Correspondence should be addressed to Eric Kaufmann, Birbeck College, University of London, Malet St., London, WC1E 7HX United Kingdom. E-mail: email@example.com
We provide a cohort-component projection of the religious composition of the United States, considering differences in fertility, migration, intergenerational religious transmission, and switching among 11 ethnoreligious groups. If fertility and migration trends continue, Hispanic Catholics will experience rapid growth and expand from 10 to 18 percent of the American population between 2003 and 2043. Protestants are projected to decrease from 47 to 39 percent over the same period, while Catholicism emerges as the largest religion among the youngest age cohorts. Liberal Protestants decline relative to other groups due to low fertility and losses from religious switching. Immigration drives growth among Hindus and Muslims, while low fertility and a mature age structure causes Jewish decline. The low fertility of secular Americans and the religiosity of immigrants provide a countervailing force to secularization, causing the nonreligious population share to peak before 2043.
Sociologists of religion typically focus on the attractiveness of denominations in the religious marketplace. Yet the main source of religious recruits is the children of communicants. The religious are a population that can be analyzed demographically. “People enter, exit, and move within religion, just as they are born, will die, and migrate, in life” (Voas 2003:94). Religious beliefs are powerful determinants of demographic events such as marriage, divorce, and childbearing (McQuillan 2004; UN 1973). The teachings of most major religions regulate partnership, sexuality, and fertility and can affect demographic patterns both explicitly—as with religious leaders’ injunctions against contraceptives and their promotion of early marriage, which is related to higher fertility outcomes—and indirectly (e.g., socialization into a group where there is strong emphasis on childbearing).
Important differences can also be found between and within major religions. Among white Christian Americans, Catholics once had a significant fertility advantage over Protestants, but this waned in the second half of the 20th century (Jones and Westoff 1979; Sander 1992). Evangelical Protestants continue to have higher fertility rates than those from more liberal Protestant groups (Hout, Greeley, and Wilde 2001; Lehrer 1996; Roof and McKinney 1989). The same is true for Mormons (Sherkat 2001). American Jews have lower fertility than other ethnoreligious groups (Mosher and Hendershot 1984), primarily because of later onset childbearing and a higher investment in human capital accumulation. Lehrer's work with the 1995 National Survey of Family Growth (NSFG), for instance, finds that the probability of marriage by age 20 is 2 percent for Jews, 9 percent for mainline Protestants, and 17 percent for fundamentalist Protestants and Mormons (Lehrer 2004; Thornton, Axinn, and Hill 1992).
Immigration is the demographic engine of religious change, and tends to increase the religious diversity of a country and challenge dominant denominations. In the United States, immigration from largely Catholic Latin America—notably Mexico—helped to mask net defections from Catholicism to Protestantism and secular nonaffiliation (Sherkat 2001). The younger age structure and higher fertility of Latino Catholic immigrants to the United States as compared to Protestants has endowed Catholicism with an additional demographic tool with which to combat its relative disadvantage in the American religious marketplace. As we shall see, both fertility and immigration will play a significant role in the recasting of America's religious composition in the 21st century. Our work builds on previous research on religious market dynamics and religious demography (Hout, Greeley, and Wilde 2001; Sherkat 2001).
Projections of Religious Composition
The U.S. Census Bureau (USCB) carries out regular projections of the American population by race to 2050, greatly advancing our knowledge of impending social changes. Far from speculations, these provide solid portrayals of the future, which have largely been borne out by subsequent developments, though the “majority minority” point has been revised forward from 2050 to 2042. Projections are only partly susceptible to change from shifting fertility and migration parameters. The characteristics of the average American of 2050, for instance, can be largely read off the youngest American cohorts of today. For this reason, demographic projections provide the most accurate predictions in the social sciences.
The absence of a census question on religious affiliation prevents the USCB from making religious projections. Even so, the availability of quality survey data from 1972 in the form of the General Social Survey (GSS; Davis and Smith 2009) renders such a study feasible. Nonetheless, no projection of America's religious composition utilizing the cohort-component approach has, to our knowledge, been carried out. The oft-cited World Christian Encyclopedia (WCE) predicts the size of religious groups (including seculars) from baptismal data, but does not take into account variation in the demographic variables of age structure, fertility, and immigration, nor the sociological dynamics of religious switching and it does not use scenarios to reflect uncertainty about future trends (Barrett, Kurian, and Johnson 2001).
