Sociologists of religion often measure religiosity along three dimensions, known informally as the three B’s—belonging to or identifying with a religious group; religious behavior such as membership in a congregation and attendance in religious services; and belief in God. Edgell (2012: 258) asserts that “religion’s influence on an individual’s attitudes, beliefs, and actions may … vary across time or social location.” This paper concentrates on the American context in the early 21st Century. Its contributions are fivefold: First, to segment the U.S. population into eight groups based on their yes or no answers on each of the three dimensions of religion by creating a 2x2x2 matrix of belonging/behavior/belief. Second, to place those eight groups along a spectrum to create a single unified scale of religiosity. Third, to examine shifts over time in the ways the population is divided into the eight groups. Fourth, to identify socio-demographic factors that are associated with a person’s position on the spectrum. Fifth, to test the power of this new measure by examining its ability to predict how people answer questions on contentious societal issues, using belief in evolution as a case study.

The main data sets utilized in the paper, the American Religious Identification Survey series, do not track the same individuals over time but rather look at increases and decreases in the size of population groups. The large national data sets are the best currently available to create a unified religiosity scale based on the three B’s and to test its application. This paper focuses on the experiences of individuals rather than congregations or denominations. Its contribution is primarily empirical.

Background and Hypotheses

Conceptualizing and Operationalizing Religiosity

Patterns of religiosity in America have been extensively researched. Some empirical studies have found stable levels of religiosity (Presser and Chaves, 2007), while others have found changes over time (Putnam, 2000; Schwadel, 2013). The different conclusions can be attributed in part to the sources of data used and in part to which aspect of religiosity the authors focus on: belonging, behavior, or belief.

To some scholars who focus on behavior, modernization, consumerism and economic constraints explain the decline in religious practice, as non-religious activities compete for limited and precious time (Robinson and Godbey, 1999; Gruber and Hungerman 2008). Disillusionment with religious institutions and their leaders, and erosion of confidence and trust in these institutions, may also distance people from religious community. The General Social Survey (GSS) showed a decline in public trust in all institutions (besides the military) since 1976. Organized religion did not fare well. Gallup (2013) showed that while 43% of Americans in 1973 expressed a “great deal” of confidence in “the church or organized religion,” by 2012 only 25% did. Other scholars argue that the rise of the Nones has been counterbalanced by a corresponding rise in the super-religious—in other words, they see an increase in polarization (Finke and Stark, 2005; Hout et al., 2001; Stark, 1999; Stark and Finke, 2000; Wuthnow, 1988). Yet others, faced with evidence of declines in belonging and behavior, have asserted that religiosity overall has not declined, because belief remains intact. Hout and Fischer (2002), for example, examined three possible explanations for the rise of the Nones (those who report no religious affiliation, regardless of their belief or behavior). Their three explanations are secularization, demographics, and politics. They reject the secularization explanation mainly because the majority of American Nones believe in God and an afterlife. They argue that the demographic explanation cannot wholly explain the rise of the Nones because demographic processes are too slow to account for the sudden change in religiosity. They conclude that politics was the primary factor. The growing detachment from organized religion in their view is associated with a backlash against the political powers of the religious right.

Proponents of the privatization hypothesis assert that even if fewer Americans report a religious affiliation, they continue to believe in God and have a spiritual life (Berger, 1967; Bibby, 1987; Luckmann, 1967; Swatos and Christiano, 2000). Chaves (2011) argues, for example, that “There is more to religious involvement than participation in organized religion.” Likewise, Idler (2001) says that private engagement in prayer and meditation is more prevalent than public attendance in religious services. Davie (1994) described the pattern in Britain as “Believing without belonging.” But this is not the only possible pattern. Day (2011) titled her recent book Believing in Belonging. Based on data from Europe and North America, she describes how people choose religious identification to complement other social and emotional experiences of ‘belongings.’ Using GSS data, Sherkat (2011) describes a pattern of belonging without believing. Sherkat argues that in the United States, unlike in other countries, there are more people who belong to religious groups who do not believe in God than there are people who believe but do not belong to religious groups.

Measuring the Three Dimensions of Religiosity

Discourse on the first B, belonging to or identifying with a religious group, has become contentious, specifically with regard to the rise of American Nones. This rise is a new phenomenon. As recently as 2000 Swatos and Christiano (2000: 8) suggested, “Virtually no empirical research supports the prediction of a societal slide from a peak of sacrality into a valley of secularity.” Since then there has been a substantial increase in the percentage of Americans who report no religious affiliation (i.e., Nones; see Hout and Fischer, 2002; Kosmin et al., 2009; Putnam & Campbell, 2010; Pew, 2012). Pew Research reports that the number of unaffiliated totaled 36 million adults in 2007 and grew to 47 million in 2012 (Pew 2012).

The second B, behavior, is measured by whether or not one chooses to devote time and financial resources to membership in a religious congregation. (This contrasts with the first B, belonging, which is strictly a measure of self-identification.) The public is almost evenly split between members and non-members of religious institutions. Some people join a congregation because it ties them to a religious community, which also provides social connections and support (Wuthnow, 1999). However, this involves commitment on the part of the individual or the family. Generally, participation in religious activities requires both economic and social commitments, payment of donations or fees, and conformance to the policy, schedule and set of rules of the congregation and its leaders. Given the voluntary nature of active religious participation and membership, only half of the population opts in.

To minimize risk of social desirability bias inherent in self-report data collection (Hadway, et al., 1993; Presser and Stinson, 1998), Presser and Chaves (2007) looked at the Bureau of Labor Statistics American Time Use Survey (ATUS). The ATUS records daily activities of a random sample of Americans 15 years and older. By asking respondents to report all types of activities, hour by hour, ATUS tracks Americans’ religious behaviors without directly asking them questions that might lead them to exaggerate. The estimates for Sunday attendance based on ATUS data1 for “religious and spiritual activities” range between 26% and 28% from 2003–2007 (Presser and Chaves, ibid.).

There is contradictory evidence on stability versus change in religious behavior. Presser and Chaves (2007) found stable weekly attendance of about 40% of the adult population at religious services from 1990 to 2006 based on General Social Surveys and American National Election Studies. The GSS asks, “How often do you attend religious services?” But the General Social Survey data from 1972–2008 showed an increase in the percentages of Americans who never attend religious services from about 13% in the early 1990s to 22% in 2008 (Chaves, 2011). That decline points to a shift toward less religious activity in the past two decades, corroborating our findings.

