Introduction

The Pew Research Center’s Jewish Americans in 2020 [Pew Research Center (PRC) 2021a] represents the latest national sociodemographic study of the US Jewish population (cf. Massarik 1974; Kosmin et al. 1991; Kotler-Berkowitz et al. 2004a; PRC 2013). Pew’s study estimated the size of the US Jewish population and examined characteristics of the population, including intermarriage, Jewish child-rearing, engagement in Jewish communal life, and a range of religious and social attitudes. The study estimated that the total US Jewish population as of 2020 is 7.5 million adults and children (Hackett and Ausubel 2021; PRC 2021a), more than 12% greater than their 2013 estimate (PRC 2013). The present paper assesses the validity of these estimates. The population size, and who is included in the estimate, is not only an important statistic in its own right but also serves as the basis for understanding levels of Jewish engagement and differences across subgroups (Saxe 2022).

To develop their estimates of the size and characteristics of American Jewry, Pew employed a sophisticated design that relied on address-based sampling (ABS). This design decision reflected a departure from prior national studies that sampled households using random digit dial (RDD) telephone survey methods.Footnote 1 Not only was it the first national Jewish population survey to use ABS, but it was also the first to collect data primarily online.

Background

Because the US census does not include questions about religious or ethnic identity, sociodemographic studies of the US Jewry must rely on alternative sources of data. Demographic studies of American Jewry have evolved over the past decades in ways that reflect broader trends in social research. Understanding how methods have changed over time is critical for understanding the design decisions of the Pew 2020 survey and their effect on the results. Specific challenges to these studies included the problem of population undercounts, the lack of baseline population data, and survey coverage.

Table 1 displays a summary of the methods and results for five major surveys of the US Jewish population conducted between 1970 and 2020. The 1970 National Jewish Population Study (NJPS; Massarik 1974) was the first modern comprehensive study. Previously, the main source of systematic data was from a survey conducted by the US Bureau of the Census (1958) which had been designed as a test of the feasibility of including a question about religion on the decennial census. The Census Bureau decided not to include the religion question on future surveys, forcing researchers of religion—and those interested in policy related to religion such as government funding of parochial and religion-based charter schools—to design their own independent studies.

Table 1 Methodological characteristics and population estimates for national Jewish population surveys 1970–2020
Full size table

The NJPS 1970 employed what were then best practices in survey research: in-person surveys in which households were identified for inclusion in the sample through extensive fieldwork. Identification of Jewish households out of all possible households in the United States would have been impossible given the time and cost constraints. To conduct more cost-effective surveys, researchers used geographically stratified samples to focus on areas where they were more likely to find members of the target group. To develop a sample plan like this, however, required that one knew the percent of the population that was Jewish in any given region. One of the designers noted the dilemma attendant to the lack of US census data:

We sought from this survey that very piece of information required to design the survey creating a sort of circular situation with the connecting link missing. (Lazerwitz 1974, p. 2).

As a substitute for census estimates, researchers divided US counties into 39 regions on the basis of data about known Jewish communities as reported in the American Jewish Year Book (AJYB) (Chenkin 1969). Only 10% of the counties in the United States were included in the sample, with a majority of counties represented by what were estimated to be similarly sized Jewish communities. The 1970 survey yielded an estimate of the size of the Jewish population of 5.4 million (Massarik 1974). Despite improvements in estimation that arose from the use of scientific probability sampling, under-enumeration was considered to be a “universal problem” of sample surveys (Bershad and Tepping 1969), in particular for small sub-groups. Although a reanalysis of the data adjusting for sources of bias in the estimates indicated that the population estimate might have been as high as 6 million (Lazerwitz 1978), absent an independent benchmark, the extent of under-counting could not be determined.

The 1970 survey highlighted two challenges that each subsequent national survey, including Pew 2020, sought to remedy:

  1. (1)

    The lack of census data on the distribution of the population nationally necessary to develop efficient, cost-effective sampling plans

  2. (2)

    The universal problem of undercounts of small sub-groups in sample surveys

When the next national survey—NJPS 1990 (Kosmin et al. 1991)—was conducted, the cost-effectiveness of phone-based surveys provided an alternative to in-person household surveys. Although phone survey methods were more efficient, the developers contended with the same design dilemma as the 1970 survey, noting:

The best alternatives are surveys in which information on religious identification is collected. Three types of such surveys are relevant to our concern: (1) national and local omnibus surveys; (2) local studies of the Jewish population; and (3) a national Jewish population survey. (Goldstein et al. 1988, p. 3–4)

The use of national and local omnibus surveys that assess religious identification was dismissed in planning NJPS 1990 because researchers believed that any single survey would be insufficient to provide detailed descriptions of the population. The use of multiple sources of data was, at the time, too difficult to implement. Since 1990, however, data aggregation methods that enable modeling across different sources of data have evolved and become commonplace (Cooper et al. 2009; Lohr and Raghunathan 2017). As described below, the American Jewish Population Project (AJPP; Saxe et al. 2021) applied these latest methods to the assessment of religious identification, with a specific focus on Jewish identification.

