EIGHT DECADES OF AMERICAN PUBLIC OPINION ON REFUGEE ADMISSIONS - - PDF document

Page 1 of 11 EIGHT DECADES OF AMERICAN PUBLIC OPINION ON REFUGEE ADMISSIONS Mariano Sana1 Submitted to the 2017 International Population Conference, Cape Town, Oct 29-Nov 4 Introduction and research goals Questions about attitudes toward refugees in the United States have been asked ever since the dawn of scientific polling in the 1930s. The archives of the Roper Center for Public Opinion Research at Cornell University allow us to track Americans’ views on refugees as specific refugee crises unfolded. There has been no shortage of these in the eight decades since the 1930s. The list includes: 1) Jews escaping the pre-World War II expansion of Nazism; 2) Post-World War II displaced Europeans; 3) Hungarians who fled after the failed revolution of 1956; 4) South Vietnamese escaping the Communist regime after the Vietnam War; 5) Indochinese refugees (dubbed “boat people”) in the late 1970s; 6) Cubans leaving from the Port of Mariel (the “Mariel Boatlift”) in 1980; 7) Cubans and Haitians landing on the Florida coast in the summer of 1994; 8) Ethnic Albanians from Kosovo in 1999; 9) Syrians fleeing their war-torn country in the last few years and at present. Other refugee flows proceeded in a steadier manner and did not make prominent news headlines. Accordingly, they did not prompt questions in opinion polls, since these tend to be sensitive to the topics covered by the media. This study reviews American attitudes toward refugee admissions since the beginning of opinion polling. The overall picture is well known (Newport 2015). Ever since pollsters began asking, Americans have generally been reluctant to admit refugees, just as they have consistently preferred lower to higher immigration flows (Fussell 2014). There are some exceptions, but what I present here is first and foremost an analysis of responses to surveys questions rather than an exploration of temporal swings toward more or less support. In fact, the analysis will prompt a fair deal of skepticism over the usefulness of opinion polls to infer valid estimates of Americans’ support toward refugees. Responses to survey questions can be influenced by the wording of the question, the order in which the questions are asked, the choices given for the responses, and other factors (Fowler 2002, Groves et al.

1 Department of Sociology, Vanderbilt University. Email: mariano.sana@vanderbilt.edu. This is work in progress

and a new version is forthcoming. Please ask the author about updates.

Page 2 of 11 2009). If the analyst could remove the effect of those influences, one could presumably access the “real” information, be it preferences, ideology, or “the forces shaping individual’s cognition and behavior” (Krosnick & Abelson 1994:177). This is not an easy task, and it may be nearly impossible, as attitudinal questions are notoriously more affected by the factors listed above than fact-based

• questions. It follows that a review of Americans’ attitudes toward refugees, as reported in surveys,

would be incomplete without an inquiry into the meaning of the questions asked. While we cannot ask respondents how exactly they interpreted the questions they were asked, we can at least look for patterns linking features of question wording to the distribution of answers. Specifically, I analyze the distribution of responses to these types of questions: a) Questions that elicit a yes/no answer; b) Questions that evoke values; c) Numeric vs. non-numeric questions; d) Questions that mention specific government action; e) Questions on whether refugees would be welcomed in the respondents’ communities. Data I use the Public Opinion Research Archives of the Roper Center for Public Opinion Research, housed at Cornell University. The iPOLL database has over 600,000 questions and their answers from U.S. national adult sample polls. Of these questions, over 6,000 are included under the topic “Immigration.” The Archive also holds over 12,000 U.S. and 8,000 international datasets. In this paper, I only work with single questions and univariate statistics (the distribution of responses to each question) from polls fielded between 1938 and 2016. I use the terms “poll” and “survey” interchangeably, as the distinction between them is trivial (Schuman 1997). I selected all policy questions referring to whether refugees should be accepted, how many should be/have been accepted, and whether they would be welcomed in the respondent’s community. Almost all questions referred to specific refugee flows, mostly excluding questions addressing policy toward refugees in general. In a few cases, questions were discarded due to their flawed design.2

2 Consider the following question from an August 1994 PSRA (Princeton Survey Research Associates) / Newsweek poll: “Which

• ne of the following would you most like to see the United States do in response to the Cuban refugees? Increase pressure on

Cuba to force Castro to give up power, do what's necessary to keep the refugees from entering the United States, let the refugees come to the United States, or take steps to improve U.S. relations with Cuba and living conditions in that country?”

