Religion and Attitudes toward Childlessness in the United States

Abstract

Attitudes toward childlessness have received little attention from social scientists even as childlessness as a family form has become more popular. One key predictor of childlessness attitudes is religious commitment, though few studies have examined this association in-depth. Using data from two recent, national datasets—the National Survey of Family Growth and the Survey of U.S. Catholics and Family Life—we assess the relationships between individuals’ religious characteristics and various attitudes about their own and others’ childlessness, as well as how these factors may vary across gender. We find strong associations between religious affiliation, religious attendance, and religious salience, and each of the outcomes such that religious commitment tends to be associated with more negative attitudes toward childlessness. Interactions across gender were rare and the direction of the interactions was mixed. Religion is a pivotal factor in perpetuating pronatalist attitudes in the United States among both men and women.

Keywords

Religion Childlessness Family Attitudes Childfree Voluntary Childlessness Fertility Gender & Family

Introduction

Scholars have paid a great deal of attention to the diversification of families in the United States in recent years, but still relatively little research has focused on people who remain childless, whether involuntarily or by choice, or people’s attitudes toward the childless (Blackstone, 2014). Childless adults, however, comprise a significant minority of adults in the United States. According to data from the Current Population Survey, about 15% of 40–44-year-old women are childless (United States Census Bureau, 2018), and data from the National Survey of Family Growth (NSFG) suggest 20% of men ages 40 to 44 are childless (Martinez et al., 2018). Moreover, about 6% of 40–44-year-old women are voluntarily childless (Martinez et al., 2012). Viewing childlessness as an acceptable and fulfilling way of life is even more widespread. In the National Survey of Families and Households (1987–1988 and 1992–1994), 22% of respondents disagreed that it is better to have a child than remain childless, and 43% agreed men and women can have a fulfilling life without children (Koropeckyj-Cox & Pendell, 2007a). Acceptance of childlessness appears to be increasing over time (Thornton & Young-DeMarco, 2001).

Understanding attitudes toward childlessness is important for at least three reasons. First, attitudes about childlessness are clearly linked to one’s own fertility intentions (Koropeckyj-Cox & Pendell, 2007a), which are clearly linked to one’s own fertility outcomes (Schoen et al., 1999). Second, attitudes toward childlessness shed light on broader cultural norms, such as the persistence of a “motherhood mandate” (Russo, 1976) or its cultural successor, the “superwoman,” who is expected to successfully combine work and parenting (Gillespie, 2001). Relatively low levels of desired childlessness and of acceptance of others’ childlessness may reveal the perpetuation of what has been described as a strong pronatalist culture in the United States (Park, 2005). This pronatalist culture is characterized by the belief that childbearing is salutary for the well-being of individuals, families, and society (Heitlinger, 1991; Park, 2005).

Third, low levels of desired childlessness and acceptance of others’ childlessness may have salient implications for the lives of childless adults. Couples, and especially women, who experience infertility or choose to remain childless report being stigmatized or the target of negativity (Jansen & Saint Onge, 2015; Jordan & Revenson, 1999; McCutcheon, 2018; Meyers, 2001; Park, 2002; Parry, 2005). Indeed, the voluntarily childless, or childfree, report being perceived as selfish, cold, and materialistic (Park, 2002), among several other negative stereotypes (see Blackstone & Stewart, 2012, for a review). Parents, alternatively, are viewed as warmer, kinder, more caring, and more nurturing than nonparents, especially women who are voluntarily childless (Koropeckyj-Cox et al., 2018). Understanding the prevalence of correlates of stigmatizing or negative attitudes could help childless adults have a better sense of how common their own attitudes are, and how their childlessness is viewed by others.