Absent of census data for stratification and methods for data aggregation, the NJPS 1990 defined strata based solely on the distribution of the US population—for which there were census data. The assumption of this approach is that the probability of reaching a Jewish household when calling a representative sample of all households would be a measure of its representation in that region.

Ultimately, population estimates derived from the 1990 survey were based on an amalgamation of a year’s worth of weekly and biweekly surveys which were conducted as part of an ongoing omnibus market research survey. The omnibus survey was used to screen for Jewish respondents, who were recontacted for participation in an extended interview. The results indicated that there were 5.5 million Jewish adults and children in the United States. It was widely acknowledged that the 5.5 million in 1990 was likely an underestimate given the tendency to undercount small populations (Choldin 1994; Finke and Scheitle 2005), as well as a propensity for not reporting Jewish heritage in surveys (Kosmin et al. 1991, p. 39), and difficulties associated with counting recent immigrants (Kosmin 1991, p. 2).

The NJPS 2000–01 survey (Kotler-Berkowitz et al. 2004a) tried to improve the efficiency of RDD survey sampling by stratifying US counties on the basis of AJYB estimates, similar to the 1970 survey. A series of methodological problems undermined the utility of the findings. These included problems with the response rate (less than half that of the 1990 survey), likely an issue because the screener question about Jewish identity was open-ended and the hand-off from the screener to the full survey lost many respondents, and weighting for nonresponse was hampered because of a loss of screening data (Kadushin et al. 2005). Kadushin et al. concluded that while analyses of the relationship among measures in the survey were potentially valuable, the population estimates could not be considered reliable.

Each national Jewish population survey attempted to use the then latest methods to improve estimation, but in fact, the same challenges remained: a lack of pre-existing data on the population necessary to create an efficient, cost-effective sampling frame, and the proclivity toward undercounting in sample surveys. In addition, new challenges arose, such as changes in the systems for data collection, which made the collection of representative data even more difficult. The Pew 2013 survey was the first to attempt stratification based on estimates that were independent of known Jewish community data reported in the AJYB. Stratification was based on the frequency of Jewish respondents in prior Pew surveys. To reduce costs, Pew excluded nearly half of US counties where it was expected that there would be a low or no likelihood of Jewish households. Although some adjustments were made for the purposes of generating a total population estimate, the data applied only to a portion of the US population. Pew’s total population estimate was 6.7 million, substantially higher than all previous survey-based estimates. When compared, however, with what might have been the upper estimate of 6,000,000 in NJPS 1970, Pew’s estimate represented a very small change of just 12% over a 40-year period that included significant immigration from the former Soviet Union (FSU), Israel, and other countries, as well as the multi-generational impact of the post-war “baby boom.”

Pew’s 2020 survey employed new methods that had evolved in response to the ever-changing landscape of survey research. It was the first national Jewish population survey to design sample stratification using completely independent model-based estimates of the distribution of Jewish adults in counties from preliminary results of the AJPP (Tighe et al. 2019), with additional stratification by ZIP Codes using their own pre-existing data.

To address low response rates and potential bias with RDD phone surveys, Pew used address-based sampling (ABS). The US Census Bureau has long employed ABS using non-public address lists from the decennial census, while other surveys use the United States Postal Service’s Computerized Delivery Sequence (CDS) file of mailing addresses. Use of the CDS for sampling has greatly reduced the costs associated with household sampling. Advancements in the use of ABS for sampling have also been accompanied by changes in modes of administration. With fewer people relying on landline telephones and greater adoption of call-blocking technologies, multi-mode surveys are becoming more common. Participants are contacted by mail and invited to complete the survey online or by other means (mail/phone) (AAPOR 2016). The Pew 2020 survey made use of all of these advancements, stratifying postal addresses by counties and ZIP Codes based on independent Jewish population estimates, which covered known Jewish communities and areas outside of known Jewish communities. Invitations to participate in the survey were mailed to households who could choose to complete the survey online or request a paper questionnaire. Pew’s approach yielded the highest single survey estimate to date: a total population of 7.5 million Jewish adults and children.

Definition of Who is Jewish

The validity of this estimate and its comparability to prior estimates turns, in part, on whether the survey assessed the same underlying population as earlier studies. The key question is who was included in the population estimate and its relationship to prior operationalizations.

