- Journal List
- HHS Author Manuscripts
- PMC4315336
No Money, No Honey, No Church
Abstract
Purpose
We examine trends in religious attendance by educational group, with an emphasis on the “moderately educated:” individuals with a high-school degree but not a 4-year college degree.
Methodology
We conduct multivariate ordinary least-squares (OLS) regression models using data from the General Social Survey (from 1972 to 2010) and the National Survey of Family Growth (from 1982 to 2008).
Findings
We find that religious attendance among moderately educated whites has declined relative to attendance among college-educated whites. Economic characteristics, current and past family characteristics, and attitudes toward premarital sex each explain part of this differential decline.
Implications
Religion is becoming increasingly deinstitutionalized among whites with moderate levels of education, which suggests further social marginalization of this group. Furthermore, trends in the labor force, American family life, and attitudes appear to have salient ramifications for organized religion. Sociologists of religion need to once again attend to social stratification in religious life.
The declining economic fortunes of the American working class have attracted considerable academic attention over the last two decades (Fischer & Hout, 2006; Hacker, 2006; Levy, 1998). But the possibility that other features of working class life—such as the religious life of the American working class—are also in distress has received less scholarly attention. This is a particularly important issue to explore both because the American working class attended religious services at rates that were similar to those of the college-educated middle class for the middle decades of the twentieth century, and because churches have been an important source of social solidarity for the working class (Hoge & Carroll, 1978; Verba et al., 1995; Wilcox, 2010). Indeed, given the rise of the post-industrial economy and government fiscal retrenchment, such that the market and the state provide less financial security to the working class than they once did, civic institutions—including religious congregations—might be one of the few institutional sectors working class Americans can turn to for social, economic, and emotional support in the face of today’s tough times. But the American religious sector is not likely to be of much help to working class Americans if they are increasingly disengaged from the life of their local religious congregations.
Accordingly, relying on an analysis of data from the General Social Survey (GSS) and the National Survey of Family Growth (NSFG), this paper first explores the changing religious fortunes of the white working class, both absolutely and in relation to lower-class and middle-class Americans, from the 1970s to the 2000s. We focus on whites because black and Latino religiosity is less likely to be stratified by class, given that churches have been an important vehicle for solidarity, community action, and political activity for blacks and Latinos of varying class backgrounds (Ellison & Sherkat, 1995; Lincoln & Mamiya, 1990; Figueroa Deck, 1989; Schwadel, McCarthy, & Nelsen, 2009). We focus on the working class—defined here as “moderately educated” Americans who have a high school degree or some college education but not a 4-year bachelor’s degree—because what little research that has been done recently on religion and stratification has focused primarily on how lower-class Catholics are less likely to attend church, compared to their middle-class and more affluent fellow believers (Schwadel, McCarthy, & Nelsen 2009), and has not considered the possibility that patterns of working class religiosity may now be diverging from those found among their more educated and affluent peers. According to GSS data, however, moderately educated whites make up 60 percent of the adult white population (aged 25–44) in the contemporary United States. In brief, the evidence we present here suggests that the middle is dropping out of the American religious sector, much as it has dropped out of the American labor market.
We then turn to a consideration of the economic, demographic, and cultural correlates of the religious disengagement of moderately educated whites. We do so from a broadly institutional perspective, recognizing both that religion is not only a social institution that supplies norms, beliefs, and rituals that pattern social behavior, or moral logics, but also an institution that depends on social and cultural structures from other institutions to sustain these moral logics (Friedland & Alford, 1991). In particular, we explore the possibility that working class disengagement from the institutions of work and marriage (Cherlin, 2009; Wilcox, 2010) are strongly associated with recent declines in religious attendance among white working class Americans. We view these two institutions as particularly important objects of inquiry because American religion has both legitimated and been bolstered by an “American Way of life” marked by stable employment and marriage over much of the last century (Edgell, 2006; Herberg, 1955). Thus, if moderately educated whites are now less likely to be stably employed, to earn a decent income, to be married with children, and to hold familistic views, they may also be less likely to feel comfortable or interested in regularly attending churches that continue to uphold conventional norms, either implicitly or explicitly (Edgell, 2006; Wilcox, 2004). We also view the institutions of work and marriage as important sources of social and normative integration that link Americans to religious institutions (Schwadel, McCarthy, & Nelsen, 2009). For these reasons, this paper relies on the GSS and NSFG to explore the links between declines in working class religiosity and patterns of employment, income, family structure, sexual behavior, and attitudes toward premarital sex.
The religious disengagement of working class whites is important for at least three reasons. First, religious institutions typically supply their members with social and civic skills, and often a worldview that motivates them to engage the political or civic spheres, that increase their civic and political participation (Putnam, 2000; Verba et al., 1995; Wuthnow, 1995). Second, religious institutions appear to foster higher levels of physical and psychological health among their members, both by providing social support and by furnishing people with a sense of meaning (Ellison, 1991; Ellison & Levin, 1998). Third, and most important for our perspective, some research suggests that least- and moderately-educated Americans are especially likely to benefit from the social support and civic skills associated with religious institutions. The non-college-educated often lack the degree of access to social networks and civic skills that the college-educated have; and religious activity can compensate for this deficit. For instance, after reporting that more educated Americans enjoy more resources, including civic skills such as leading a meeting and income that they can use to make a donation, that provide them with a relative advantage when it comes to civic engagement and political participation, Verba et al. (1995: 18) find that “only religious affiliations… provide a counterbalance to this cumulative resource process.” This, then, is why our paper examines recent trends in religious attendance, and their social and cultural causes, among white working class Americans.