On the third B, belief in God, most surveys have failed to capture the complexity and nuance of Americans’ belief. This is not—or should not—be a simple yes-or-no question. The conventional wisdom that the United States is overwhelmingly a nation of believers (Wuthnow, 1988; Kosmin and Lachman, 1993; Lipset, 1996; Newport, 2012), while correct, obscures more than it reveals. The empirical evidence created by the basic question asked in the Gallup Polls since 1944, “Do you believe in God?” shows that well over 90% of Americans answer “yes”—affirming their belief in God. However, the percentage answering “no” rose from 1% in 1967 to 7% in 2011 (Gallup, 2011). As I will show, the sharp rise in “no” answers is just one indicator of a widespread degradation of religious belief in America.

This standard question on belief in God unintentionally overestimated the intensity of belief (Bishop 1999), as it masked what people actually think regarding the existence of God. In fact, levels drop drastically when Americans are asked more specific belief questions. For instance, when asked how certain they are about their belief in God, only 72% said “absolutely certain” in 1996. Furthermore, refined responses help distinguish between types of beliefs about God. As illustrated by an example from a GSS belief question asked in 1993, 1994 and 1998, a person who chose “I don’t believe in God” could be classified as an atheist; “I don’t know whether there is God, and I don’t believe there is any way to find out” as an agnostic; “I don’t believe in a personal God, but I do believe in a higher power of some kind” as a deist; “I know that God exists and I have no doubts about it” as a theist and a firm believer (only 63% in 1998). There are two additional optional responses in the GSS of less committed believers: “While I have doubts, I feel that I do believe in God” and “I find myself believing in God some of the time, but not at others.” (For full comparisons between levels of beliefs generated by different types of survey questions, see Bishop, 1999). Differences in questions asked over the years make it difficult to determine whether belief in God has diminished over time. Wuthnow (1988), citing the Gallup polls, asserted that most Americans claim some belief in God; however, fewer Americans than in the past are certain about their belief. Those who are absolutely sure God exists dropped from 66% of adults in 2003 to 54% in 2013 in Harris Poll surveys (2013).

Belief in the divine is the most common way American people express their religiosity. As pointed out by Bishop (1999), question wording regarding belief affects how respondents understand and respond. Agreeing that “God exists,” one might argue, is effortless (Beit-Hallahmi and Argyle, 1997). In fact, by following the majority, people dissociate themselves from the outcast atheists. In the U.S., after all, atheists are the most distrusted religious group (Edgell, et al., 2006). In a recent public opinion poll four out of ten Americans (43%) said that they would not vote for a presidential candidate who is an atheist,2 compared with 5% who would not vote for a Catholic candidate, 6% for a Jewish candidate, 18% for a Mormon candidate, and 40% for a Muslim candidate (Gallup, 2012).

Belief in the divine is the most persistent of the three B’s. It is the stronghold of religiosity. But this stronghold rarely stands alone. Anthropologist T.M. Luhrmann (2012) argues that belief is socially situated: It follows from rather than causes religious belonging and behavior. If this is so, then it should be expected that people will lose belief after ceasing belonging and behavior, and conversely gaining belief after affiliating with a religion and joining a congregation.

The next step is to build a religious-secular continuum that combines the three B’s into a single scale. This allows us to extract meaning and draw conclusions that aren’t obvious from regarding each of the three B’s separately. Movement in one direction along the continuum represents secularization, while movement in the other represents deepening of religiosity. This approach is not without controversy. Pearce et al. (2013) argue against measuring religiosity along a continuum from low to high. The authors chose to create five latent classes based on 12 indicators of religiosity in their study of American adolescents, naming the classes abiders, adapters, assenters, avoiders, and atheists. However, this approach has its own difficulties. Aside from the least religious and most religious groups—“atheists” and “abiders”—the classes are hard to distinguish from one another. They overlap significantly in beliefs, experiences and practices.

Despite intense interest and debate, the lack of an agreed-upon metric of secularization before now has complicated social scientists’ ability to reach consensus on the extent of the phenomenon. Fortunately, relevant data are available in the American Religious Identification Surveys (ARIS) of 2001 and 2008 and the General Social Surveys (GSS) of 1988, 2000, and 2010. It is impossible to track the religious journeys of individuals in surveys that collect only aggregate data. What is possible is to see the process of secularization unfold through the decline and growth of different population clusters, from the fully religious by the three measures of religiosity to the fully secular.

This leads to the first hypothesis:

Hypothesis I

I hypothesize that for society as a whole, the first of the three B’s to decline is religious behavior. It is followed by a decrease in religious identification, and only then by a decline in belief in the divine. Therefore the transformation from the most religious to the least religious pattern on the religious-secular spectrum will generate certain combinations of the three B’s that are more common than others. For example, the cell of the matrix representing people who lack belief but display religious behavior and belonging is predicted to be quite small.

What variables explain movements along the religious-secular spectrum? Demographic traits and religious upbringing help explain religious and non-religious behaviors and attitudes. One of the most notable features of the rising Nones is their demographic profile—young, male, and highly educated (Kosmin and Keysar, 2006; Baker and Smith, 2009). The Pew survey (2012) found that 32% of young Americans have no religious affiliation, while according to ARIS 2008 29% of Americans 18–29 years old professed no religion (Kosmin, and Keysar, 2009). Religious leaders have long observed the demographic gap in their pews. Many young people find organized religion irrelevant and outdated (Religious News Service, 2012). For psychologists youth is associated with rebellion and rejection of organized religion (Beit-Hallahmi and Argyle, 1997; Keysar, 2007). Older Americans are more likely to be sure of God’s existence (Keysar, 2007).

Gender is another important demographic attribute associated with religious and secularity gaps. Intensity of religiosity and worldviews is gendered. Persistently different patterns for men and women have been observed in Western societies (Baker and Smith, 2009). Women are well known to be more religiously active than men (Beit-Hallahmi and Argyle, 1997). In Western European and Scandinavian societies, men were found more likely to be non-believers in an afterlife who ‘never pray’ (Furseth, 2010).

