https://orcid.org/0000-0003-4094-1369
Abstract
It is often asserted that religion is changing rather than declining and that the usual indicators give a misleading impression of growing secularity. The evidence presented here suggests that on the contrary, social measures like affiliation and attendance may understate the weakening of personal religiosity. The apparent persistence of belief in God conceals erosion in its substance, strength, salience, and stability. The levels of religiosity and spirituality among the religiously unaffiliated are declining, notwithstanding the rapid expansion in their numbers. Moreover, belief has not just diminished among Americans who no longer belong to religious organizations: it is fading from one generation to the next among people who continue to belong. Americans are becoming less confident in the existence of God, less persuaded that God is active and judgmental, less inclined to see God as important, and less likely to express consistent conviction. Contrary to claims that apparent secularization masks enduring (if unorthodox) invisible religion, implicit religion, diffused religion, lived religion, and so on, the evidence shows that continuing religious involvement disguises waning religiosity. The research is based on six waves of the Baylor Religion Survey (2005–2021) in conjunction with the General Social Survey (1972–2022).
Introduction
Background
Religious involvement is falling in the United States, according to indicators such as declining church attendance and the remarkably rapid growth in the proportion of people who identify with no religion (Pew Research Center 2024). Many commentators maintain, however, that beneath the surface Americans are as religiously (or spiritually) minded as before (Levin et al. 2022). They may be drifting away from conventional practice and institutional membership, but even non-attenders and the “nones” (the non-affiliated) continue to hold religious or at least quasi-religious beliefs. The suggestion is that they are unchurched believers: “nothing in particular” rather than nothing at all. Contemporary tolerance leaves them free not to identify with a denomination or go to church every week. A related claim is that social scientists are measuring the wrong things. People who are “spiritual but not religious”—a designation so familiar that it has an acronym (SBNR)—are replacing orthodox churchgoers (Fuller 2001).
This study finds evidence that most people who appear to be non-religious are just that. Their levels of theistic belief are low compared to their religiously involved neighbors, and these levels are declining, notwithstanding the rapid expansion in their numbers. Nor is religion being replaced by spirituality. A dwindling share of nones qualifies as SBNR.
It is possible to go further, however. Belief has not just diminished among Americans who no longer belong to religious organizations: it is weakening among Americans who still appear to be religious. Contrary to claims that apparent secularization masks persistent (if unorthodox) invisible religion, implicit religion, diffused religion, lived religion, and so on, continuing religious involvement disguises an erosion in theism.
The substance, strength, salience, and stability of belief
Belief in God is far from being a yes/no question. There is a large gap between having a judgmental, constantly active God at the center of one’s life and not believing in God at all. The substance, strength, salience and stability of belief can vary. To take these issues in turn, starting with substance: what sort of God are we talking about? People have very different images of God, even in the United States. As for the strength of belief, some believers have convictions that feel intense and certain; others are subject to severe doubt. Salience is a matter of how important our beliefs are, both personally and to society. Some people might build their lives around them; for others, their beliefs (even if strong) make little difference.1 Finally, belief—like identity and practice—can vary from week to week or be stable over many years.
The substance, strength, salience and stability of belief have changed over the past 200 years or more. The evidence that follows shows that many Americans do not believe that God routinely intervenes in human affairs. He2 is less important than before to what we do, how we make sense of the world, how and by whom we are governed, and so on. Theism itself can no longer be taken for granted. In short, God has come to seem less personal, less involved, more abstract and more distant. And as people view God as indifferent to them, they become indifferent to God.
These changes amount to a kind of invisible secularization. People may continue to identify with a religion, accept that God exists, and attend collective worship. Underneath the surface of religious continuity, however, a substantial change has occurred. It may only become manifest when some people defect from religion, but it affects the ambient levels of religiosity across society.
Previous research
Religious belief in the United States
The growing share of the American population that is religiously unaffiliated—the “rise of the nones”—has inspired a mountain of popular and scholarly commentary. The phenomenon can be viewed as evidence that the United States is following the pattern of secularization seen in other highly developed countries (Voas and Chaves 2016; Inglehart 2021). This interpretation is often disputed, however. A common objection is that “many unaffiliated individuals, atheists, and agnostics attend religious services, pray, meditate, worship God or a higher power, and believe in heaven, hell, and miracles … The use of words and phrases such as none, no religion, and not religious to describe this group of unaffiliated individuals is thus inappropriate, inaccurate, and misleading” (Levin et al. 2022, 19). The idea that people are “believing, not belonging” has long been popular; one even encounters the claim that “religious belief is inversely rather than directly related to belonging. In other words, as the institutional disciplines decline, belief not only persists but becomes increasingly personal, detached and heterogeneous and particularly among young people” (Davie 2002, 8). The data are held to show “not the decline of religion but that of churched religion, that is, of a particular and peculiar historical form of religion” (Gauthier 2000, 312).
Not all of the criticisms are so radical. Schnabel and Bock (2017) argue that the United States remains exceptional because intense religiosity persists. Cadge and Babchuck (2023) point to chaplaincy and other forms of religious activity that escape general notice. Knoblauch (2008) maintains that “popular religion” (or spirituality broadly conceived) endures. The notion that people are pursuing individualized forms of religion, or following alternative spiritual paths, is frequently encountered (e.g., Ammerman 2014; Watts 2022). The claim can be traced back at least to Luckmann’s (1967) The Invisible Religion, but many terms have been devised for such pathways, including implicit religion, diffused religion, lived religion, and a number of others. Most of the concepts are intended to capture the supposed durability of personal religiosity. This “religious individualization thesis … has established itself as the chief adversary of the secularization theory” (Pollack and Pickel 2007, 604). Steve Bruce points out that Bryan Wilson (an early exponent of the secularization thesis) was already aware of the “change not decline” argument in 1966 (Wilson and Bruce 2016, xiii).
