This article aims at reconstructing the size of US Jewry between 1945 and 2010, looking ahead to 2020, using the disciplinary concepts and tools of demography (see also DellaPergola 2005). Population size is a sensitive indicator of a broader array of demographic, socio-economic, and cultural processes. The focus of this paper in no way wishes to minimize or undervalue other possible analytic paths to the study of the US Jewish population. We rather aim at outlining the logical processes and factual conclusions that can be reached by using the tested, transparent, and replicable tools of demographic research.

This paper also wishes to stress a necessary transnational/global analytic perspective in the study of contemporary Jewish communities by contextualizing the demography of Jews in the United States within a broader set of comparable processes. Such broader outlook is admittedly quite unusual in a literature that has predominantly watched American Jewry from the inside, somewhat disjointed from its longer-term historical, social and cultural origins and affinities, and incidentally, with relatively little attention paid to the nature and effects of changes in the general US demographic environment.

In this spirit, at the outset we review three research issues, fundamental but often neglected in scientific and public discourse alike: (a) the nature of documentation available, (b) the nature of the investigated population, and (c) the nature of demographic processes at stake. We then review some of the main demographic trends that have affected American Jewry over the past decades in comparison with two other relevant populations: the US total population and the Jewish population of Israel—the world’s other major Jewish community. We summarize the main competing estimates on US Jewish population size; offer a new critical reading and consistency check of some of the better known among these sources; and finally suggest one or more scenarios for US Jewish population size in the present and short-term future.

Sources of Data

Since the end of World War II, the debate on American Jewish demography has been built on a great variety of data sources (DellaPergola 2005; Goldstein 1981; Rosenwaike 1980) (see Fig. 1). Each source came with its specific advantages and disadvantages, and a specific trade-off between quality and cost. In an American constitutional environment that does not allow a question on religion in its decennial census, the only national study undertaken by the Bureau of the Census including information on religion was the 1957 Current Population Survey (CPS) (United States Census Bureau 1957, 1968). The survey offered the high reliability of a very large sample assembled by the top professional authority, at no cost to the Jewish community at a time when Jews could be effectively and quite unequivocally defined through the simple (and gender-insensitive) question: “What is his religion?” But governmental involvement in a census-like, large-scale survey including a direct question on religion was not politically correct or feasible, and was never repeated. Public-sponsored bodies have occasionally undertaken surveys on specific topics that included information about religion, based on smaller samples.

Fig. 1
figure 1

Main approaches and sources in the study of US Jewish population

Full size image

Special national surveys undertaken by Jewish initiatives, such as the three National Jewish Population Surveys (NJPS) (Kosmin et al. 1991; Kotler-Berkowitz et al. 2003; Lazerwitz 1978; Massarik 1974) sponsored by the US central Federations organization in 1970, 1990, and 2000–01 (here for brevity 2000), or the 2001 American Jewish Identity Survey (AJIS) (Mayer et al. 2001) paid for by private funds, have constituted the main poles of reference in the recent assessment of American Jewish demography. These efforts had the advantage of obtaining large samples of Jews approached with consistent procedures and through detailed questionnaires, but involved growing investments of Jewish money and statistical biases of various and not always controllable types and amounts. Among the latter, response rates were relatively low and reflected respondents’ self-selection by degrees of Jewish community involvement. All in all, while probably technically inferior to its predecessor NJPS 1990, even the much debated NJPS 2000 (Kadushin et al. 2005) was described by a panel of senior independent referees (Schulman 2003) as a usable source, albeit one that “may have produced an estimate of the Jewish population that is slightly lower than that reported by…other surveys.” The reviewers also recommended “follow-up research to gauge the extent to which the Jewish sample may skew toward Jews who are more religiously identified and who reside in completely Jewish households.” Yet, “these issues will likely have little impact on the analysis of relationships between variables in this dataset. Analysis of these relationships will provide valuable insights.” Interestingly, the simultaneous and competing AJIS provided readings of American Jewry very similar to NJPS, and such consistency should have received far greater attention than it actually did (Perlmann 2007b).

Among the main alternatives that have been suggested to national Jewish population surveys, one is culling country-wide Jewish population estimates drawn from compilations of local estimates. This was undertaken across nearly a century by the American Jewish Year Book. Some of those local estimates lacked true research foundations. When grounded on actual Jewish community studies (Sheskin and Dashefsky 2010), these relied on quite different survey methods, namely sampling randomness and population definitions, choices of variables, survey personnel, fieldwork timing, and quality of investment. Documentation and insights on local contexts was sometimes better than that obtained from national studies, but it traded-off with the local studies’ lack of synchronization, conceptual consistency, and national coverage. Coverage of diverse urban areas at different times tended to generate problematic double counts given the intense inter-city and inter-state mobility typical of Jews in the United States (Groeneman and Smith 2009; Rebhun and Goldstein 2006). Focus on larger Jewish communities missed the peculiarities of smaller communities, thus introducing significant biases on national Jewish population profiles. The local-studies approach, therefore, provided important insights on cross-country diversity but insufficient foundation for national population estimates.

Another research direction is meta-analysis utilized by the Steinhardt Social Research Institute at Brandeis University (SSRI). This important project uses a large database of national and regional surveys, each including relatively small subsamples of Jews from which a large merged Jewish sample can be obtained (Saxe et al. 2006; Tighe et al. 2005, 2009b, 2011). This approach, too, synthesizes sources of diverse quality and scope, but lacks time synchronization and contents homogeneity, usually focusing on adult respondents with scant information on child population, and little or no depth on Jewish identification variables. The latter omission critically weakens the ability to establish the boundaries and size of a relevant Jewish subpopulation which tends to become increasingly associated with multi-dimensional constructs (see below). Synthesis of many sources requires very lengthy and complex data processing where statistical assumptions and procedures tend to overwhelm the direct survey evidence. Closer scrutiny of the meta-analysis unveils a systematic bias of over-coverage of Jews in general sample surveys (Tighe et al. 2011, Appendix Table B4). Out of 131 surveys analyzed for the years 2000–08, 105 reported initial over-sampling of Jews, while 19 reported under-sampling, and 7 about-right sampling. These acknowledged biases can and are only partially corrected by survey undertakers, their sample-weighting being based on the known characteristics of the total population—mostly sex, age, and US general geography—but not on the less-known characteristics of the Jewish population. To the extent that such surveys carry inherent biases regarding Jewish population characteristics and size, the conglomeration of numerous sources into one comprehensive database does not reduce errors but rather compounds them. Further challenges in using and understanding these data for national Jewish population estimates are specified below.