Literature Review
In this paper we first examine how and why these economic changes may have led to increasing disengagement from institutional religion among the moderately educated working class. Our attention to religion is particularly important and timely because a longstanding tradition of research on religion and stratification (e.g., Demerath, 1965; Lenski, 1961; Niebuhr, 1929) has largely fallen dormant since the 1970s. This older tradition found that working class Americans tended to be less religious in the early part of the twentieth century, at least when it came to church attendance, than their middle class peers. But by the 1970s social scientists had largely come to believe that the impact of social class on religion in the U.S. was no longer of much consequence (e.g., Hoge & Carroll, 1978; Mueller & Johnson, 1975), in large part because the fortunes of American religion rose for both the working and middle/upper classes in the post World War II era to the point where there were not marked class differences between these two groups.
But, given growing income inequality, labor force instability, and the deteriorating economic position of non-college-educated men and their families, we believe that there are good reasons to expect that American religion is once again becoming heavily stratified. Indeed, recent research suggests an emerging class divide when it comes to civic engagement more generally (Wuthnow, 2004).
Why might religious participation have declined more among moderately educated Americans than among their college-educated peers? We argue that the transformation of the economy and the resulting decline in marriage have played a central role in eroding the structural and cultural connections between religious institutions and moderately-educated men and women. We further hypothesize that these social changes have been more important when it comes to religious attendance among non-Hispanic whites than among African Americans and Hispanics. Our basic argument is that shifts in economic opportunities and in family formation over the last four decades have made many of the moral logics associated with American religious institutions both less realizable and less desirable among moderately educated whites.
Lamont (2000) has argued persuasively that steadily employed white “working class” men, nearly all of whom are moderately educated, construct a moral world centered on what she calls “the disciplined self.” It is based on the value of working hard every day, often with one’s hands, and of being a steady provider to one’s wife and children. This work ethic and sense of responsibility becomes a central part of a worker’s self-worth, she writes, and workers use it to differentiate themselves from middle-class professionals, who are seen as lacking personal integrity and sincerity. Steadily employed black workers, Lamont writes, also construct a moral sense of self, but it centers not on discipline but rather on sharing with others in need—what she calls “the caring self.”
The white working class ideal of the disciplined self has become less attainable in today’s postindustrial economy. Over the last forty years, the U.S. has witnessed a precipitous decline of job opportunities in manufacturing for working-class individuals due to globalization and automation (Fischer & Hout, 2006). The result is a sharp decline in economic opportunity for young adults who would have followed their parents’ footsteps into working-class occupations: The wages of men without college degrees have fallen since they peaked in the early 1970s, unemployment and underemployment have risen markedly among moderately educated men since the 1970s, and the wages of women without college degrees have failed to grow (Mishel, Bernstein, & Shierholz, 2009; Wilcox, 2010). In fact, high-school educated young men today may be the first generation in memory to earn less than their fathers did (Levy, 1998).
These economic developments have made it less possible for working class Americans to live up to their own ideals for economic success and to model the bourgeois lifestyle—steady work, decent income, and upward mobility—that has long been associated with mainstream religion in America (Herberg, 1955). Thus, moderately educated Americans may feel less attracted to churches that uphold the bourgeois virtues—delayed gratification, a focus on education, self-control, etc.—that undergird this lifestyle. As importantly, working class whites may also feel uncomfortable socializing with the middle and upper class whites who have increasingly come to dominate the life of religious congregations in the U.S. since the 1970s (Schwadel, McCarthy, & Nelsen, 2009), especially as they see their own economic fortunes fall. In brief, the declining economic position of white working class Americans may have made the bourgeois moral logic embodied in many churches both less attractive and attainable.
Not only have white churches in the United States functioned as bulwarks of bourgeois respectability, they have also promoted a family-centered moral logic that valorizes marriage and parenthood for much of the last century (Christiano, 2000; Edgell, 2006). When moderately educated white men and women can attain strong and stable marriages, they can find reinforcement for the lives they lead from their churches. In other words, white married couples attend church with their children partly as a way of displaying to their fellow congregants, who are often their neighbors and friends, their sense of responsibility and their commitment to familism—and also to gain reinforcement for their moral view of the world.
But since the 1970s, stable marriage has become harder to attain for moderately educated Americans (Cherlin, 2009; Wilcox, 2010). They are now markedly less likely to get and stay married as adults, compared to college-educated adults. This is partly for economic reasons, with increasing spells of unemployment and underemployment, along with declining real wages, making working class men less attractive marriage partners.
But, in all likelihood, the retreat from marriage among moderately educated Americans is also rooted in cultural changes that have gone hand in hand with the economic changes affecting this sector of American life. There is some evidence that suggests a divergence in familistic attitudes by educational attainment, with moderately educated Americans becoming less familistic relative to their college educated counterparts since the 1970s (Martin & Parashar, 2006; Wilcox, 2010). If this is the case, we might expect the familistic values promoted by many religious institutions create increasing amounts of cognitive dissonance for the moderately educated (relative to the most educated). Certainly there is an endogeneity issue here as changes in religious participation may lead to less familistic attitudes, but shifts in attitudes may also lead some to withdraw from religious institutions. The fact that less-educated Americans are now less likely to embrace a marriage-minded mindset, in turn, has been linked to the declines in the percentage of moderately educated Americans who are in their first marriage (Wilcox, 2010).
