than simple partisanship.
To measure ideology, we use DW Nominate scores (Poole and Rosenthal 1985). Higher scores indicate greater conservatism. For the president, these scores are, in chronological order: Reagan (0.693), H. W. Bush (0.557), Clinton (-0.438), G. W. Bush (0.693), and Obama (-0.354). For Congress, we calculate the median Nominate scores for each chamber for each Congress and then calculate the midpoint between the medians. This approach follows those taken by leading scholars of Congress (Gray and Jenkins 2017). These scores range from -0.1935 to 0.2848.
For the multivariate analysis, we include a host of control variables suggested by the literature on asylum and refugee admissions in the United States. We control for the preferences of political elites as noted. Varying levels of attention to immigration in the public and among political elites might alter the extent to which preferences can operate to shape refugee and asylum flows. We therefore also introduce controls for the number of mentions of immigration in the president’s State of the Union address, the number of congressional hearings on immigration, and the number of articles in the New York Times on refugee or asylum issues.
We do not have any a priori expectations about how the attention of elites and the public to immigration issues will affect the number and characteristics of those admitted in either system. Nevertheless, inclusion of the attention variables is important because it helps account for exogenous shocks to the asylum and refugee system (Wong 2017).9 In addition, we include two control variables in our refugee regressions to account for changes in the decision-making process for resettling refugees in the United States. First, we include a dummy variable for the switch to UNHCR involvement in the process in 1994 (0 before and 1 after). Second, we include a dummy variable to account for the creations of the Refugee Corps in 2006 (0 before and 1 after). Both changes were made in an attempt to refocus the refugee system on the need to protect the persecuted. We therefore expect that both will increase the repression scores of those admitted in the refugee system. The equivalent change in the asylum system is the creation of asylum officers for interviewing affirmative asylum applicants in 1991 (modern asylum system). This change should also work to increase the average level of repression for those admitted into the system.
Finally, to help account for change over time not otherwise controlled for in our regressions, we include a counter of elapsed time that starts at one in fiscal year 1982 (or 1983 in the asylum system) and adds one for each additional year. Where necessary, we also include lags of the dependent variable to account for potential autoregression in our multivariate regressions. This is a concern only in a few of the asylum regressions, as indicated by testing using the Durbin-Watson alternative statistics. Table A1 in the appendix includes descriptive information on the included variables.
RESULTS
Table 3 presents the results of the descriptive analyses (for additional results and models, see the appendix). We divide our descriptive analyses into two parts, the asylum system and the refugee system. For each system, we display results for the three characteristics of interest and the number of immigrants admitted in each category by partisan control of the presidency and Congress. To aid with interpretation, we present the difference between the average score (or number admitted) for Republicans and Democrats; signs for each category indicate whether we expect Republican control to lead to higher or lower numbers than under Democratic control. Each category corresponds to a described hypothesis; for each, we indicate the expected direction (positive or negative) of the difference between the Republicans and Democrats. We test the differences across each category for statistical significance using a t-test of the means, p-values reported in parentheses. Finally, we shade the comparisons that are consistent with our hypothetical expectations.
The takeaway from table 3 is the considerable evidence that the preferences of the president predominate in refugee admissions, and that congressional preferences matter for the number of refugees admitted but not the characteristics of the flow. Presidential preferences seem to affect the number and characteristics of asylum admissions; congressional preferences do not matter with respect to the characteristics of those admitted as asylees. Evidence suggests that approximately eleven thousand fewer asylees are admitted under Republican administrations than under Democratic administrations (see table 3, panel A, top rows).