Religious projections have recently been carried out for several countries. Goujon, Skirbekk, and Fliegenschnee (2007) present census-based religious projections for Austria and Switzerland1 and find the Christian share to be shrinking in both. The Swiss were more than 95 percent Christian in 1970, but this figure sank to 75 percent in 2000 and will fall to between 42 and 63 percent by 2050. In Austria, the long dominant Roman Catholic population decreased to 75 percent in 2001 and is expected to comprise less than half the population by mid century. In both cases, Christian decline is mainly related to secularization, but is also linked to the growth of non-Christian religions, notably Islam. Statistics Canada (2005) has carried out projections of the religious composition of Canada, which accounts for fertility and mortality differentials as well as rates of intergenerational religious transmission. But these do not take religious conversion into consideration and only cover the period to 2017, too short a span to capture most demographic effects.
We project the size of America's main ethnoreligious groupings to 2043, taking into account the impact of religion on fertility and the way immigration affects religious composition. We also account for conversion and secularization by age and sex as well as the intergenerational transmission of religious affiliation. We find that the United States in 2043 will remain majority Christian, but with a different ethnoreligious composition. Hence the share of Hispanic Catholics, Muslims, Hindus, Buddhists, and seculars increases, while the mainly “white” religious groups—liberal, moderate, and fundamentalist Protestants as well as non-Hispanic Catholics—declines proportionally. Smith and Kim (2004) recently found that the Protestant share of the American population will fall below 50 percent. We envisage a further decline in the Protestant total, notwithstanding Protestant gains from Catholic switching.
This research relies on a cross-pollination of census and survey data. The principal data source is the GSS that has been conducted annually (1972–1993) with an interview sample of around 1,500 and biennially since 1994 with a sample of 2,800.2 Respondents were asked about their current religious affiliation as well as their affiliation at age 16, enabling a measure of religious conversion. It has been used extensively by scholars who have examined longer-term trends in the American religious marketplace (i.e., Hout, Greeley, and Wilde 2001; Sherkat 2001). The GSS classifies largely white (non-African American) Protestant denominations as “fundamentalist,”“moderate,” or “liberal” according to a schema developed by Smith (1986). It also aggregates denominations into larger religious affiliation categories such as Protestant, Catholic, or other non-Christian. In all cases, we adopt the classifications used by the GSS. This yields 11 major ethnoreligious groups for analysis: Fundamentalist Protestants excluding Blacks (PFU), Moderate Protestants excluding Blacks (PMO), Liberal Protestants excluding Blacks (PLI), Black Protestants (PBL), non-Hispanic Catholics (CAT), Hispanic Catholics (CHI), Jews (JEW), Hindus and Buddhists (HBU), Muslims (MUS), Other Religions (OTH), and No Religion (NOR). Note that the non-Hispanic Catholics, nonblack Protestants, and no religion groups are overwhelmingly white but not exclusively so. For instance, there are significant numbers of Hispanic Protestants and black Catholics. Similarly, the small Asian-American population contains Protestants, Catholics, and those of no religion as well as Hindus, Buddhists, Muslims, and others.
Of course, Muslim, Buddhist/Hindu, and other religions are extremely small categories (i.e., 1 percent or less), and can be undercounted by the GSS, so we rely upon a set of recent Pew Forum on Religion and Public Life surveys that provide precise estimates of their size (Pew 2007, 2008). The GSS also likely undercounts African Americans and Hispanics, so we weight our data to reflect their share in the U.S. census. The PSU (primary sampling units) that the GSS uses at different points in time can also affect the size and characteristics of certain groups. To ensure that the GSS data are fairly representative we compare its findings to alternative surveys that have been conducted since the year 2000 (Table 1). In general, although the estimates differ, the various surveys present a broadly consistent picture, with about half the population Protestant, a quarter Catholic, and about one in eight without religion, with a scattering of smaller groups (Jews, Hindus, Muslims, and other religious groups).