The demographic and cultural shifts that have taken place among moderately educated Americans are important because a disproportionately high percentage of active adherents in American churches are married with children (Stolzenberg, Blair-Loy, & Waite, 1995). Moreover, churches tend to be cultural bulwarks of familism, with markedly higher levels of adherence to norms against divorce and premarital sex found among regular churchgoers compared to the population at large (Wilcox, 2004). Thus, insofar as working class whites are less likely to abide by a familistic moral logic—both in practice and belief—they may be less attracted to religious congregations that tend to valorize conventional family life.
In sum, then, changes in the institutions of the labor market and the family appear to have undercut many of the socioeconomic and cultural resources that had until recently enabled many working class adults to identify with the moral logics of bourgeois respectability and familism that have long been upheld by mainstream religious institutions in the United States (Edgell, 2006; Herberg, 1955).
Black churches, however, emphasize marriage less than white churches, relative to qualities such as shared struggle and perseverance (Cherlin, 2009; Ellison & Sherkat, 1995; Lincoln & Mamiya, 1990). For instance, when it comes to family life, they speak of parent and child, of broader networks of kin, and of the fictive kinship to be found among one’s brothers and sisters in church. It is possible, then, that African Americans could achieve the caring sense of self, even in an unfavorable economy and without benefit of marriage, and find support at church. This suggests that declines in church attendance among the moderately educated should be less for blacks than for whites.
Assessing trends over the past few decades in attendance among Hispanics is difficult because of changes in the composition of the Hispanic population. Given that issues related to immigration, discrimination, and incorporation into American society loom large for churches serving Hispanics (Figueroa Deck, 1989), we suspect that Hispanic churches are less focused on family structure and employment, and more focused on providing a sense of solidarity and practical support to their members, than are non-Hispanic white churches. Moreover, there is less class heterogeneity among Hispanics, who tend not to be college-educated or affluent; this probably affords working class Hispanics a sense of comfort in the churches they attend (Schwadel, McCarthy, & Nelsen, 2009). Thus, we would expect that employment difficulties and lower incomes would be less likely to influence the church attendance of Hispanics than non-Hispanic whites.
Accordingly, we hypothesize the following:
- H1Religious service attendance will have declined more precipitously among moderately educated Americans than among college-educated Americans since the 1970s.
- H1aThis decline will be evident for Whites, but not for Blacks or Hispanics.
- H2This decline will be associated, at least in part, with the disparate economic fortunes of the moderately- and college-educated during this time, namely their employment status, income level, and job stability.
- H3This decline will be associated, at least in part, with the disparate family structures (including family structure of origin, since parental religiosity is highly predictive of offspring religiosity) of the moderately- and college-educated during this time, namely their marital and parenthood statuses.
- H4This decline will be associated, at least in part, with changing attitudes toward premarital sex, as moderately-educated views have moved in a more liberal direction and college-educated views have stayed about the same (or perhaps become slightly more conservative) since the 1970s on this issue.1
Data and Methods
We use data from the General Social Survey (GSS), a study of American adults conducted by the National Opinion Research Center at the University of Chicago, and the National Survey of Family Growth (NSFG), conducted by the National Center for Health Statistics. The GSS is an ongoing nationally-representative survey that has been conducted on an annual or biannual basis since 1972. We pool the data from all of the surveys from 1972 to 2008. We limit our sample to 25–44 year-old respondents in order to focus on the stages of the life course most closely associated with career development and family formation, and to remain consistent with the NSFG data, which—due to their focus on fertility—only include respondents up to age 44. We impose the lower age limit of 25 so that most individuals will have finished their educations. Our sample size for our baseline model is 16,620. For models including the mediating variables, the sample is 5,805 due to missing data. Cases with missing values were deleted listwise except in the case of household income, where the mean value was imputed for missing values and a dummy variable added to flag cases that were imputed.
The National Survey of Family Growth (NSFG) is a series of national surveys of women of childbearing age, 15 to 44. In this paper, we pool data from the 1982, 1988, 2002, and 2006–2008 surveys. (Two earlier waves, 1973 and 1976, only included married women. The NSFG added interviews with men to the waves conducted in the 2000s, but it has no information on men prior to that.) The main purpose of the NSFG is to obtain detailed information about fertility, such as children ever born and family planning methods used; but it also obtained marriage and cohabitation histories. It is also one of the few government surveys that asks about religious attendance and beliefs. The baseline sample size is 10,851; missing values were imputed by the NSFG staff.
Dependent variable
Our dependent variable in the GSS is a nine-category measure tapping respondents’ frequency of attendance at religious services. The GSS asked, “How often do you attend religious services?” Response categories ranging from never to more than once per week were provided. We code this variable on a scale from 0 (never) to 8 (more than once a week). The NSFG also asked about religious attendance, but the response categories varied from survey to survey. For the comparative analyses of this paper, we were able to construct a four-category dependent variable: never (coded as 1), less than once per month (2), once per month but less than once per week (3), and once per week or more often (4).
Key independent variables
Our key independent variables are educational attainment, survey year or decade, and multiplicative interaction terms between survey year or decade and educational attainment. In both datasets, educational attainment is dummy coded as less than high school degree, high school degree or some college, and the reference category of four-year college degree or more-. In the GSS “year” refers to the actual year of the survey with 1972 coded as 0 and 2008 coded as 36. For the NSFG, “decade” is a dummy variable coded as 1 if the observation is from the 2002 or 2006–8 rounds, and 0 if it is from the 1982 or 1988 rounds. Thus, the NSFG variable for year compares observations from the decade of the 2000s with observations from the decade of the 1980s. In both datasets, two interaction terms are added to the models: year times less than high school education, and year times high school degree or some college education.