Debating whether the glass is half empty or half full—for example, whether the story lies in the dramatic decline of religious affiliation or the continuing spirituality of many nones—seems fruitless. Both observations can be true: what we want to know is the direction of change. No one disputes that many of the non-religious still believe in God or think of themselves as spiritual. The more interesting questions are how they compare to religious affiliates, and whether the differences between the religious and non-religious are growing, stable or shrinking over time and across generations. Recent reports from the Pew Research Center (2023, 2024) on Spirituality among Americans and Religious “Nones” in America are balanced and informative, but they work better as snapshots of the present than as moving pictures covering the past and prospective future.
The other issue is whether the worldviews of religious people are changing, and if so, how. Publications of the findings from the first two waves of the Baylor Religion Survey show the influence of the late Rodney Stark, who believed that the project would help to debunk the view that religion is weakening in America (Baylor Institute for Studies of Religion 2006; Stark 2008). To that extent the survey is particularly appropriate for present purposes, if new findings suggest the opposite. The investigation is primarily concerned with belief in God, which is therefore the focus of the remainder of this review.
Divine engagement
Theologians recognize that the concept of God has changed (or should change) over time (Clayton 2000; Cone 1970; Kaufman 1972; McFague 1982; Wildman 2017). Changes in substance lead to changes in strength and salience. If God is not active and judgmental, it is less clear what he might be, and in any event the growing distance and abstractness of God have diminished the importance of belief.
In late modern societies, a change of consciousness “has driven the awareness of the supernatural to the periphery. Science has not ‘disproved’ religion. … But it is almost impossible in the West today to experience the supernatural in the taken-for-granted manner of the past” (Gifford 2019, 41). By contrast, supernaturalism continues to be pervasive in sub-Saharan Africa, in ways that would once have been familiar elsewhere but now seem exotic (Gifford 2016).
The issue of how people see God has long occupied psychologists, starting with Freud. A good overview can be found in Tung et al. (2018). The work tends to be directed to ends such as investigating mental health, however, rather than explaining unbelief. Similarly, the much smaller body of work in the sociology of religion typically looks at the effect of God’s image on delinquency, social attitudes, political affiliation, and so on. Greeley (2000, 43) developed a four-item scale to measure God’s grace, asking respondents whether they pictured God as mother or father, lover or judge, spouse or master, and friend or king.
Christian Smith found that “moralistic therapeutic deism” (MTD) is replacing old-fashioned theism among teenagers in the United States. God’s function is to serve, not to be served. MTD is “belief in a particular kind of God: one who exists, created the world, and defines our general moral order, but not one who is particularly personally involved in one’s affairs—especially affairs in which one would prefer not to have God involved” (Smith 2005, 164). Returning to the respondents in the National Study of Youth and Religion ten years on, researchers found that “The God of emerging adults has become increasingly remote from their everyday concerns and rarely enters their thinking or occupies any real place in their lives” (Denton and Flory 2020, 7).
The Pew Research Center carried out a survey in 2018 on the concept of God in the United States. They report that “young adults are far less likely than their older counterparts to say they believe in God as described in the Bible” (Pew Research Center 2018, 16); college graduates are also less likely to believe in an active, engaged deity.
Differences in God concept can be linked to the success or failure of religious transmission within families. “An ideal-typical religious conservative will believe in a personally engaged God exercising ultimate and immutable moral authority,” and there is good evidence that “Religious transmission is strongest among religious conservatives” (Smith 2021, 335, 353).
The contribution of Paul Froese and Christopher Bader (2015, but originally published in 2010) to the study of images of God is particularly significant. They produced a striking analysis of four different concepts of God, derived from answers to two questions: “To what extent does God interact with the world?” and “To what extent does God judge the world?” In the Afterword to the updated edition, they note that the Baylor Religion Survey shows atheism in America rising from 5 to 9 percent between 2005 and 2014, with belief in a distant God (as they define it) going from 24 to 30 percent.
Divine judgment
Much research on images of God relates to what one might call the folk theory of secularization: religious belief is disrupted by modern science and technology. But unbelief is not just about facts: values figure as well. People not only stop believing in a divine watchmaker, they also stop believing that God is the source of morality and that religious organizations tell us what we must do to gain his approval and avoid his wrath.
A good deal has been written in recent years about the connection between value change and religious decline. Voas (2025) provides an overview of this work. The key hypothesis has been stated most explicitly by the late Ronald Inglehart: “if pro-fertility norms come to be seen as outmoded and repressive, their rejection also brings rejection of religion” (Inglehart 2021, 11). Inglehart suggested that while in the past religious adherence promoted traditional values (especially around gender, family formation and sexuality), today the rejection of traditional values undermines religious commitment. The effect is not merely a rebellion against organized religion: it is also a change in how people think about God. Many people have stopped accepting that the church, or indeed any institution, should be in the business of defining social norms and compelling obedience to them. The insistence that God has opinions on gender roles, contraception, premarital sex, cohabitation, divorce, homosexuality, abortion, and assisted dying has created a problem for religion, because there is a growing consensus that these are matters of personal choice. Individuals are increasingly confident that they know what counts as good or bad, right or wrong.