Page 3 of 11 The construct: refugee The first step in the design of survey questions is to make sure that constructs are clear and their understanding is similar across respondents (Groves et al. 2009). When necessary, definitions can be provided to the respondent. In the case of refugees, survey questions rarely, if ever, clarify what a refugee is. A glimpse into the public understanding of the term is provided by a 1984 survey fielded by the United States Committee for Refugees (a now defunct not-for-profit corporation) and the firm Kane, Parsons and Associates, Inc. A question in that survey asked:

Let's think more specifically about three different terms used to describe people coming into the U.S.: refugees, legal immigrants and illegal immigrants. Which of these groups (refugees, legal immigrants, or illegal immigrants) do you think could best be described as (people who) are fleeing persecution for their political beliefs?

“Refugees” was selected by only 43% of respondents. Another question asked:

Which of the following statements best describe your knowledge and attitude toward refugees?

The distribution of responses was as follows:

a) You don't know enough to really have meaningful opinions on the issue of refugees: 25% b) You don't know very much about refugees but you do have opinions on these matters: 26% c) You know something about refugees but are not sure what policies our country should adopt: 36% d) You are well informed about the issue and have opinions about what should be done: 9%

e) Not sure: 4% Unfortunately, there has not been a more recent survey asking similar questions, but the situation in 1984 was striking. Most people did not know what refugees are or how they are different from regular migrants or from undocumented migrants. Furthermore, only one in ten reported to both be well informed and have opinions concerning policy. Considering the way questions about policy are sometimes worded—seemingly assuming that the respondent is a policy wonk—this precedent alone establishes quite a bit of uncertainty surrounding the interpretation of questions about refugee admissions and policy.

Only 1% of respondents chose “let the refugees come.” Given the choices the respondents were given, concluding that only 1%

• f the population supported accepting Cuban refugees would be the wrong conclusion.

Page 4 of 11 Overall support for refugee admissions (yes/no questions) Table 1 compiles questions that were formulated as a dichotomy between either admitting or not admitting refugees3. Specific question wording can be found in the appendix, but some examples are given below. As it can be seen, the questions refer to support for or opposition to specific refugee flows, but there is a great variety of ways in which these questions were asked.