Key mediating variables
Our key mediating variables are as follows:
Employment status
For the GSS, a series of dummy variables with working fulltime as the reference category, and a dummy variable marking unemployment over the last 10 years (1=ever unemployed over last 10 years). For the NSFG, a dummy variable coded as 1 if employed and 0 otherwise.
Income
For the GSS, household income in constant 1986 dollars (logged to account for skew to the right). For the NSFG, household income relative to the poverty line, a series of dummy variables: 100% to 199% of the poverty line, 200% to 299%, 300%–399%, 400% to 499%, and 500% or more, with less than 100% of the poverty line as the reference category. (In the 1982 NSFG no income or poverty information was obtained. NSFG analyses that include this variable exclude the 1982 sample.)
Current family structure
For both datasets, a series of dummy variables: married with children, married without children, unmarried with children, and the reference category of unmarried without children. For the GSS, a dummy variable for whether the respondents’ parents were divorced or separated at age 16. For the NSFG, a dummy variable for whether the respondent lived with both biological or adoptive parents from birth to age 18.
Cultural factors
For the GSS, we include a dummy variable for whether respondents oppose premarital sex. The NSFG has no relevant attitudinal variables, but we have included age at first intercourse, given that premarital sex has been linked to declines in religious attendance (Uecker, Regnerus, and Vaaler 2007).
Control variables
We include GSS controls for region of residence, age, age squared, and gender. For the NSFG, we include age and age squared (all respondents are women, and region of residence is not available). (Note: for descriptive statistics for the GSS and NSFG variables, see, respectively, Appendix A and Appendix B.)
Appendix A
Descriptive Statistics for Variables in Regression Models, General Social Surveys 1972–2010, a Non-Hispanic White 25–44-Year-Old Respondents
| Full sample (N = 17, 077) | Reduced sample (N= 5,959) | |||||
|---|---|---|---|---|---|---|
| Mean | SD | Range | Mean | SD | Range | |
| Religious service attendance | 3.49 | 2.64 | 0–8 | 3.55 | 2.66 | 0–8 |
| Survey yearb | 18.29 | 10.84 | 0–38 | 13.88 | 10.58 | 0–36 |
| Least educated | .11 | 0, 1 | .12 | 0, 1 | ||
| Moderately educated | .61 | 0, 1 | .61 | 0, 1 | ||
| Most educated (ref.) | .28 | 0, 1 | .27 | 0, 1 | ||
| Least educated*Survey yearb | 1.53 | 5.67 | 0–38 | 1.21 | 4.79 | 0–36 |
| Moderately educated*Survey yearb | 11.11 | 12.20 | 0–38 | 8.38 | 10.55 | 0–36 |
| Working part time | .11 | 0, 1 | ||||
| Temporarily not working | .02 | 0, 1 | ||||
| Unemployed, laid off | .03 | 0, 1 | ||||
| Retired | .00 | 0, 1 | ||||
| School | .02 | 0, 1 | ||||
| Keeping house | .16 | 0, 1 | ||||
| Other | .01 | 0, 1 | ||||
| Working fulltime (ref.) | .64 | 0, 1 | ||||
| Household income (logged) | 10.57 | .75 | 6.00–12.10 | |||
| Household income missing | .05 | 0, 1 | ||||
| Ever unemployed in last 10 years | .39 | 0, 1 | ||||
| Parents divorced/separated at age 16 | .13 | 0, 1 | ||||
| Unmarried with no children (ref.) | .19 | 0, 1 | ||||
| Unmarried with children | .15 | 0, 1 | ||||
| Married with no children | .10 | 0, 1 | ||||
| Married with children | .56 | 0, 1 | ||||
| More conservative premarital sex attitudes | 1.98 | 1.16 | 1–4 | |||
| Lives in New England (ref.) | .05 | 0, 1 | .06 | 0, 1 | ||
| Lives in Middle Atlantic | .15 | 0, 1 | .15 | 0, 1 | ||
| Lives in East North Central | .20 | 0, 1 | .21 | 0, 1 | ||
| Lives in West North Central | .08 | 0, 1 | .08 | 0, 1 | ||
| Lives in South Atlantic | .17 | 0, 1 | .17 | 0, 1 | ||
| Lives in East South Central | .06 | 0, 1 | .06 | 0, 1 | ||
| Lives in West South Central | .08 | 0, 1 | .08 | 0, 1 | ||
| Lives in Mountain | .07 | 0, 1 | .06 | 0, 1 | ||
| Lives in Pacific | .13 | 0, 1 | .13 | 0, 1 | ||
| Age | 34.30 | 5.71 | 25–44 | 34.14 | 5.73 | 25–44 |
| Age-squared | 1209.15 | 394.19 | 625–1936 | 1198.12 | 394.69 | 625–1936 |
| Female | .54 | 0, 1 | .55 | .50 | 0, 1 | |
Appendix B
Descriptive Statistics for Variables in Regression Models, National Survey of Family Growth 1982–2008, Non-Hispanic White Female 25–45-Year-Old Respondents
| Full sample (N = 10,851) | Reduced sample (N= 8,682) | |||||
|---|---|---|---|---|---|---|
| Mean | SD | Range | Mean | SD | Range | |
| Religious service attendance | 2.59 | 1.15 | 0–4 | 2.58 | 1.15 | 0–4 |
| Survey year (1982–88) (ref.) | .48 | 0, 1 | .40 | 0, 1 | ||
| Survey year (2002–08) | .52 | 0, 1 | .60 | 0, 1 | ||
| Least educated | .11 | 0, 1 | .11 | 0, 1 | ||
| Moderately educated | .63 | 0, 1 | .63 | 0, 1 | ||
| Most educated (ref.) | .26 | 0, 1 | .26 | 0, 1 | ||
| Least educated*Survey year | .06 | 0, 1 | .07 | 0, 1 | ||
| Moderately educated*Survey year | .25 | 0, 1 | .30 | 0, 1 | ||
| Employed | .74 | 0, 1 | ||||
| Below poverty line (ref.) | .06 | 0, 1 | ||||
| 100 – 199% of poverty line | .16 | 0, 1 | ||||
| 200 – 299% of poverty line | .16 | 0, 1 | ||||
| 300 – 399% of poverty line | .19 | 0, 1 | ||||
| 400 – 499% of poverty line | .22 | 0, 1 | ||||
| 500% or more of poverty line | .21 | 0, 1 | ||||
| Parents divorced/separated at age 14 | .28 | 0, 1 | ||||
| Unmarried with no children (ref.) | .18 | 0, 1 | ||||
| Unmarried with children | .11 | 0, 1 | ||||
| Married with no children | .17 | 0, 1 | ||||
| Married with children | .54 | 0, 1 | ||||