Data and analytic approach
Baylor Religion Survey, 2005–2021
The Baylor Religion Survey aims to explore American religious beliefs, values, and behaviors from nationally representative samples. To date there have been six waves, conducted in 2005, 2007, 2010, 2013, 2017, and 2021. Data collection was carried out by Gallup via telephone and mailed questionnaires in waves 1–3 and mailed questionnaires thereafter, with an option for web completion in wave 6. Waves 1 through 5 each have 1,500 to 1,725 respondents; wave 6 has 1248 in the archived dataset. The total across all six waves is 9,404. Cases were weighted using the variable provided. Data and documentation are freely available from the Association of Religion Data Archives (ARDA).
The survey gives extensive coverage to images of God. The questions were put to all respondents in wave 1, but in later waves, respondents who do not believe in God were routed away from this section. For consistency, wave 1 atheists are excluded from the analysis. In wave 6, agnostics (8.7 percent of valid cases) as well as atheists skipped the questions, so the responses on God will be more theistic than in previous waves. Trends that indicate decline in conventional belief will therefore be conservative estimates.
The analysis by Froese and Bader (2015) of waves 1 and 2 provides the theoretical point of departure. Their starting point was the creation of two additive scales, intended to capture God’s degree of engagement and judgment respectively. The present study uses questions that were asked in all six waves. The extent to which God is engaged is based on four variables:
Based on your personal understanding, what do you think God is like?
Concerned with the well-being of the world.
Concerned with my personal well-being.
Directly involved in world affairs.
Directly involved in my affairs.
(Strongly agree, agree, disagree, strongly disagree)3
The extent to which God is judgmental is measured using five items:
Based on your personal understanding, what do you think God is like?
Angered by human sin.
Angered by my sins.
(Strongly agree, agree, disagree, strongly disagree)4
How well do you feel that each of the following words describe God?
Punishing.
Wrathful.
Critical.
(Very well, Somewhat well, Not very well, Not at all)
Waves 1 and 2 included “undecided” as a response option, with 10–15 percent of respondents giving this answer. Treating “undecided” as a missing value accentuates the decline in traditional theism across the six waves. One can reasonably argue, however, that this version of “don’t know” is informative and should be regarded as a valid response. For the sake of scholarly caution, the answer is coded as midway between “agree” and “disagree.”
Cronbach’s alpha is high for both the engaged and judgmental God scales (0.902 and 0.827 respectively). These two scales are fairly highly correlated (0.456). Four concepts of God are defined by a 2×2 table of high or low values for engagement and high or low values on the judgmental scale.5 The God that is both engaged and judgmental is Authoritative; engaged but not judgmental is Benevolent; not engaged but judgmental is Critical; neither engaged nor judgmental is Distant.
Froese and Bader employed the “Four Gods” as a categorical variable in their discussion of American public opinion. But they assigned respondents to these categories using scale measures derived from the items mentioned above, and for some purposes it is helpful to treat the key “authoritative God” concept as continuous. The ten6 variables on the image of God that appear in all six waves can be combined into a single additive scale (0–30) measuring the extent to which God is seen as both active and judgmental; it has a Cronbach’s alpha of 0.88.
Some of the findings that follow make use of religious tradition, following Steensland et al. (2000) in modified form.7 Frequency of prayer is also mentioned below; the question is “About how often do you spend time alone praying outside of religious services?”, with six options from “Never” to “Several times a day.”
General Social Survey, 1972–2022
The General Social Survey (GSS) in the United States is run by the National Opinion Research Center at the University of Chicago. Prior to the Covid pandemic, data were collected through face-to-face interviews with a random sample of non-institutionalized adults aged 18 and over. The survey was conducted annually from 1972 until 1994 (except in 1979, 1981, and 1992) and every other year since then. Sample sizes for each survey range from approximately 1500 to approximately 3000. Fieldwork for the planned 2020 survey was largely carried out in 2021 via self-administered questionnaires; in 2022, the GSS was conducted using in-person interviews, online surveys, and telephone interviews.8
Schnabel et al. (2024) warn that the change in survey mode during the pandemic resulted in the most intensely religious segment of the population being substantially less well represented than previously. While potentially important in some contexts, this bias should have minimal influence on analyses of non-religious respondents, who are the focus in much of what follows.
The GSS used a rotating panel design during the period 2006 to 2014. Approximately 2,000 respondents in 2006, 2008, and 2010 were followed up two and four years later to produce three separate panels: 2006/2008/2010, 2008/2010/2012, and 2010/2012/2014. Full details can be found in Smith and Schapiro (2017).
Questions on religious affiliation and attendance at services have featured in every survey year. The following question on belief has been included on the GSS since 1988:
Please indicate which statement below comes closest to expressing what you believe about God:—I don’t believe in God;—I don’t know whether there is a God, and I don’t believe there is a way to find out;—I don’t believe in a personal God, but I do believe in a Higher Power of some kind;—I find myself believing in God some of the time, but not at others;—While I have doubts, I feel that I do believe in God;—I know God really exists and have no doubts about it.
A yes/no question on belief in life after death (the variable postlife) has been included in most GSS years.
In 1998 and then in each survey since 2006, GSS respondents have been asked “To what extent do you consider yourself a spiritual person? Are you. .. very spiritual, moderately spiritual, slightly spiritual, not spiritual at all?” A parallel question asks whether the respondent is a religious person.
Finally, the findings here also make use of three items from the religion modules produced by the International Social Survey Program (ISSP) and included in the GSS in 1991, 1998, 2008, and 2018. The question on belief in life after death has four options, from “yes, definitely” to “no, definitely not.” Respondents were asked how far they agree with the statements that “There is a God who concerns Himself with every human being personally” and “To me, life is meaningful only because God exists” (five options from “strongly agree” to “strongly disagree”).