Despite these implications, few studies have examined attitudes toward childlessness with respect to one’s own childlessness or with respect to childlessness among others (Koropeckyj-Cox & Pendell, 2007a), and there is a particular dearth of studies explicitly on attitudes toward voluntary childlessness. Many recent studies that do examine childlessness attitudes in-depth rely on samples of undergraduate students that likely do not generalize to the population (McCutcheon, 2018). Two exceptions to these criticisms are a pair of studies by Koropeckyj-Cox and Pendell (2007a, 2007b). They examined attitudes about others’ childlessness, but they use data that are now decades old, and they did not examine attitudes toward voluntary childlessness explicitly.

Moreover, the recent literature often overlooks or glosses over an essential piece of the puzzle: how religion is associated with attitudes about childlessness. This is curious given the long-acknowledged role of religious groups in promoting procreation within marriage (Veevers, 1980). Scholars have frequently noted that many religious communities have teachings on family values and childrearing that encourage their congregants to have children (Hackett, 2008; Hayford & Morgan, 2008; Heaton, 1986; McQuillan, 2004). These religious messages have the potential to shape attitudes toward those who choose not to have children. But religion and family research frequently overlooks the issue of childlessness per se, focusing instead on outcomes like fertility rates (Hackett, 2008; Hayford & Morgan, 2008; Mosher et al., 1992; Zhang, 2008) or fertility timing (Pearce, 2010; Pearce & Davis, 2016). The few studies that do examine religion and childlessness are dated, exploring data from National Survey of Families and Households (collected in the late 1980s and early 1990s) and the 1994 General Survey (e.g., Heaton et al., 1992; Koropeckyj-Cox & Pendell, 2007a, 2007b), and do not consider attitudes toward voluntary childlessness.

This study serves as a contemporary investigation into the relationship between religion and a wide variety of childlessness attitudes. We use two nationally representative datasets to examine how religious characteristics are associated with a variety of attitudes related to one’s own childlessness, including (a) one’s own childlessness intentions, (b) the degree to which individuals who did not intend to be childless would be bothered if they did not have children, (c) whether childbearing is necessary for people’s happiness, and (d) how acceptable voluntary childlessness is as a way of life. We also consider how these associations may vary across gender. Before so doing, we first explain theoretically why we would expect differences in attitudes about childlessness to exist across religious characteristics.

Religious Commitment and Childlessness Attitudes

There is a long tradition of investigating fertility differences among religious groups (Lynch, 2006). Historically, these investigations have fallen under what has been termed the particularized theology hypothesis (Goldscheider, 1971), according to which fertility differentials are attributed to doctrinal differences across religious groups. These doctrinal messages can be conceived of as schemas that individuals use to determine how and why to act, as well as to evaluate the actions of other people (Johnson-Hanks et al., 2011; Sewell, 2005).

For example, large, hierarchical religious traditions in the United States have clear statements about childbearing that could shape the way their adherents think about childlessness. The Catholic church is perhaps the prime example of such a tradition. Among many authoritative statements on the topic, the Catechism of the Catholic Church (2011, p. 2201) makes childbearing and childrearing a central component of marriage: “Marriage and the family are ordered to the good of the spouses and to the procreation and education of children.” Although authorities in the Catholic church may unambiguously proclaim these pronatalist notions, Catholics have been much less enthusiastic about adapting them themselves (Curran, 2018; Westoff & Jones, 1979). Acceptance of doctrine regarding childbearing among Catholics is thus not guaranteed; indeed, Catholics in the late 1980s and early 1990s did not differ from other groups in their attitudes toward others’ childlessness (Koropeckyj-Cox & Pendell, 2007a).

The Mormon church features a pro-family doctrine that integrates the importance of raising children in its theology and promotes a family-centered lifestyle (Tark, 2003). In the LDS church’s 1995 “Proclamation to the Family,” church authorities state, “The first commandment that God gave to Adam and Eve pertained to their potential for parenthood as husband and wife. We declare that God’s commandment for His children to multiply and replenish the earth remains in force.” LDS fertility is well above the national average (Hackett, 2008; Heaton, 1986).