3 At the moment, the table is slightly incomplete. Omissions will be resolved as work progresses.

Table 1. Support for refugee admissions, 1938 - 2016 Refugee flow Date Yes No DK/not sure N 1 German and Austrian political refugees (a) May 1938 23 67 9 5151 2 10,000 refugee children from Germany (b) Jan 1939 26 67 7 750 3 Bring British women & children by sea (c) Aug 1940 63 37 8 1500 4 Jews & other Europeans, proportionally (d) July 1946 40 50 10 1500 5 Jews & other Europeans, proportionally (e) Sept 1946 27 59 13 1500 6 Displaced by communists, 240,000 (f) May 1953 47 48 5 1291 7 Displaced by communists, "limited number" (g) Apr 1955 52 42 5 1226 8 Hungarians and poles fleeing communism (h) Nov 1956 26 57 17 1502 9 130,000 Vietnamese (i) May 1975 37 49 14 1428 10 Vietnamese, American foundational values (j) May 1975 49 40 11 1428 11 15000 Indochinese (k) Jul 1977 31 57 12 1515 12 Indochinese, American foundational values (l) Aug 1977 44 39 17 1515 13 Indochinese, American moral obligation (m) Aug 1977 17 71 12 1515 14 Indochinese ("boat people") (n) Aug 1979 32 57 11 1571 15 "Most" Cubans from the Mariel boatlift (o) Jun 1980 25 71 4 1517 16 Escaping communism (p) Jun 1986 51 39 10 1618 17 Haitians escaping military dictatorship (q) Feb 1993 27 63 10 1009 18 Cuban rafters, 1994 (Guantanamo) (r) Sept 1994 15 80 5 1161 19 Haitian rafters, 1994 (Guantanamo) (r) Sept 1994 19 77 4 1161 20 GSS, "political repression" (s) Feb 1996 42 23 36 1452 21 Kosovars, "as many as possible" (t) Apr 1999 32 60 8 750 22 Iraqis escaping Saddam Hussein's regime (u) Feb 2003 40 49 11 1078 23 Syrians, "humanitarian act" (v) Sep 2015 49 46 5 1001 24 Syrians, "increase" number of refugees (w) Sep 2015 51 45 4 1502 25 "At least" 10,000 Syrians (x) Nov 2015 28 67 5 1016 26 "At least" 10,000 Syrians (y) Nov 2015 37 60 3 1013 27 Syrian refugees, unspecified number (z) Dec 2015 42 52 6 1453 28 Prevent Syrian refugees from entering (ab) Dec 2015 42 50 9 1006 29 Ban Syrian refugees "at least temporarily" (cd) Apr 2016 43 49 8 572 * Sample sizes are sometimes approximate. In the January 1939 poll, the question was asked to a subsample

• nly and the reported sample size is that of the subsample. Same for GSS 1996.

Notes (a) through (cd) in the Appendix. They specify the exact question wording and the source for each.

Page 5 of 11

Row 2: “It has been proposed that the government permit 10,000 refugee children from Germany to be brought into this country and taken care of in American homes. Do you favor this plan?” (Gallup Poll (AIPO), January 1939). Row 9: “Do you favor or oppose allowing 130,000 Vietnam refugees to come to live in the United States?” (Harris Survey, May, 1975). Row 11: “There are still about 100,000 refugees who have fled communist governments in their home countries in Southeast Asia and do not have any country where they can go and resettle. Most of them, who are now in camps in Thailand and on boats without permission to resettle anywhere, would like to come to the U.S. to live. President Carter has proposed that 15,000 of these refugees be allowed to come here, in addition to the 15,000 already here. Do you tend to favor or oppose 15,000 more Indochinese refugees coming to live in this country?” (Harris Survey, Jul, 1977).

The cells highlighted in yellow show the most common response, which in the large majority of cases is a majority of responses. Two cells are highlighted in green to note that the “Yes” and “No” options swap meaning, since the questions were asked in terms of opposition rather than support—thus, in the two questions about Syrian refugees, the most common answer is in support of the refugee flow. Overall, it can be seen that, in the eight decades under consideration, Americans have been much more prone to

• ppose than to support the admission of refugees into the country.

Questions that evoke values From Table 1 one could quickly conclude that Americans do not generally support refugees, but a closer look at some questions raises the possibility of question wording exerting a large influence on the

• responses. After all, as shown earlier, average respondents are unlikely to know exactly what a refugee

is (and what their predicament is) and they also are, if the 1984 survey is still a good indication, unlikely to have formed opinions on refugee policy. If so, many of them are making up their minds when asked, providing an answer on the spot. Admittedly, many (perhaps most) of them may have been exposed to the issue by watching or reading the news, but if they have not formed a solid opinion, question wording can have an impact. The questions in rows 6, 7 and 16 in Table 1 explain that the refugees in question “have left their own homes because their countries were taken over by the communists” (row 6 and 7, Foreign Affairs Survey, May 1953 and April 1955) or “are escaping from political persecution in communist countries” (CBS/New York Times poll, June 1986). Undoubtedly, by explicitly referring to refugees as victims of communism during the Cold War years, these questions elicited more support toward the refugees than

Page 6 of 11 if they had been formulated without this reference. In turn, the question in row 20 (General Social Survey, February 1996) refers to refugees as having “suffered political repression in their own country,” while the question in row 23 asks about taking in “Syrian refugees as a humanitarian act” (Bloomberg Poll, September 2015). In these two cases, the questions evoke the defense of political freedom and compassion, values cherished by many respondents. In four of these five questions, respondents favor refugee admissions more than they oppose them, with the fifth being a statistical tie between the two

• ptions.