Women tend to seek comfort and emotional support in religion, while men express a greater tendency to disbelieve and reject authority (e.g. Beit-Hallahmi, forthcoming). Socialization is one factor. Many women are raised to be mothers and to take care of their families (Douglas, 1977), and be nurturing and submissive, qualities that are associated with greater levels of religiousness (Thompson, 1991). Voas and Crockett (2005: 24) studied religious patterns in Britain and asserted that “daughters are generally influenced by their mothers – and there are signs in the data that maternal influence on religiosity is stronger for female than for male teenagers – the gender gap may be partly self-perpetuating.”

Miller and Stark (2002) challenged the religious socialization theory, which holds that the different societal roles assigned to men and women help explain the differences in their religiosity. In a cross-cultural and cross-religion exploration they offered risk-taking as an alternative approach, claiming that non-belief and irreligious behavior involve risk-taking. “Failure to conform in terms of beliefs and practices, or the commission of ‘sins,’ can result in serious consequences, such as going to hell” (ibid. 1404). Men are known to be risk-takers in other domains, such as drinking, drug use, smoking and adultery, as well as in financial decision-making. These risky behaviors may also apply to irreligious behavior. According to Miller and Hoffman (1995) reluctant risk-takers were more religious, effects that held within each gender. Why are women less likely to be risk-takers? Miller and Hoffman (1995) attributed gender gaps in risk taking to differential socialization, as boys are encouraged to be courageous and aggressive while girls are encouraged to be passive and gentle. Miller and Stark (2002) concluded that differences in risk-taking probably have a biological rather than sociological explanation. Comparing the United States and Japan using World Values Survey data (1995), they showed that on measures of religious affiliation, practice and belief in the supernatural, gender differences were by far smaller and less significant for the Japanese than Americans, explaining the narrower gender difference by the lower risk of being irreligious in Japan.

Economists of religion also challenge the importance of religious socialization, arguing that while it applies to traditional societies it is less relevant in modern industrialized economies. Economists of religion apply a human capital framework (Iannaccone, 1990). They say that in societies with a traditional familial division of labor, religion is regarded as a household commodity, which belongs to the female sphere. As women make strides in higher education and increase their participation in the labor force, economic realities overwhelm traditional religious socialization. As women take upon themselves tasks that were previously assigned to men, traditional religious practices are also challenged, and women’s religious involvement and worldviews may change as well. In pursuit of empowerment, some women join new religious movements and alternative spiritual communities (Berger, 1998; Finley, 1994); some try to feminize their religious community; others simply follow their male counterparts and reject organized religion (Furseth, 2010).

Geography also plays a role. The geographic clustering of American non-identifiers, non-members and non-believers in the West fits its classification as the “none zone” in the Religion by Region series (Killen and Silk, 2004).

This leads to the second hypothesis:

Hypothesis II

I hypothesize that one’s place on the 2x2x2 matrix of belonging/behavior/belief will differ systematically by demographic characteristics, such as gender, age, geography, and religion. Specifically, I expect men and young Americans to be more secular, and those who reside in the South to be more religious.

The third hypothesis examines the connection between religiosity and public opinion. The American public is divided on societal issues, which often develop into political rifts. Fiorina and Abrams (2008: 584) described the polarization of the American public along lines of religiosity as well as political membership, asserting, “A significant degree of sorting has occurred, however—most clearly between members of the two parties, but also along lines of religion and possibly geographic location.”

Some of the debates are over legalizing same-sex marriages and abortions. Other discourses challenge the scientific community over stem cell research and school curriculum, mainly science education. These debates, often characterized as culture wars, reflect people’s religiosity and secularity (Keysar and Kosmin, 2008; Kosmin, 2013). For methodological reasons and availability of data, this paper focuses on attitudes relating to science, which are at the heart of societal debates on public policy issues.

One important issue in the culture wars is evolution. The American public is deeply divided over accepting or rejecting human evolution3 (Miller and Pennock, 2008). This division directly influences battles over science education and contributes to the emergence of ideological politics (Mooney, 2005; Pennock, 2005). A great deal has been written on Americans’ attitudes toward scientific issues, ranking the U.S. nearly last in acceptance of evolution (only Turkey ranks lower among 34 surveyed nations;4 see Miller, Scott and Okamoto, 2006). Evolution challenges fundamentalist Christians’ teachings about God’s powers and the belief that human beings are the direct creation of God (Pennock, 2007). The creationist movement’s battles over scientific curriculum, inflamed by religious motivations, are seen by some as “part of a broad assault on rationality and on secular institutions” (Blackburn, 2008: 45). Religious fundamentalist worldviews clash with science over the issue of human evolution; as Miller and Pennock (2008) explain, “The antipathy that so many Americans feel toward the idea of being related to animals is part of the reason that the religious right has used evolution as a political wedge issue” (17).

These attitudes towards scientific issues are significantly related to Americans’ self-identification with a religious group. When asked in 2008, “Do you think that human beings, as we know them, developed from earlier species of animals?” 17% of Americans in general said they ‘definitely’ accept human evolution while 36% said ‘definitely not.’ However, Americans who profess no religion in 2008 exhibited notably different worldviews: 33% ‘definitely’ accepted human evolution while 17% said ‘definitely not.’ (Kosmin, et al. 2009). Attitudes towards human evolution have never before been tested against the combinations of the three B’s.

This leads to the third and fourth hypotheses:

Hypothesis III

I hypothesize that one’s position on the religious-secular spectrum is related to one’s attitudes towards current social debates, specifically towards human evolution. Moreover, I hypothesize that Americans who reject the theory of human evolution will cluster in the religious end of the spectrum.

Hypothesis IV

I hypothesize that the religious-secular scale is a useful tool to assess people’s perceptions on ideological and educational issues. Specifically, I believe that the most religious and the most secular remain at opposite poles on attitudes towards human evolution even after demographic and religious variables are controlled for.


This paper utilizes several sources of data. For the exploratory analysis and creation of the religious-secular scale, I utilize the American Religious Identification Surveys (ARIS). To validate the scale generated by the ARIS data, I supplement it by creating similar grids utilizing the General Social Surveys (GSS). For the multivariate regression analysis, I use the ARIS 2008 data.

I conduct both descriptive and multivariate statistical analyses. Four regression models test shifts along the religious-secular spectrum. Similar to hierarchical steps, in each model I add more variables. I start with socio-demographic factors, add geography, then religious indicators, and lastly worldviews as predictors for shifts along the religious-secular spectrum from the most religious point to the most secular point.