| Age at first sexual intercourse | 18.85 | 3.25 | 3–42 | 18.86 | 3.31 | 3–40 |
| Age | 34.04 | 5.69 | 25–45 | 34.97 | 5.70 | 25–45 |
| Age-squared | 1190.65 | 392.38 | 625–2025 | 1192.49 | 393.94 | 625–2025 |
We first present graphs displaying trends in attendance by education, race, and Hispanic ethnicity for Americans age 25–44. We then present a series of ordinary least-squares (OLS) regression models predicting religious service attendance among White Americans ages 25–44. Our first model (Model 1) uses all available cases to establish a significant interaction between educational attainment and survey year (for the GSS) or survey decade (for the NSFG). We present a second model (Model 2a) for the GSS that parallels the first but is restricted to the cases for which there is nonmissing information for all the mediating variables. We do not present Model 2a for the NSFG because missing values were imputed by the NSFG staff. We then enter the mediating variables in blocks: Model 2b enters the economic variables, Model 2c enters the family structure variables (and removes the economic variables), Model 2d enters the cultural variables (and removes the economic and family structure variables) and Model 2e enters all of the independent and mediating variables. This approach allows us to examine the mediating effect of economic, family structure, and cultural factors separately, as well as together.
Results
Figure 1 displays trends in the percentage of women aged 25–44 who attended religious services once per month or more often, based on the GSS and NSFG data, by race and Hispanic ethnicity, from the 1980s to the 2000s. Let us begin with Panel A for non-Hispanic Whites. Looking first at the two left-hand sets of bars for the 1980s, we can see that in both data sets the moderately educated (the middle category) were nearly as likely to attend monthly or more as were the college-educated. Both datasets also show that the least educated had markedly less religious attendance than did either of the two more-educated groups. Turning to the two right-hand bars for the 2000s, we can see that the level of attendance of the college-educated had diverged from the level among the moderately educated. In the GSS, all three educational groups showed a decline, but the decline (8 percentage points) was greatest among the moderately educated. In the NSFG, attendance among the college-educated had increased slightly, while decreasing among the moderately educated and the least educated (4 percentage points for both groups). Overall, both datasets show a pattern in the 1980s of comparable attendance among the moderately educated and the college educated, whereas the pattern in the 2000s is of sharper differences between the higher attendance of the college-educated and lower attendance among both the moderately educated and least educated.
Percentage of women, age 25 to 44, who attended religious services once per month or more often, pooled 1980s compared to pooled 2000s (GSS) or pooled 1982 and 1987 compared to pooled 2002 and 2006–8 (NSFG), by race and Hispanic ethnicity.
In contrast, attendance for non-Hispanic Blacks (Panel B) shows less decline. Among the moderately educated, in particular, attendance declined only 2 percentage points in the GSS data from the 1980s to the 2000s and it did not change at all in the NSFG data. Among Hispanics (Panel C) the educational gradient in attendance is not visible at either time period, according to the GSS data. (We do not present data for Hispanics from the NSFG because of small sample sizes in the 1980s waves.) Thus, as predicted, the attendance declines of the moderately educated relative to the college-educated are small or non-existent among non-Hispanic Blacks and Hispanics. Consequently, we focus our data analysis on non-Hispanic Whites.
Table 1 reports coefficients from OLS regression models predicting religious service attendance in the 1972–2010 General Social Surveys. Model 1 shows that there have been differential declines in religious service attendance across education categories. Religious service attendance has been in decline for all groups across this time period, but the decline has been steeper among those with no high school degree and those with a high school degree but no college degree. While college-educated Whites saw their attendance go down by about .016 units per year (see the coefficient for survey year, which for the college educated is not modified by either interaction term), this decline is more than twice as steep for less-educated Americans. Summing the main effect of year with the respective interaction terms—which are significant at p < .001 for the less-than-high-school and moderately educated groups—we find that religious service attendance declined on the order of .039 units per year for the least educated and .034 per year for the moderately educated. Although a .034 unit decrease for the moderately educated may seem small, the cumulative effect of this difference over the 38-year study period is quite sizable: 1.292 units, or about .49 standard deviations in religious service attendance. Moderately educated White Americans were not significantly less likely than the most educated to attend religious services in 1972 (b = −.110; p = .23), but they clearly were less likely to do so in 2010. Ancillary analyses show that in 2010, religious service attendance was .785 units lower (p < .001) among the moderately educated than among the most educated.