In general, more conservative religions, including conservative Protestants, seem to emphasize the importance of childbearing more so than their mainstream counterparts and have higher rates of fertility (Hackett, 2008; Wilcox, 2004). Perry and Schleifer (2019) find that even though conservative Protestants’ number of children are no longer distinct from others, they are still more likely to have at least one child. Baptists and Fundamentalists in the late 1980s and early 1990s also reported more negative attitudes toward childlessness (Koropeckyj-Cox & Pendell, 2007a). Mainline Protestants tend to be oriented toward traditional family forms, as congregation-level messages often outweigh the progressive rhetoric at the denominational level (Edgell, 2006). Shain (2019) found that Orthodox Jews have significantly higher fertility rates than Reform Jews, even when controlling for mother’s education. Jewish Americans in the late 1980s and early 1990s had more negative attitudes toward childlessness than did others (Koropeckyj-Cox & Pendell, 2007a).

Given the unique messages, adherents of a denomination may receive from religious authorities, we would expect that attitudes toward childbearing would differ based on religious affiliation. However, as other scholars point out, embeddedness in a religious community and the importance individuals place on religion in their daily lives may be more crucial than one’s religious identity and the accompanying family-related doctrine (Hackett, 2008; Hayford & Morgan, 2008; Zhang, 2008).

Attendance at religious services is regularly shown to be significant in predicting fertility behaviors and attitudes (e.g., Mosher et al., 1992; Hackett, 2008; Perry & Schleifer, 2019), including attitudes toward childlessness (Koropeckyj-Cox & Pendell, 2007a). Embeddedness in a community that espouses pronatalist messages and has pronatalist norms likely shapes individuals’ perceptions about what constitutes a “normal” family size and the appropriateness of remaining childless (Hackett, 2008). Religious communities may also imbue childbearing with sacred significance through rituals like baptism (Wilcox & Wolfinger, 2008).

As faith becomes more individualized, however, we may expect the role of religious service attendance to decrease in importance vis-à-vis personal religious salience (Stark & Bainbridge, 1985; Wuthnow, 1988). In other words, religion may not have decreased in importance, but the focus has turned from congregational to individual as much of American religion has come to value one’s personal relationship with God over and above adherence to doctrine and moral codes, or membership in a religious community (Bellah et al., 1985; Luhrmann, 2012; Madsen, 2009; Wuthnow, 1988). Using personal religiosity to determine fertility behaviors has grown increasingly popular in recent fertility scholarship (Hayford & Morgan, 2008; Pearce & Davis, 2016; Zhang, 2008). Religious messages and norms may not matter for attitudes and behaviors unless they are internalized by the individual and serve as a guiding schema for how to think and act (Zhang, 2008).

At the same time, personal religiosity that is an individualized religiosity—unmoored from the teachings of religious traditions—may not be impactful as one’s religiosity is privately fashioned and can include or exclude religious teachings that do not suit the individual (Chaves, 1994). Indeed, religious individualism appears to be associated with more progressive family-related attitudes (Uecker & Froese, 2019).

Gender, Religion, and Childlessness Attitudes

Several scholars critique research on religion and fertility that focus on women’s attitudes and ignore men’s (Lehrer, 1996; Zhang, 2008). In this article, we consider the role of religion among both women and men in shaping their attitudes on childlessness. Beyond that, we consider how religion may differentially impact men’s and women’s attitudes toward childlessness. Specifically, we suggest that religion may matter more for men’s attitudes than it does for women’s.

Considerable evidence suggests that religion is an especially important factor leading men to be invested in family life. Christianity, and particularly conservative Protestantism, is associated with men engaging in more “warm, expressive [styles] of fatherhood” (Wilcox, 2004, p. 107). Tichenor et al. (2011) find that higher religiosity and non-egalitarian gender attitudes are associated with men’s higher perceived importance of fatherhood. Although religiosity is also associated with women’s reported importance of motherhood (McQuillan et al., 2008; McQuillan et al., 2012; McQuillan et al., 2015; Tichenor et al., 2017), there are fewer norms and institutions pushing men to invest in their families, while women—irrespective of religiosity—have to manage norms of intensive mothering (Hays, 1998). Thus, religion may be one of only a few institutions that clearly promotes family formation among men, whereas women receive those messages from a variety of sources.