Analytic approach
The investigation will start with the four facets of belief: substance, strength, salience and stability. The best (and almost only) source of data on the substance of theistic belief—respondents’ image of God—is the Baylor Religion Survey. The analyses use the Froese and Bader typology of America’s four gods, the scale measure of belief in an authoritative God, and single item measures of the image of God.
The focus remains on the BRS in the discussion of strength—the intensity of belief—and in the initial consideration of salience: the perceived importance of God in everyday life. The BRS shows the association between belief in a caring God and prayer. The ISSP/GSS then comes to the fore with an item on God and meaning in life. Finally, stability of belief is studied using the GSS panels.
The next question is whether belief in a Christian God has been replaced by other forms of supernaturalism—or more broadly, whether religiosity has given way to spirituality. The GSS is the main source of data on this issue. And lastly, the question of whether the growing unaffiliated or non-observant population is in fact still believing is tackled using the GSS.
Tests of statistical significance are provided in the supplementary appendix, along with some additional tables and figures. Further details on any of the findings are available on request.
The data show that Americans are increasingly less likely to believe in the God of Christian tradition; the substance, strength, salience and stability of that belief as measured in various ways has eroded; that decline is associated with reductions in religious affiliation and practice; this waning of conventional religion is not being offset by a rise in spirituality; and the growing numbers of Americans who say they have no religion are neither “believing without belonging” nor becoming “spiritual but not religious.” Thus, the secular turn in the United States is genuine and not merely a matter of people rejecting denominational labels.
Findings
The substance of belief
If the Baylor Religion Survey presents an accurate picture, there has been a remarkable alteration in the substance of American theism (fig. 1). The share of the “authoritative God” (engaged and judgmental) has fallen 17 percentage points between 2005 and 2021; the “critical God” (judgmental but not engaged) has fallen 3 points. Three quarters of the combined 20-point gap is filled by the rising share of Americans whose God is distant or non-existent. The benevolent God—engaged but not judgmental—picks up the difference.
Belief in an authoritative or critical God is declining; God is increasingly seen as benevolent or nonexistent. Source: Baylor Religion Survey, waves 1–6. Note: See Supplementary Table A1 and fig. A1 for the underlying data and confidence intervals.
The change is occurring even among the most religious, contrary to suggestions that this segment of the population has been spared from decline. In the first two waves, 11 percent of respondents scored in the top quarter on both indicators; by the last two waves, that share had dropped to 5 percent. (See Supplementary Tables A2a and b.)
The authoritative God scale (which runs from 0 to 30) offers a similar picture. In Wave 1 of the Baylor Religion Survey, 41 percent of respondents who did not identify themselves as atheists scored 20 or higher. In Wave 6, even with agnostics as well as atheists excluded, only 30 percent did so. The entire distribution of how God is seen has shifted away from the traditional view of an active and judgmental deity. The change is most obvious among the most intensely religious: the share of theistic respondents who scored 25 or higher on the 30-point scale fell from 17 percent in 2005 to 8 percent in 2021. (See Supplementary Table A3.)
The image of God is changing across generations, even with non-believers excluded. Pooling all six waves of the Baylor Religion Survey, the mean value on the “authoritative God” scale drops for Pre-war, Baby Boomer, Gen X and Millennials/Gen Z respondents from 17.9 to 17.6, 17.3, and 16.1 respectively. The largest gap is that separating the youngest generation from the earlier ones. And looking across the whole typology by decade of birth, it is evident how much of the trend over time is driven by recent cohorts, with a sharp decline in the authoritative God, stagnation in the benevolent God, and an increasing likelihood of God appearing critical, distant or non-existent (see Supplementary fig. A2).
Belief in an engaged and judgmental God has declined across all major Christian traditions. Figure 2 shows levels of strong agreement that God is directly involved in world affairs, and angered by my sins, for evangelical Protestants and Catholics. (The graph for black Protestants is very similar to that for evangelicals, and the picture for mainline Protestants is close to that for Catholics.) Along with the overall trends and the difference in levels for evangelicals and Catholics, what stands out is how much less fearsome God is becoming for many evangelicals.
Strong agreement that God is “angered by my sins” and “directly involved in world affairs” has declined across major Christian traditions. Source: Baylor Religion Survey, waves 1–6. Note: See Supplementary Table A4 for significance tests.
Strong agreement with the statement that God is “directly involved in world affairs” is associated with feeling very religious, and hence it is no surprise that levels of self-described religiosity are falling. In 2007, 47 percent of evangelical Protestants described themselves as very religious; in 2021, only 32 percent did so. The share of Catholics calling themselves very religious dropped from 27 percent to 14 percent over the same period.
The strength of belief
It is tempting to think that changes in religiosity could be transitory. The evidence from around the world, however, is that religious affiliation, attendance and belief decline from one generation to the next, while average levels of religious involvement are relatively stable over adulthood within each birth cohort. That means that old generations show us the past: young ones show us the future.
This pattern holds in the United States as well as in Canada, Europe, Australia, New Zealand and elsewhere. “The overall level of belief in God is being eroded as people born early in the 20th century are replaced in the population by members of subsequent generations with weaker religious convictions” (Voas and Chaves 2016, 1546). Among members of the “Greatest Generation” (born in the first quarter of the 20th century), three quarters stated that “I know God really exists and I have no doubts about it.” Only a third of Gen Z (born since 1997) do so.