Data Sources

American Religious Identification Surveys (2001 & 2008)

The data are drawn from the findings of the American Religious Identification Survey series, in particular ARIS 2001 and ARIS 2008. Both ARIS 2001 and ARIS 2008 are nationally representative random-digit-dialed telephone surveys, each with more than 50,000 adult respondents: ARIS 2001 with 50,281 and ARIS 2008 with 54,461. The data are based on self-reporting in response to an open-ended question: What is your religion, if any? Religious self-identification is not based on whether established religious (or non-religious) bodies, institutions, churches, mosques, or synagogues consider them to be members. To the contrary, the surveys sought to determine whether the respondents regarded themselves as adherents of a religious community. The open-ended approach generated a significant number of responses or categories of religious groups, faiths, and denominations (Kosmin and Keysar, 2006). Similarly, other segments of the American population, those who self-identify as “atheist,” “secular,” or “humanist,” chose these options voluntarily.

For a more detailed inquiry into religious beliefs, ARIS 2001 incorporated a nationally representative sub-sample of 17,911 adults, collecting information on beliefs in God and membership in religious institutions, while ARIS 2008 utilized a nationally representative sample, albeit smaller, of 1,015 adults who were asked more detailed questions on life cycle religious behaviors and beliefs in the divine.

Two methodological adjustments were introduced in the data collection of ARIS 2008. In 2008 the traditional landline interviews were supplemented with a national cell phone survey to include the growing segment of the population who use cellular telephones exclusively. In addition, interviews were conducted both in English and Spanish. The use of cell phones interviews and inclusion of Spanish interviews increased the coverage and reduced the non-representation bias.

General Social Surveys (1988, 2000, 2010)

The General Social Survey (GSS) is an annual or biennial probability sample of households conducted by the National Opinion Research Center (NORC) of the University of Chicago. Face-to-face interviews are conducted with one randomly selected adult in selected households. Since 2006 interviews in Spanish have been added. The GSS is highly regarded due to its high response rates (between 70–80%) and its wide range of demographic, behavioral and attitudinal questions, including special societal and political topics. Many of the core questions have remained unchanged since 1972, allowing investigators to conduct trend studies. Unlike the ARIS series, the GSS main question on religion provides closed options to choose from: What is your religious preference? Is it Protestant, Catholic, Jewish, some other religion, or no religion?

Measuring the 3 B’s—Descriptive Analysis

1. Belonging to, or Identifying with, a Religious Group

Belonging is measured by answers to the open-ended question: What is your religion, if any? For purposes of this article, survey respondents are categorized as “belonging” if they self-identify with a religious group, whether or not they are official members or belong to a particular congregation or attend services. One important measure of secularity is professing no religion. People who profess no religion include those who replied “none,” “agnostic,” “atheist,” “secular” and “humanist” to the open-ended religious identification question. They numbered 14 million adults in 1990, 29 million in 2001 and over 34 million in 2008 (ARIS 2008).

In all, the responses aggregated as “No Religion” amounted to 8.2% of the American adult population in 1990, 14.1% in 2001 and 15.0% in 2008. These findings were corroborated by the General Social Survey (GSS) in 2004. The GSS asked: “What is your religious preference?” and 14.3% said “none.” A survey by Pew found that 16% of adult respondents said they were religiously unaffiliated in 2007 and just under 20% in 2012 (Pew, 2007, 2012).

Of those people who reply “none” to a survey question on religious identification, only a fraction self-identifies as “agnostic” or “atheist.” Nevertheless, these small groups have almost doubled in number from 1.9 million in 2001 to over 3.6 million adults or 1.6% in 2008.

2. (Religious) Behavior

Another obvious social manifestation of being connected to religion is one’s behavior, and one way this can be indicated is by membership in a religious congregation. The reasons for not being a member of a religious congregation vary widely, from ideological attitudes to physical access issues. Although the population of those who are members of a religious congregation is very large, ARIS 2001 found that at the beginning of the 21st century, 46% of American adults, nearly 100 million people, did not regard themselves as or claim to be members of religious congregations.

It is hard to assess to what extent members of houses of worship are religiously active. Therefore, ARIS 2008 looked at participation in religious services and did not simply ask about congregational membership. Despite the limitations, one can assume that those who attend religious services regularly are also members. In general, ARIS 2008 showed that 55% of American adults said they attend once a month or more often, 20% of Americans said they never attend religious services and 23% said they attend only a few times a year. (As noted above, these numbers are subject to over-reporting.) The latter group could be intermarried, leading the nonreligious partner to join his/her more religious family member occasionally at services. Putting these latter two groups together, 43% of Americans say they do not attend religious services regularly5.

3. Belief in God

A new belief question was introduced in ARIS 2008 to better understand people’s perceptions regarding the existence of God. While 70% “definitely” believe in a personal God, an additional 12% believe that there is a higher power but no personal God. These two categories combined (82% of respondents) are approximately the same as the single category from the 2001 survey, those who believe in the existence of God, which comprised 95% of respondents. The remaining categories from the 2008 survey are hard to describe as believers. They are “there is no such thing” (2%), “there is no way to know” (4%), “I’m not sure” (6%) and refused (6%).

The Complexity of the Three B's

The three dimensions of religiosity and secularity reveal the complexity of slicing the American religious/secular pie. Each dimension when looked at separately shows a different pattern revealing the asymmetric relationship of the religious/secular in each dimension:

  • Belonging (Identification):
    • In 2001 & 2008 there was an 80/20 split between identifiers and non-identifiers.
  • Behavior (Membership/Attendance):
    • In 2001 & 2008 there was a 50/50 split between members and non-members in religious organizations.
    • In 2008 there was a 55/45 split between attenders and non-attenders in religious services.
  • Belief in God:
    • In 2001 there was a 95/5 split between believers and non-believers in the existence of God.
    • In 2008 there was an 82/18 split between believers and non-believers in a personal God or higher power.

Creating a Scale: Belonging, Behaving and Believing 2001 & 2008

To capture all three dimensions together, I created a scale which takes into account for each respondent whether s/he 1) identifies with a religious group; 2) holds household membership in a religious institution (2001) or attends religious services regularly (2008); and 3) believes in God, i.e., agrees that “God exists” (2001) or believes in a personal God or higher power (2008).