Sociologically, there is no official definition of who is Jewish in the United States. There is, however, substantial consensus among social scientists who study Jewish populations. The 2020 Pew survey included the same groups that have been included in US Jewish population surveys dating back to 1990. Specifically:

  • US adults who identified their religion as Jewish and who were either born Jewish or converted

  • US adults who identified culturally or ethnically as Jewish, were born Jewish (had at least one Jewish parent), and when asked about their religion, indicated that they were atheist, agnostic, or otherwise not affiliated with a religion

The inclusion of both groups—those who identify their religion as Judaism and those who identify ethnically or culturally as Jewish, but are non–practicing—is broadly recognized in the social-scientific study of Jews (cf., Hartman and Sheskin 2012; Rebhun 2004; Sasson et al. 2017; Sklare 1971). Judaism extends beyond the conception of religion as a set of beliefs and practices to include the idea of peoplehood—having a common ethnic and cultural identity. As described by Marshall Sklare (1971):

Being Jewish involves two complementary aspects: membership in the ethnic group and membership in the religious community. The extent to which the individual exercises his prerogatives by participating in the ethnic group affairs, and particularly in the life of the religious community, is a matter of choice. But no matter to what extent the prerogatives of birth are exercised, all Jews are essentially equal members of the ethnic group and of the religious community. (p. 26)

Thus, for the purpose of understanding the American Jewish community, sociodemographic studies consider all Jews “equal” and distinguish them from those who are non-Jewish. It is recognized that Jewishness exists on a continuum and identification can depend on the salience of a variety of social and cultural factors (cf. Horowitz 1998), but dichotomization into those who identify as Jewish and those who do not allows one to compare and contrast the attitudes and behavior of those who are Jewish-identified. To ensure that comparisons over time and across Jewish communities are meaningful, the basis for identifying Jews should remain stable and reflect an accepted construct.

To ensure such stability, Pew gathered input from a panel of experts (PRC 2021a, p. 5). Questions were designed in a way that would reduce survey error, in particular, sampling error associated with including only a portion of the population. The survey was conducted in two stages. In the first stage, members of the larger population were screened to identify eligible members of the target population. In the second stage, eligible respondents were asked detailed questions about their Jewish identification and the identification of household members.

Because the US Jewish population is ~ 2.5% of the total US population of 330 million, large numbers of screening interviews were needed to identify a sufficient number of eligible participants to be able to describe the population. Screening large numbers to identify members of a “rare” group can lead to problems of “false positives” (people who screen into the survey as Jewish who are not) as well as “false negatives” (people who screen as not eligible who actually are Jewish). The problem of screening people into the survey who are not actually Jewish is remedied by including more detailed questions about identification in the extended interview. The problem of false negatives is more problematic. Screening that is too stringent and excludes eligible members of the group can lead to undercounting the overall population or underrepresenting certain groups. To avoid false negatives, standard practice is to cast the widest net in the initial screening and use the detailed interview questions to remove false positives (Kalton 2009). Pew 2020 employed this method.

Another challenge associated with the screening process is the changing nature of religious identification in the US, which is the primary question used in screening. This is reflected in how questions about religion have changed over time. Akin to the observation from Herberg (1955) of the “triple melting pot,” it was common for questions about religion in the USA to be expressed simply as “Is your religious preference Protestant, Catholic, or Jewish?” (Fig. 1). Today, the question is likely to be asked similarly to how it is in Pew surveys: Not only are there more options for religion, there are also multiple ways to express “no religion.” Thus, in the United States, where it is increasingly acceptable to say that one is not religious (Thiessen and Wilkins-Laflamme 2020), and there are different varieties of “no religion” to choose from, secular Jews are likely to choose one of the "no religion" options.Footnote 2 This trend is similar to that noted by Himmelfarb (1980), who predicted that the proportion of Jews not identifying denominationally would increase with each generation. For studies of the US Jewish population to accurately reflect the two complementary aspects of ethnic and religious identity, screener questions cannot limit identification to only a question about religious identification.

Fig. 1
figure 1

Change in religious preference question in surveys, 1948 to 2020. Note: ANES, 1948 (Campbell and Kahn 1948); Growth of American Families, 1955 (Freedman et al. 1959); General Social Survey, 1972 (Davis 1973); Pew Research Center, 2020 (PRC 2021b).

Full size image

Pew’s screening consisted of three questions (PRC 2021b):

  • What is your present religion, if any?

  • Anyone who identified as Jewish when asked about their religion was screened into the long interview.

    Anyone who did not indicate that their religion was Jewish was asked:

  • Aside from religion, do you consider yourself to be any of the following in any way (for example, ethnically, culturally, or because of your family’s background)? Jewish/Catholic/Mormon/Muslim

  • Anyone who indicated that they considered themselves to be Jewish was screened into the long interview.