This difference stems from ex post (after arrival in the United States) enforcement discretion, most likely manifested through directives to IJs and asylum officers. Such enforcement is the common way the executive is able to influence immigration policy (Rodríguez 2010). Evidence also indicates that asylees fleeing U.S. military allies fare better under Democratic presidential administrations—69 percent of the asylum flow comes from these countries, 28 percentage points higher than in the Republican administrations. This fits with the notion that Republican administrations might seek to use the humanitarian immigration system to further U.S. geopolitical interests. Together, the evidence from the number of asylees admitted being higher in Democratic administrations and Republican administrations disfavoring asylees fleeing military allies supports the notion that presidential preferences shape the asylum flow.10
Although differences are significant in the average number of economic sanctions and human rights repression between Democratic and Republican administrations in the asylum system, the direction of these effects is inconsistent with our expectation for how presidential preferences will matter. Therefore, evidence in favor of presidential preferences conditioning asylum admissions is mixed. The second set of rows in panel A indicate limited evidence that congressional preferences are important determinants of asylee admissions. Although a number of differences across partisan congressional control are statistically significant, none is consistent with theoretical expectations. For instance, we do not expect that Republican control of Congress will lead to an increase in the number of asylum seekers admitted. Indeed, these results disappear in the regression analyses.
Panel B of table 3 presents results for the refugee system. Here, evidence for the role of executive preferences in shaping refugee immigration to the United States is striking. In each category, we observe statistically significant and theoretically consistent differences between Republican and Democratic administrations. On average, we expect that almost twelve thousand fewer refugees will be admitted in Republican administrations than in Democratic administrations.11 Democratic administrations are also more likely to admit refugees fleeing military allies, are less likely to seek to embarrass states under U.S. economic sanctions, and provide relief to refugees who are, on average, fleeing regimes that are more repressive.
All of these differences are consistent with our expectations, as represented in hypotheses 1a through 1d. We also find evidence that partisan control of Congress is consequential for the number of refugees admitted. When Republicans control both the House and the Senate, we expect almost twenty-six thousand fewer refugee admissions than when the Democrats control both houses. This difference is about twice as large as the effect for changes in partisan control of the executive. It may be that a conditional effect between executive and congressional control applies, which we explore further in the context of multivariate regression. In regard to Congress, this is the only difference consistent with our theoretical expectations, despite significant and theoretically inconsistent differences in the number of refugees from military allies and average number of U.S. economic sanctions.
That congressional preferences only seem to affect the number of refugees admitted, but not the underlying characteristics of those admitted, supports hypothesis 3. It is consistent with the notion that the president is required to consult with Congress on the number of refugees admitted and the notion that members of Congress should care about refugee admission numbers to the extent the number of refugees admitted affects the availability of resources for constituents. The clear evidence supporting the importance of presidential preferences in the refugee system relative to the more mixed evidence in the asylum system tends to support hypothesis 2.
As noted, to test the robustness of the descriptive results presented in table 2, we also estimated a series of multivariate regressions. Table 4 presents the asylum models and table 5 presents the refugee models.12 The most important takeaway from these regressions is that they support our descriptive analyses on the refugee system but undermine the statistically significant descriptive results for asylum admissions. Namely, we continue to find significant evidence in favor of the importance of presidential preferences in the refugee system and limited evidence of the importance of Congress, excepting the number of refugees admitted. In addition, we find evidence that changes in the decision-making apparatus in both the asylum and refugee systems had an important impact on how many and what kinds of humanitarian immigrants are admitted to the United States independent of the preferences of political elites.
In regard to presidential preferences, we find some theoretically consistent evidence for effects in the asylum system, namely, conservative presidents decreasing the number of asylees admitted. Moving from the most liberal to the most conservative president in the data decreases the number by three thousand. This effect is only marginally significant at the p<.10 level, an effect contradicted in the descriptive analysis. Although the coefficient for presidential ideology is statistically significant for military aid, it is incorrectly signed, indicating that conservative control of the White House increases the likelihood of admitting refugees from military allies, an effect we expect in Democratic administrations, not Republican ones.
This result is also contrary those in the descriptive analysis, in which Democratic administrations were more likely to admit those fleeing military allies. The coefficient for congressional control is significant for human rights repression, but the sign of the coefficient is not consistent with our theoretical expectations about the effects of conservative control. Furthermore, these results are unsupported by the descriptive evidence presented in table 3. For the most part, neither public nor elite attention to immigration and asylum issues affects either the number of asylees admitted or their characteristics. The exceptions are that an increase in the number of New York Times articles mentioning asylum decreases the number of asylees admitted and an increase in the mentions of asylum in the SOTU increases the portion of asylees admitted from military allies.