Further sources of data, such as samples grounded on commercial directories or other address compilations like the Knowledge Networks’ KnowledgePanel® (Knowledge Networks 2010) may include abundant data on Jewish behaviors and attitudes but tend to be highly unrepresentative of other socio-demographic and attitudinal characteristics. These studies may be useful in assessing relationships between variables, but their respondents’ structure is so biased and imperfect that in no way can they serve the purpose of estimating Jewish population size. For example, a 2010 Survey of American Jews (Tighe et al. 2011, pp. 26–27) found a US Jewish population of which an exceedingly high 63 % did not indicate a preferred Jewish denomination, with only 17.5 % Reform, 13.6 % Conservative, 5.1 % Orthodox, and 0.8 % Reconstructionist; 68.1 % were not synagogue members; and 42.5 % of the total married had a non-Jewish spouse, which would correspond to a rate of out-marriage around 70 % among younger marriage cohorts. Such a profile would probably fit a population broadly defined in terms of Jewish origins, inclusive of large sections of the uncommitted, uninvolved, uninterested, and unidentified (see below)—in any case not comparable with Jewish population profiles according to other major surveys mentioned above.

Some other sources and techniques of data collection look promising, but they have not yet been tested, such as a general national survey with a significantly enhanced Jewish subsample (possibly being considered by the Pew Research Center for 2013), or a national longitudinal panel involving repeated interviewing of the same panel, or of a periodically renewed sample (Perlmann 2007a). It is to be hoped that these new research paths will be explored and pursued soon.

To complete the picture, data on Jewish population or on selected sections of it were periodically collected through compilations of vital or migration statistics (HIAS annual), or of people included in specific institutional situations, like children in Jewish schools (e.g., AVI CHAI 1994). These sources could provide competent observation of the specific topics investigated but were not designed to provide a picture of the whole Jewish population, its size, and trends. Finally, the lowest level in the knowledge pyramid is reliance on local observers, outside any methodology or control—sometimes used in past estimates of local Jewish population size.

As noted, each data collection method carries its advantages and disadvantages, but when it comes to national Jewish population estimates, one point should be stressed unequivocally. Some sources, like the CPS or the various NJPS surveys, were initially designed to produce, among other things, a national estimate of the US Jewish population, while some other sources were not designed for that purpose. No matter how sophisticated, the effort to extract national Jewish population estimates from data never intended to provide that piece of information, though valid for other purposes, cannot solve the problem nor can claim consensus among investigators.

Definitions

As noted, a crucial bone of contention in Jewish population research concerns the definition of target population—both conceptually and in terms of data-processing. Multiple and overlapping identities and group allegiances normally characterize all individuals, but on top of that, in highly open and mobile societies, of which the United States provides the ultimate paradigm, group identities and boundaries tend to become increasingly fluid and porous. When it comes to symbolic as well as personal and social identities, enhanced proximity among different groups and other interactions—namely intermarriage—enable more frequent coexistence (or syncretism) of competing identities within the same person. This is particularly true in the case of Judaism whose nature is inherently multilayered, much beyond its conventional reduction to a religious or ethnic category. The consequences for quantification cannot be underestimated. One serious limit of most Jewish-sponsored population surveys is that they tend to investigate Jewish identification in some detail but fail to devote attention to the coexisting non-Jewish identifications of the same individuals. General social surveys mostly ignore these subtleties and allow a growing grey area of unknowns, refusals, or uninterested in the ethno-religious identification fields. It follows that group boundaries, hence subpopulation sizes, are increasingly determined by the active involvement and definitional assumptions of investigators.

The three constructs of the core, enlarged, and Law of Return Jewish populations have been reviewed elsewhere in greater detail (DellaPergola 2010a). The core concept, originally introduced by the 1990 NJPS analysts (Kosmin et al. 1991), addresses the self-declared or otherwise identified aggregate of persons of Jewish origin who do not hold an alternative religious identification. The enlarged is the total household population of core Jewish individuals, regardless of current personal identification. Israel’s Law of Return—determining the virtual pool of people eligible for immigration and immediate citizenship—provides a definitional concept inclusive of the total of first, second, or third-generation descendants of Jews and their spouses, regardless of personal identification (see Fig. 2). An even broader identificational conceptual circle includes anyone whose ancestors ever were Jewish. Such cases may be unveiled in studies of population genetics that retrieve DNA profiles and mutations back in time to the original ancestors (Adams et al. 2008; Behar et al. 2010; Hammer et al. 2000; Nebel et al. 2001) or in anthropological or legal studies of descendants of former converts out of Judaism (Corinaldi 2001).

Fig. 2
figure 2

Configuring contemporary Jewish populations

Full size image

A closer and more relevant resolution of the core population addresses and incorporates (a) those who identify as Jews by religion, (b) those for whom Jewish identity is declared but defined in terms other than religion, and (c) those who do not initially declare themselves Jewish while being fully eligible for the core criterion but can subsequently be incorporated for analytic purposes—like the Persons of Jewish Background (PJBs) in 2000 NJPS (Kotler-Berkowitz et al. 2003).

Some people may decide to permanently change their group allegiance, thus creating a difference between those born Jewish and those Jewish now, the balance being accessions of Jews by choice and secessions of Jews opting out. But the more recent experience also points to increasingly numerous and frequent back-and-forth lifecycle passages of the same individuals across several sub-categories of the core Jewish population, and more intriguingly, across the whole gamut of the core-non-core typology (Horowitz 2003).