Table 1. Religion data from GSS compared with alternative data sources (age 18+) as percent of sample
Baylor Religion Survey, 2005 (Collected by Gallup), N = 1,721
The variation in findings between the surveys probably stems from a number of factors: the relatively small sample sizes, which can lead to random error; variations in survey approach, most notably stratification procedure and sampling units; the representativeness of the immigrant and minority sample, which may depend on the language of the survey. Some of the variation in estimated group size likely is caused by somewhat different definitions of religious intensity (i.e., Protestant “fundamentalism”). Steensland et al. (2000) advocate the categories Unaffiliated, Other, Jewish, Catholic, Mainline Protestants, Evangelical Protestants, and Black Protestants. We choose to use the somewhat more fine-grained distinction between fundamentalist, moderate, and liberal Protestants employed by the GSS. Both the demography and religious views of these groups differ, thus a dichotomous categorization may be too crude. Finally, we include Hindu/Buddhists and Muslims as separate groups since they have their own dynamics, which are obscured in the “Other” category.
No religious affiliation
Our starting year data (2003) are drawn from the GSS for the years 2000–2006. These years were pooled together in order to increase sample size for the base population (N = 12,674) and they are the only available survey years that include both minority religions (notably Hinduism and Islam) and a separate Hispanic category. Figure 1 shows the ethnoreligious composition of our base population in 2003.
Immigration forms a crucial part of the projection, and we introduce an annual addition to each religious group, broken down by age and sex, based on observed immigration. Annual immigration figures come from the Population Estimates Program of the U.S. Census Bureau (2007). The religious affiliation of immigrants is based on CIA data on source country religious composition (CIA 2008). Immigrant age structure is derived from a standard schedule (Rogers and Castro 1981). We assume immigrants are randomly selected in terms of religion in their country of origin, though we accept that there are instances where immigrants are unrepresentative of their homeland's religious composition. Basing our analysis on census data can help to avoid problems related to survey response bias. An alternative data source on affiliation (although not intensity), the New Immigrant Survey (NIS) (e.g., Jasso and Rosenzweig 2006), produces fairly similar proportions for most religious groups (NIS estimates deviate from ours by up to 4.3 percent), although NIS suggests a considerably higher Protestant share (15.7 rather than 7.2 percent). Nevertheless, sensitivity analysis has shown that using NIS estimates would not affect the main results of our projections.3
The aforementioned sources provide us with information regarding base population, age structure, fertility, conversion behavior, and immigration. These provide the inputs we need to undertake population projections. For the United States, the long period available in the GSS (1972–2006) allows us to observe a time series run of conversion and fertility behavior analogous to annual immigration statistics. These are scenario-based multistate cohort-component projections, carried out with the use of PDE projection software,4 a multistate population projection program. We use initial population by age, sex, and ethnoreligious denomination, age- and religion-specific fertility rates, age- and sex-specific mortality rates, and age-, sex-, and religion-specific net migration numbers. In addition, a central input into any multistate projection is the religious conversion rate, such as the secularizing trend from Christianity to no religion, or conversion from Catholicism to fundamentalist Protestantism. Questions are asked about denominational affiliation at age 16, which we cross-tabulate with current denominational affiliation to produce an estimate of conversion flows by sex and age band. We employ both expected and alternate scenarios based on varying fertility, conversion, and immigration assumptions.
Fertility differences by religion in the United States were estimated from GSS data on children ever born to women aged 40 to 595 for the period 2000–2006. The differentials were then proportionally adjusted and applied to the national TFR reported for 2003 by the U.S. Census Bureau. The data were not sufficient to estimate the age-specific schedules of fertility rates. Hence all religious groups follow the age-specific fertility schedule as observed at the national level. The estimated religious fertility differentials are given in Table 2.6 Hispanic Catholics and Muslims have the highest fertility (2.8 children per woman), while Jews have the lowest (1.4 children). Among Protestants, black Protestant fertility is the highest, at 2.4 children per woman. The two largest “white” religious categories, non-Hispanic Catholics and fundamentalist Protestants, have replacement fertility (2.1 children) while others and the “no religion” groups have much lower TFRs of around 1.65 children per woman. The relatively low fertility of Hindus and Buddhists may be attributable to very selective migration from India and the Far East.7
Table 2. TFR (total fertility rate) by religion, 2003
Hispanic Catholics (CHI)
Black Protestants (PBL)
Fundamentalist Protestants excluding blacks (PFU)
Non-Hispanic Catholics (CAT)
Moderate Protestants excluding blacks (PMO)
Liberal Protestants excluding blacks (PLI)
No religion (NOR)
U.S. population average
Mortality cannot be estimated for each religious group, so we assume a single value for each age group and sex following the estimates of the National Center for Health Statistics (NCHS), available in Kung et al. (2008). While we acknowledge that mortality differences could have a minor effect on religious composition in 2043, this does not affect fertility, migration, or the number of women of various denominations in their childbearing years. It is also not clear whether higher religiosity would be positively or negatively associated with life expectancies. Some studies find religiosity to be associated with better health outcomes (for a review, see Hummer et al. 2004). On the other hand, the more religious tend to have lower education levels (Glaeser and Sacerdote 2008; Inglehart and Baker 2000) that may lower health outcomes since education is clearly positively related to health (Groot and Maassen Van DenBrink 2007).