Table 1
Coefficients from Ordinary Least Squares Regression Models Predicting Religious Service Attendance, General Social Surveys 1972–2010,a Non-Hispanic White 25–44-Year-Old Respondents
| Model 1 | Model 2a | Model 2b | Model 2c | Model 2d | Model 2e | |
|---|---|---|---|---|---|---|
| Key Independent Variable and Constitutive Terms | ||||||
| Least educated* Survey year | −.023*** | −.014 | −.008 | −.007 | .003 | .006 |
| Moderately educated* Survey year | −.018*** | −.016* | −.012† | −.010 | −.011† | −.008 |
| Survey year | −.016*** | −.018** | −.015* | −.009 | −.013* | −.007 |
| Least educated | −.990*** | −1.250*** | −1.206*** | −1.401*** | −1.577*** | −1.593*** |
| Moderately educated | −.110 | −.127 | −.138 | −.298* | −.296* | −.379** |
| Mediating Variables | ||||||
| Working part time | .548*** | .283** | ||||
| Temporarily not working | −.153 | −.144 | ||||
| Unemployed, laid off | −.001 | .085 | ||||
| Retired | −.231 | .428 | ||||
| School | .722** | .584** | ||||
| Keeping house | .602*** | .134 | ||||
| Other | .040 | .042 | ||||
| Household income (logged) | .189*** | .054 | ||||
| Household income missing | −.032 | −.247† | ||||
| Ever unemployed in last 10 years | −.557*** | −.338*** | ||||
| Parents divorced/separated at age 16 | −.554*** | −.377*** | ||||
| Unmarried with children | .126 | .231* | ||||
| Married with no children | .342** | .178 | ||||
| Married with children | 1.171*** | .637*** | ||||
| More conservative premarital sex attitudes | 1.039*** | .975*** | ||||
| Control Variables | ||||||
| Lives in Middle Atlantic | .066 | .225 | .236 | .205 | .068 | .084 |
| Lives in East North Central | .329** | .374* | .418** | .327* | −.047 | −.025 |
| Lives in West North Central | .541*** | .533** | .592** | .528** | .098 | .129 |
| Lives in South Atlantic | .479*** | .575*** | .622*** | .579*** | −.089 | −.026 |
| Lives in East South Central | 1.046*** | 1.105*** | 1.168*** | 1.060*** | .245 | .289† |
| Lives in West South Central | .716*** | .734*** | .792*** | .745*** | .072 | .143 |
| Lives in Mountain | .071 | .183* | .263 | .201 | −.334† | −.265 |
| Lives in Pacific | −.549*** | −.429* | −.374* | −.339* | −.623*** | −.541*** |
| Age | .151** | .212** | .167* | .056 | .106 | −.018 |
| Age-squared | −.002* | −.002* | −.002† | .000 | −.001 | .000 |
| Female | .611*** | .636*** | .389*** | .593*** | .442*** | .341*** |
| Constant | .408 | −.799 | −1.797 | 1.489 | −.167 | .783 |
| R-squared | .069 | .071 | .093 | .111 | .264 | .284 |
| N | 17,077 | 5,959 | 5,959 | 5,959 | 5,959 | 5,959 |
Reference groups are most educated, working fulltime, unmarried with no children, and lives in New England.
The least educated have experienced faster rates of decline than even the moderately educated, and they began at an even “lower” starting point (b = −.990; p < .001), meaning the gap between the least educated and most educated is even larger than the one between the moderately educated and most educated.
Models 2a–e of Table 1 attempt to shed light on the mechanisms driving these differential rates of change. Because our potential mediating variables introduce large amounts of missing data, we present another model similar to Model 1 but with the reduced sample (Model 2a). Then we consider the mediating effects of economic factors (Model 2b), family structure factors (Model 2c), cultural factors (Model 2d), and all three types of factors (Model 2e). Model 2a reveals a similar story as Model 1, though the interaction effect for year times least educated is no longer statistically significant and is smaller in size than the interaction for year times moderate education. The interaction effect for the moderately educated is slightly smaller than in Model 1 and significant at p < .05. Model 2b suggests the changing economic fortunes of the moderately and most educated Americans are associated with the gap in religious service attendance. In particular, those with higher incomes are more likely to attend religious services more frequently, and those who have experienced unemployment at some point over the last 10 years are less likely to attend (Schwadel, McCarthy, & Nelsen, 2009). Moreover, those who “keep house” are much more likely to attend religious services, a group that has declined more rapidly among the moderately educated than among the most educated. These economic factors reduce the year-moderate education interaction effect by about 25 percent, and its statistical significance is only marginal.
Model 2c looks at how changes in family structure are associated with this growing divide in religious service attendance. Growing up with divorced parents has a strong negative association with religious service attendance as an adult, and being married—especially being married with children—has a positive effect. These variables reduce the difference in the rate of decline between the moderately and most educated by about 38 percent and make it no longer significant statistically. Model 2d looks at associations with changing attitudes—measured here by attitudes toward premarital sex. Those who say premarital sex is wrong in more circumstances are much more likely to go to religious services more often than are those who are more accepting of premarital sex. Accounting for these different attitudes reduces the year-moderately educated interaction effect by about 31 percent, and makes it statistically significant only at p < .10. In the final model (Model 2e), the year-moderately educated interaction is reduced by 50 percent and is no longer significant. Certainly there are endogeneity issues with some of these mediators, and we caution against any strong causal conclusions from these data, but it does appear that differences in economics, family structure, and culture account for some of the difference in declining religious service attendance between moderately educated Americans and their college educated counterparts.