The questions in rows 10 and 12 evoke American foundational values. These questions ask respondents whether they agree with the following statements:

Row 10: “This country began with people of all races, creeds, and nationalities coming here to escape religious or political persecution, so we ought to let the refugees from Vietnam in” (Harris Survey, June 1975). Row 12: “This country was built on the humane principle of being a place where oppressed people could come to live, and Indochinese refugees should not be an exception” (Harris Survey, August 1977).

Many respondents would not be indifferent to the direct reference to the principles upon which the country was founded. The narrative of the origins of America as a safe haven for oppressed people has a solid appeal in the population and may explain the survey results in these cases, where more respondents supported rather than opposed the refugee flow. Consider, by contrast, the July 1977 question in row 11 (quoted at the beginning of the previous section). The refugee flow that question refers to is the same as in the question in row 12, but the question in row 11 (its reference to fleeing communist countries notwithstanding) did not invoke American foundational principles. Support for the refugee flow in row 11 is, consequently, much lower, with a solid majority in opposition. An interesting case is the question in row 13, where support for that same refugee flow (“boat people”) and in the same poll as the question in row 12, collapses to a meager 17% while opposition climbs to a staggering 71%. This question was formulated as follows:

Row 13: “Now let me read you some statements about the U.S. admitting more Indochinese refugees into this country. For each, tell me if you tend to agree or disagree... The U.S. was at least partly responsible for non-Communist people wanting to leave their home countries, so we should let all 100,000 of these refugees come to this country to live.” (Harris Survey, August 1977).

Page 7 of 11 This in an example of a two-barrel question (Groves et al. 2009) that detracts from its own validity, as the respondent is asked to agree with both an acknowledgement of U.S. responsibility for the consequences of the Vietnam War and support for refugees. Logically, one could deny U.S. responsibility while supporting the refugee flow. As the question is formulated, disagreement with either component of the question will lead to a “No” answer. Arguably, the questions in rows 10 and 12 above are also double-barrel questions. However, as very few respondents would disagree with the initial premise of both questions, “No” responses can still be interpreted solely as opposition to refugee admissions. I should note here that I am not arguing that there is a “true” response and question wording biases the respondents in a particular direction. This is certainly possible, in these examples and in any survey, but a case can be made for question wording helping to provide the respondent with more information or factors to consider. In the presence of questions that force the respondents to make up their minds on the spot about issues they are not quite knowledgeable about, a more elaborate question wording could play a positive role in eliciting a more thoughtful answer. Whether the “true” response is the knee-jerk response or the response given after more thought is perhaps an open question. Numeric questions and government action Survey respondents are sometimes asked to evaluate the magnitude of refugee flows. These questions generally offer a specific number of refugees to be admitted or that have already been admitted, asking respondents whether this number is “too many,” “about right” or “not enough.” Table 2 shows these cases.4 Again, the highlighted cells show the most common response. Notice, however, that in order to gauge total support for the refugee flow we should add the “about right” and “not enough” columns. This is shown in the column with the header “AR+NE”, where the public shows support for five out of seven refugee flows. One of the questions (row 3, on Vietnamese refugees) refers to a process already finalized, as it asks the respondents to evaluate the flow of Vietnamese refugees that arrived “in the years following the Vietnam war.” The rest of the questions refer to what were ongoing refugee crises at the time of the

4 I may have missed some cases here. They will be added as work progresses.

Page 8 of 11 surveys, except for the question on row 4. All the questions, except again for that on row 4, refer to specific government policies, which explains the specific numbers quoted in the questions. The question

• n row 4 (“In general, should the United States admit more political refugees, fewer, or about as many

as it does now?” from a CBS News/New York Times Poll from June 1986) refers to neither a then

• ngoing crisis nor government policy toward a specific flow.