The number of immigrants since the 1980s has reached levels unseen since the immigration peak of the early 20th century. Immigration is therefore a key factor in the changing religious landscape of the United States. Yet there are two major difficulties in estimating immigration differentials by religion. One is inherent to the immigration process in the United States where illegal flows from across the Mexican border play an important role. We do not take illegal immigration into account, though a substantial share of legal immigration consists of formerly undocumented immigrants who have been granted amnesty. The second difficulty has to do with the lack of data on the faith of immigrants. We obtained the differentials in the religious affiliation of the immigrants from the starting year (2003) as follows. First, we retrieved the number of persons obtaining legal permanent resident status by region between 2003 and 2006 (U.S. Department of Homeland Security 2007). We selected the countries of birth of most persons acquiring legal permanent status (all above 5,000 persons per year during the 2003–2006 period).8 Next, we obtained the shares of the population by religion from the CIA World Factbook (CIA 2007). Some adjustments were made to fit the CIA data to our specific categories. We treat Latin American Protestants and East Asian Protestants as 90 percent fundamentalists, 5 percent moderates, and 5 percent liberals. We treat European Protestants as 50 percent moderates, and 50 percent liberals. For Canadians, Protestants are divided equally between fundamentalists, moderates, and liberals, reflecting the intermediate position of Canadian Protestantism between British and American denominational traditions. These rates were then applied to the number of persons obtaining legal permanent resident status for the main countries of birth between 2003 and 2006 and aggregated by the 11 religious categories reported in Figure 2. These categories were then applied to the net number of immigrants for the period 2000–2005 and distributed by age and sex (according to model age schedules of migration).
Transition rates reflect conversion flows between religions. We based our estimate of transition probabilities between religions (110 possible flows between the 11 religious categories) on comparing religion retrospectively reported for age 16 with current religion. Since the GSS does not provide the age at which the switch to another religion occurred, we distributed the transitions equally across three age groups: 15–19, 20–24, 25–29. This is in line with switching patterns observed in other countries (Goujon, Skirbekk, and Fliegenschnee 2007). We further assume that men are 6 percent more likely to switch out of their own religion than women. This is based on gender differences among apostates: the proportion that were members of a religion at age 16 but now report being nonreligious. Moreover, women who were religiously unaffiliated at age 16 are 29 percent more likely to adopt a religion than men from the same (secular) background.
Table 3 reports the observed transition probabilities. For example, 15.1 percent of those without religion at age 16 became fundamentalist Protestants as adults and 11.7 percent of those raised fundamentalist Protestant transited the other way. Note the substantial losses to secularism (NOR) across all religions, the relative retentive power of the more “ethnic” Jewish, black, Hispanic, and Muslim groups, and the comparative deficit of mainline Protestants (PMO, PLI) and white Catholics (CAT) in exchanges with fundamentalist Protestants (PFU). This confirms existing scholarship pertaining to religious marketplace trends, as well as insights from the “strict church” hypothesis (Hout, Greeley, and Wilde 2001; Iannaccone 1994; Sherkat 2001).
Table 3. Matrix of total transition probabilities: Religion at age 16 versus current religion
Sources: Authors’ calculations based on GSS 2000–2006.