    Anyone who did not indicate that they considered themselves Jewish was asked:

  • Please indicate whether you were raised in any of the following traditions or had a parent from any of the following backgrounds. Jewish/Catholic/Mormon/Muslim

  • Anyone who indicated that they were raised Jewish or had a Jewish parent was screened into the long interview.

Similar screening questions were used in prior national surveys, including NJPS 1990 (Kosmin 1991), NJPS 2000–01 (Kotler-Berkowitz et al. 2004b), and Pew’s survey of Jewish Americans 2013 (PRC 2013), as well as local Jewish community sociodemographic studies (see, e.g., Aronson et al. 2021; 2022). Although the screener questions opened the Pew 2020 survey to a broad group, not all of these respondents were included in population counts. Only those who identified as Jewish, were raised Jewish or had at least one Jewish parent, and were not currently practicing another religion were included. Those who had a Jewish parent and did not identify as Jewish were considered separately as people of Jewish background, as were those who identified as Jewish but were practicing another religion. All of the national surveys used these criteria for who was included in the population statistics, with the exception of NJPS 2000, which included, in addition to those who were practicing no religion, a small number who were practicing non-monotheistic religions such as Buddhism.

An exception to this definition was introduced by DellaPergola, who has criticized the Pew estimates (both 2013 and 2020; see DellaPergola 2022a, 2022b; Rosner 2021). DellaPergola argues that Pew broadened the definition of who is included in the US Jewish population estimates. This criticism was first introduced in response to Pew’s estimate of 6.7 million Jewish adults and children in 2013 (PRC 2013); DellaPergola argued that Pew included a “previously not empirically tested category of partly Jewish” in their total population estimate (DellaPergola 2015). This was based on Pew 2013 having included a “partially” option to the second screener question about whether one considered themselves Jewish. Pew provided three options: “Yes,” “Yes, partially Jewish (includes “half-Jewish”), and “No.” These options, however, have been used before. NJPS 2000–01 used the same options with the same screener question: “Yes (definitely)”, “Yes (half/partially),” “Yes, Other (Specify),” and “No” (Kotler-Berkowitz et al. 2004b, p. 5). In NJPS 2000–01, the core Jewish population included those who chose the “half/partially” option as well as those who chose the “definitely” option. It is unclear why DellaPergola revised population estimates from the Pew 2013 survey to exclude all those who said partially (and their children) when previous studies and population estimates had included them as part of the Jewish population.

DellaPergola argues that those who chose the “partially” option were individuals who considered their Jewish identity “part of a broader composite cluster of two or more sub-identities” (DellaPergola 2022a, p. 380). He defines the core on the basis of the concept of a “mutually exclusive identification framework” in which those who have converted or practice another religion are not included in the core population; in addition, only those who identify ethnically or culturally as Jewish aside from religion are excluded if they identify with another ethnicity. These groups are considered to be part of what he calls the Enlarged Jewish population. The exclusion of those who have converted or are practicing a religion different from the main Jewish population statistics is common in social scientific studies of the population. The exclusion of those who might identify with multiple ethnicities is without precedent. It is also never directly assessed by DellaPergola.

In response to Pew’s 2020 estimate of 7.5 million US Jewish adults and children, DellaPergola operationalized the idea of the mutually exclusive identification framework as excluding those who identify with another religion—as has been common—and, only for those who identify as Jewish aside from religion, excluding those who have only one Jewish parent, with the assumption that they, therefore, do not have a mutually exclusive identification framework. Applying this more restrictive definition, DellaPergola revised estimates of the US Jewish population to 6 million, 20% smaller than Pew’s estimate.

Traditional Jewish law (halacha) follows matrilineal descent, and thus, there is no precedent for excluding those with Jewish mothers (half of those excluded by DellaPergola), with the exception of those who practice another religion. With respect to those with Jewish fathers (absent a Jewish mother), the largest religious denomination of American Jews (Reform) accepts patrilineal descent. Thus, excluding those with a Jewish parent even though they self-identify as Jewish and are not practicing another religion is counter to current Jewish practice. There is no precedent for doing so in sociological surveys. It not only distorts the snapshot provided by Pew of the contemporary Jewish population but makes it difficult to compare with past surveys of the US Jewish population. DellaPergola does not provide revised estimates of the Jewish population from earlier studies using his novel and more limited definition of who is a Jew.