With all due caution and caveats, Jews can be counted at any given moment in time, and they can be counted through mutually exclusive definitional criteria that avoid the double count of multiple identities. This is the underlying hypothesis of the present paper and one of the leading clues to understanding disagreements between different analysts. One is well aware that the constant turnaround of individuals across identificational categories tends to create much broader aggregates of individuals ever pertaining to any given definition, or pertaining to more than one definition at a given point in time (e.g., carefully compare the Jewish population typologies in New York, as defined by Cohen et al. 2012). The sum of the respective population sizes would determine a grand total far larger than the total existing population in the geographical unit of reference, such as the United States. Ideally, having already noted the lack of genuine longitudinal sources, comparisons of population size and characteristics over time should relate to findings about the same persons at different points in time and under the same definitions—no matter how complex—rather than be an artifact of technical decisions taken by analysts and not directly by the surveyed actors themselves.

Demographic Drivers

One quite surprising conclusion that emerges from perusal of the recent literature on American Jewish demography is the scarce reference to the broader social context within which Jewish demographic trends unfold. De-contextualization is one of the major weaknesses of the current debate where sometimes a “new number” appears without reference to either previous population trends within the group investigated, or the general demographic patterns of the surrounding society. Entire generations of literature sensitive to these issues seem to have been erased from discourse (just to quote the work of the profession’s dean, see Goldstein 1969, 1981, 1989, 1992), as if a new American Jewry had been created out of the blue, suddenly becoming the subject of assessment.

As against this pattern, one of the fundamental tenets of demography is that population change depends on a limited number of well-defined drivers, each of which reflects a non-random array of process-specific determinants (DellaPergola 2011) (see Fig. 3). Obviously, US Jewry is not to be understood as a purely physical entity but rather reflects complex interactions between biological, demographic, social and cultural determinants and consequences. The assessment of the main direction and plausibility of social and demographic trends cannot be severed from the prime frame of reference of occurrences within American society at large, and from the complementary outlook of global Jewish society of which American Jewry is one very peculiar manifestation. Population patterns, the explained variable, in turn feed back into the continuing process as one additional explanatory variable.

Fig. 3
figure 3

Processes affecting Jewish population size and composition

Full size image

Within these relevant contexts, Jewish population transformations unavoidably derive from the initial size and structure of the community (under a certain set of definitions) and from the intervening changes over a given period of time due to: (a) the vital balance of births and deaths, (b) the mobility balance of incoming and outgoing migration, and (c) the identificational balance of those who choose to join and leave the Jewish group. Each of these factors responds to a broad variety of social transformations affecting the specific group at stake as well as all other sectors of society; and to intervening (or proximate) variables, including policy interventions, which operate within each process-specific constraint. This is why total population growth and decline cannot be adjudicated unless the mode of operation of each of the several components of change is carefully understood within its own logic—partly unique to the population group studied, partly shared with, if not generated by interactions with, other population groups. Demographic trends will not be dealt with without devoting special attention to the role of gender (Hartman and Hartman 2009).

Nature and Consistency of Past US Jewish Demographic Trends

Under these binding terms of reference, we briefly re-examine the main processes at work in US Jewish population change over the last decades. For reasons that will become clearer later, throughout this review we stress parallel trends occurring in Israel, the other major concentration of Jews worldwide.

International Migration

International migration historically functioned as the main engine in the foundation, quantitative growth, and socio-cultural characterization of American Jewry. After World War II, albeit comparatively less than in earlier periods, Jewish immigration continued to play a significant role in the United States, reflecting major out-migration mostly from European countries, especially during the initial stages of the major exodus from the (former) Soviet Union (HIAS annual). The latter was eventually curtailed by more stringent immigration policies in the United States (Tolts 2007). Overall, Jewish migration constituted a sharply declining share of total US immigration, especially after the new legislation introduced during the mid-1960s significantly expanded the catchment of American immigration to Latin American and Asian countries (Miron 2009). Periodical Jewish influxes also came from nearly all Latin American countries, South Africa, Western Europe, the Middle East, namely Iran, and Israel (Cohen 2009). When compared with the parallel influx of Jewish migrants to Israel, the United States overall received far lesser quantitative impact (Fig. 4). At the same time, self-selection brought to the United States comparatively better educated Jewish immigrants who had the additional economic advantage of an age composition with lower dependency ratios (DellaPergola 2009).

Fig. 4
figure 4

Jewish international migration balance; US and Israel—1948–2010. Source: Author’s processing of NJPS 1970, NJPS 1990, NJPS 2000; DellaPergola (2005), HIAS annual, Israel Central Bureau of Statistics (2011)

Full size image

The timing of major Jewish migration waves was quite similar in both main receiving countries, primarily related to international migration policy shifts in sending countries like the Former Soviet Union (FSU) which in turn reflected major changes in the global geo-political system. On the other hand, Israel experienced higher emigration rates than the United States—whose aliyah propensities were the lowest of any country worldwide (DellaPergola 2011)—but Israel’s immigration rates were significantly higher. Hence, unquestionably, the Jewish net migration balance was significantly larger in Israel than in the United States.

Fertility

With diminishing Jewish international migration, the balance of births and deaths gained enhanced relevance for population growth and size. Birth and death rates reflect, respectively, intrinsic levels of fertility and survivorship in a population, and—sometimes crucially—its age composition. In turn, age composition tends to reflect demographic trends among one or more previous generations. Jewish fertility in the United States was clearly tuned to the massive highs and lows of American fertility in general, only at a consistently lower level. During the most recent period, total American fertility tended to be stable or moderately declining, but it mostly owed its resilience to the input of population groups among which Jews were very little represented: Latinos and African Americans, with Whites and Asians at much lower levels (United States Census Bureau 2012). Jewish families tended to catch and interpret these general changes of mood, lifestyle, and norms earlier and faster than most of the total population, the reasons being related in part to different socio-economic structures, in part to different value systems and psychological perceptions. A further important factor affecting the effectively Jewish fertility rate, the number of Jewish children born to their parents, was the affiliation of children of intermarriage (Barack Fishman 2004; Phillips 1997; Dashefsky and Heller 2008). Most of the time, this further eroded an already scarce level of intergenerational replacement. Several scholars, using different data and tools, all demonstrate that current US Jewish fertility, measured through the Total Fertility Rate (TFR) stood (DellaPergola 1980; Schmelz and DellaPergola 1983) and stands (Barack Fishman 2004; Goujon et al. 2012; Hurst and Mott 2006) at around 1.5 children. Only for a short period at its peak during the late 1950s was US Jewish fertility higher than Jewish fertility in Israel at its lowest in the early-mid-1990s (Fig. 5).