The transition to the modern asylum system seems to have had an important effect on the number of asylees admitted. Our estimates suggest that moving to the modern system, with asylum officers, increased the number of asylees admitted by 3,560. Considering that the mean number of asylum seekers admitted across our period of study is 18,656, this is a 19 percent increase in the number of people receiving relief. Evidence also indicates that the advent of the modern asylum system increased the likelihood that asylum seekers fleeing U.S. military allies receive admission—increasing the percentage admitted from allies by 21 percentage points.
In table 5, we find evidence consistent with the influence of presidential policy preferences in the refugee system. The coefficients for presidential administration are statistically significant and consistently signed in each of the four models (refugee admissions, military aid, economic sanctions, and human rights repression). Similarly, the coefficient for congressional ideology is statistically significant and consistently signed in the refugee admissions regression, supporting the evidence presented in table 3. Moving from the most liberal to the most conservative president decreases the number of refugees admitted by 15,334. Similarly, it decreases the percentage of the refugee flow coming from military allies by 16 percentage points, increases the number of economic sanctions countries being fled are under by 0.30 (about 30 percent), and decreases the level of repression in the flow by 0.33 points (about 9 percent). However, the effects for refugees fleeing military allies are not robust to the exclusion of the Cold War era, signaling that since the early 1990s, this particular strategic use of the refugee system may have diminished.
Last, moving from a very liberal Congress (10th percentile of the data) to a very conservative one (90th percentile of the data) decreases the number of refugees admitted by 32,228. This effect is twice that of presidential ideology on the numbers of refugees admitted and suggests strong interest from Congress over concerns about how many are admitted as refugees. That we find no other effects for congressional ideology implies little concern, however, over who is admitted as a refugee.
Public and elite attention to immigration and refugee issues appears to matter a good deal to the number of refugees admitted. As State of the Union mentions of immigration increase, a proxy for public attention, so do the number of refugees admitted.13 Similarly, as the number of congressional immigration hearings increase, so do the number of refugees admitted. As the number of New York Times articles on refugees increase, a proxy for elite attention, the number of refugees admitted decreases. Varying the number of articles from the 10th percentile (0) to the 90th percentile (153) decreased expected refugee admissions by 23,562. Attention to refugee issues in the New York Times is accompanied by an increase in the level of human rights repression in the refugee system. An increase from the 10th percentile to the 90th percentile in the number of articles is accompanied by an increase in the average level of repression by 0.46 points. In reference to the average repression score for the refugee system as a whole (see table 1), this represents an 11-percentage point increase in the level of human rights protection.
Taken together, the results for the attention variables suggest that increasing elite attention increases the number of refugees admitted but increasing public attention decreases the number admitted. This divergence mirrors the sense that many immigration scholars have that elite opinion has been insulated from popular opinion (Tichenor 2002, 246). Notably, it may be that the Trump presidency has changed this dynamic, given that Trump has made immigration restrictions perhaps the central policy component of his administration (Gimpel 2017).
Mirroring findings in table 4 for the asylum system, we find evidence that changes in the decision-making apparatus of the refugee system had important consequences. The movement in the mid-1990s to incorporating greater UNHCR input decreased the number of refugees admitted considerably, by about thirty thousand. This result holds if we omit the 9/11 outlier year, though the size of the effect decreases to about twenty-six fewer refugees. The introduction of the Refugee Corps in 2006 also lead to a decrease in the number of refugees admitted, about thirty-one thousand fewer. But the effects of the Refugee Corps are more far-reaching than those of the UNHCR, given the additional evidence that creating the Corps reduced the likelihood that refugees fleeing military allies were admitted, and increased substantially the protection the refugee system provides against human rights repression.
