Direct all correspondence to Susan Welch. Professor Welch will share all data and coding for replication purposes. The original surveys and weights are located in the Roper Center’s data archives with the survey numbers given in the paper.
*For his considerable help with data management, I thank Adam Berinsky, MIT, who facilitated my acquisition from the Roper Organization of the weighted data sets analyzed here. I also thank Rebekah Young for her help with data analysis and Ron Filippelli, Alan Booth, and two SSQ reviewers for their comments on an earlier draft. Keith Gaddie was very helpful and encouraging.
Objectives: We investigate Americans’ opinions about European and American Jews between 1938 and 1945, the period from the height of Nazi domestic power to the end of the war in Europe. Methods: Several surveys of U.S. public opinion between 1938 and 1945, reweighted to reflect national population parameters, were examined to uncover both aggregate patterns of responses and predictors of pro-and anti-Jewish sentiment.
Results: We find that individuals’ social status, gender, partisan learning, and to some extent region affected their views on Jewish Americans and on European Jews.
Conclusions: Roosevelt’s policies of speaking out against Hitler’s atrocities, but yet doing nothing to facilitate more Jews to enter the U.S. as refugees, reflected the complexities of Americans’ opinions about Jews here and abroad but led to failure to provide a safe haven for those thousands of Jewish refugees who might have fled before the war.
American Public Opinion about Jews during the Nazi Period Passionate debate continues whether the U.S. should or could have done more to save European Jews from the Nazi murderers in World War II (recent publications on the topic include Beir with Josepher, 2013; Breitman and Lichtman, 2013; Medoff, 2013; Plaud, 2007; Rosen, 2006; Leff, 2005). That debate largely focuses on the beliefs and actions of the Franklin Delano Roosevelt, key members of Congress and the State Department, the press, and important lobby groups and other opinion leaders, from Father Coughlin and the America First Committee to the American Jewish Congress.
Analysis of public attitudes of the time has been based on inference from the views of these elite groups and on diaries, letters, and newspaper articles. And yet, scientific survey research in the U.S. began in 1935, alongside the debate over how to deal with Nazi Germany and the Nazi victims.1 Beginning in 1938, a few of these surveys asked Americans about the treatment of Jews in Germany and opinions about fellow Americans of Jewish descent.
These opinions have received very little systematic attention even though they are crucial to understanding the American public during the Nazi period.2 Longitudinal studies of American public opinion usually start in the 1950s. Analysts rejected earlier surveys because of flaws in sampling design and question wording (for a review see Berinsky, 2006, esp. pp. 500-505; 508-509).
Using these surveys with new weights that allow a much more accurate reflection of population paramenters (see Berinsky, et al, 2011), this paper investigates Americans’ opinions about Jews asked in seven national surveys between 1938 and 1945, the period when the Nazis were at the height of their domestic power to their time of defeat. We will examine whether elite opinion about Jews, as measured by public debates, reflect the views of the public. And, we are interested in how socio-demographic patterns of tolerance and intolerance map to those that we have seen in the post-war era.
Americans’ Attitudes toward Jews
It was not until after the war, when the horrors of the Nazis were more fully revealed and digested, that research on anti-Semitism began in earnest. Indeed, the interest in the part that “average” people played in the Holocaust has stimulated decades of significant research on anti- Semitism and its relationship to other forms of intolerance. Questions raised in the post-war period still influence the work of social scientists who study intolerance and prejudice. Adorno and his colleagues’ classic work, The Authoritarian Personality (1950), using the now famous F scale (Fascism scale) linked anti-Semitism to education, personality types, and a variety of intolerant attitudes toward other out groups and nonconformists. A few years later, another classic, Communism, Conformity and Civil Liberties: A Cross Section of the Nation Speaks its Mind (Stauffer, 1955) showed again that social and psychological factors helped explain intolerance for non-mainstream ideas. Generations of social scientists since then have examined and re-examined factors that these seminal works discovered. The link among anti-Semitism, education, and other forms of out-group prejudice has held up well (see Raden, 1999).
Specialized studies of Americans’ opinions about Jewish Americans are fewer. The author of an early compilation of survey results from dozens of survey questions about attitudes toward Jews before 1966 predicted that anti-Semitism would disappear as a problem (Stember, 1966). Later studies were not so sanguine, noting the persistence of anti-Jewish stereotypes (Selznick and Steinberg, 1969; Rosenfield, 1982; Wilson, 1996). Where predictors of these attitudes were examined, analysts found that anti-Jewish attitudes were most prevalent among those who were less educated and older and had lower incomes (Rosenfield, 1982) and, in more recent findings, were male or African American (Sigelman, 1995).