Table 1 and Figure 1 present the overall distribution of these eight (2x2x2) options for both 2001 and 2008. They show how the frequency of the eight combinations of belonging, behaving and believing have changed over time. Although some of the measurements were modified in the two surveys, they remain similar enough that changes can still be monitored. The challenge is to create a linear scale for the degree of religiosity using dichotomous variables of qualities that are not directly comparable. In the regression analysis (below), clusters are created based on the number of “yes” answers a respondent gave on the three B questions: 0, 1, 2, or 3. The ordering within each cluster was based on the level of religiosity of the different combinations as discussed below.

Table 1

Religious Identification, Membership/Attendance and Belief U.S. Adult Population, 2001 and 2008 (weighted). Sources: American Religious Identification Survey, 2001 n = 17,911; American Religious Identification Survey, 2008 n = 1,015

Identification Membership/Attendance Belief in God Share of Population 2001 Share of Population 2008

No No No 2.8% 6.0%**

No No Yes 8.2% 9.5%**

Yes No No 1.0% 4.1%**

No Yes No 0.5% <1%

Yes Yes No 0.7% 1.1%*

No Yes Yes 2.2% <1%**

Yes No Yes 30.0% 24.7%**

Yes Yes Yes 54.5% 53%**

** < .01; * < .05

Legend (excluding ‘don’t know’ and ‘refuse’ in both surveys)


Identification: Yes = Identifies with a religion; No = ‘None,’ ‘Atheist,’ ‘Agnostic’ ‘Secular,’ ‘Humanist’

Membership: Yes/No Is anyone in your HH currently a member of a church, temple, or mosque?

Belief that God Exists: Yes = Agree strongly/Agree somewhat; No = Disagree somewhat/Disagree strongly


Identification: Yes = Identifies with a religion; No = ‘None,’ ‘Atheist,’ ‘Agnostic’ ‘Secular,’ ‘Humanist’

Attendance: Yes = Attend more than once a week/Once a week/Once or twice a month; No = A few times a year/Never

Belief in God: Yes = There is definitely a personal God/There is a higher power but no personal God; No = There is no such thing/There is no way to know/I am not sure

Figure 1 

Changes in Belonging, Behavior, and Belief. Source: ARIS 2001 and ARIS 2008

Legend: See Table 1

Because this is not a longitudinal study, it is impossible to follow individuals over time to observe their movement between the eight groups. I can only infer individuals’ movements from changes in the relative sizes of the groups. I make the parsimonious assumption for this study that when people shift, it is to a neighboring cluster, such as from (yes, yes, yes) to (yes, no, yes). Support for this assumption is provided by a longitudinal study of young Conservative Jews from 1995 to 2003, which showed a large number of students ceasing to attend services while still belonging and believing, but few or none jumping from fully religious in one survey wave to fully secular in the next (Keysar and Kosmin 2004). Certainly there are some individuals who do so, but they are outliers.


1. Using ARIS Data

Two results emerge: Wide variation in degrees of religiosity, and a trend toward secularity at all points along the spectrum. There is a decrease in the share of Americans in the most religious groups and a corresponding increase in the least religious groups. The most religious group shrank slightly from 54.5% to 53% but still constitutes the largest sub-group of the adult population. It represents adults who report identifying with a religion, holding religious membership or attending services regularly, and believing in God (the most religious category – (yes, yes, yes)).

More noticeably at the other end of the scale, there is a doubling of the most secular combination: people who do not identify with a religious group, do not belong to a congregation or never attend religious services (or go just a few times a year), and are atheists or agnostics with regard to a personal God (no, no, no).

Changes also occur in the middle points of the scale where the three dimensions are not the same (in other words, not all ‘yes’ or all ‘no’). On the secular side of the spectrum are those who belong yet do not believe or behave religiously (yes, no, no). Although this is a small minority, it is four times its share of the American population compared with 2001.

On the religious side of the spectrum there is a significant decrease (from 30% to 25%) in those who belong, do not participate regularly in religious services, yet believe in a personal God (yes, no, yes). However, this remains the second biggest group of Americans, outnumbered only by the most religious group (yes, yes, yes). Although I am not tracking the same people over time, ceasing regular participation in religious services remains a common choice for Americans, presumably when they leave the (yes, yes, yes) group. The fact that this group’s share of the population shrank from 2001 to 2008 suggests that some people who were once in the (yes, no, yes) group have taken a further step away from religiosity, to (no, no, yes) or (yes, no, no). Those two groups together grew by 4.4 percentage points from 2001 to 2008—coincidentally, about the same as the decrease in the (yes, no, yes). Clearly the transition from a religious to a more secular society is not a drastic change from one extreme (yes, yes, yes) to the other extreme (no, no, no).

Further evidence for the importance of membership and attendance as a bulwark of religiosity is that people who do attend services are highly unlikely to drop either of the other two B’s. It is rare to find a person who attends but does not believe (yes, yes, no)=1%, or who attends but does not have a religious identification (no, yes, yes)<1%, or who attends but does not believe or have a religious identification (no, yes, no)<1%.

Beyond a shift towards more secularity, Figure 1 illustrates the particularity of American religion and non-religion. There are persisting combinations, while other combinations along the religious-secular spectrum do not survive. Most notably, the pattern of believing without belonging (no, no, yes), constituting one in every ten Americans, is more common than belonging without believing (yes, no, no), supporting the claim that belief is the last bastion of American religion. This latter group, although small (only 4%), seems to have gained adherents at the beginning of the 21st Century. These are people who are probably attached to a religious group nominally. They do not attend religious services regularly nor do they express belief in a personal God. Affiliation with a religious group might reflect family loyalty and nostalgic sentiments toward their religious upbringing (Beit-Hallahmi, 1997).

It is quite revealing to document which religious-secular combinations are rare or even endangered, namely, attending religious services without belonging to a religious group or believing in God (no, yes, no). Some people might feel lost in houses of worship without an attachment to a specific religious group or strong belief in the divine. Likewise, the share of Americans who do not belong yet behave or believe is shrinking (less than 1%). In other words, I could hardly find any Americans who attend religious services regularly and believe in a personal God yet do not identify with a religious group (no, yes, yes), or non-believers who identify with a religious group and participate at services regularly (yes, yes, no), perhaps because the latter as non-believers feel uncomfortable in religious congregations and communities.