This Study

We examine the role of religion in shaping attitudes towards one’s own childlessness and the suitability of childlessness in general. We consider four attitudes: (a) the fertility intentions of those who are childless, (b) how bothered childless individuals who intend to have children would be if they did not, (c) whether having children is necessary for people’s happiness, and (d) the acceptability of voluntary childlessness as a way of life. We use data from the 2006–2017 NSFG continuous interviews, as well as the Pew Research Center’s 2015 Survey of U.S. Catholics and Family Life (SUSCFL), a representative sample of U.S. adults (not only Catholics). We consider the role of religious affiliation, religious service attendance, and religious salience in shaping these attitudes and explore the potential for differential religious effects across gender.

Data, Measures, and Methods

Data

National Survey of Family Growth

The 2006–2017 NSFG is a longitudinal, repeated cross-section dataset that includes information on family life, marriage and divorce, pregnancy, infertility, use of contraception, and general and reproductive health. The NFSG is conducted by the National Center for Health Statistics and Centers for Disease Control. Originally intended to be used to measure fertility among women, the NSFG was expanded to include both men and women in 2002. The NSFG switched to continuous interviews in 2006 and includes 53,397 respondents from 2006 to 2017. Historically, the NSFG included 15–44-year-old respondents but expanded to include 45–49-year-olds in 2015. For our purposes, we exclude 15–17-year-old respondents in order to focus on adults. We also exclude the 45–49-year-olds in order to have consistent samples across time. The sample size after these restrictions is 45,905.

Survey of U.S. Catholics and Family Life

The Pew Research Center’s 2015 SUSCFL is a representative sample of U.S. adults. Despite its name, it does not focus on Catholics exclusively, but includes many items related to the Catholic faith (see Pew Research Center, 2015). The sample size is 5,122. Interviews were conducted either over a landline phone or via cell-phone and were administered in English or Spanish. The data include adults of all ages (beyond age 44) and include a number of religion variables as well as measures of attitudes regarding certain family arrangements, including voluntary childlessness.

Measures

Dependent variables

Our study includes four dependent variables that measure different attitudes about childlessness, including one’s own childlessness and others’ childlessness. Three of the four dependent variables are from the NSFG. The first is a binary variable measuring whether the respondent intends to be childless. Here the sample is limited to childless adults who are not pregnant and do not have a pregnant partner. This variable is based on the NSFG-recoded variable INTENT, with a 1 indicating that they intend to be childless. After applying sample restrictions (age 18–44 and childless/(partner) not pregnant), the sample for this outcome is 20,084. After listwise deletion of missing values, our working sample is 19,998.

The second dependent variable is based on whether the respondent says they would be bothered if they did not have children (CHBOTHER). Response options include “not at all,” “a little,” “some,” and “a great deal.” If the respondent answered “not at all,” “a little,” or “some,” they are coded 0, and responses of “a great deal” are coded 1. After applying sample restrictions (age 18–44, childless/(partner) not pregnant, and intends to have a child), the sample for this outcome is 15,424. After listwise deletion of missing values, our working sample is 15,368.

The third dependent variable is whether the respondent agrees that “people can’t really be happy unless they have children” (CHUNLESS). Respondents who agree or strongly agree are coded 1; respondents who neither agree nor disagree (a volunteered response), disagree, or strongly disagree are coded 0. This variable was only asked from 2007 to 2015. After applying the age restriction (age 18–44), the sample for this outcome is 33,002. After listwise deletion of missing values, our working sample is 32,542.