Jews in the Nazi Era
World War II and the Holocaust are still much in the news as we honor the disappearing “Greatest Generation,” including survivors of Holocaust horrors. A brief review of Nazi and U.S. policies will help situate our discussion of Americans’ attitudes during this time (the literature on World War II is voluminous, and so is the subset on the Holocaust. For more details on the latter see classics: Dawidowicz, 1975; Gilbert, 1985; Yahil, 1990; Mazower, 2008, Synder, 2010).
Nazi persecution of the Jews, begun with the Nazi control of the state in 1933, was codified in the Nuremburg laws of 1935 defining what a Jew was and prohibiting intermarriage of Jews and others. Beginning in 1933, German laws gradually excluded Jews from every aspect of economic and community life in Germany and from any legal protections, what Kaplan (1998) calls a “social death” (see the remarkable diaries of Klemperer, 1998). In 1938, Germany imposed similar anti-Semitic policies and actions when they annexed Austria (the Anschluss) and the Sudetenland.
With a few exceptions, the U.S. media only infrequently covered escalating Nazi outrages against Jews (Seelye, 2006; Leff, 2005; Lipstadt, 1995). 3 Though the New York Times published more than 1100 articles on some aspect of the Holocaust between 1939 and 1945, only six were on the front page, and many did not focus on the destruction of the Jews as a special target (Leff, 2005). Kristallnacht, theNovember 1938 Nazi destruction of thousands of Jewish businesses, homes, schools, and synagogues, murder of nearly 100 Jews, and arrest of 30,000, was the one instance of extensive media coverage of Nazi persecution of Jews (Friedlander, 1997: 270). Kristallnacht was extensively covered by the press and brought condemnation in many parts of the world, including from President Roosevelt and many U.S. Christian as well as Jewish congregations. These and other acts led Time magazine to choose Hitler as 1938’s “Man of the Year,” because he was “the greatest threatening force that the democratic, freedom-loving world faces today4.” The story included only two sentences about torture, robbery, economic dispossession, and street violence against Jews, and these sentences were in the context of persecutions of other groups. Hollywood, too, down played Nazi terrorism in this period (Kafka, 2013; Doherty, 2013; Urwand, forthcoming).
Though Hitler was increasingly frightening to many Americans, anti- Semitism, in the State Department, Congress, and among elements of the vocal public, played an important role in pre-War American foreign policy (see Wyman, 1984, and Rosen, 2004 for a taste of the anguished debate over why America did not do more for European Jews during this period; for a treatment of the issue in popular culture, see America and the Holocaust, 1992;United States Holocaust Museum, undated). America had a number of prominent home grown Nazis and anti-Semites, including the radio personality, Father Charles Coughlin, whose show reached millions of listeners (Stember, 1966: 111 States Holocaust Museum, undated). 5
Many members of Congress expressed strong isolationist views and wanted nothing to do with Europe’s conflicts. Massive unemployment during the Depression and then later fear of spies and saboteurs partly shaped anti immigrant views. Some elected officials and other public officials held anti-British or pro-German sentiments, but many expressed anti-Semitic views. Some prominent figures, like Charles Lindbergh and Joseph P. Kennedy, accused American Jews of pushing the U.S. toward war.6 These views insured that quotas for German immigrants, mostly used by Jews escaping from increased terrorism there, would not be raised. In 1939 Congress defeated a bill to allow 20,000 Jewish children into the country above the national quotas then in place, partly because of isolationist and anti-immigrant sentiment generally but also anti-Jewish sentiment particularly (Stember, 1966: 149). Indeed, observed one commentator, “Anti-Semitism was no stranger on Capitol Hill...it was, in fact, an important ingredient in the sharp hostility to refugee immigration that existed in Congress.... (Wyman, 1985: 14-15; see also Breitman and Kraut, 1987)”
Anti semitic views were also rife in the higher levels of the State Department (see for example Larson, 2011; Dodd and Dodd, 1941).The State Department constructed numerous roadblocks to potential immigrants fleeing from Nazi persecution (such as requiring a certificate of good conduct from the German police before processing visa applications from German Jews), and tightened requirements as time went on.
The minimal attention to the status of Europe’s Jews continued after the war began and throughout the conflict (Leff, 2005; Seelye, 2006; Novick, 1999; Lipstadt, 1995). War conditions, especially in the East, gave Germany free rein to pursue their extermination policies and made it difficult to detect even mass murders. Following the German invasion of Poland in September, 1939, squads of troopers whose mission was to kill Communists, Polish nationalists, and Jews followed the army through the countryside (Browning, 1992). After the German invasion of Russia in 1941, this slaughter culminated in the establishment of death camps whose sole purpose was to murder as many Jews as possible in the quickest time possible. The “social death” of Germany’s Jewish population turned into physical annihilation when Germany’s remaining Jewish population was sent to the camps in late 1941 and early 1942 (see Kaplan, 1942), followed by the Jewish populations of Holland, France, and Poland and its neighbors in 1942 and 1943, and ending with the deportation of most Hungarian Jews in 1944.