Table 2 reports coefficients from OLS regression models predicting religious service attendance in the pooled 1980s and pooled 2000s National Surveys of Family Growth. The results are, in general, similar to the GSS. The coefficients in Model 1 suggest that religious attendance has dropped more for the moderately educated than for the college educated, although the significance levels are not as strong as in the GSS: The key interaction between moderate education and decade is negative, as predicted, and significant at the p < .10 level; the interaction between low education and decade is negative but not significant. Models 2b and 2c show significant associations with religious attendance that reduce the magnitudes and significance of the key interaction terms for education and decade. There is a curvilinear relationship between increases in income (relative to the poverty line) and religious service attendance, with attendance highest among those at 200–299% of the poverty line. Being married with children is also, as in the GSS, associated with more frequent religious service attendance. Age at first intercourse, which was obtained in the NSFG but not the GSS, shows a significant association between an older age at first intercourse and more frequent religious attendance. But age at first intercourse does not appear to impact the association between the key interaction terms for education and decade. Moreover, we must caution here against strong causal inferences because it is just as plausible that frequent religious attendance is the cause of an older age at first intercourse rather than the effect. Except for employment, all of the mediating variables retain their statistical significance in the final Model 2e.
Table 2
Coefficients from Ordinary Least Squares Regression Models Predicting Religious Service Attendance, National Survey of Family Growth, Pooled 1982 and 1987 Waves Compared to Pooled 2002 and 2006–08 Waves, Non-Hispanic White 25–44-Year-Old Respondents, Women Only
| Model 1 | Model 2a | Model 2b | Model 2c | Model 2d | Model 2e | |
|---|---|---|---|---|---|---|
| Key Independent Variable and Constitutive Terms | ||||||
| Least educated* Decade | −.079 | — | −.006 | −.043 | −.109 | −.017 |
| Moderately educated* Decade | −.115† | — | −.067 | −.003 | −.116† | .014 |
| Decade | .014 | — | −.039 | .004 | .124* | .080 |
| Least educated | −.477*** | — | −.611*** | −.431 | −.225** | −.297*** |
| Moderately educated | −.074† | — | −.176*** | −.105 | .056 | −.050 |
| Mediating Variables | ||||||
| Employed | −.154** | −.023 | ||||
| 100–199% poverty line | .169† | .266* | ||||
| 200–299% poverty line | .352*** | .072*** | ||||
| 300–399% poverty line | .463*** | .345*** | ||||
| 400–499% poverty line | .368*** | .257*** | ||||
| 500% or more of poverty line | .329*** | .184** | ||||
| Parents together through age 18 | .247*** | .181*** | ||||
| Married with no children | −.375*** | −.246*** | ||||
| Unmarried with children | −.444*** | −.404*** | ||||
| Unmarried with no children | −.535*** | −.628*** | ||||
| Age at First Intercourse | .062*** | .278*** | ||||
| Control Variables | ||||||
| Age | .018 | — | −.017 | −.073 | .016 | −.095* |
| Age-squared | .000 | — | .000 | .001† | .000 | .001* |
| Constant | 2.144 | — | 2.624 | 3.743 | .983 | 2.671 |
| R-squared | .018 | — | .040 | .064 | .049 | .114 |
| N | 10,851 | — | 8,682 | 10,851 | 10,851 | 8,682 |
Note: Reference groups are most educated, below poverty line, and married with children.
Discussion
This paper finds evidence that religious life among the moderately educated – which may be the closest analogy to the “working class” today – is becoming increasingly deinstitutionalized, much as working class economic and family life have become increasingly deinstitutionalized. Using repeated cross-sectional surveys from two national data collections programs, the GSS and the NSFG, we find that religious attendance among whites has declined most precipitously among whites without college degrees, including moderately educated whites—that is, whites with a high school degree or some college but no bachelor’s degree. By contrast, we do not find a decline among moderately educated blacks; and we do not find a monotonic educational gradient at all among Hispanics.
Our results suggest that the bourgeois and familistic moral logics that have long been linked to religious institutions are now less powerful in the lives of working class whites than they used to be. Specifically, in the last forty years, white working class income, employment, marital stability, and cultural conservatism have all declined—and markedly more so than they have for college-educated whites (Cherlin, 2009; Wilcox, 2010). Indeed, our results suggest that these bourgeois and familistic factors may account for a substantial share of the relatively large decline of working class church attendance. Within the limits of observational data, we think that our results suggest that the erosion of the labor market and cultural structures associated with the bourgeois and familistic moral logics in American life may have played an important role in accounting for recent declines in religious attendance among working class whites.
While we recognize that not everyone wishes to worship, and that religious diversity can be valuable, we also think that the existence of a large group in the middle of the American stratification system that is increasingly disconnected from religious institutions is troubling for our society. This development is especially troubling because it only reinforces the social marginalization of working class whites who are also increasingly disconnected from the institutions of marriage and work (Cherlin, 2011).
Moreover, our results suggest that it is important for the sociology of religion to once again attend to social stratification in religious life. The broadly shared prosperity of the mid-twentieth century may have diminished class differences in religious experience in the middle of the twentieth century to the point that they were unimportant. That at least would seem to be the conclusion drawn by leading scholars during the latter half of the twentieth century. But two great forces of change have widened the differences since the 1970s: the bifurcation of the labor market due to globalization and automation and the great cultural changes in family life that have made non-marriage-based family patterns acceptable. Studies of religion in twenty-first century America will need to take these class-based differences into account.