This pattern of generational decline in strength of belief is found even when we exclude people with no religion, who constitute a growing proportion of each successive birth cohort. Of people who say that they have a religion, three quarters of those born 1915–24 have no doubts about God’s existence; fewer than half (46 percent) of those born 1995–2004 say the same. (The figures come from pooled GSS data since 1988, when the question on God first appeared.)
This phenomenon is also apparent in the Baylor Religion Survey. Figure 3 gives mean values for belief in God by survey year and generation.9 Although the time span is much shorter than that available from the GSS, the picture is the same. “The trend lines show steady generational drift away from unwavering belief, and there are few signs of change with age or over time” (Voas and Chaves 2016, 1545). The similarity between figure 3 and comparable values from the GSS helps to give us confidence in the BRS.
Declines in belief in God are mostly generational. Source: Baylor Religion Survey, waves 1–6. Note: The belief in God scale runs from 0 for atheism to 5 for belief with no doubts. See Supplementary Table A5 for significance tests.
The salience of belief
Although it is valuable to understand what and how strongly people believe, a key issue is how much these beliefs affect their lives. The importance of God or gods to an individual cannot be inferred from how confidently theistic the person claims to be. The evidence suggests that people who are neither especially religious nor wholly secular—a large proportion of the population—do not see God as important in life (Voas 2009).
Frequency of private religious observance may be a proxy for importance of God to an individual. Belief in an engaged God is highly correlated (>0.5) with praying often. The association is particularly high for belief that God cares about me rather than just the world in general. Respondents were asked to agree or disagree that “God is concerned with my personal well-being.” Among respondents who strongly disagreed, a majority (even with non-believers excluded) never pray. Of those who strongly agreed that God is concerned, nearly half pray several times a day (fig. 4).
People pray more often when they believe that God cares about them. Source: Baylor Religion Survey. Note: See Supplementary Table A6 for significance tests.
One can only speculate about causality, but it seems plausible that feeling confident that God is concerned about you leads to more prayer, rather than the reverse. But whatever explains the connection between belief and prayer, there is an ongoing intergenerational erosion, as evident from pooled GSS data covering the past two decades (2004 to 2022). Among respondents born 1995–2004, 33 percent never pray, while 30 percent pray every day. Among respondents born 1915–24, by contrast, only 8 percent of never pray, and 72 percent pray every day. (See Supplementary Table A7).
Unfortunately the GSS provides only limited data on the importance of God. A question on the importance of religion has appeared just once, in 2021. The findings strongly suggest that the importance of religion has been declining across generations. Among people born in the decade 1935–44, only 15 percent say that they have no religion, and of those with a religion, more than half (51 percent) say that it is very important in their life. Among people born between 1995 and 2004, nearly half (48 percent) are nones, and even of those who do have a religion, only just over a quarter (27 percent) say that it is very important. (See Supplementary Table A8).
A more direct question about the personal importance of God appears in the ISSP modules on religion: “To me, life is meaningful only because God exists.” Agreement in the United States came in at 44 percent in 2018, which is much higher than in most of Europe (14–16 percent for Germany, Spain, Switzerland, France and Great Britain; 6–10 percent in Sweden, Denmark and the Czech Republic), though considerably lower than in Turkey (84 percent), the Philippines (77 percent) and South Africa (67 percent). The importance of God in providing meaning is declining from one generation to the next in the United States. Figure 5 shows that agreement with the statement has fallen from around 60 percent among people born before the Second World War to half that for people born since 1980. As with other indicators of religiosity, most of the change is occurring between rather than within generations (Voas and Chaves 2016).
Agreement that “To me, life is meaningful only because God exists” has declined across generations. Source: General Social Survey, 1991, 1998, 2008, 2018. Note: See Supplementary Table A9 for significance tests.
Agreement that God makes life meaningful is highly predictive of strong identification with a religious denomination and attendance monthly or more often. The levels of these three variables are very similar within subgroup, for example those defined by region or birth cohort; see Supplementary Table A10.
The stability of belief
We tend to suppose that while unconventional worldviews can be volatile, orthodox religious belief is relatively stable in adulthood. In each of the three-wave GSS panels starting in 2006, 2008 and 2010, the distribution of responses to the question on strength of theistic belief (with six options from not believing to believing with no doubts) was essentially unchanged across the three surveys covered. That is what one would expect if religiosity—including belief in God—is stable over the adult lifecourse. On closer examination, however, many respondents did not give the same answer each time they were asked. Across the three panels, only 59 percent of people participating in all three waves gave the same response on every occasion.
This inconsistency might suggest that beliefs are more volatile than we realize and really do change from one survey year to the next. Alternatively, however, it could be that beliefs are stable but respondents struggle to express their views using the options available. One might see this phenomenon as liminality, to use the term popularized by Lim et al. (2010) in relation to religious identity: they are on the fence and will come down sometimes on one side and sometimes on the other. The absence of an overall trend, combined with the fact that most individual changes are just a single step up or down on the scale, tends to support this idea.
The modal response to the GSS question on God has always been the strongest option (“I know God really exists and I have no doubts about it”), though only a minority of respondents now select that statement. One might expect this choice to be quite stable, both because of the ceiling effect (no stronger statement is available) and because that degree of conviction seems unlikely to fluctuate. In fact, however, a substantial number of these respondents give a different answer in other waves. Among people giving this most confidently theistic answer at least once, 34 percent (across the three panels) give some other answer in one or both of the other waves.