Fig. 5
figure 5

Current total fertility rates: US Jewish and total population; Israel Jewish—1947–2010. Source: Author’s processing of NJPS 2000, and interpolation 2001–2010; DellaPergola (1980, 2005), Schmelz and DellaPergola (1983), United States Census Bureau (2012), Israel Central Bureau of Statistics (2011), Goujon et al. (2012)

Full size image

Given the pivotal role of fertility and the birthrate in the demographic equation, it is particularly important to scrutinize the available Jewish evidence over the last decades. An important caveat here is that all calculations of detailed rates rely on relatively small sample sizes which are subject to significant statistical variation. Looking at data within their confidence intervals may substantially alter the analytic perceptions of demographic trends. Therefore it is eminently important to submit the data to a consistency check by repeated observation of the same variables and their relations across successive data sources, namely the three NJPS surveys of 1970, 1990, and 2000. The lifetime number of children reported by women belonging to the same birth cohort at three different points in time not only strengthens our perception of the substantive process, but also provides a reliability test of the data sources themselves (Fig. 6). The display of highs and lows in US Jewish fertility (regardless of Jewish identification of children) in three successive NJPS surveys is extraordinarily consistent, which offers much wanted proof that NJPS 2000 covered the very same American Jews included in 1990 and in 1970 and effectively described their unique demographic patterns. With the already noted exception of the baby-boom years, final family size ranged between moderate to low levels mostly below the two children required for generational reproduction. Evidently, the youngest cohorts had not completed their reproduction course at the time of the last observation in 2000. However, their initial performance, including a diffused propensity to postpone marriage, was so slow and low that in order to approach anything close to two children, practically every Jewish woman aged 30 to 40 would need to have a child during the next ten years (2000–10). This would not be a realistic scenario knowing what we know about the status of the American economy during that decade and the actual decline of US fertility toward the end of that period (United States Census Bureau 2012).

Fig. 6
figure 6

Lifetime fertility of Jewish women (including all marital statuses by Birth Cohort, US, 1970–2000. Central values of sample variability range). Source: author’s processing of NJPS 1970, NJPS 1990, NJPS 2000; Kotler-Berkowitz et al. (2003)

Full size image

On top of this, longevity and life expectancy among Jews in Israel (82 years of life expected at birth in 2010 among Jews—Israel Central Bureau of Statistics 2011) tended to be consistently higher than among the total population in the United States (78 years for the total population—United States Census Bureau 2012). In an earlier past, evidence on American Jewish-non-Jewish mortality differentials pointed to a conspicuous advantage among Jews, related to several health-related, cultural, and lifestyle factors. But over time these differentials tended to diminish if not disappear (Goldstein 1986). Consequently, fertility-mortality differences, compounded with a much younger age composition among Israelis, determined much lower population growth rates for American Jews.

Conversions

In the United States the balance of accessions to and secessions from Judaism is much debated among those who stress the steady and enriching inflow of new adepts to the community and those who admonish about the constant erosion of the existing stock through assimilation. Much of this discussion reflects the normative positions of the discussants. Inasmuch as empirical evidence existed, it quite consistently pointed to a negative balance between total accessions and secessions (Smith 2009). One recent source, the 2007 Survey of Religion (Pew Forum on Religion & Public Life 2008), compared the percentages of those raised Jewish with those currently Jewish among the US total population (see Fig. 7). At least in terms of Jews by religion, the lifetime balance was unequivocally negative and equaled about 0.2 % of the country’s total population. Assuming the same effects among children as among adults, this would amount to a net lifetime loss of about 600,000 individuals identifying as Jews by religion, or well above 10 % of a total Jewish population estimated at between 5 and 6 million (see below). It is true that some of these passages occurred from/to the unknown/unreported/agnostic group, rather than from/to another specified religion group. But such data disprove the assumption of a significant ongoing passage from the more peripheral toward the more central areas of the Jewish identificational typology (outlined in Fig. 2) that would be fueling an increase in the declared Jewish population.

Fig. 7
figure 7

Adults joining and leaving Jewish group as percent of US total population, 2007. Source: Pew (2008)

Full size image

In Israel, in spite of the high profile debate about conversions procedures, since the late 1990s to date over 50,000 persons—largely Ethiopians but also from the FSU and other countries—were converted to Judaism by Orthodox rabbinical authorities (DellaPergola 2011).

All in all, the demography of American Jewry must be interpreted not only on its own but against relevant comparisons. In the United States, total population increased by 13.1 % between 1990 and 2000, and by 9.7 % between 2000 and 2010. Of the total increase between 2000 and 2009, 64 % was due to the balance of births and deaths, and 36 % to the international migration balance. Among total Whites, the contribution of major components of population growth was quite the same—63 % and 37 %, respectively (United States Census Bureau 2012). In Israel repeated immigration waves and strong natural increase plus conversions produced steady and persistent Jewish population growth—25.5 % between 1990 and 2000, and 17.1 % between 2000 and 2010. Of the latter, 88 % was contributed by natural increase, 11 % by net international migration, and 1 % by conversions (Israel Central Bureau of Statistics 2011).