Though the full horror of the death camps was not revealed until after the end of the war, throughout 1942 increasing numbers of accounts of murder and deportation found their way to the U.S (see Wyman, Chapters 2 and 3; American Jewish Committee, 1942; 1943). In November 1942, a compelling account, finally accepted as true by the State Department and released to several major news outlets , confirmed that deportations were culminating in mass murder. Even the estimation that two million Jews had been killed already was reported only in small stories on the inside page (cf. Washington Post, November 25, 1942). In 1943 advocates for the potential victims still alive purchased advertisements in newspapers and gave pageants at several locations to publicize the murders and the need for action. Most Americans said that reports of mass killings of Jews did not cause any change in their own attitudes, but of the 15-20% who said it did, most were more sympathetic to the Jews (Stember, 1966: 143).
The murder of European Jews reached its peak in 1942 and early 1943 but continued to the last days of the war. Knowledgeable scholars believe that “In mid-March 1942, some 75 or 80 per cent of all victims of the Holocaust were still alive, while 20 or 25 per cent had perished. A mere eleven months later…, the percentages were exactly the reverse (Webber, 2009: 71).”
Data and Methods
Were Americans outraged by what they knew of the acts of the Nazis, did they agree, or did they not care? Did views change once the war began? To help answer these questions we examined the seven surveys tapping American’s attitudes toward Jews, a small subset of the more than 400 surveys conducted between 1936 and 1945 by the Gallup Polls (American Institute of Public Opinion), the Roper polling firm, the Office of Public Opinion Research run by Hadley Cantril, and the National Opinion Research Council (NORC).7 These first surveys were conducted using quota-controlled sampling. Rather than drawing a random sample from a carefully delineated population, quota sample designers decided a priori how many interviews they would obtain from different groups of people (men or younger people, for example) and then the interviewers could choose whom to interview as long as met their category quotas (see Berinsky, 2006 for a thorough analysis of this methodology). Quota sampling yielded an imperfect representation of the population from which the sample was drawn because it was impossible to adequately represent many groups and subgroups in this way (for example, the sample would include men in proportions reflecting their population, but this technique could not ensure appropriate representation of old men, or black men, or married men, or any other categories not part of the defined quota.)
In addition to errors introduced by quotas, the survey design introduced further errors by letting the pollsters choose whom to interview. Together, these methods led to a bias toward inclusion of more educated and better off individuals who were more willing to be interviewed or with whom the pollsters felt more comfortable. Gallup’s polls introduced a further bias because they sought to reflect not the adult population but the population of voters, thereby seriously underrepresenting women, southerners, and blacks, who voted at lower rates than others of that era.
Because of these sampling issues, most subsequent scholars have ignored these surveys despite their historical importance. But recently analysts (Berinsky, 2006; Berinsky and Schickler, 2011) have demonstrated the utility of these surveys if properly weighted to achieve an accurate reflection of population parameters. The polls used in their survey employ the weights developed to reflect the national population (Berinsky, 2006; Berinsky and Schickler, 2011). The cell weighting factors vary from survey to survey, depending on what variables are available.8
These weights increased the proportion of women in the early samples from the low 30 percent to around 50 percent; decreased the representation of people owning phones from around 50 percent to 40 percent, and increased the proportion in the sample who were pro-Roosevelt by 3 to 5 percent. The reweighting also reduced anti-Jewish proportions by 3 to 5 percent. These changes were consistent in direction over all the samples examined here. Though the changes in distribution of the dependent variables were not large, in combination with the changes in the distribution of the independent variables, they did produce some differences in findings between the weighted and unweighted data. The greatest advantage in using these weights is that we can be confident that the findings better reflect population parameters.
The Suvey Questions about JewsIn these early surveys, pollsters asked about opinions on Jews in America and in Europe only infrequently, and only two questions were repeated in more than one survey. We examine both pre-war questions, the first asked in 1938, (tables 1and 2) and those during the war (tables 3 and 4).
Though obviously we cannot go back to change the question wording in any secondary analysis, by contemporary standards these questions are flawed. For example, modern survey researchers who are trying to elicit rather than influence opinion balance both question stem options and answer options. So, for example, question C in Table 1 offers an unbalanced question stem: “Would you support a widespread campaign against Jews in this country?” Scientific polling would offer alternatives in the question, perhaps rephrasing as “Would you support or not support a widespread campaign against Jews in this country?
These questions also do not measure up to modern standards because there is little attempt to understand what the respondent knows before asking for an opinion about it. Questions A and B in Table 1 and question A in Table 3 illustrate this problem. Respondents are asked to evaluate policies of which most people probably know very little (persecution of Jews in Europe, Nazi treatment of Jews in Germany, and Hitler’s taking away power from the Jews). News coverage of what was going on was sparse, so most individuals’ knowledge was probably also very limited. Without attempts to measure knowledge, responses to questions like these probably measure overall sentiment toward Jews as well as opinions about policy.