Acknowledgments
The third author received support from the grant 5 T32 HD007168, Population Research Training, awarded to the Carolina Population Center at The University of North Carolina at Chapel Hill by the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
Footnotes
1It may also be that this trend merely reflects the rise of religious nones (Hout & Fischer, 2002). Controlling for growing up with no religion, however, did not affect our results significantly.
Contributor Information
W. Bradford Wilcox, University of Virginia.
Andrew J. Cherlin, Johns Hopkins University.
Jeremy E. Uecker, University of North Carolina at Chapel Hill.
Matthew Messel, Johns Hopkins University.
References
- Cherlin AJ. The marriage-go-round: The state of marriage and the family in America today. New York: Alfred A. Knopf; 2009.
- Cherlin AJ. Between poor and prosperous: Do the family patterns of moderately-educated Americans deserve a closer look? In: Carlson M, England P, editors. Social class and changing families in an unequal America. Stanford: Stanford University Press; 2011. pp. 68–84.
- Christiano KJ. Religion and the family in modern American culture. In: Houseknecht SK, Pankhurst JG, editors. Family, religion, and social change in diverse societies. New York: Oxford University Press; 2000. pp. 43–78.
- Deck AF. The second wave: Hispanic ministry and the evangelization of cultures in the United States. Mahwah, NJ: Paulist Press; 1989.
- Demerath NJ. Social class in American Protestantism. Chicago: Rand McNally & Company; 1965.
- Edgell P. Religion and family in a changing society. Princeton, NJ: Princeton University Press; 2006.
- Ellison CG. Religious involvement and subjective well-being. Journal of Health and Social Behavior. 1991;32:80–99. [PubMed]
- Ellison CG, Levin JS. The religion-health connection: Evidence, theory, and future directions. Health Education and Behavior. 1998;25:700–720. [PubMed]
- Ellison CG, Sherkat DE. The “semi-involuntary institution” revisited: Regional variations in church participation among black Americans. Social Forces. 1995;73:1415–1437.
- Fischer CS, Hout M. Century of difference: How America changed in the last one hundred years. New York: Russell Sage; 2006.
- Friedland R, Alford RR. Bringing society back in: Symbols, practices, and institutional contradictions. In: Powell WW, DiMaggio PJ, editors. The new institutionalism in organizational analysis. Chicago: University of Chicago Press; 1991. pp. 232–263.
- Hacker JS. The great risk shift: The assault on American jobs, families, health care, and retirement and how you can fight back. New York: Oxford University Press; 2006.
- Herberg W. Protestant, Catholic, Jew: An essay in American religious sociology. Garden City, NY: Doubleday; 1955.
- Hoge DR, Carroll JW. Determinants of commitment and participation in suburban Protestant churches. Journal for the Scientific Study of Religion. 1978;17:107–127.
- Hout M, Fischer CS. Why more Americans have no religious preference: Politics and generations. American Sociological Review. 2002;67:165–190.
- Lamont M. The dignity of working men: Morality and the boundaries of race, class, and immigration. New York: Russell Sage; 2000.
- Lenski G. The religious factor: A sociological study of religion’s impact on politics, economics, and family life. Garden City, NY: Doubleday; 1961.
- Levy F. The new dollars and dreams: American incomes and economic change. New York: Russell Sage; 1998.
- Lincoln CE, Mamiya LH. The black church in the African American experience. Durham, NC: Duke University Press; 1990.
- Martin SP, Parashar S. Women’s changing attitudes toward divorce, 1974–2002: Evidence for an educational crossover. Journal of Marriage and Family. 2006;68:29–40.
- Mishel L, Bernstein J, Shierholz H. The state of working America 2008/2009. Ithaca, NY: ILR/Cornell University Press; 2009.
- Mueller CW, Johnson WT. Socioeconomic status and religious participation. American Sociological Review. 1975;40:785–800.
- Niebuhr HR. The social sources of denominationalism. Gloucester, MA: Henry Holt and Co; 1929.
- Putnam RD. Bowling alone: The collapse and revival of American community. New York: Simon & Schuster; 2000.
- Schwadel P, McCarthy JD, Nelsen HM. The continuing relevance of family income for religious participation: U.S. white Catholic church attendance in the late 20th century. Social Forces. 2009;87:1997–2030.
- Stolzenberg RM, Blair-Loy M, Waite LJ. Religious participation in early adulthood: Age and family life cycle effects on church membership. American Sociological Review. 1995;60:84–103.
- Uecker JE, Regnerus MD, Vaaler ML. Losing my religion: The social sources of religious decline in early adulthood. Social Forces. 2007;85:1667–1692.
- Verba S, Schlozman KL, Brady HE. Voice and equality: Civic voluntarism in American politics. Cambridge, MA: President and Fellows of Harvard College; 1995.
- Wilcox WB. Soft patriarchs, new men: How Christianity shapes fathers and husbands. Chicago: University of Chicago Press; 2004.
- Wilcox WB. When marriage disappears: The retreat from marriage in middle America. Charlottesville, VA: The National Marriage Project and Institute for American Values; 2010.
- Wuthnow R. Learning to care: Elementary kindness in an age of indifference. New York: Oxford University Press; 1995.
- Wuthnow R. Saving America? Faith-based services and the future of civil society. Princeton, NJ: Princeton University Press; 2004.
Facebook
Twitter
Google+