Between 1950 and 2010, the US Jewish population experienced a significantly lower total volume and yearly rate of increase than the US total population, and even more so, than Israel’s Jewish population. This reflected less positive or negative balances of the three components of demographic change: international migration, vital statistics, and—where applicable—conversions (or more broadly accessions/secessions). In the United States, periods of more rapid Jewish population growth following higher birthrates in the ten to fifteen years following World War II, and again during the years of enhanced immigration during the late 1970s and early 1990s, were interspersed with periods of near stagnation due to low Jewish birth rates, rising intermarriage rates and assimilation, less immigration, and population aging.

Reassessing and Comparing Data Sources

The foregoing review provides the necessary background when turning to compile and compare national Jewish population estimates in the United States. Before looking at the scene after World War II (see Fig. 8), a few quick observations are worth mentioning regarding the preceding period. American Jewry dates back to the 17th century, but its growth to global prominence started in the 1880s with the inception of massive migration waves from Eastern Europe. By 1900, with more than 1 million Jews, the United States had the third largest Jewish community in the world after the Russian and Habsburg Empires. During the 1920s and 1930s demographic growth continued even after the imposition of stringent immigration quotas and the rapid decline of fertility levels, thanks to the persistence of a young Jewish population composition. It should be noted, however, that the propensity to exaggerate Jewish population estimates was already present in the main documentation source of the interwar years, the US Census of Religious Bodies, which provided decennial estimates mostly based on the unchecked testimonies of local informants (Schwartz et al. 2002). More rigorous survey methods were only making initial steps in relatively small local Jewish communities (Robison 1943).

Fig. 8
figure 8

Population estimates, US Jewish and total; Israel Jewish—1945–2010 Source: Various sources and author’s (DP) processing

Full size image

The total US Jewish population in 1945 was realistically assessed by Pennsylvania demographer Ira Rosenwaike at 4.4 million (Rosenwaike 1980). Between then and 1990, across some high-low gaps, all the main available estimates tended to agree on the general direction and speed of change. Relatively rapid growth until the late 1960s was followed by slower growth during the following twenty years, and stagnation or incipient decline soon after. As against this broadly shared understanding, the subsequent fifteen years between 1995 and 2010 were characterized by widely different population estimates and perceptions of the direction of change.

The whole set of available sources relies on very different approaches to estimating population. National compilations of local Jewish communities historically gathered by the American Jewish Year Book, and largely based on local reports, regularly stood at the top of the range, through periodical downward corrections that reflected newly accumulated evidence. The independent critical reviews by Rosenwaike until the mid-1970s (Rosenwaike 1980), and subsequently by the Institute of Contemporary Jewry (ICJ) (DellaPergola 2010a), regularly provided somewhat lower estimates that also manifested smoother change over time.

A highly coherent time sequence was provided by several forward–backward projections that tried to find whether the various national surveys could be logically related to each other through a set of assumptions inferred from the findings of the same surveys on international migration, age composition, marriage, and fertility (DellaPergola 2005). Thus, in light of the then ongoing and expected demographic trends, the over 5 million found in the 1957 CPS quite accurately predicted the approximately 5.4 million of the 1970 NJPS, which in turn predicted the over 5.5 million in the 1990 NJPS. Both the 1970 NJPS (Schmelz and DellaPergola 1983, 1988) and 1990 NJPS (DellaPergola 2005; Rebhun et al. 1999) predicted the 2000 NJPS and AJIS results. Both these surveys initially assessed the US Jewish population at 5.2–5.3 million within reasonable confidence intervals and margins of statistical error not only for the total Jewish population but also for each birth/age cohort.

Such population projections were obtained through very detailed matrixes compounding all changes occurring in each five-year period, within each sex and five-year age group, for each of the relevant components of population change: incoming and outgoing international migration, fertility rates of women at relevant ages, death rates, and accession/secession rates. The crucial baseline for such a procedure needs to be a detailed age-sex composition of the population at stake, which, as noted above, is available for only some of the data sources about US Jewry. In the absence of such a detailed baseline, no reliable projection is feasible. In probabilistic terms, these models have to be understood as suggestive of a more likely central value out of a broader range of higher or lower estimates of lesser likelihood (see below).

The puzzle of a growing discrepancy between population estimates during the more recent period reflects several problems. First and foremost is the inconsistency of Jewish population definitions adopted by different sources. For example, the 2001 HARI study (Tobin and Groeneman 2003) clearly used a broader definition than NJPS and AJIS in the same year. AJIS used the same definition as NJPS 1990; NJPS 2000 used a broader definition. On the other hand, the three ARIS surveys (Kosmin et al. 2001; Kosmin and Keysar 2009), after incorporating estimates of the child population and a proportional allocation out of the steadily growing share of those among the US total population with religion none, unknown, or not reported, provided comparatively lower Jewish population estimates. However, the direction of change over time was the same as other national estimates, pointing to decline toward and after 2000. A similar trend appeared from comparing over time repeated returns of the NORC General Social Survey (Smith 2011).

Here we will note again that Jewish national surveys, with their detailed information on individual identification characteristics, offered good opportunities to assess the grey zones around the more clearly declared Jewish population core. In Jewish-sponsored surveys, along with generally lower response rates, significantly fewer than in general surveys would readily admit their Jewishness when defined in terms of religion. On the other hand, quite a few respondents who in the first place would not seem to belong to the core Jewish population could be recovered and incorporated through detailed reading of personal family and life histories. General surveys, based on population classification by religion, do not offer the same maneuvering opportunity—hence resolution of the non-declared parts of the Jewish core becomes largely conjectural. A sure mistake would be to attribute in general surveys the same rate of non-response/unknown/agnostic as found in Jewish surveys.

Facing all other sources, the Brandeis SSRI meta-analysis estimates provided a different time series, starting at a level similar to NJPS in 1990, with an estimate higher than NJPS in 2000, and one definitely higher than most other estimates in 2010 (Tighe et al. 2011). Likewise, the most recent national compilations of local Jewish community estimates pointed to a rapidly expanding American Jewry (Sheskin and Dashefsky 2010). That the latter increase was very implausible (at least under constant population definitions) is demonstrated by a comparison of these higher estimates with the abovementioned pace of growth of the US total population and of Israel’s Jewish population—represented in Fig. 8 on different scales. Following our discussion of the different components of change in the United States—among Jews and in general—and in Israel, it does not stand to reason that between 2000 and 2010 American Jewry would be growing faster than the US total population and at a pace similar to that of Israel’s Jews, as implied by the higher estimates.