Why does support for refugee admissions increase when respondents are asked to assess specific magnitudes or the question refers to specific government policies? I offer two possibilities, one linked to the response choices offered and the other one to question wording. Evaluation questions offer respondents an attractive choice: the middle road—in most of these questions, by reporting that the number of refugees is “about right.” This middle option, which by default supports the status quo (Krosnick 1991), gives those respondents who should otherwise consider the “don’t know” option a way to respond noncommittally without appearing ignorant or misinformed. This middle response effect is not new (Bishop 1987; Schuman & Presser 1996). The second possibility pertains to question wording. I would argue that, when people are simply asked whether they support or oppose refugee admissions, with no information about the refugees, uncertainty leads many to withhold support. If this is true, additional information could elicit more support, even if this additional information is as limited as to only quantify the flow. Interestingly, the actual number in the evaluation questions does not seem to matter. Ten thousand Syrians in 2015; 20 Table 2. Assessment of the magnitude of refugee flows. Refugee flow Too many About right Not enough Admit none DK AR+NE Total* 1 Dec 1956, Hungarians 35 48 11 7 59 101 2 June 1980, 60,000 Cubans 63 22 11 4 33 100 3 Apr 1985, >600,000 Vietnamese 36 34 15 15 49 100 4 June 1986, political refugees 39 43 7 11 50 100 5 Apr 1993, Haitians# 43 27 12 18 39 100 6 Apr 1999, 20,000 Kosovars 31 40 19 11 59 101 7 Sept 2015, 10,000 Syrians 17 35 21 24 3 56 100 * Totals may differ from 100 due to rounding. # The question here refers to how much the president "compromised", not to the number

• f refugees admitted.

AR+NE: About right plus not enough.

Page 9 of 11 thousand Kosovars in 1999; or 600,000 Vietnamese in the years after the Vietnam War, all get support from respondents. In addition, these questions make it clear to the respondent that what is being evaluated is a specific plan or policy, often already being implemented, and some respondents who would have otherwise voiced opposition may feel reassured by this certainty. In addition, the specificity of the questions informs the respondent that there is a plan in place, or a policy in action even when a number is not specified. In this respect, the question on row 1 is quite

• interesting. The question is: “Do you feel that the United States is letting in too many refugees from

Hungary, about the right number, or not enough?” (Foreign Affairs Survey, December 1956). In this case, the respondent takes notice that the United States has been taking refugees from Hungary already and is not given a number. Overall, Table 2, when compared to Table 1, suggests that, once a plan is in place, a number of respondents will tend to switch from opposition to support. The presence of a plan alone seems to reduce uncertainty, even if the respondent cannot possibly evaluate such plan on the

• spot. This suggests certain level of trust in government for the handling of these matters, contrary to

the general trend toward less trust in government seen in many aspects of public life in the United States (Pew Research Center 2015). Welcomed in local communities? The previous analysis is consistent with the results of a question, asked in several surveys, on whether the refugees would be welcomed if they settled in or near the respondent’s community. The question has been formulated in various ways, but the wording has been fairly similar across surveys, with the typical options being “welcomed” and “not really welcomed.” Table 3 shows the cases where this question was asked alongside the general question on support of/opposition to the specific refugee flow. In every single case, the respondents estimate that the refugees will be more welcomed than the level

• f support for their admission into the country suggests. This may just be showing that the respondents

take pride in their communities’ hospitality, but it is also consistent with the idea that part of the reported rejection of refugees stems from the uncertainty posed by point-blank yes/no questions with no additional information.