Inclusion of those who are “partially” Jewish, or have just one Jewish parent, is not unique to studies of the US Jewish population. The national Jewish community survey of Australia, “The Gen17 Australian Jewish Community Survey,” (Graham and Markus 2018) included all those who saw themselves as “Jewish in any way at all” (p. 22). This is true, as well, for population estimates, which, in Australia, are based on the census which collects data on religion and ancestry, with the latter—along with language spoken at home—indicative of ethnic identification (Graham 2021). All those who identify as Jewish by religion or by ancestry or language are included in population estimates (Graham 2014; Markus et al., 2021). The national Jewish community surveys of the United Kingdom (Graham et al. 2014) and South Africa (Graham 2020) also included all those who identified as Jewish in any way. In addition, in a nine-country study of European Jews, the “core” Jewish population included all those who identified themselves as Jews, or were identified as Jewish by a household member, and did not currently identify with another religion (Graham 2018).

Review of the conceptualization of the construct and how it was operationalized indicates that Pew’s definition is consistent with past national social surveys of Jews. There is no indication that Pew extended the definition. Nonetheless, DellaPergola’s revised estimates serve as the official statistics for worldwide estimates of the US Jewish population.

The Jewish Identity of Jews Aside from Religion

The premise, from NJPS 1990 to present, is that asking about identity aside from religion is a necessary and more direct method to assess the Jewish identification of those for whom membership in the ethnic group has precedence over membership in the religious community. The inclusion of questions about identity aside from religion also reduces survey error by ensuring that these members of the population are not excluded from sampling. DellaPergola has suggested that the majority of those who identify as Jewish in response to the religion question hold secular views; therefore, asking additional questions beyond religion only serves to include people who “lack the bravery (or interest) to concede they are just Jewish [sic]” (2022a, p. 361).

To assess the validity of the group Pew identifies as Jewish aside from religion, we compared this group directly with those who responded that their religion was Jewish and identified as secular/not practicing when asked about denominational identification. The group that identifies their religion as Jewish and denomination as secular/not practicing, definitionally, most directly represents the similar construct of individuals for whom Jewish identification is not based on its religious aspects. This group’s Jewish identification in the religion question might be attributable to a holdover from the concept of the triple melting pot, where even though they identify as secular or not practicing, it has been ingrained in American culture to identify as Jewish when asked one’s religion even if one is not practicing.

Among the 3,825 US adults who completed the extended interview after having screened into the survey by reporting that their religion was Jewish, only a very small percentage reported that they were not practicing, secular, or culturally Jewish (1.4%), or that they were atheist or agnostic (0.2%) (Table 2). Although some who identify with a denomination might hold secular views, those who self-identified as secular/not practicing in their denomination should represent the same underlying construct as those who identify as Jewish aside from religion. Both groups, those who self-identified as secular/not practicing in their denomination and those who identify as Jewish aside from religion, should represent those for whom Jewish identification is less a matter of religion than ethnicity or culture.

Table 2 Distribution of self-identified secular Jews among those who say their religion is Jewish
Full size table

Figure 2 displays the results from analysis of the question about whether one sees being Jewish mainly as a matter of religion, ancestry, or culture, or a mixture. Responses were recoded into those who saw it mainly as a matter of culture or ancestry and those who saw it mainly as a matter of religion, or a mix of religion and culture or ancestry.

Fig. 2
figure 2

Being Jewish is mainly a matter of culture and/or ancestry for Jews aside from religion compared with Jews by religion who are secular/not practicing and other Jews by religion (Orthodox, Conservative, Reform, and other denominations)

Full size image

As expected, the vast majority of those who identified their religion as Jewish and indicated that they were secular/not practicing (80%) and those who identified as Jewish aside from religion (85%) agreed that being Jewish was mainly a matter of ancestry and/or culture, which is consistent with the definition of this group representing those for whom Jewish identification is mainly about membership in an ethnic group. As well, a majority of those who identified their branch as non-denominational (69%) and Reform (54%) also indicated that being Jewish was mainly a matter of ancestry and culture.

Analysis of the question “How important is religion in your life?” yielded similar results (Fig. 3). Nearly all who self-identify as secular/not practicing (97%) or Jewish aside from religion (92%) indicate that religion is not very or not at all important in their life. Conversely, the majority of all denominationally identified groups report that religion is important in their life, with the highest among those who identify as Orthodox (96%) or Conservative (80%).

Fig. 3
figure 3

Importance of religion by Jewish identity

Full size image

Figure 4 displays engagement in ritual practices on a scale from low to high across four common ritual practices: attending or holding a Seder on last Passover, fasting on the last Yom Kippur, marking Shabbat in a meaningful way, and keeping a kosher home. The scale is cumulative, meaning those in the highest category engage in all four of these practices.

Fig. 4
figure 4

Ritual engagement scale by Jewish identity

Full size image

The majority of those who said their religion was Jewish and are secular/not practicing and those who identified as Jewish aside from religion were likely to engage in no practices or just attending a Passover Seder (57% and 73%, respectively). There were, however, individuals across all groups who engaged in no, or few, ritual practices. Conversely, there were individuals across all groups who engaged in many or all of the practices.