These inconsistently firm believers are less orthodox on each dimension of theism than respondents who expressed no doubts in all three waves. The contrast is most obvious in the substance of belief: only 28 percent strongly agree that “there is a God who concerns Himself with every human being personally,” versus 73 percent among the consistently convinced. The salience of belief is also substantially lower: only 7 percent strongly agree that “to me, life is meaningful only because God exists,” versus 39 percent among the constant non-doubters. The gap is similarly high for attendance at services weekly or more often (20 percent versus 43 percent), strong affiliation (24 percent versus 55 percent), and so on. The question “To what extent do you consider yourself a religious person?” could be regarded as an alternative measure of strength of belief, and only 8 percent of respondents whose certainty about God wavers describe themselves as “very religious,” versus 31 percent among those who never express doubts.10
In short, the high levels of agreement with the statement “I know God really exists and I have no doubts about it” give a misleading impression of strong and resolute theism in America. A third of respondents who select this option make a different choice when asked again: it appears that they do have doubts after all. Insipient secularity is hidden beneath a veneer of strong theism.
Moreover, it appears that the instability of belief is rising in more recent generations. Among people who expressed a firm belief in God in one wave, 71 percent of those who were born between 1925 and 1949 did so in all three waves (pooling the panels that started in 2006, 2008, and 2010). The corresponding figure is just 57 percent for those born between 1975 and 1988.11
Is spirituality replacing religion?
Although the drift away from religious affiliation, practice and belief is hard to deny, some scholars maintain that these declines are offset by a rise in alternative spirituality. The claim is difficult to contest, because the “holistic milieu” is so varied and diffuse that it is always possible for critics to argue that one is measuring the wrong things. A recent effort by the Pew Research Center (2023) is impressive, but their data are currently insufficient to identify the direction of travel.
Although the subject warrants a much longer treatment, some basic findings from the GSS are relevant here. One of the few variables that addresses this issue directly is the following: “To what extent do you consider yourself a spiritual person? Are you. .. very spiritual, moderately spiritual, slightly spiritual, not spiritual at all?” A parallel question asks whether the respondent is a religious person. Most Americans are willing to say that they are at least moderately spiritual, and the term is readily accepted by people who also call themselves religious. The issue is whether the “spiritual but not religious” share of the population is growing, especially among people who have no religion. Figure 6 shows that SBNR numbers in the whole population appear to have leveled off since the Baby Boomer generation, while there is gradual decline among nones.
Self-described spirituality has plateaued; nones are increasingly neither religious nor spiritual. 6a full sample; 6b nones. Source: General Social Survey, 2006–2022. Note: See Supplementary Table A11 and fig. A3 for the underlying data and confidence intervals.
Another possible approach is to look for trends in belief in the non-theistic supernatural, for example angels, demons, ghosts, or life after death. The Baylor Religion Survey had items on each of those beliefs (measured on a four-point Likert scale) in waves 1–4. The GSS has not generally covered this territory, except for belief in life after death (unfortunately with a yes/no question that gives us little clue as to the substance, strength, salience or stability of belief). The surveys show virtually no change over the years, which suggests that supernaturalism holds up well—at least initially—even as conventional religiosity declines. Some caveats are in order, as explained in a well-known article by Greeley and Hout about belief in life after death:
The meaning of the question itself could have changed over time. We do not know much about changes in peoples’ images about the afterlife, but detailed questions from the 1983 and 1984 GSSs give us a cross-sectional glimpse. … The results did reveal that it is difficult to compare people who have a religious preference with persons who have no religious affiliation. Unaffiliated persons gave answers to the supplemental questions about what an afterlife would be like that differed from the responses of Christians and Jews. … These differences … are so great that we set aside those with no affiliation for future analysis. (Greeley and Hout 1999, 816).
In any event, nones are much less likely than people with a religious affiliation to say that they believe in life after death; the levels have hardly changed over the 50 years between 1973 and 2022, with 75 percent of affiliates saying that they believe as opposed to 45 percent of nones. Slight upward trends are just about detectable (see Supplementary fig. A4). The high average conviction of affiliates is distilled by the removal of the least committed, while the relatively low average belief of nones is diluted by recent switchers out of religion, who tend to be less secular than long-standing nones.
Similarly, a large majority of theists also believe in life after death, while a bare majority of the rest do not, as shown in figure 7. There appears to be some distillation among theists (as the least believing migrate out of that category), but the shift away from belief in God does not seem to have diluted skepticism about life after death.
Belief in life after death by survey year and belief in God. Source: General Social Survey, 1988–2022. Note: Belief in God is with or without doubts. See Supplementary Table A12 for significance tests.
Answers to the life after death question from the ISSP modules in 1991, 1998, 2008 and 2018 show a noticeable generational shift from the “definitely” to the “probably” options, suggesting that although the belief remains popular, young people are open to the idea rather than being committed to it. Here too there is a large gap between people with a religion (who overwhelmingly believe) and the nones (only one in five of whom say that they definitely believe in life after death).
Finally, the supernatural belief items in the Baylor Religion Survey are highly correlated with belief in God. A partial exception is belief in ghosts, where the correlation is lower (0.204 in wave 1). Only 4 percent of respondents believe in ghosts more than they believe in God, however. Americans continue to hold supernatural beliefs at levels that are both high and relatively steady, but the association between theism and these beliefs raises the question of whether those levels will persist indefinitely.
Theism among religious nones
As is well known, nearly 30 percent of people in the United States now say that they have no religion. This figure is a dramatic increase from the single-digit levels that prevailed until the mid-1990s. Scholars who are skeptical about secularization in the United States often argue that the nones are just “nothing in particular,” to use a term from Pew questionnaires. It is plausible that the rise of the nones reflects a weakening of denominational identities rather than declining religiosity, as well as a growing willingness among the least religious people to say that they have no religion. The fact that a significant proportion of the unaffiliated attend church at least sometimes and agree that God exists offers some support for this view. The key question is whether the profile of nones is changing, so that the atheists and secularists who might have been dominant in times past are now diluted by unchurched believers.