In order to validate the methodology used for the projections, we applied it to historical GSS data to see if we could fit our model to observed data. This was performed for the five main religious categories that were found across all years for which GSS data are available. Simulations using a six-year moving average are deployed to produce data from 1975 to 2000. The fertility differentials between religions as well as the religious composition of the immigration flow are based upon those observed in the base year of our projection. However, we make some adjustments for historical data. Assumptions for total fertility (estimated at 1.81 for 1975–1980), mortality, and migration (350,000 per annum during 1975–1980) follow the historical data available from the U.S. Census Bureau (for fertility and mortality) and from the United Nations (2006—for migration). Transition probabilities were calculated in the same way as mentioned in the previous section, based on comparing religion retrospectively reported for age 16 with current religion for two periods: 1972–1978 and 1992–1998. The results are shown in Figure 3. Our model performs quite well against observed data, projecting the trend toward a relatively less Protestant and more secular nation. It also shows that GSS data fluctuate significantly around the trend, reflecting period and sampling effects.
Scenarios In addition to our expected scenario (H0) based on current trends, four alternative scenarios were developed; they diverge by the net number of immigrants, the fertility rates of the 11 religious categories, and the conversion rates between religions. Table 4 summarizes the assumptions made in the five scenarios for our 2003–2043 projections.9
Table 4. Scenarios matrix
Constant Fertility Differentials
Converging Fertility Differentials
We consider two alternative fertility assumptions, constant and converging. Constant fertility (scenarios H0, H1, H2) holds fertility within each religion constant at the level observed in the base year, 2003 (see Table 2), consistent with the U.S. Census Bureau's constant ethnic fertility differences (Day 1996). Note that the overall American TFR changes as a result of religious compositional effects. Thus the constant fertility assumption raises the aggregate American TFR from 2.08 in 2003 to 2.2 in 2043 as high fertility Latino Catholics increase their share of the total. Converging fertility (scenario H3) assumes that fertility by religion converges to a TFR of 2.1 children by 2033–2038 and remains constant thereafter. This TFR is slightly lower than the medium variant of the U.S. population projection that envisions this figure increasing to 2.19 in 2050 (U.S. Census Bureau 2004). We further assume that children have the same religion as their mothers, regardless of the type of union, monoreligious, or mixed. This is a problematic assumption in Europe, where mixed unions often lead to secularism, but not in the American case.
With respect to immigration, there are three possible pathways. Constant migration (scenario H0) involves the net number of immigrants to the United States remaining constant at 1.2 million per year until the end of the projection period (value from UN 2006 for 2005–2010). Double migration (scenario H1) assumes that net immigration doubles from the start, resulting in an annual influx of 2.4 million per annum between 2003 and 2043. The current congressional debate over immigration reform may lead to legislation that reduces the number of immigrants entering the country. Accordingly, the zero migration scenarios (H3, H4) gauge the impact of immigration against other drivers of projection outcomes. In all immigration scenarios, the share of immigrants by religious denomination stays constant at the levels estimated for the starting period.
As regards conversion between religions, there are two options. The first is a constant conversion assumption that fixes adult switching probabilities at the levels observed during 2000–2006 with children inheriting the religious category of their mothers and summarized in Table 3. The second is zero conversion, which assumes no adult religious switching between groups over the life course.
Our five scenarios produce significantly different total fertility rates for the American population. We expect an upward trend in fertility over the projection period as more fertile religions expand. The national TFR varies significantly—between 2.10 and 2.16—due to changes in the religious composition of the population. U.S. population size is first and foremost affected by immigration (Figure 4). If immigration remains constant, the population size reaches 420 million in 2043. Notice that this is in line with official projections for 2043 from the USCB and Social Security Administration.10 If immigration doubles, we project a U.S. population of 495 million while zero immigration results in a population of 342 million, 78 million less than in the constant immigration scenario. Immigration also affects the population size through its effect on fertility levels since the religious composition of the immigrants differs from that of the resident population. This is mainly due to the increase in the proportion of high fertility Hispanic Catholics in the population. Conversion similarly affects population size partly because Hispanic Catholics convert to lower fertility secular or Protestant groups—hence in the absence of secularization and conversion (H2), there will be 2 million more Americans in 2043 than under our constant conversion (H0) scenario that fixes secularization and conversion rates at base-year levels.