Although engagement in ritual practices is not definitional for Jewish identification, these comparisons—along with attitudes about the importance of religion and the meaning of being Jewish—highlight how Jewish identification being more a matter of ethnicity or culture than religion can be seen across all branches and is typified by Jews aside from religion. Further, the similarities of this group to those who identified as Jewish when asked about religion and then indicated that they were secular/not practicing makes clear that inclusion of questions about identification beyond religion provides a better representation of the portion of the population for whom being Jewish is more a matter of ethnicity than religion.

Comparison to Recent Population Estimates

Additional evidence that Pew’s assessment of US Jewry accurately reflects the underlying construct as recognized by the social science community is demonstrated through convergent validity—the extent to which the results compare with other estimates of the population obtained through other methods. The Pew 2020 estimate of a total US Jewish population of 7.5 million is consistent with other contemporaneous estimates from independent researchers (Table 3). These include the AJPP, which estimates a total US Jewish population of 7.6 million (Saxe et al. 2021) and the AJYB’s most recent estimate of 7.3 million (Sheskin and Dashefsky 2022).

Table 3 Pew 2020, AJPP 2020, and AJYB 2021 estimates of the US Jewish population
Full size table

Rather than a single survey, the AJPP estimate of 7.6 million is based on a statistical analysis of primary data from hundreds of surveys of the US population (Magidin de Kramer et al. 2018; Saxe and Tighe 2013; Saxe et al. 2021; Tighe et al. 2021). Each survey includes a question on religious identity that is consistent with the religion screener question in Jewish-focused surveys. Population estimates are based on responses from over 1.3 million US adults about their religious identity and are supplemented by analyses of recent local Jewish community studies. The AJPP approach provides an independent check on the validity of Pew’s estimate. The Pew 2020 estimate of 7.5 million, although slightly lower than AJPP’s, is within the “margin of error” of the AJPP estimate, represented by the 95% confidence intervals. The Pew 2020 95% confidence interval ranges from 7.3 to 7.8 million. The AJPP 2020 95% Bayesian credible interval ranges from 7.47 to 7.8 million.Footnote 3

Sheskin and Dashefsky’s AJYB 2021 estimate of 7.3 million is slightly lower than Pew 2020 and AJPP 2020, but the difference is explicable. AJYB’s estimate is based on a simple summing of reported population counts primarily from federated Jewish communities throughout the United States. This estimate should be somewhat lower than the Pew and AJPP estimates since Pew and AJPP estimates are designed to cover the entire United States, whereas the AJYB estimates are focused on known Jewish communities and do not include systematic assessment of areas outside of known Jewish communities. The timing and frequency of the studies used to create the AJYB estimate varies widely. The editors provided some adjustments to the community estimates for population change over time, but most of the community studies were conducted more than a decade ago. Because the AJYB estimate is a simple aggregation of published population totals, which does not take into account the sample variance in each study, the margin of error is unknown.

Although Pew and AJPP appear very similar in their overall estimates of the US Jewish population (7.5 versus 7.6 million), there are differences that can affect conclusions about the size of the population. In particular, Pew estimates a smaller percentage of US adults who identify their religion as Jewish (1.7%) compared with the AJPP estimate (1.9%). Pew also estimates a larger percentage of US adults who identify as Jewish “aside from religion” and a greater number of Jewish children. The total population appears similar because the gains and losses of these subgroups balance out. If Pew underestimated the percentage of US adults whose religion is Jewish, then the total population would be larger than the 7.5 million reported. If the AJPP underestimated the percentage of US adults who are Jewish aside from religion, their 7.6 million would be larger.

There are no independent sources of data with which to validate estimates of Jews aside from religion or the number of Jewish children. Absent independent sources, AJPP researchers adopted a conservative position and developed estimates of these two groups from secondary analysis of recent local Jewish community surveys. Where no recent community studies had been conducted, AJPP used estimates from the Pew 2013 survey, which was the most recent national source of data available when the analyses were conducted.Footnote 4

The major benefits provided by AJPP are the detailed data it yields about the majority of the US Jewish population who are identified through the religion screener question, and its ability to describe differences sub-nationally, such as between states and metropolitan areas. The AJPP was specifically designed to address the ongoing challenge associated with studies of the US Jewish population because of the lack of census data (Saxe and Tighe 2013; Saxe et al. 2014). Here, we examine responses to Pew’s screener question and whether there might have been bias in who chose to answer the screener after receiving the invitation to participate in the survey addressed to them personally at their home address.