In fact, however, nones are less religious now than in the past, despite the rapid expansion of the “no religion” category. There has been a decline in theism, as indicated by agreement with either “I know God really exists and I have no doubts about it” or “While I have doubts, I feel that I do believe in God.” The average among nones between 1988 (when the question was first asked) and 2000 was 35.7 percent; in the period 2006 to 2018 it was 31.4 percent, and the average from the 2021 and 2022 surveys is 25.8 percent; see Supplementary Table A13. Churchgoing tells the same story.12
As an alternative way of considering the religiosity of the seemingly nonreligious, we can look at the average belief in God among people who do not go to church. Whatever their age, frequent churchgoers show similar levels of conviction in the existence of God, although a small generation gap has opened between Millennials and earlier birth cohorts (fig. 8). Among non-attenders and even infrequent attenders, however, belief is much lower in the young than in the older generations. There is strong evidence that these differences are cohort rather than age effects (Voas and Chaves 2016). Although the share of non-attenders is rising in each successive birth cohort, these people seem increasingly unreligious, neither believing nor belonging.
Among non-attenders, belief in God has declined across generations. Source: General Social Survey, 1988–2022 pooled. Notes: Attendance classified as rarely = less than once a year, once or twice a year; sometimes = several times a year, about once a month, two to three times a month; frequently = nearly every week, every week, several times a week. The belief in God scale runs from 0 for atheism to 5 for belief with no doubts. See Supplementary Table A14 for significance test.
The Pew Research Center provides further evidence that nones are genuinely nonreligious, not simply people who feel no attachment to a particular denomination. Their latest survey suggests that about two thirds of nones defected because “they question a lot of religious teachings or don’t believe in God” (Pew Research Center 2024, 9, 24–6). Only 13 percent believe in God as described in the Bible, although a majority (56 percent) believe there is some other higher power or spiritual force in the universe. Traditional theism is strongly associated with age and lower education (Pew Research Center 2024, 46–7).
In forecasting future levels of religiosity, it is important to note that people who were not brought up in a religion are less religious on average than people who disaffiliate. “The very fact of having been affiliated in childhood leaves an imprint on adults irrespective of childhood religiosity. Thus, even among those who were only nominally Christian in childhood, childhood religious belonging remains significant in determining adult religiosity, with disaffiliates more religious than lifelong nones” (Beider 2023, 64). As more people are raised with no religion, the average religiosity of nones will fall further.
The experience in other countries is that nones become less and less likely to pray, attend or believe. And although many are indifferent to religion, a growing proportion positively rejects it (Voas and Bruce 2019, 23–4).
Discussion
It is often asserted that religion is changing rather than declining and that signs of growing secularity disguise elements of invisible religion. The evidence presented here suggests that on the contrary, what we find is invisible secularity: social measures (affiliation, attendance) may understate the decline in personal religiosity, and specifically the substance, strength, salience and stability of theistic belief. Continuing affiliation and practice can disguise an erosion of intrinsic religiosity.
The claim that social scientists are measuring the wrong things is hard to dismiss categorically, because the range of beliefs and practices that might be called “spiritual” is so broad that proponents of this view can always point to something not covered. The evidence for religious decline is strong, though, and the “spiritual but not religious” are not filling the gap (see Kasselstrand 2022 on Europe). The burden of proof is on the critics to show that growth in spirituality is making up for waning religiosity. In any case, a shift from organized religion—with membership in major institutions, collective practice, and beliefs that motivate important social, political and personal decisions—to more idiosyncratic worldviews associated with none of those things looks very much like a decline in the social significance of religion/spirituality (Wilson and Bruce 2016).
Even a society that seems relatively religious (like the United States) has been overtaken by a kind of secularity that is not easy to see. Beneath the surface and often out of sight, Christians as well as nones are becoming less intrinsically religious. Supernaturalism has a smaller role in their worldviews, they are less inclined to see spiritual forces at work, and they are less likely to rely on external authority in matters of personal morality. Churchgoers today are more secular than their great-grandparents were. There are many other reasons that religious identity and churchgoing are in decline, but the erosion of belief is arguably one of them.
All else being equal, people will gravitate toward hopeful or undemanding beliefs and away from those that are depressing or constraining. If people feel free to decide what God is actually like, they will show an aversion to the idea of divine anger. One possible consequence is deciding that God is understanding and forgiving and will support us whatever we do. An alternative is the belief that God is distant, uninvolved, abstract and indifferent. In the first case, God is available to serve us, but we do not really need him; in the second, it is unclear whether he even exists.
The journey from theocentrism to unbelief normally takes more than a lifetime: few individuals go from one extreme to the other. It is a social and cultural shift that occurs over many decades and generations. The first step away from the most intense religiosity may be a slight moderation in the God-centered worldview, but conversely the last step away from religion is generally the abandonment of theism. People who are fully convinced that the wrath of God comes upon the sons of disobedience (Ephesians 5:6) will not skip church on Sunday morning. A weakening of belief seems likely to precede reduced commitment to attendance, prayer, rites of passage and other forms of observance. But even when the various forms of belonging have gone, some residual supernaturalism typically remains, like the last embers of a once-blazing fire.