The fourth dependent variable, from the SUSCFL, is a measure of the appropriateness of voluntary childlessness. Respondents were asked, “Overall, what is your opinion about a husband and wife who choose not to have any children?” Possible responses included: (a) acceptable and as good as any other way of life, (b) acceptable but NOT as good as some other ways of life, (c) not an acceptable way of life, (d) depends (voluntary response), or (e) this way of life is the best/better than others (voluntary response). The first and fifth responses were coded 1 to indicate full acceptance of voluntary childless as an acceptable and good way of life, while the second, third, and fourth options, as well as those saying they “don’t know,” were coded 0 to indicate at least some equivocation about the appropriateness of voluntary childlessness. The working sample for this outcome, after listwise deletion of missing values, is 4,782.

We ran ordered logit regression models for the second and third dependent variables—level of bothered-ness and agreement that childless people cannot be happy—but all models violated the parallel regression assumption per Brant tests. We also ran ordered and multinomial logit models for the voluntary childlessness outcome. Ordered logit models again violated the parallel regression assumption. Multinomial models are available upon request from the first author, but—in addition to being unwieldy and complicated to describe—they suffered from small cell sizes for some of the outcome categories, making the estimates unreliable.

Key independent variables

For both datasets, our key independent variables are measures of religious affiliation, frequency of religious service attendance, and religious salience. The NSFG includes a recoded RELTRAD variable following the scheme explicated in Steensland et al. (2000). Because the NSFG only includes response categories with at least 100 cases in each of its data releases, the RELTRAD variable does not include LDS/Mormon or Jewish respondents as separate categories. Rather, they are included with the “other religion” category. Thus, the NSFG religious affiliation measure includes conservative Protestants, mainline Protestants, Black Protestants, Catholics, those with an “other” religious affiliation, and those with no religious affiliation. Frequency of religious service attendance in the NSFG is grouped into seven categories: never, once or twice a year, 3–11 times a year, once a month, 2–3 times a month, once a week, and more than once a week and coded 1–7, with higher numbers indicating more frequent attendance. Religious salience is measured with the question, “Currently, how important is religion in your daily life?” Responses included not very important, somewhat important, and very important. These are coded 1–3 with higher numbers indicating higher importance. Importantly, those who indicated they did not have a religious affiliation were skipped out of the religious salience question. We code those respondents as 1 (“not very important”) on religious salience.

The SUSCFL also included these three religion variables. The religious affiliation variable in SUSCFL is coded according to the Steensland et al. (2000) scheme, but because of lack of specificity among Protestants we used a born-again/evangelical self-identity variable to sort conservative and mainline Protestants, and the race variable to identify Black Protestants (i.e., Protestants who are Black are considered Black Protestants). We thus have an approximation of the Steensland et al. (2000) categories, including born-again/evangelical Protestants, Protestants who are not born-again or evangelical, Black Protestants, Catholics, LDS/Mormons, Jews, adherents to other religions, and those with no religious affiliation.

Frequency of religious service attendance aside from weddings and funerals was reverse coded and ranges from (0) “never” to (5) “more than once a week.” Religious salience was measured with the question, “How important is religion in your life?” and was reverse coded such that responses range from (0) “not at all important” to (3) “very important.”

Control variables

We include a similar set of control variables for both datasets. In the NSFG, we include continuous measures of age (18–44) and year of interview, dichotomous measures indicating gender and whether the respondent was born outside the United States, and systems of dummy variables for race-ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic other race), marital status (married, cohabiting, divorced/separated, widowed, never married), and highest degree obtained (less than high school degree, high school degree, some college but no degree, associate’s degree, bachelor’s degree, and higher than bachelor’s degree). For the analysis that includes both childless adults and parents (i.e., the “people really can’t be happy unless they have children” outcome), we include a continuous measure of number of children (capped at 5).

The control variables are nearly identical for the SUSCFL analysis. There are differences in question wording, but response categories are similar for all variables with the exception of age, which is coded in 14 (mostly) five-year groupings (24 or younger, 25–29, 30–34, . . ., 85 or older). There is no year of interview variable in the SUSCFL because all interviews were conducted in 2015.

Descriptive statistics for all variables in both datasets are included in Tables 1 (NSFG) and 2 (SUSCFL).

Table 1.

Descriptive Statistics for All National Survey of Family Growth Variables (N = 19,998, 15,368, and 32,542).