2. Corroboration from GSS

NORC’s General Social Survey allows us to create a similar religious-secular grid of identification/attendance/belief for 1988, 2000 and 2010, when the same questions were administered by the GSS.

GSS data corroborate the ARIS trend data by showing how the two polarities, the religious and secular, change over time. As the share of the most secular segment goes up, the share of the religious part is going down. Furthermore, the earlier starting point for the GSS allows for a longer time span than ARIS. The results show a greater drop in the share of the most religious segment of American society (yes, yes, yes) from 63% in 1988 to 54% in 2010, and an increase in the most secular segment of American society (no, no, no) from 2% in 1988 to 7% in 2010. The share of believers who neither identify with a religious group nor attend religious services (no, no, yes) increases from 5% in 1988 to 10% in 2010, according to GSS (See Figure 2).

Figure 2 

‘Corroboration from GSS: Changes in Identification, Attendance, and Belief. Source: GSS


Identification: Yes = all religions; No = no religion

Behavior:How Often Attend Religious Services?

Yes = Attend several times a year/once a month/2–3 times a month/nearly every week/every week or more; No = Never attend religious services/less than once a year/once a year

Belief:Respondent’ Confidence in the Existence of God?

Yes = Some higher power/believe sometimes/believe but doubts/know God exists; No = Do not believe/no way to find out

Beyond portraying similar trends, the GSS and ARIS data establish remarkably similar estimates on both ends and the middle of the religious-secular spectrum.

Looking at several cohorts in the GSS, Sherkat (2011) also finds a growing segment of Americans who neither believe nor belong to religious organizations. ARIS revealed the initial dimension – participation in religious services – decreased when society experienced shifts in religiosity. Both ARIS and GSS data provide support for hypothesis I by showing that most of the change on the religious-secular grid to date has been in religious behavior, with a smaller decrease in religious identification, and only small changes in belief. However, aggregate data cannot fully establish the validity of this hypothesis. That will require a longitudinal study showing the sequence of movements of individuals between the different cells of the 2x2x2 matrix.

ARIS and GSS data show that the most prevalent secular Americans are people who do not identify or behave religiously but do believe in God (no, no, yes) and those who totally distance themselves from all three dimensions of religion (no, no, no). But another group has gained in numbers: identifiers who do not participate in religious services and do not believe in the divine (yes, no, no). There is also a group that has almost disappeared: (no, yes, yes). These are people who do not identify with a religious group but do attend religious services regularly and believe in God.

3. Socio-demographics

A. Gender

In 2001 men dominate (72%) the small group of non-identifiers who are involved with a religious institution and don’t believe in God (no, yes, no). Indeed, the top three most male groups (at least 2/3 male) are non-believers. The fourth group from the top, again a small one, consists of those who don’t identify with a religious group, yet are members of religious institutions and believe in God. Females, on the other hand, dominate only two groups, the two most religious combinations, which are the largest groupings, representing almost 80% of the adult population. In both of these combinations people believe in God and self-identify with a religious group (See Figure 3).

Figure 3 

Religious Identification, Membership and Belief Scale by Gender in 2001. Source: ARIS 2001

The relatively small sample in 2008 precludes a similar detailed presentation of the eight-point scale by gender. Even so, similar patterns persist. Most striking, both in 2001 and in 2008 men are twice as likely as women to be in the most secular group, (no, no, no), and also among the believers who do not belong and do not attend services, (no, no, yes).

About 10% of men in 2008 are found at the extreme (no, no, no) category. An additional 10% follow the (no, no, yes) pattern. That means that one in every five males is in these two secular groupings. Similarly, 47% of males versus 58% of females are found in the most religious category (yes, yes, yes). Moreover, the (yes, no, yes) grouping, the group of non-attenders, which plummeted in 2008, decreased mainly among men, from 30% in 2001 to 22% 2008 (and from 32% to 26% respectively among women).

B. Age

Two groupings stand out because of their large number of young people. More than half of the (no, yes, yes) category consists of young adults under age 35, and 47% of the (no, no, yes) category is under age 35. As expected, young people are found in the three non-believer groups: (no, no, no), (no, yes, no), and (yes, no, no). In each about 42% of group members are adults under age 35. In 2008, young people are over-represented among non-identifying adults, who don’t attend services regularly, yet believe in a personal God (no, no, yes), and among those who belong, but do not behave or believe (yes, no, no). Two groups stand out in 2008, the believers who do not belong or behave religiously, the (no, no, yes) and the most religious one, (yes, yes, yes). While the first group is relatively young, the last one, the most populous, is far older.

C. Geography

The two polarities of the religious-secular spectrum are concentrated in different geographic regions. While the growing, yet small, secular polarity (no, no, no) is most likely to be found in the West (about 40%), the somewhat dwindling religious polarity (yes, yes, yes) is most likely to be found in the South (again about 40%). Similarly, the other religious segment (yes, no, yes) is most likely to be found in the South.

D. Upbringing and Life Cycle Behavior

Among people who had a religious initiation, such as a baptism, christening, circumcision, confirmation, bar mitzvah or naming ceremony, indicating some religious upbringing, 60% believe, behave and belong (yes, yes, yes) as adults, compared with only 42% of those who did not have a religious initiation. Those who did not have an initiation are by far more likely to be in the most secular grouping (no, no, no): 13% compared with 5% of those who had an initiation ceremony. Although these results are highly significant they are correlations and do not indicate a causal relationship.

I hypothesized (II) that the type of dimension (belonging/behavior/belief) one loses might be different systematically by demographic characteristics and geography. That was supported. Americans who reside in the South are more religious in all three dimensions. The evidential gender gaps clearly support hypothesis II whereby more men are clustered in the secular parts of the religious-secular grid.

E. Attitudes toward Human Evolution

The most non-religious (no, no, no) people overwhelmingly accept human evolution. In contrast, people at the other extreme (yes, yes, yes) overwhelmingly reject it. In between, believers, the (no, no, yes), are split almost evenly in their opinions on human evolution, while identifiers, the (yes, no, no) people, tilt toward accepting human evolution. But the pattern reverses toward rejection of human evolution among the (yes, no, yes) grouping. In other words, believers who identify with a religious group but do not participate in religious services regularly tend to adopt a religious view toward human evolution.