Figure 5 shows the projected trend for five meta-religious groups. Under all scenarios, Protestants, Catholics, those from other religions, and the nonreligious are expected to grow in absolute terms, while the Jews, due to low immigration and low fertility, are expected to decline slightly. In terms of the religious composition of the American population in 2043, the constant (immigration, fertility, conversion) scenario (H0) projects that Protestants will decrease from 47 to 39 percent as Catholics rise from 28 to 32 percent. Those of other religions will almost double, from 6 to 11 percent, the unaffiliated “secular” population increases slightly from 16 to 17 percent while the Jews decline but remain above 1 percent of the population.
The difference between H0 (constant conversion rates) and H2 (no conversion) shows that today's conversion trends mainly benefit the Protestant and secular groups. Religious conversion reduces the number of Catholics (relative to no conversion) by 15.5 million and those from other religions by 2 million. Conversely, seculars increase by 3 million through conversion and Protestants by 12.5 million. If fertility differentials and immigration remain at today's levels, but there is no religious conversion (H3), the Catholic population would exceed that of Protestants—a symbolic moment in American history! Even under our constant assumption (H0), Catholics in the youngest age cohorts will outnumber their Protestant counterparts by 2043 and take over some time in the second half of the 21st century. This would principally be due to higher Hispanic Catholic fertility and immigration. If immigration continues at today's pace (H0), there will be 35 million more Catholics in 2043 than would have been true without immigration (H3). Protestants, by contrast, gain only 9 million adherents through immigration in the same period. Other religions gain 20 million and seculars 12 million through immigration.
We have largely discussed trends in ethnoreligious change, but we know that denominations are theologically diverse (Hoffmann and Miller 1998). In particular, the religious restructuring or “culture wars” perspective highlights the importance of transdenominational processes like secularism or traditionalism (Guth et al. 2006; Hunter 1991). This theory predicts a continued hollowing out of the religious center in favor of fundamentalism and secularism. However, while seculars do grow as projected, we find powerful demographic limits to secularism under the constant (H0) scenario. In spite of considerable gains through the secularization (conversion) of members from religious groups, the share of the population comprised of secular nonaffiliates plateaus before the end of the projection period. In effect, low secular fertility is sufficient to reverse the secularization process at the aggregate level! This is an extremely important result in that it demonstrates the power of demography to reverse secularization even in developed societies (Kaufmann 2008). This may lead us to question the widely shared view that secularization is an inevitable handmaiden of the modernization process.
We now move beneath meta-religious groups in Figure 5 to consider the relative position of our 11 ethnoreligious categories. Figure 6 sets out our projections based on the constant (conversion/immigration/fertility differentials) scenario (H0). The most rapid changes take place among Hispanic Catholics, who almost double from roughly 10 percent in 2003 to 18 percent in 2043. Along the way, they surpass the two largest “white” religious groups, fundamentalist Protestants and Catholics. “White” (i.e., non-Hispanic) Catholics decline in the same period from 19 to 15 percent. In addition, all Protestant groups—fundamentalist, moderate, liberal, and black—lose market share toward the end of the projection. The secular proportion of the population, as noted, peaks in 2033 and declines somewhat toward the end of the period as the long-term effects of low secular fertility kick in.
Nevertheless, we find that the most committed parties in the “culture wars” that divide America, fundamentalist Protestants and those without religion, trade places over this period. Fundamentalist Protestants, 78 percent of whom supported George W. Bush as president in 2004, decline from 19.5 percent to 16.7 percent. The decline in Protestant fundamentalism is driven by a more mature age structure, which increases the loss of members through mortality, despite their relatively high fertility. Also, a mere 5 percent of immigrants are fundamentalist Protestants. In spite of their relative decline, we expect Protestant fundamentalists to increase in absolute terms under all scenarios.