To sharpen this point, the population data in Fig. 8 were translated into decennial percents of growth (see Fig. 9). The growth rates suggested by the SSRI meta-analysis and, to a lesser extent, local community summations were clearly anomalous, and contradicted any empirically grounded Jewish demographic or identification process. The possible contention that more Jews were now “coming out of the cabinet” was disproven by empirical evidence (Pew Forum on Religion & Public Life 2008). Because of their combinative nature, both meta-analysis and local community summations risked amassing significant amounts of errors and biases all along the way when trying to estimate national Jewish population—a product they never were designed to supply in the first place.

Fig. 9
figure 9

Decennial percent population growth, US Jewish and total; Israel Jewish—1950–2020. Source: Fig. 8

Full size image

To further clarify the obstacles met in the process of data evaluation, we now review step-by-step some selected data sources, demonstrating how much caution should be exerted if the information is to be correctly read and understood. Jews clearly constitute a very small fraction of the US total population, inherently exposed to large sampling errors in general surveys. Pedantic precision is required when projecting survey percentages to population estimates; whereas 1 % of the US total population is 3 million individuals, 0.1 % is 300,000 individuals, and every 0.01 % is 30,000 individuals. Most rounded percentages of Jews out of the total population produce quite rough estimates, when a difference of a few hundred thousands can be significant in the context of the current analytic debate. For example, any figure in the range between 1.75 % and 1.84 % may be rounded as 1.8 %, but the difference between the higher and the lower percent is nearly 300,000 individuals. More significantly, many quite crucial Jewish-non-Jewish demographic differentials are neglected when data for a sample of adults are routinely projected for the population of all ages. Disregarding possible structural differences, namely the share of children in households, or variation in personal and spouse’s religious identifications in household size estimates and projections, may cause the incorporation of some non-Jews in Jewish population estimates.

Interestingly, the initial release of the SSRI meta-analysis (Tighe et al. 2005) provided results highly consistent with the majority of previous and contemporaneous national survey data. Based on a compilation of 74 general social surveys conducted over the period 1990–2005 (whose central year would be about 1997), the median percent of Jewish adults aged 20 and over by religion was 1.94 %. Allowing for the observed lower share of Jewish persons under 20, the ratio of Jewish to total population of all ages was 130/139 = 0.935. The corrected share of Jews among total population hence was: 1.94 % * 0.935 = 1.814 %. The 2000 US total population was 281,421,906. Consequently, the median US Jewish population (by religion) in 2000 was: 1.814 % * 281,421,906 = 5,104,993. The average survey response rate on religion was: .95. Therefore the central Jewish population estimate adjusted for non-response/no religion was: 5,104,712/.95 = 5,373,677, plus or minus standard deviations within known confidence intervals.

As noted, later versions of the SSRI meta-analysis, based on a much expanded pool of data sources, suggested much higher Jewish population estimates. Among the more recent general surveys addressing religion, the one with by far the largest national sample was the 2007 Pew Survey (Pew Forum on Religion & Public Life 2008) which included 682 unweighted Jewish cases, becoming 1,515 after weighting, out of a total sample of 88,292. This is a large enough database to allow for more intense scrutiny. The percent Jewish by religion out of the total unweighted sample was 1.919 %, but it was 1.716 % out of the weighted sample—a proof of typical Jewish over-coverage in general sample surveys.

Further processing of the 2007 Pew Survey unveiled that along with 1,515 Jewish cases out of a total weighted sample of 73,360 with a declared religion (2.065 %), there were 261 weighted cases who reported no or unknown religion or refused the question, and were raised Jewish in their childhood, out of a total of 14,932 cases with unreported or unknown religion (1.716 %). Thus the total number of persons currently Jewish or raised Jewish with no current religion was 1,776, or 2.0115 % of the total population above age 18. The religiously non-declared of Jewish origin represented 14.7 % of total Jews—declared or undeclared—while the share of unaffiliated/unknowns/refusals out of the total sample was 16.9 %. This seems to disprove the diffused assumption that the percent of Jews not declaring a religion is similar or higher than that among the total population. Out of a total US population above age 18 of 227,240,000 in 2007, 2.0115 % would correspond to an adult Jewish population of 4,570,933. With a Jewish population under age 18 estimated at 18.3 % of the total—based on NJPS 2000 population corrections and projections, see below—as against 24.7 % among the total US population, the total Jewish population of all ages would be assessed at 5,594,777. With a confidence interval of ±4.5 %, a total Jewish population range would obtain of 5,343,012–5,846,542.

In the 2007 Pew Survey in addition to the RDD landline telephone sample, interviews were completed with 500 respondents who used a cellular telephone and who did not have a landline telephone in their household. An analysis of the data revealed no significant differences in the religious makeup of the sample that included cell-only respondents and the full sample based solely on respondents from landline households. As a result, cell-only respondents were excluded from the analyses (Pew 2008, p. 113). While this is interesting for general survey evaluation, it bears accrued significance for the evaluation of the Jewish population or of any other minority group. Marital status composition of the contacted households may play a crucial role in determining the actual probability of an individual to be contacted, hence his/her weight in the total sample and projected population. Jews—declared and non-declared together—included 55 % of married and another 6 % of adults living together, for a total of 61 %, versus 60 % among the total population. Among declared Jews who were married, 69.2 % had a Jewish partner, as against 14.7 % among the non-declared of Jewish origin, and a total weighted of 63.3 % in married Jews. This means that combining Jews of single, widowed, divorced and separated marital status with Jews living with a non-Jewish partner, only 38.4 % of households with a Jewish person included more than one Jewish adult, whereas 60 % of all US households held more than one adult. As a consequence landline phones, which are usually shared within a household, yielded very different probabilities to hit more than one person at the given phone number—a generic respondent if looking for a generic respondent, and a Jewish respondent if looking for a Jewish respondent. This made the share of Jews reached in the survey more similar to percent of Jewish households among total households than to percent of Jewish individuals among total individuals. The former is higher because of smaller Jewish household size. The implications for data weighting and for population estimates are momentous. Estimated Jewish populations tend to become quite overweighted, hence exaggerated, and should actually be reduced. These considerations apply to the 2007 Pew Survey and to any other surveys which address one respondent per household and not the full roster of household members.