The effect of the Refugee Corps on the level of human rights protection the system offers is significant. Before the Refugee Corps, we estimate the average level of repression in the system at 3.31, which increased to 4.2 after implementation. The introduction of asylum officers resulted in a shift from an average level of repression equivalent to Egypt in FY 2005 to Uganda in FY 2007 (straddling the introduction of the Refugee Corps in FY 2006). The introduction of the Refugee Corps therefore increases the human rights protection afforded by the refugee system while also seeming to increase the geopolitical uses of the system (by decreasing the likelihood of a refugee from a military ally gaining admission). This effect disappears, however, if we exclude the Cold War era from the calculation, so it should be treated with caution.
Given the finding in the descriptive analysis that both presidential and congressional preferences seem to influence the numbers of refugees admitted, we return to the possibility that presidential and congressional preferences interact to influence the number of refugees admitted to the United States. To better understand whether the preferences of one actor condition the effect of the other, we interacted our measure of presidential ideology with our measure of congressional ideology and entered it into the regression in table 4 for refugee admissions.
Figure 3 presents the results of this model (full model results presented in the appendix). Under conservative Congresses (90th percentile of ideology), differences between liberal and conservative administrations are considerable, liberal control of the White House leading to approximately seventy-eight thousand refugee admissions, and under GOP control of the presidency and the legislature some thirty-seven thousand (95% confidence intervals are represented by the vertical bars). Yet under Democratic control of Congress, we estimate no difference in the number of refugees admitted, no matter who controls the White House—approximately eighty-five thousand under both Democratic and Republican administrations.
The pattern here suggests that under Republican control of Congress, presidential discretion on the numbers of refugees is at a maximum, but is more constrained when Democrats control Congress. It is not immediately clear why this should be the case. The key here is to understand the movement of GOP presidents, who change from admitting some thirty-seven thousand refugees under GOP control of Congress to admitting more than twice as many under Democratic control. It might be that Republican presidents are happy to admit fewer refugees when their copartisans in Congress seek such a reduction but are also happy to increase the number of refugees admitted, satisfying more liberal Congresses, as long as they maintain control over the characteristics of admitted refugees. This discussion throws relief on the notion that an inherent trade-off between humanitarian and strategic uses of the refugee system is not necessarily the case, because it is at least theoretically possible that an administration could seek to protect allies while also admitting those fleeing truly repressive situations.
DISCUSSION AND CONCLUSION
Evidence demonstrates that presidents exert a good deal of control over the refugee system in the United States. More specifically, presidents can influence who comes in to the country as a refugee, beyond just the number of refugees admitted. Three separate aspects of the countries that refugees flee reflect this: the level of human rights repression, whether the country is a military ally, and whether the country is under U.S. economic sanctions. Further, the ideology of both the president and the Congress influence the number of refugees admitted. We believe that the focus for Congress is mostly on the potential costs of the refugee program.
Contrasting the results for the refugee system with those for the asylum system, we find less influence for the president and Congress on asylum admissions, consistent with our a priori expectations—excepting presidential influence on the number of asylees admitted. This relative lack of influence reflects the fact that the asylum system is significantly more insulated than the refugee system from outside influence. Judge-like actors make asylum decisions in a court-like setting and federal courts are much more involved in asylum decision making than they are in regulating refugee admissions. However, under President Trump, conflict is heightened between the executive and the IJs, through numerous DOJ policies and other executive decisions.
Notably, much of the system we describe here may have changed with the election of Donald Trump, the first modern president to make immigration the central issue of his campaign (Gimpel 2017). Evidence for this comes in the dramatic drop in refugee admissions for fiscal year 2018, when just twenty-two thousand refugees were admitted. For the first time since 9/11, the number of refugees has declined below the number of asylees. In a sense, Trump may be seen as a president who is not as insulated as most prior modern presidents from popular opinion on immigration.
Given bipartisan public support for admitting refugees (Daniller 2019), such a dramatic reduction in the number admitted may be surprising. Yet our evidence suggests that presidential control of the refugee system makes it an easy example for Trump to use in bolstering his immigration bona fides to his base. Control over the asylum system will remain more difficult for the president as long as other branches of government—particularly the judiciary—are willing to intercede in asylum decision making. However, the Trump administration has also appointed significant numbers of judges, nearly four in ten IJs as of July 2019 (Taxin 2019), which may significantly influence individual asylum outcomes in the long term if the administration is appointing judges that may be predisposed to deny asylum. This is particularly true given that Trump has also successfully appointed a large number of federal judges (Petersen and Szafir 2019). The Trump presidency may therefore increase the importance of the asylum system relative to the refugee system in providing relief for those fleeing persecution in the short term.