Sociologists tend to focus on structural conditions, such as pluralism, prosperity, and social complexity. Once we come into contact with different beliefs and cultures, it is hard to take it for granted that we know the truth and everyone else is wrong. Affluence leads to individualism in beliefs and behavior. There are now secular specialists to heal us, teach us, marry us, counsel us, judge us, and so on. But perhaps—after an early period in which scholars reflected on advances in science and technology and the nature of plausibility structures—we have paid insufficient attention to changes in belief. Secularization is not only or even mainly about the percentage of people who say that they have no religion.
Why religiosity wanes remains controversial, as is the degree to which less institutional expressions of supernaturalism are important: in particular alternative spirituality, but also “lived religion,” “invisible religion,” “implicit religion,” “diffused religion,” “believing without belonging,” and so on. Scholars have been tempted to hunt for signs of residual religiosity; it would arguably be more fruitful to attend to how little difference religion makes in the daily lives of most Americans. Lived secularity is the everyday reality.
In future research, we should try to understand the evolution, in individuals and society, of the connections between the image of God, the importance of God, and the intensity of belief in God. For example, does thinking that God is distant lead to indifference and doubt, or do the doubts come first? The next step is to investigate the connection between these aspects of belief and other expressions of religiosity, such as prayer, attendance and affiliation.
Alan Wolfe showed characteristic insight two decades ago in discussing the transformation of American religion. His conclusion was reached by another path, but it can now be reinforced through survey research. “God is not dead in America, but the way he lives and breathes has nothing in common with the old-time religion dramatized in Inherit the Wind. … God has met and struggled fiercely against American culture—and the culture has won” (Wolfe 2003, inside cover).
About the author
David Voas is Emeritus Professor of Social Science at University College London, where he led the Social Research Institute from 2016 to 2020. His research is mainly concerned with religious change in modern societies, based on quantitative analysis of large survey datasets. Key contributions include work on the secular transition, the relationship between religious diversity and involvement, fuzzy fidelity, conspirituality, the gender gap in religiosity, the sociology of attitudes, and secularization in the United States.
Acknowledgements
I am grateful to Paul Froese, director of the Baylor Religion Survey, for his generous assistance with the data. The paper was much improved by comments from Ozan Aksoy, Steve Bruce, Conrad Hackett, Jörg Stolz, Wesley Wildman, and anonymous reviewers.
Funding
None declared.
Conflicts of interest
None declared.
Data availability
The Baylor Religion Survey data are available from the Association of Religion Data Archives at www.thearda.com. The General Social Survey data are available from NORC at the University of Chicago at gss.norc.org.
Endnotes
A couple of examples might help to distinguish strength from salience. For some people the belief that they will be reunited with loved ones in heaven is of the utmost importance, but they cannot help feeling doubt about the afterlife. Conversely, some Christians might be completely convinced of the truth of various points of doctrine, without regarding all the issues as important to them.
“He” is a traditional representation; the gender, nature and existence of God are of course matters for theology, not sociology.
More people think that God cares about the world than about their personal wellbeing. By contrast, more people think that God is directly involved in personal than in world affairs.
More respondents strongly agree that God is angered by human sin than that he is angered by mine. It would be interesting to know whether they see a distinction between general and personal sinfulness, or simply feel that they themselves have done nothing to make God angry.
The typology was created in essentially the same way by Froese and Bader (2015), but they defined “high” and “low” as above or below the mean value, rather than above or below the midpoint on each scale. They thereby achieved a more even distribution across the four categories. For present purposes it seemed clearest and most appropriate (given the changes over time) to consider whether people tended to agree or disagree with the relevant statements about God, rather than agreeing more than average.
In addition to the nine listed in the text, a tenth item appears in all six waves: the appropriateness of the word “forgiving” to describe God. The overwhelming majority of respondents believe that God is forgiving, though intriguingly, these responses are positively correlated with the scale measuring the degree to which God is seen as angry and punitive. For many respondents, God may be wrathful, but he is also forgiving.
The findings below make use of religious tradition, with Protestants divided into mainline, evangelical and Black. Steensland et al. (2000) defined non-denominational Christians as evangelical Protestants if they attended monthly or more often; otherwise they were treated as unclassified (missing) for the purposes of their “Reltrad” variable. This decision has been criticized (for example by Burge and Djupe 2021). In most Baylor Religion Survey datasets, all respondents who selected “Non-denominational Christian” rather than any more specific label were assigned a Reltrad code (usually 1 = Evangelical Protestant). In BRS4, it appears that infrequent attenders were treated as missing. In BRS6, those who reported that they never attend services are classified as “non-affiliated” (none), which is thus a third approach to the issue. For the purposes of this study, Reltrad was modified to create a consistent version across all waves. Reltrad was set to missing in cases where affiliation was given as non-denomination Christian but the respondent never attends.
Cross-sectional analyses use the 1972–2022 cumulative data file, release 3a. The weights are wtssps for years through 2002 and wtssnrps from 2004 onwards.
Pre-war, Baby boomer, Gen X and Millennials/Gen Z cover birth years up to 1945, 1946–64, 1965–80, and 1981 onwards respectively.
The figures in this paragraph come from the 2008 panel, the only one that includes the variables on a personal God and God making life meaningful. Attendance, affiliation and self-described religiosity are averaged across 2008, 2010, and 2012 for respondents participating in all three waves.
This generation gap persists in the two-wave panel a decade later in which some respondents from 2016 and 2018 were re-interviewed with the 2020 questionnaire.
During the half century that the GSS has been in existence, approximately a quarter of people with no religion have said that they attend religious services at least once a year. The trend between 1972 and 2018 is a very gradual decline, followed by sharp falls since 2020.