Those without religion, just 28 percent of whom backed Bush, increase slightly from 17 to 17.4 percent, surpassing fundamentalist Protestants in 2033. Hispanic Catholics lean Democratic by a 48:20 two-party ratio in the 2003 GSS, thus the increasingly secular and Hispanic-Catholic America of 2043 should favor the Democrats in the coming decades (Guth et al. 2006). A glimpse of what may transpire comes from California, whose trends tend to foreshadow those of the nation as a whole. During 1980–2003, rapid ethnodemographic change transformed the state from white (non-Hispanic) majority to white minority. Along the way, it changed from a finely balanced battleground state into a “natural” Democratic one. Demographic change was only part of the story, but played a significant role in the process (Korey and Lascher 2006:58, 61). However, while seculars and Hispanic Catholics are Democrat-leaning, the latter are far more conservative in their social attitudes. In this sense, Hispanic Catholics resemble white working-class Democrats and reinforce the median American political profile that has held since 1954: conservative but Democratic (Box-Steffensmeier and De Boef 2001). Though ideology has been coming into alignment with partisanship in recent decades as the main parties distinguish themselves more clearly along ideological lines (Abramowitz and Saunders 2006), Hispanic immigration will work in the opposite direction to maintain dissonance between ideology and partisanship. This was demonstrated in 2008 by the passage of Proposition 8 (opposing same-sex marriage) in California on the strength of minority support.
Our projections further indicate that Muslims, Hindus/Buddhists, and other non-Christian faiths will increase their share of the population throughout the projection period. The balance between Muslims and Jews (Figure 7) is especially noteworthy in view of their differing views on American foreign policy. Should current immigration and fertility patterns continue, we expect Muslim Americans to overtake Jews by 2020 within the population and 2028 within the electorate, although both are expected to remain small minorities in the United States.11 American support for Israel is largely attributable to extra-Jewish forces such as Christian Zionism or partially Jewish ones like neoconservatism (Mearsheimer and Walt 2007) and also derives from the substantial presence of Jews within the American elite. This may insulate it from demographic change. Even so, Muslim America's eclipse of Jewish America will register in the nation's consciousness and could affect America's foreign policy calculus.
Whites are disproportionately represented in the American electorate, media, and power structure. They thereby merit closer scrutiny. Figure 8 shows trends within the white (non-Hispanic) population. We begin by noting the relative strength of liberal Protestants and seculars within the white, as compared to the total, population. Whites are affected least by immigration but most by secularization. During the projection period, seculars increase their share of the white population substantially. Moderate and fundamentalist Protestants retain their positions, while Jews, Catholics, and liberal Protestants decline. These trends owe something to the religious restructuring that is polarizing liberal Protestants and white Catholics toward secularism and fundamentalism. Low Jewish and liberal Protestant fertility also account for some of the trend. We may surmise that these patterns will enhance the secular tint of the American white elite and may deepen the divisions between religion and secularism, which characterize the so-called culture wars (Fiorina, Abrams, and Pope 2005; Hunter 1991). Curiously, relaxed immigration, a liberal cause celèbre, actually works to curb secularizing tendencies in the population at large.
The U.S. Census Bureau has, for some time, published projections of the racial composition of the American population to 2050, which show that a majority of Americans will be non-“white” by 2050.12 This so-called browning of America has entered the public lexicon, but we have no similar awareness of what is happening with religion because of the lack of a census question on the subject. This study provides the first ever cohort-component projection of the main religious groups in the United States. Largely based on the General Social Survey, census immigration statistics and Pew small religious group data, it projects the size of 11 American religious groups to 2043. Though our projections, like those of the Census Bureau, depend on immigration and fertility assumptions that can fluctuate, demographic projections are the most certain of any in the social sciences. This is because the characteristics of future populations are heavily constrained by the age-structural features of the current population.
We find considerable stability of religious groups over time, but there are some important shifts. Hispanic Catholics experience the strongest growth rates to 2043. Immigration, high fertility, and a young age structure will enable this group to expand from 10 to 18 percent of the American population between 2003 and 2043, despite a net loss of communicants to secularism and Protestantism. This will power the growth of Catholics as a whole, who will surpass Protestants by mid century within the nation's youngest age groups. This represents a historic moment for a country settled by anti-Catholic Puritans, whose Revolution was motivated in part by a desire to spread dissenting Protestantism and whose population on the eve of revolution was 98 percent Protestant (Huntington 2004; Kaufmann 2004). Another important development concerns the growth of the Muslim population and decline of the Jews. High Muslim fertility and a young Muslim age structure contrasts with low Jewish childbearing levels and a mature Jewish age structure. Barring an unforeseen shift in the religious composition and size of the immigrant flow, Muslims will surpass Jews in the population by 2023 and in the electorate by 2028. This could have profound effects on the course of American foreign policy. Within the non-Hispanic white population, we expect to see continued liberal Protestant decline due to low fertility and a net loss in exchanges with other groups. White Catholics will also lose due to a net outflow of converts. Fundamentalist and moderate Protestant denominations will hold their own within the white population, but will decline overall as the white share of the population falls.