	Mean/SD(Childless sample)	Mean/SD(Childless but intending sample)	Mean/SD(Full sample)	Min–Max
Does not intend to have child	.229			0, 1
Bothered “a great deal” if can’t have child		.368		0, 1
Agree people can’t be happy if childless			.082	0, 1
Religious affiliation
Conservative Protestant	.167	.173	.203	0, 1
Mainline Protestant	.151	.157	.153	0, 1
Black Protestant	.081	.086	.097	0, 1
Catholic	.204	.216	.229	0, 1
Other religious affiliation	.097	.101	.080	0, 1
No religious affiliation	.302	.267	.238
Frequency of religious service attendance	3.081/2.072	3.213/2.076	3.394/2.118	1–7
Religious salience	1.932/.830	1.981/.823	2.087/.832	1–3
Age	26.582/7.017	25.085/5.882	30.978/7.760	18–44
Born outside United States	.131	.139	.168	0, 1
Year of interview^a	6.794/3.178	6.730/3.183	4.932/2.408	0–11/0–8
Female	.441	.436	.510	0, 1
Number of children			1.201/1.319	0–5
Race-ethnicity
Hispanic	.166	.176	.198	0, 1
Non-Hispanic White	.634	.617	.598	0, 1
Non-Hispanic Black	.123	.126	.137	0, 1
Non-Hispanic other race	.077	.810	.067	0, 1
Marital status				0, 1
Married	.172	.152	.422	0, 1
Cohabiting	.131	.126	.151	0, 1
Divorced/separated	.031	.021	.075	0, 1
Widowed	.001	.003	.022	0, 1
Never married	.666	.701	.350	0, 1
Educational attainment
Less than high school degree	.091	.092	.145	0, 1
High school degree	.250	.244	.274	0, 1
Some college but no degree	.291	.305	.233	0, 1
Associate’s degree	.071	.069	.080	0, 1
Bachelor’s degree	.211	.213	.180	0, 1
Higher than bachelor’s degree	.084	.078	.088	0, 1

Notes. ^aYear of interview equals 0 for the year 2006 and 11 for the year 2017 for “intentions” and “bothered” outcomes.

Year of interview ranges from 2007 to 2015 for the “happy” outcome and ranges from 0 (for 2007) to 8 (for 2015).

Table 2.

Descriptive Statistics for All Survey of U.S. Catholics and Family Life Variables (N = 4,782).

	Mean	SD	Min–Max
Dependent variable
Choosing not to have a child as good as other arrangements	.766		0, 1
Religious affiliation
Protestant, born-again/evangelical	.241		0, 1
Protestant, not born-again/evangelical	.151		0, 1
Black Protestant	.083		0, 1
Catholic	.199		0, 1
LDS/Mormon	.015		0, 1
Jewish	.018
Other religious affiliation	.057
No religious affiliation	.235
Frequency of religious service attendance	2.533	1.629	0–5
Religious salience	2.210	1.040	0–3
Female	.511		0, 1
Age	5.978	3.550	1–14
Born outside the United States	.142		0, 1
Number of children	1.724	1.542	0–5
Educational attainment
Less than high school degree	.097		0, 1
High school degree	.307		0, 1
Some college but no degree	.198		0, 1
Associate’s degree	.118		0, 1
Bachelor’s degree	.145		0, 1
Higher than bachelor’s degree	.135		0, 1
Marital status			0, 1
Married	.476		0, 1
Living with a partner	.690		0, 1
Divorced/separated	.132		0, 1
Widowed	.068
Never been married	.255
Race-ethnicity
Non-Hispanic White	.654		0, 1
Non-Hispanic Black	.120		0, 1
Hispanic	.149		0, 1
Non-Hispanic other race	.077		0, 1

Methods

In Tables 3 –6, we report predicted probabilities for the religious characteristics based on logistic regression models. There is one table per dependent variable, and each table includes four models. The first model of each table includes controls and the religious affiliation variable, the second model includes controls and the attendance variable, the third model includes controls and the salience variable, and the fourth model includes controls and all three religious characteristics. There is debate about the appropriateness of reporting logistic regression coefficients or odds ratios as “effects” given the inherently interactive nature of the nonlinear logistic regression model (e.g., Buis, 2017; Mood, 2010). At the very least, probabilities are far more intuitive to interpret. Thus, we present results in the probability metric.