Page 10 of 11 Conclusions Their limitations notwithstanding, surveys give some insight into propensities to accept or reject refugees, ideological leanings, and the popular climate toward potential refugee admissions. The public has only a limited understanding of the concept of refugee, and regular survey respondents cannot be expected to be policy wonks or deeply informed about the complexities of every refugee crisis. Thus, their overall support for or opposition to specific refugee flows may be best interpreted as an indicator

• f the default political climate toward (additional) immigrant admissions. Or, in other words, the public

is reluctant to issue a blank check to government without information and reassurances that there is a plan in place to manage refugee flows in ways respondents would support. Further analysis gives support to this contention, or is at least consistent with it. First, support increases when questions evoke familiar values cherished by Americans, making the question more meaningful (and in a way, less uncertain). Second, the contrast between respondents’ support for refugee admissions and their (clearly warmer) evaluation of whether refugees would be welcomed in their communities, suggests that the uncertainty implied by point-blank yes/no questions plays a role in explaining the generally negative responses those questions get. Third, when respondents are asked questions referring to specific plans in place or in the works, or referring to refugees already admitted, responses become more positive, often completely flipping from opposition to support. If this is, as I suspect, a result of the reassurance respondents receive when given more information, I would then argue that there is plenty of room for policy-makers to formulate refugee policy and exercise leadership. Ambivalent survey responses suggest that political leaders can likely make the case

Table 3. Welcomed in or near the respondent's community? Refugee flow Question Yes No DK Total Accept 14,000/month 34 62 4 100 Would they be welcomed near you? 60 31 9 100 Relax policies to allow many in 32 57 11 100 Would they be welcomed near you? 57 30 13 100 Allow most to settle 25 71 4 100 Would they be welcomed near you? 42 48 10 100 Allow most to settle 15 80 5 100 Would they be welcomed near you? 36 55 9 100 Allow most to settle 19 77 4 100 Would they be welcomed near you? 31 62 7 100 Allow most to settle 37 60 3 100 Would they be welcomed near you? 49 46 6 101 July 1979 (Indochinese) August 1979 (Indochinese) June 1980 (Cubans) September 1994 (Cubans) September 1994 (Haitians) November 2015 (Syrians)

Page 11 of 11 for refugee-friendly policies. Conversely, the same ambivalence may turn into firm rejection of refugees if the political momentum is seized by actors hostile to immigration, who may fill the uncertainty vacuum with negative expectations. References Bishop, George F. 1987. “Experiments with the Middle Response Alternative in Survey Questions.” Public Opinion Quarterly 51(2): 220-232. Fowler, Floyd J. Jr. 2002. Survey Research Methods. Thousand Oaks, CA: Sage Publications, Applied Social Research Methods Series v. 1. Third edition. Fussell, Elizabeth. 2014. “Warmth of the Welcome: Attitudes toward Immigrants and Immigration.” Annual Review of Sociology 40:479-498. Groves, Robert M., Floyd J. Fowler Jr., Mick P. Couper, James M. Lepkowski, Eleanor Singer, and Roger

• Tourangeau. 2009. Survey Methodology, second edition. Wiley Series in Survey Methodology. John

Wiley & Sons. Krosnick, Jon A. 1991. “Response strategies for coping with the cognitive demands of attitude measures in surveys.” Applied Cognitive Psychology 5: 213-236. Krosnick, Jon A., and Abelson, Robert P. 1994. Pp. 177-203 in Tanur, Judith M., Questions about Questions: Inquiries into the Cognitive Bases of Surveys. New York: Rusell Sage Foundation. Newport, Frank. 2015. “Historical Review: Americans' Views on Refugees Coming to U.S.” Gallup. http://www.gallup.com/opinion/polling-matters/186716/historical-review-americans-views-refugees- coming.aspx (retrieved December 14, 2016). Pew Research Center. 2015. “Beyond Distrust: How Americans View Their Government.” Washington D.C.: Pew Research Center. Schuman, Howard. 1997. “Polls, Surveys, and the English Language.” The Public Perspective 8:6-7. Schuman, Howard, and Stanley Presser. 1996. Questions and answers in attitude surveys: Experiments

• n question form, wording and context. Sage Publications.