The estimates of the percentage of the US population who identify their religion as Jewish in the Pew 2020 survey and in the AJPP data synthesis are displayed in Fig. 5. For comparison, results from a sample of notable recent surveys are also displayed along with their 95% confidence intervals. This includes a sample of recent surveys conducted by Pew, such as their Religious Landscape Surveys (RLS, PRC 2008, 2015), Global Attitudes Survey (GAS, PRC 2010), and previous survey of Jewish Americans (JA, PRC 2013). In these recent Pew surveys, estimates ranged from 1.7% to 1.9%. All were RDD phone surveys and included both landline and cellphone samples. Other national surveys, such as the General Social Survey (GSS, Smith et al. 2019) and the American National Election Study (ANES 2019, 2021), have estimates of the US adults who identify their religion as Jewish that range from 1.7% to 2.1%. These surveys employed address-based sampling similar to Pew 2020 but were conducted in person.Footnote 5 In the Cooperative Election Study (CES), a web-based survey with a sample of 60,000 US adults, 2.1% of US adults identified their religion as Jewish. For nearly all of these surveys, the AJPP estimate of 1.9% and the Pew estimate of 1.7% were within the 95% confidence intervals of the estimates.

Fig. 5
figure 5

Percentage of US adults who say their religion is Jewish in recent surveys in comparison to Pew 2020 and AJPP 2020

Full size image

None of the weights in these surveys—aside from the Pew 2013 survey of US Jews—were designed for the specific purpose of US Jewish population estimation. Most include in their poststratification weighting standard variables such as age, race, Hispanic ethnicity, and educational attainment. Interactions between these variables are not included. Geographic adjustments are often based to the broadest geography of US census region. None include interactions of demographic characteristics and geography, which are needed to account for the fact that, for example, the likelihood that a Hispanic adult in the United States identifies as Jewish is not equal throughout the United States. Another example is that, although Jewish adults generally have higher levels of educational attainment than US adults, the likelihood that a college graduate is Jewish differs substantially in places like Manhattan versus areas within Brooklyn.

The AJPP model-based estimate standardizes the factors typically included in weighting to account for differences between the surveys in terms of who is represented in the samples of each survey and whether the factors are actually associated with Jewish identification. These factors include geographic distribution at and within the ZIP Code clusters as well as interactions of geography with demographics, such as age, educational attainment, and race/ethnicity (see Tighe et al. 2021, Chapter 3 for details). Such interactions typically are not possible to estimate reliably in single surveys given small sample sizes.

When differences in the samples in terms of their geographic distribution, their representation of men and women, age, race/ethnicity groups, and educational attainment are accounted for, the surveys converge on an estimate of 1.9%, with a 95% credible interval ranging from 1.9 to 2%. This suggests that the Pew estimate of 1.7% with a 95% confidence interval ranging from 1.6 to 1.8% likely somewhat underrepresented the true percentage of US adults who identify their religion as Jewish. Most importantly, Pew’s estimate of 1.7% is not an outlier relative to other surveys; rather, it is within the range of what an individual survey might be expected to produce given the vagaries of survey sampling and the challenges of using a survey to enumerate a rare population.

Another benefit of the AJPP data synthesis is that it includes a variety of survey sources and sampling methods. This enables assessment of the validity of the Pew estimates regarding whether survey mode affected their estimates of the size of the population. The Pew 2020 survey differed from previous surveys in that it was conducted primarily online, and participants were contacted by mail at their home addresses. Pew suggested that the web/mail survey might be more effective at reaching young people or others more comfortable with technology, whereas traditional phone surveys might be better at reaching groups not comfortable with technology or those that lacked access to the internet (PRC 2021a, p. 13). Although Pew conducted a mode experiment (PRC 2021b, pp. 239–247), it was designed to assess how mode affected responses to particular questions and not how mode might have affected its effects on population estimates. Because the AJPP data synthesis includes traditional RDD telephone surveys as well as web and in-person surveys, whether survey mode yields differences in estimates of the Jewish population can be directly examined. Typically, differences such as those described by Pew of survey mode affecting whether younger or older people are more likely to be reached are remedied through the poststratification weighting and adjustments for non-response. Mode, therefore, should not affect population estimates.PRC (2021) has been changed to PRC (2021a, 2021b).

Figure 6 displays unweighted, weighted,Footnote 6 and model-based estimates by the methods employed across the full sample of surveys included in the AJPP data synthesis. The unweighted data from RDD telephone surveys yielded higher estimates of the percentage of Jewish adults than cellphone surveys. Applying the weights provided with each survey, RDD landline and cellphone estimates were reduced. For example, in Pew RDD landline surveys, unweighted, 2.7% of adults were Jewish. After weighting, the estimate was reduced to 1.9%. The unweighted GSS and ANES surveys administered in-person saw the 1.7% of adults who identified their religion as Jewish increase to 1.9% after weighting. When weighting was standardized through model-based poststratification, the estimates across all sampling methods and modes of administration converged at 1.9–2%. It does not appear that Pew’s use of address-based sampling and web administration would yield a lower estimate of the percentage of US adults who say their religion is Jewish compared with RDD phone surveys.