Last, our results speak to the need to treat the humanitarian immigration system in the United States as two distinct systems. The asylum system can be characterized as adversarial in nature and driven by case-by-case decision making in a relatively well-insulated bureaucracy. On the other hand, the refugee system is premised on deciding based on groups of people through a process that is susceptible to executive influence. An open question is whether it is possible and desirable to design a more insulated refugee decision-making bureaucracy.
APPENDIX
Data and Descriptive Statistics
Refugee counts come from the Department of State and, before 1988, the UN High Commissioner for Refugees. Physical repression scores, both State Department and Amnesty International measures, are from the political terror scale (Gibney et al. 2016; see also http://www.politicalterrorscale.org). Data on countries to which we send military aid are collected from USAID Foreign Aid Explorer (http://explorer.usaid.gov). Data on economic sanctions comes from the TIES (Threat and Imposition of Sanctions) project (Morgan, Bapat, and Kobayashi 2014; see also http://www.unc.edu/~bapat/TIES.htm). Data on congressional immigration hearings comes from the Policy Agendas Project, specifically the congressional hearings dataset focusing on the major topic of immigration, which includes hearings on refugee and asylum issues, among others. These data were originally collected by Frank R. Baumgartner and Bryan D. Jones, with the support of National Science Foundation grant numbers SBR 9320922 and 0111611, and are distributed through the Department of Government at the University of Texas at Austin. Finally, data on the number of New York Times articles concerning refugees in the United States was collected by our research assistant from the New York Times online index (http://www.nytimes.com/info/contents/contents.html).
Additional Model
Table A2 displays the full regression results used to generate figure 3 in the text. A few things are noteworthy. First, the constituent terms and the interaction term are all statistically significant, as is an F-test of whether the interaction should be included in the model (F (3, 22) = 10.24 (0.000)). Most of the variables that are significant in the model for refugee admissions presented in table 5 are significant in table A2, save that presidential mentions of immigration in the State of the Union are no longer significant.
Asylum and Refugee Weights
Tables A3 and A4 present, by fiscal year, the top five countries contributing the U.S. asylee and refugee flows, noting their weights—the percentage of the total contributed by those countries. We include weights at four-year intervals in the tables. These weights are applied to create our dependent variables for military aid, economic sanctions, and human rights repression.
Appendices
FOOTNOTES
↵1. The president, through the secretary of the Department of Homeland Security, has the power to grant parole to noncitizens to allow them to enter and remain in the United States for specific reasons, determined by the executive branch. Permission to remain is granted temporarily under current law, but before the Refugee Act of 1980, the president in conjunction with Congress regularly used the power to grant permanent admission to large numbers of immigrants. A prominent example is Hungarians admitted using the parole power in the 1950s and granted status as lawful permanent residents.
↵2. This is preliminary evidence that the partisan-ideological split we describe was present from the beginning in our data.
↵3. Rebecca Hamlin (2012) recounts the fight over how to define Chinese asylum-seekers fleeing forced sterilization. Although the Clinton administration and the asylum bureaucracy initially classified these asylees as not facing state persecution, Republicans in Congress ultimately intervened to redefine state-required sterilization as persecution in the Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA) in 1996. This was done with a nod to the domestic politics of abortion.
↵4. The analysis is meant to include descriptions of the following seven factors: the nature of the refugee situation; the number and allocation of the refugees to be admitted and an analysis of conditions within the countries from which they came; the proposed plans for their movement and resettlement and the estimated cost of their movement and resettlement; the anticipated social, economic, and demographic impact of their admission to the United States; the extent to which other countries will admit and assist in the resettlement of such refugees; the impact of the participation of the United States in the resettlement of such refugees on the foreign policy interests of the United States; and any additional information as may be appropriate or requested by such members.