Unlike the Pew and other general surveys, NJPS 2000—the subject of so much debate and disagreement—did collect information on the entire roster of household members, including information on the respective religion, and is worth a new independent reading. As has already been mentioned, the NJPS 1990 data were projected ten years forward and compared with the findings of NJPS 2000 (see Fig. 10). This cohort-wise comparison provided quite crucial evaluative information. The core Jewish population according to NJPS 1990 was 5,515,000. In NJPS 2000, the Jewish population was initially estimated at 5,035,468. After imputation of people not actually covered in the survey, such as institutionalized persons in homes for the elderly or in prisons, the NJPS final estimate somewhat arbitrarily amounted to 5,200,000. Our new projection from 1990 to 2000 based on the evaluation of current migration, fertility, mortality, and accession and secession frequencies provided a higher estimate of 5,367,244.

Fig. 10
figure 10

Assessing NJPS: US core Jewish population, 1990 and 2000—actual and projected. Source: Author’s estimates. See also: DellaPergola (2005)

Full size image

Particularly sensitive were the results of birth cohort projections, described here in broader aggregates from the more detailed tabulations available (see also DellaPergola 2005). Our new projection produced results nearly identical to the actual NJPS 2000 relative to two cohorts, born in 1970–1990, and born in 1950 or before. The population actually covered fell short of the one projected by 1 % for those born in 1970–1990, aged 0–19 in 1990 and 10–29 in 2000, and 1.7 % short for those born in 1950 or earlier, aged 40+ in 1990 and 50+ in 2000. Moreover, the new projection estimate of the age group 0–9 in 2000—the product of births deriving from expected age-specific fertility rates during the inter-survey period—was 514,095 which turned out to be nearly identical to the actual number of children of the same ages found in NJPS 2000, 515,146—a discrepancy of 0.2 %. So far, then, the expected and actual data were extraordinarily consistent, besides minor corrections. However, the situation was different for the 1950–1970 birth cohort, aged 20–39 in 1990 and 30–49 in 2000. Here NJPS 2000 found 1,338,527 individuals versus a projected figure of 1,624,543—a difference of −286,016 or −17.6 %. This seems to indicate a real shortcoming of NJPS 2000—an issue that had already been critically noted (Saxe et al. 2006, 2007; Tighe et al. 2009a, 2011). Whether the significant under-coverage of this specific birth cohort/age group depended on insufficient efforts or skills at the stage of fieldwork, or on the elusive nature of Jewish identification among this particular generation of adults cannot be adjudicated with absolute certitude as available evidence runs both ways. But unquestionably, the data called for correction which was obtained by age-specifically adding a projected missing 331,776 core Jews to the original NJPS 2000 figure. The correction affected not only total Jewish population size, but also age composition with visible effects on the subsequent demographic dynamics of US Jewry. In fact, the addition of nearly 300,000 adults at ages typical for reproduction and family growth helped generate some Jewish population increase over the decade 2000–10. Projecting the corrected NJPS 2000 to 2010 resulted in a total of 5,425,000 Jews—some 150,000 higher than had been inferred in previous estimates which, it must be recalled, already upwardly adjusted the original NJPS 2000 returns (e.g., DellaPergola 2010a).

In Fig. 11 we compare our NJPS 2000 corrected results, our reading of the 2007 Pew Survey, and our 2010 NJPS-based corrected projection, displaying Jewish population estimates with the respective ranges of statistical variation due to sampling error. It is evident that these various estimates are all on the same page as each estimate falls within the expected range of variation of another estimate. In the light of the preceding analyses, this is what demographic research suggests as the most likely assessment of the current core Jewish population size in the United States.

Fig. 11
figure 11

US Jewish population: comparison of revised NJPS 2000 and 2010 projection, and Pew 2007 survey, showing average estimate and sampling error intervals. Sources: Pew (2008) not corrected for overweighting bias due to marital status and household composition (see text); author’s (DP) processing and estimates

Full size image

How Many: Yesterday, Today, Tomorrow

At this point, having critically reviewed the main evidence available, we are in a position to represent what seems to have been the most plausible course of development of the number of Jews in the United States since the end of World War II to date, and beyond until 2020 (see Fig. 12). US Jews experienced steady growth during the late 1940s and 1950s, due to the influx of post-war immigrants, and even more significantly due to a relatively young age composition and the enhanced fertility baby-boom years. While at a decreasing pace, growth was still visible during the 1970s. If there had been a 1980 NJPS, it would probably have shown a peak-ever around 5.6 million Jews in the United States, reflecting a first echo of the large baby-boom cohorts. But Jewish population was getting older through the combined effect of postponed marriage, low fertility, more frequent intermarriage, and the non-attribution of Jewish identification to high percentages of the children of one non-Jewish parent. The unavoidable consequence was the stoppage of growth and incipient decline which became more visible between 1990 and 2000. The somewhat late and incomplete entrance of the children of baby-boomers into the reproduction stage of the lifecycle generated a much weakened second baby-boom’s echo effect, visible in the corrected data of the 2000–2010 decade. Dutifully taking into account survey statistical errors and confidence intervals, the NJPS-projected Jewish population in 2010 could be no less than 5.3 and no more than 5.6 million, with a central value at 5,425,000. However, the impact of such echo-effect was in no way comparable to that of the original baby-boom, and American Jewry was again beginning to moderately shrink after 2010. The expected core Jewish population central estimate for 2020 would be 5,350,000.