When I combined three small groups, (no, yes, no), (no, yes, yes) and (yes, yes, no), into one group—an ambivalent group—there is a strong linear relationship between the support of human evolution and the religious-secular scale. The more secular a person the more likely s/he finds human evolution to be “definitely true.” Simultaneously, flat denial of human evolution peaks at 78% among the most religious group and drops sharply to 9% among the most secular group. Figure 4 shows that the “somewhat secular” (no, no, yes), believers who neither identify nor actively engage with religion, are at the crossover point: roughly equal numbers of them describe human evolution as “definitely false” and “definitely true.”

Figure 4 

Attitudes toward Human Evolution and the Belonging, Behavior, and Belief Scale 2008. Acceptance of Evolution by Religiosity: Most and Least Accepting. Source: ARIS 2008

Figure 5 shows a gap between the two poles on the religious-secular scale. The most secular group overwhelmingly believes in human evolution (78% said “definitely yes”) while a great majority of the most religious group flatly rejects it (78% said “definitely no”).

Figure 5 

Probability of Accepting or Rejecting Evolution: The Most Religious versus the Least Religious 2008. Source: ARIS 2008

The religious-secular grid is highly correlated with people’s attitudes towards a highly debated social issue, here measured by attitudes towards human evolution. This scale effectively exhibits the two polarities of the American public—the most religious versus the most secular—illustrating how they are harbingers of extremely different worldviews.

Multivariate Models

The multivariate methods aim to explore religiosity and irreligiosity and also test hypothesis IV. The goal of the analysis is to test how the various clusters of the three B’s are created and explore changes in religious intensity. One way is to monitor the number of positive answers respondents gave on the three B questions on a scale from the most secular position with no positive answer, namely category 0=no, no, no to 1 positive answer, “yes,” to 2 “yes” and finally, the most religious position with all positive answers: 3=yes, yes, yes. I predicted that if people lose the first two dimensions they will soon lose the third and be regarded as nonreligious. The models segment the population based on socio-demographics, region and religion in order to predict the likelihood of progressing along the spectrum of belonging, behaving and believing.

There is, however, one feature of the American religious landscape that needs to be taken into consideration. The belongers, behavers and believers (yes, yes, yes) constitute half of the American population, arguably not an extreme group, but rather a majority. In order to improve the explanatory power of the multivariate models, I sliced the American pie further by splitting this large religious group, and creating a 9-point scale (rather than an 8-point scale). The asterisk after the word yes signifies greater religious intensity. In the “extremely religious” (yes, yes*, yes*) group are Americans who identify with a religion, attend religious services weekly or a few times a week, and “definitely” believe in a personal God. They are about 40% of the adult population. Looking at their socio-demographics, they are overwhelmingly older and more likely to be female. The share of the ‘extremely religious’ is far more prevalent in the South than in other geographic regions, the Northeast or in the West and Midwest.

Table 2 presents hierarchical regression models for the 9-point religious-secular scale, whereby 1=(no, no, no), i.e. the most secular group of non-identifiers, non-behavers and non-believers and at the other extreme, 9=(yes, yes*, yes*), i.e. the most religious group of identifiers, weekly attenders who definitely believe in a personal God. To repeat, in the religious-secular scale, the first item represents identification with a religion, the second represents attendance at services, and the third represents belief in a personal God. The 9-point scale is defined: 1=(no, no, no); 2=(no, no, yes); 3=(yes, no, no); 4=(no, yes, no); 5=(yes, yes, no); 6=(no, yes, yes); 7=(yes, no, yes); 8=(yes, yes, yes), 9=(yes, yes*, yes*).

Table 2

Hierarchical Regression (OLS) of Belonging, Behaving, and Believing 9-Point Scale 2008

Dependent Variable BBB Scale 1 = no, no, no… 8 = yes, yes, yes 9 = yes, yes* yes*

Predictors Model 1 Model 2 Model 3 Model 4


Gender -.137*** -.138*** -.073* -.052

Age .103** .103** .082** .041


Region -.135*** -.093** -.063*

Life Cycle Religious Behavior

Initiation -.115*** -.129***

Religious Funeral -.465*** -.364***

Religious Attitudes

Human Evolution -.340***

Adjusted R2 .031 .048 .291 .385

*** < .001; ** < .01; * < .05

Note: Standardized regression coefficients are shown. The baseline are adult females, living in the South, who had a religious initiation, such as baptism, who expect to have religious funeral services, and do not believe in human evolution. Age is a continuous variable.

I chose the most religious Americans as the base categories: females, living in the South, who had a religious initiation, plan a religious funeral, and do not believe in human evolution. I introduced the various variables in stages, in all utilizing four models, to predict shifts along the spectrum from the most secular to the most religious. The first model introduces demographic predictors, the second adds geography, the third adds life-cycle religious behavior, and the last adds attitudes towards human evolution.

Model 1 shows that the demographic factors, gender and age, are significant predictors of the 9-point religious-secular scale. Being male significantly reduces the likelihood of being placed higher on the religious spectrum (negative coefficient) while being older is positively associated with greater religiosity. Model 2 indicates a significant effect of region, whereby those who live outside the South are less likely (negative coefficient) to religiously belong, behave and believe in 2008.

Model 3 adds religious initiation and end of life ritual, which elevate substantially the power of the analysis (Adjusted R2 increases sharply to 0.291). Model 4 increases the explanatory power even further (Adjusted R2 =0.385) by adding an attitudinal variable of a contentious topic, people’s perceptions towards human evolution. Those who believe in human evolution are the most secular along with those who do not expect to have a religious funeral, supporting hypothesis IV. Interestingly, gender is only marginally significant at the 0.06 level, and age is not statistically significant in this model. It raises the possibility of an interaction between age and religious attitudes.

I tested the interaction between religiosity and worldviews within each age group to assess whether the effects of attitudes toward human evolution on the religious-secular spectrum operate differently for younger or older Americans. Examining the interaction of age and attitude towards human evolution shows that people’s attitudes are statistically more important in determining their placement on the religious-secular scale than their age is. To illustrate: between 14% (those under 35) and 19% (those over 35) of believers in human evolution are in the most secular category (“no” on belonging, “no” on behavior, and “no” on belief) compared with only 1% of those who do not believe in human evolution regardless of age. In contrast, between 17% (those under 35) and 19% (those over 35) of deniers of human evolution are in the extremely religious group of identifiers, which includes weekly attenders who definitely believe in a personal God, compared with 2% - 5% of believers in human evolution.