↵5. To further clarify the distinction between humanitarian and strategic or geopolitical uses of the humanitarian immigration systems consider that the law dictates that refugee status be granted only to those fleeing a well-founded fear of persecution in their home countries, but many U.S. presidents are seen as using the refugee system in particular to advance their “particularistic foreign policy goals” (Cox and Rodríguez 2014, 119).
↵6. For instance, Iran is a strategic opponent of the United States that also typically has high human rights repression scores. Accepting refugees from Iran could be seen as satisfying both strategic and humanitarian objectives. For reference, in the period after 2005, Iran is typically under economic sanctions, does not receive military aid from the United States, and has an average repression score of 3.94.
↵7. The political terror scale levels can be interpreted as follows: 1) Countries under a secure rule of law, people are not imprisoned for their views, and torture is rare or exceptional; political murders are rare. 2) Imprisonment for nonviolent political activity is limited; few persons are affected; torture and beatings are exceptional; political murder is rare. 3) Political imprisonment or a recent history of such imprisonment is extensive; execution or other political murders may be common; unlimited detention, with or without trial, for political views. 4) Civil and political rights violations have expanded to large numbers of the population; murders, disappearances, and torture are common; terror tends to affect those who interest themselves in politics or ideas. 5) Terror has expanded to the entire population; leaders of these states place no limits on the means or thoroughness on means in pursuit of goals. See http://www.politicalterrorscale.org/Data/Documentation.html (accessed May 18, 2020).
↵8. We tested whether two lags better fit the data than the one-lag structure we use. For the refugee data, evidence from comparing the information criterion suggest the one lag models fit best. For the asylum data, only the model for human rights repression in the asylum system evidence better fit with two lags. This two-lag model suggests an even stronger positive effect for conservative congressional ideology on how much the system protects those fleeing persecution (about twice the size of the effect displayed in table 4). As we note elsewhere, this effect is contrary to our theoretical expectations.
↵9. Wong gives the example of interest in unaccompanied minor immigrants from Central America dropping once the media stopped covering the issue (2017, 7).
↵10. It also fits with evidence on the decision of immigration judges. More conservative judges are significantly less likely to admit asylees fleeing U.S. military allies and less likely to admit those fleeing highly repressive regimes (Miller, Keith, and Holmes 2015).
↵11. This difference is robust to the exclusion of the 9/11 outlier in refugee admissions.
↵12. A disadvantage of the multivariate models is that for several independent variables we do not have data beyond 2014, or 2013 for economic sanctions data. This reduces our N from 36 to 32 or 31, depending on the model. Counting the lag necessary to allow independent variables to credibly affect the dependent variable, most of the refugee multivariate models include FY 1983 through FY 2014, whereas the asylum models include FY 1984 through FY 2014. Modern record keeping for asylum admissions did not begin until FY 1983 (for more information, see the appendix).
↵13. In most years, no mention is made in the State of the Union of immigration. Using a dummy variable rather than a continuous indicator of number of mentions does not alter our conclusions about the effects of presidential attention, with one exception. Using the dichotomous indicator reduces the statistical significance of the SOTU variable in the refugee model for total admissions from p<.05 to p<.10 while increasing the coefficient to 17,702.
- © 2020 Russell Sage Foundation. Miller, Banks, Jennifer S. Holmes, and Linda Camp Keith. 2020. “The Preferences of Political Elites and Humanitarian Immigration to the United States.” RSF: The Russell Sage Foundation Journal of the Social Sciences 6(3): 150–71. DOI: 10.7758/RSF.2020.6.3.07. We acknowledge the UT Dallas Advisory Board Competitive Research Grants program for research assistance on the project and thank Marisa Durham for her skillful research assistance with much of the data gathered or created here. Direct correspondence to: Banks Miller at millerbp@utdallas.edu, Political Science Program, School of Economic, Political, and Policy Science, University of Texas at Dallas, 800 W. Campbell Rd., GR 31, Richardson, TX 75080, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.