Fig. 6
figure 6

Unweighted, weighted, and model-based estimates of percent Jewish in RDD, cell, web, and in-person surveys

Full size image

Conclusions

Estimating the size of the US Jewish population, a small percentage of the total US population, is a challenging but essential task if one is to understand the dynamics of contemporary Jewish religious and cultural engagement. Pew’s 2020 study of Jewish Americans estimated the US Jewish population to be 7.5 million. Our review indicates that Pew’s estimate is valid and consistent with other recent systematic analyses.

The questions used to screen for eligible members of the population, and identify eligible members based on follow-up questions in the extended interview, are consistent with past approaches to assess the sociodemographic characteristics of American Jews. The design of the questions also conformed to the practice of assessing the underlying construct of Jewish identification as one comprised of both ethnic and religious group membership, regardless of the extent to which identification is exercised in overt behaviors related to the group.

The findings document that those who identified as Jewish aside from religion represent the same underlying construct as those who said their religion was Jewish and then explained that they are secular/not practicing. There was no difference between these two groups in the belief that being Jewish was mainly a matter of ancestry or culture. Those who identified as Jewish aside from religion were also similar to those who identified as Jewish by religion and were secular/not practicing in their level of engagement in ritual practices. Further, regardless of whether one identified as Jewish by religion or as Jewish aside from religion, there was a range of Jewish engagement from none/very little to very much across all of the groups.

Contrary to the suggestion that Pew changed and expanded the definition of who is a Jew (DellaPergola 2022a), the definition has remained consistent. The implications of applying DellaPergola’s new definition not only result in a smaller Jewish population but also yield a very different narrative of American Jewish identity and engagement (Saxe 2022). A reanalysis of Pew’s 2020 data, excluding non-religious adults who have only one Jewish parent, results in an increase in the proportion of Jews by religion to 87% of the total population (compared with 73% in original analysis) and a decrease in Jews aside from religion (13% compared with 27% in original analysis). Were one to accept DellaPergola’s exclusion of children of intermarriage who do not identify religiously, American Jewry would appear as highly religious and Jewishly engaged, while excluding a large number of people who identify as Jews and participate in Jewish home and community life. That narrative is at odds with how most Jewish communities view their constituents and the diversity of engagement with Jewish life (Aronson et al. 2019).

Although not fully realized in NJPS 1990, the switch to screening representative samples of US adults to identify eligibility on the basis of religion and on the basis of identification aside from religion paved the way to circumvent the circularity inherent in Jewish population surveys. Although there is no independent source of data on the US Jewish population available to inform the design of and evaluation of survey samples, there is an abundance of data on the religious identification of US adults. With the majority of US Jewish adults identifying by religion, these data are highly useful. In this case, the comparison of the independent model-based estimates of the percentage of US adults who identify their religion as Jewish based on the analysis of hundreds of independent samples of US adults provided evidence that the percentage of American Jews who identify their religion as Jewish observed in the Pew 2020 survey was somewhat low.

It is possible that the lower percentage reporting their religion as Jewish might simply represent sampling error. However, it also may be more substantive. The survey made clear that it was about religious identification and participation. Even a broad screening process designed to reduce undercounting cannot remedy the possibility that some members of a rare group who consider their membership sensitive will deny or refuse to answer questions about group membership (Kalton 2009). One of the key findings from the Pew 2020 survey must be considered: 75% of Jewish Americans said there is more antisemitism in the United States than there was five years ago (PRC 2021a, p. 122), and 53% said they felt less safe as a Jewish person in the United States than they did five years ago (PRC 2021a, p. 124). The share was even higher among those who identified their religion as Jewish. Concerns about antisemitism might have been particularly important in the Pew study because of the use of ABS and the fact that respondents were contacted at their home address. This is a very different context than the general population surveys included in the AJPP data synthesis, in which the focus of the surveys was on a variety of topics and religious identification was one of several standard demographic questions included (typically at the end of the survey).

That Pew’s population findings, confirmed by multiple other sources of information and analytic approaches, point to population growth is important. However, demographic facts will not determine the future of the American Jewish community. That future will be shaped by a host of endogenous and exogenous factors. The findings from Pew 2020 are a reminder that, similar to past surveys, there are limitations to relying on a single sample. Having an external frame of reference, such as the AJPP data synthesis, both for the design of the survey and for evaluating the representativeness of the sample, will be critically important. At the same time, local community studies will continue to be essential for understanding variations at the state and county level, and the ways in which specific constituencies are affected by developments affecting the broader Jewish population.