Question C in Table 1 poses a similar problem. Each person interviewed probably has a different view of what a “widespread campaign against Jews in this country” means. Would the ends be further discrimination, segregation, prohibiting the practice of the Jewish religion, exile, or murder? Would supporting such a campaign, however defined, mean being sympathetic to it, giving money to a group organizing it, participating in rallies, burning down synagogues, or participating in violence and murder? Question D also reflects this problem in that there is no attempt to ask whether the respondent has any idea of how many Jewish exiles were currently entering the United States.
A third limitation is that the questions lack baselines against which to measure opinions about Jews. For example, in question B in Table 3, what does it mean that 45 percent of the sample think that Jews have too much power? Would views of other religious groups be similar? Other nationalities? Other races? We can only guess. Modern survey research would ask about a few other religious groups, mainstream and minority, perhaps about immigrants, and about racial groups, all to provide a context for views on any one group.
Despite these limitations, these questions are worthy of study because they tap important historical attitudes not accessible any other way. For the first time, researchers used scientific methods to learn what average Americans thought. Moreover, President Roosevelt himself followed the findings of the Gallup and Roper polls as one of the ways to learn what Americans were thinking, and, an advisor reports, used them to develop his own persuasive arguments (Gelderman, 1997).
Predicting Opinions about Jews In addition to examining overall responses to the questions about Jews, we want to see who is likely to be more and less sympathetic. Are those individuals who are part of the Democratic coalition (urbanites, Southerners, working people) more or less sympathetic? How do findings about attitudes during the Nazi years comport with post-war findings about anti-semitism that showed that education is associated with tolerance?
We have relegated details of our methods to the appendix, but here provide the outlines of our approach. We use probit, a technique allowing us to look at the simultaneous impact of a number of factors on an attitude. In terms of the factors we could examine, we ran into limitations in these early surveys just as there are limitations in the items used to tap views about Jews. By contemporary standards, these were very short survey instruments omitting much of what we now consider standard background information.
For example, although “party identification” questions are a staple of modern polling, in the 1930s this was not a meaningful concept to survey researchers. Surveys did ask about past presidential votes and in some cases projected presidential choice. However, these voting items resulted in many missing cases from those who did not vote, could not remember their vote, or were ineligible. In the pre-war years, a measure of support for Roosevelt was included, and we used that in our analyses. Unfortunately, the question was not asked in the wartime surveys we used.
And, in reverse, education was not asked about in any of the pre-war surveys used here, though it was assessed in the wartime surveys. Time and again, education has been found to be the key to more tolerant attitudes (cf. Smith, 1993; Schuman et al, 1992; Sullivan, et al, 1981; Stouffer, 1955), and we expect it to be related to sympathetic attitudes toward Jews here.
More professional occupations and higher income also are related to tolerance. Income was not included in any survey, but we were able to assess whether the head of household was in a professional or semi-professional occupation (the latter was used but not well defined). We also used the presence of a telephone in the home as another measure of economic status. The rate of phone ownership increased during this period, but even in 1945 fewer than half of households owned a phone (Fischer, 1992: 93; 107; 112). Phone ownership was, then, an indicator of economic status and nonfarm residence and perhaps is also a measure of social integration.
We expected that men would be more anti- Semitic than women. Though Stouffer showed that women are more intolerant of Communists and atheists, later research has found women to be more tolerant of other kinds of groups, including minorities, homosexuals, and those in need (Herek, 2002; Welch, et al, 2001; Wilson, 1996;Sigelman, 1995).
We also expected younger people to be less anti-semitic than older ones. Younger adults are more open to change on most social issues and are more tolerant than older people (see longitudinal research on changing attitudes on race, for example Hyman and Sheatsley, 1956; 1964; Welch and Sigelman, 2011; or contemporary polls on attitudes toward gays and gay rights such as http://www.gallup.com/poll/163730/back-law-legalize-gay-marriage-states.aspx).
We also created a measure designed to tap possible familiarity with Jews and Jewish issues. In the absence of questions focusing on that knowledge, our measure distinguished those individuals who lived in a community of 500,000 in one of the six states with a significant proportion of Jews (Linfield, 1941).9 The size of community variable, along with that of telephone ownership, might also be thought of as crude indicators of exposure to news.
Regional differences are endemic to understanding many political and cultural attitudes. We examined differences among three regions, broadly defined: the south, the non-Pacific north, and the Pacific coast. We were interested in the latter because of special hostility to Japanese Americans there that culminated with the internment of Japanese Americans during the war. We expected southerners and Pacific Coast residents to be less supportive of Jewish immigrants and southerners to be less sympathetic to Jews in general.