Four limitations should be considered when interpreting the findings presented. First, the wording of questions about behaving and belief changed slightly from 2001 to 2008. (There was no change in question wording for belonging.) While this limitation must be acknowledged, it is highly unlikely to have affected trend estimates. In 2001, the behavior question asked whether anyone in the respondent’s household was currently a member of a church, temple, or mosque. In 2008, the question was how frequently the respondent attended religious services. The respondent was categorized as a “yes” for behavior if he or she attended more than a few times a year. That is similar to answering, “yes” in 2001 to being a member of a religious institution. As for belief, in 2001 people were asked if they agree (strongly or somewhat) that God exists. In 2008, the question was more detailed. A “yes” on belief was given to people who answered that there is definitely a personal God or who said there is a higher power but no personal God. Regrouping categories helped us minimize the effects of different question wordings.

A second drawback is the relatively small sample in 2008 compared with 2001. Nevertheless, ARIS 2008 is a representative national sample of the American adult population. A third issue concerns limited availability of data critical to understanding social divisions and culture wars. The only available data was on attitudes towards human evolution. Clearly a battery of questions on highly contested issues that the American public grapples with, such as abortion, climate change, gun control, immigration, and same-sex marriage, would help determine the correlation between religiosity and worldviews. To compensate for the limitations of the ARIS data, GSS findings were also included. The corroboration from the GSS validates the ARIS findings by showing similar patterns and trends. Also, the GSS questions were identical in each of the three years, strengthening the significance of the observed trend.

Fourth and finally, neither the ARIS series nor GSS were designed as longitudinal studies and they do not track the same individuals over time. To capture the dynamic and characteristics of shifts along the religious-secular spectrum researchers would have to follow people over their life course—from adolescence to young adulthood and further as they develop romantic partnerships, get married and become parents. These life events involve decision-making on religious behavior and participation. Following the same individuals over time would also shed light on the role of socio-demographic factors and help predict future trends. A longitudinal design could shed light on the sequence of transitions from a religious to secular society. which trigger people to alter their religious behavior, identification and belief.


The ‘de-Christianization’ of American society, namely the sharp decline of those identifying with Christian denominations from 86% of the population in 1990 to 76% in 2008 (ARIS 2001 & ARIS 2008) and to 73% in 2012 (Pew 2012), was not offset by a rise in new religious movements (as predicted by Stark, 2000; Barker, 1989) but rather by a rejection of all organized religions and the marked increase in the ‘none’ population to 16%-20% of the adult population. Of course, this still leaves a great majority of people retaining a religious identity, though many of them don’t belong to a religious institution or participate in religious services.

The notable rise of the Nones, people who profess no religion, began in the 1990s (Kosmin and Keysar, 2006; Putnam and Campbell, 2010; Chaves, 2011) but remained under the radar for a long time. Even the ARIS 2001, which showed the growth of Nones from 8% in 1990 to 14% in 2001, was met with skepticism. In a USA Today article in March 2002, for example, American sociologist of religion Rodney Stark said, “People aren’t really saying, ‘I have no religion.’ …they are just unchurched.” It took a few years until corroborating results – 14% of Nones found by GSS 2004; 15% by Pew (2007); and more recently 20% (Pew 2012) – stirred the public and religious leaders to grapple with the secularization trend (Religious New Service, 2012).

This paper demonstrates that the growing secular segment of the American population is becoming heterogeneous. On the eight-point religious-secular spectrum, the last two groups on the grid (yes, yes, yes & yes, no, yes) are classified as ‘religious America’ and the six other groups, 22% of the population, as ‘secular America.’ Obviously these latter 55 million adults are not all similar to each other. They include men and women, young and old, poor and rich. They live in the East, West, North and South. For religious America there is more stability among the (yes, yes, yes) than among the (yes, no, yes). The latter segment seems to lose adherents. Furthermore, all combinations indicating ‘yes’ to attendance in services are losing adherents. It is possible that the path from a religious to a secular mode is first manifested by a drop in religious service attendance. Durkheim made this point a century ago. In The Elementary Forms of Religious Life (1915:44) he wrote that “the idea of religion is inseparable from the idea of a Church” and “religion must be an eminently collective thing.” The “Church” is far weaker than it was in Durkheim’s time. Rather than participation in religious activities following from belief, it may be more common that belief follows from religious activities. As more Americans turn away from religious services, I expect religious uncertainties and disbelief will grow.

The most secular segment on the grid, the (no, no, no) group, has doubled in the first decade of the 21st Century. Some deliberately depart from the majority and from the traditional American norms, while others drift into this minority status. Their youth and high educational level hint at possible further growth and their potential influence on the general population beyond their relative minority. In this project I showed that the secular extreme is growing at the same time as the religious extreme is declining.

Norris and Inglehart (2004: 18) asserted that,

The United States remains an outlier among postindustrial societies, having a public that holds much more traditional worldviews than that of any other developed country except Ireland. But even in America, there has been a lesser but perceptible trend toward secularization; the trend has been partly masked by massive immigration of people with relatively traditional worldviews…but when one controls for these factors, even within the U.S. there has been a significant movement toward secularization.

The findings presented here support the conclusion on contemporary trends of American religion by Chaves (2011: 110): “no indicator of traditional religious belief or practice is going up. There is much continuity and some decline.” And, “If there is a trend, it is toward less religion.”

The current paper demonstrates the growth of the most secular, and the decline in the religious segments and groups in the middle. The process is complex but for the most part consistent. Remarkably, the surveys show a wide-ranging migration in the population from (yes, yes, yes) toward (no, no, no). This paper has explained the “what.” Further research is required to establish the “why.”

The religious-secular scale also demonstrates that the most religious and the most secular are at opposite poles on at least one critical ideological and educational issue, evolution. As illustrated in Figure 5, their opinions regarding human evolution pull them in opposite directions – the most religious definitely reject human evolution while the most secular definitely believe in human evolution – and create sharp social divisions. This 8-point scale is not the final word. It may apply to some religious groups better than others. Still, by providing a simple metric of religiosity and secularity, it could become a useful tool for social scientists. Further research should be conducted to determine whether the scale is useful in predicting attitudes toward other controversial matters.

Since the religious majority is only slowly eroding and it is the young who are most likely to be in the growing secular minority, it is likely that culture wars over polarizing issues, like reproductive rights and human evolution, will be with us for a long time to come.