The finding that Protestant fundamentalism may decline in relative terms over the medium term contrasts with a prevailing view that envisions the continued growth of “strong religion” (Stark and Iannaccone 1994a). This is the result of an older age structure, which increases loss through mortality, and immigration, which reduces the size of all predominantly white denominations—all of which are poorly represented in the immigration flow. Fundamentalists’ relatively high fertility and net surplus from the religious marketplace is not sufficient to counteract the effects of immigration. Obviously, this could change if significant immigration begins to arrive from more Pentecostalist source countries such as Guatemala or parts of sub-Saharan Africa.
Our work also sheds light on the religious restructuring paradigm, though we do not find a clear victor between secularism and fundamentalism. The secular population will grow substantially in the decades ahead because it has a young age structure and more people leave religion than enter it. The sharpest gains for secularism will be within the white population, where seculars will surpass fundamentalists by 2030. On the other hand, there are important demographic limits to secularism, demonstrating the power of religious demography. The relatively low fertility of secular Americans and the religiosity of the immigrant inflow provide a countervailing force that will cause the secularization process within the total population to plateau before 2043. This represents an important theoretical point in that demography permits society to grow more religious even as individuals tend to become less religious over time.
Haug and Wanner (2000) also projected future religious denominations for Switzerland but only up to 2020 and excluded those without religion, the fastest growing group.
The only exceptions are the years 1979, 1981, and 1992 (a supplement was added in 1992).
For instance, the H0 (constant) scenario would lead to a difference in the size of religious categories of −.5 to 1.1 points by 2043 depending on the category.
For Muslims, we base the differential on the 35 to 59 population to increase sample size.
Fertility differentials are very close (the average difference is .02 child) to those computed by Chandra et al. (2005) using the 2002 National Survey of Family Growth. Their classification is based on children ever born and total children expected from women aged 15–44 for the following religious categories: none, fundamentalist Protestant, other Protestant, Catholic, and other religion.
Due to selective migration and a younger age structure, Indians living in the United States have a high education level, and higher education tends to be related to lower fertility (Skirbekk 2008). More than 58 percent have college degrees (compared to 25 percent of the general U.S. population and 6 percent of the population of India), and they also possess higher than average wealth and income levels (Lutz et al. 2007; Kiviat 2005).
The countries of birth of most persons (82 to 85 percent across the 2003 to 2006 population) acquiring legal permanent status are: Bangladesh, Bosnia-Herzegovina, Brazil, Canada, China, Colombia, Cuba, Dominican Republic, Ecuador, Egypt, El Salvador, Ethiopia, Germany, Ghana, Guatemala, Guyana, Haiti, Honduras, India, Iran, Jamaica, Japan, Kenya, Korea, Mexico, Nigeria, Pakistan, Peru, Philippines, Poland, Russia, Taiwan, Trinidad and Tobago, Ukraine, the United Kingdom, Venezuela, and Vietnam.
As mentioned in the preceding section, mortality is not considered separately for each religious category. The life expectancy assumptions to 2043 are interpolated based on assumptions used by the U.S. Census Bureau (2004).
The SSA 2003 estimated population was interpolated by the authors between the values provided by SSA for 2000 and 2005.
Note that there are almost certainly more “ethnic” Jews than Muslims among those raised with no religion, though secularization rates are similar among adults of both groups. The low Jewish fertility rate may also increase as the Orthodox share of the American Jewish total rises. For instance, while only 2 percent of secular Jews and 3–5 percent of Reform and Conservative Jews had more than two children in their household in 2001, 25 percent of Orthodox Jews did. Thirty-three percent of Jews aged 18–34 are Orthodox, whereas just 8 percent of Jews over 75 are (Ament 2005:16).
“White” here excludes the roughly 50 percent of Hispanics who identify as racially white on the census.