Fig. 12
figure 12

Estimated US Jewish population, 1950–2020. Source: Author’s estimates

Full size image

One final important consideration brings us to a reassessment of the necessary relationship between population estimates and the underlying definitional constructs; which brings us back to the unavoidable trade-off between the size of a Jewish population and the degree of its involvement with the particular community’s culture and relational networks. Clearly, the more involved constitute a quantitatively more limited group, and the more extended a group becomes, the less one can expect of the participation, interest, or knowledge of its members. The core Jewish population that has constituted the main topic of this review admittedly is a research construct, somewhat abstract when related to the real world of Jews and other persons and institutions. The theoretical construct, if correctly measured and adjusted where necessary, has the great advantage of offering a consistent meter for comparisons over time as well as across countries. Moreover, population estimates can be evaluated for a variety of other definitional constructs, each with its own observable patterns of growth or decline (see Fig. 13).

Fig. 13
figure 13

US Jewish population: alternative definitional criteria, 2010. Source: NJPS 2000; author’s processing and estimates

Full size image

The total of individuals ready to respond Jewish (either by religion or by other definitional criteria) to an American survey stands today at a level several hundreds of thousands to a million lower than the core Jewish population of somewhat over 5.4 million. The total aggregate of those affiliated with any Jewish organizations probably does not reach three million, about half of whom are actively involved in those frameworks. In the light of recent survey evidence, all of these inner population circles, from the core down to the actively involved, have been slowly shrinking. On the other hand, the more religious, here subsumed under the Orthodox definition, constitute a much smaller but surely growing circle. This particular current’s trend to growth is shared with all the external segments of the overall possible configuration, namely the total persons with direct Jewish ancestry regardless of current religious or other mode of Jewish identification estimated at 6.8 million, the about 8 million total members of households with at least one core Jew, the possibly up to 12 million eligible for the Law of Return legal construct, or the ever-expanding and unknown number of those who ever had a Jewish ancestor. It is important to realize, therefore, that a growth/decline dynamic can and does simultaneously operate in the broadest possible assessment of Jewish population trends. But the distinction between which parts are growing and which are shrinking, and as a function of precisely what determinants and processes, as well as the adjudication of what is a meaningful and coherent definition of the Jewish collective, are crucial steps in the investigative effort and its implications.

Concluding Remarks

When trying to understand contemporary American Jewry, population size may not be the most significant aspect. Some may find it more relevant to focus on the potential revitalization of the community generated by educational programs such as Taglit (Saxe et al. 2011). To be sure, such extraordinary and quite successful investments in Jewish formal and informal education were initially prompted by the dismal findings of NJPS 1990, namely an alleged 52 % intermarriage rate, clearly exemplifying the importance of empirical research for policy development. But, in the short-term, population size is only moderately affected by such potential changes and cannot be neglected as such. Judged by demography’s research tools and constructs, the US core Jewish population amounted to no more than 5.7 million and no less than 5.2 in 2010. Among nearly a million of these, the relevance of Jewish identity was so low that they would not take care to declare themselves Jewish when asked in the first place. On the other hand, an additional 1.2 million currently non-Jewish Americans would readily recognize direct Jewish ancestry and/or a past personal Jewish belonging, and therefore significant personal linkages to Jewish persons and groups—nonetheless not relevant enough to declare themselves Jewish.

As against the widespread feeling that too much ink is spilled on demography at the expense of other more significant aspects of the quality and intensity of Jewish identification, it has been argued that:

…excluding from the analysis those who might at first be considered the weaker, more assimilated or indifferent sections of the Jewish collective unveils a singular analytic dissonance. Indeed, when the discussion focuses on the demography and total size of US Jewry, great efforts are displayed to include these more marginal fringes, and their quantitative extent is the object of high profile negotiations…but when these ‘peripherals’ tend to weaken the overall intensity of the identification profile, they are readily forgotten on the grounds that the topic at stake is not assimilation but a more substantial aspect of Jewishness. One cannot have it both ways: If American Jewry must be larger, it is because of the inclusion of more of the peripherals; and if American Jewry must be more coherently identified, than its size must necessarily become smaller. (DellaPergola 2010b)

Facing this trade-off, the different aspects should be dealt with in conjunction. From demography’s perspective, definitional and analytic rules in the study of American Jewry need to follow as broad as possible a spectrum of theoretical and empirical pathways, but they cannot seriously elude the following two basic constraints: (a) Jews in the United States integrally pertain to American society, and consequently significantly shared and responded to changing socio-economic, cultural and political stimuli in their country. Jews often anticipated national trends in the realms of demographic and socio-economic change, and occasionally they were late joiners in those trends; (b) American Jews, inasmuch as they are part of an historical and cultural global Jewish collective, belong to a transnational entity significantly sharing and affected by unique and crucially important commonalities and processes. Jews in the United States on some accounts manifested definite patterns of exceptionalism vis-à-vis Jews in other parts of the world, but on other accounts they were part of the same broader processes affecting other Jews as well.

Being part of a large Jewish collective in a powerful, dynamic, and decentralized country like the United States helped to create economic opportunities, relational networks, and emotional experiences unique to American Jews. But to the extent that it was a shared trait, minority status deeply influenced those same opportunities, networks, and experiences among American Jews in directions that could be observed among Jews in other countries as well. Still other Jews, in Israel, experienced the majority-minority relation in quite different modes, which in turn profoundly affected their demographic and socio-economic dynamics.

These various constraints determined the past, and will delimit the future range of the possible and of the plausible regarding the demographic, economic, and cultural trends and patterns of Jews in America and in other countries. It is to be hoped that those interested in learning and clarifying the crucial issues regarding these matters and their longer-term implications can eventually agree on a broad multiplicity of shared approaches and coherent definitions, and will thus be able to coalesce into one epistemic community—beyond the several possible narratives and empirical specializations—for the sake of the social scientific study of Jewry.