Table 3.

Predicted Probabilities from Logistic Regression Models Predicting Does Not Intend to Have a Child among Those with No Children and (Partner) Not Pregnant (N = 19,998).

	Model 1:Affiliation	Model 2:Attendance	Model 3:Salience	Model 4:All Religious Characteristics
Religious affiliation
Conservative Protestant	.191^f			.201^f
Mainline Protestant	.198^f			.200^f
Black Protestant	.196^f			.204^f
Catholic	.188^f			.191^f
Other religious affiliation	.213^f			.219^f
No religious affiliation	.304^a,b,c,d,e			.289^a,b,c,d,e
Frequency of religious service attendance
Never		.266		.242
Once or twice a year		.246		.235
3–11 times a year		.227		.228
Once a month		.209		.221
2–3 times a month		.193		.214
Once a week		.177		.208
More than once a week		.162		.201
Religious salience
Not very important			.277	.231
Somewhat important			.222	.229
Very important			.176	.227

Notes. Predicted probabilities are average adjusted predictions, which set all covariates at their observed values.

The underlying logistic regression models control for age, race/ethnicity, nativity status, marital status, year of interview, gender, and educational attainment.

Significance tests are tests of difference in the average discrete change.

For religious affiliation, superscript letters denote differences among groups as described in the key below.

For religious service attendance and religious salience, underlining reflects a significant average discrete change in the probability at p < .05 when increasing the variable from its observed value by one-unit, setting all covariates at their observed value.

significantly different from conservative Protestant at p < .05.

significantly different from mainline Protestant at p < .05.

significantly different from Black Protestant at p < .05.

significantly different from Catholic at p < .05.

significantly different from other religious affiliation at p < .05.

significantly different from no religious affiliation at p < .05.

Table 4.

Predicted Probabilities from Logistic Regression Models Predicting Would Be Bothered “a Great Deal” If Remained Childless, among Those with No Children and (Partner) Not Pregnant Who Intend to Have a Child (N = 15,368).

	Model 1:Affiliation	Model 2:Attendance	Model 3:Salience	Model 4:All Religious Characteristics
Religious affiliation
Conservative Protestant	.422^b,c,d,f			.394^c
Mainline Protestant	.371^a,f			.361
Black Protestant	.340^a,e			.316^a,d,e
Catholic	.384^a,f			.379^c
Other religious affiliation	.422^c,f			.400^c
No religious affiliation	.308^a,b,d,e			.352
Frequency of religious service attendance
Never		.343		.387
Once or twice a year		.354		.379
3–11 times a year		.365		.371
Once a month		.377		.363
2–3 times a month		.388		.355
Once a week		.400		.347
More than once a week		.412		.339
Religious salience
Not very important			.311	.308
Somewhat important			.368	.368
Very important			.429	.432

Notes. Predicted probabilities are average adjusted predictions, which set all covariates at their observed values.

The underlying logistic regression models control for age, race/ethnicity, nativity status, marital status, year of interview, gender, and educational attainment.

Significance tests are tests of difference in the average discrete change.

For religious affiliation, superscript letters denote differences among groups as described in the key below.

significantly different from conservative Protestant at p < .05.

significantly different from mainline Protestant at p < .05.

significantly different from Black Protestant at p < .05.

significantly different from Catholic at p < .05.

significantly different from other religious affiliation at p < .05.

significantly different from no religious affiliation at p < .05.

Table 5.

Predicted Probabilities from Logistic Regression Models Predicting Agreement with the Statement, “People Can’t Really Be Happy Unless They Have Children” (N = 32,542).

	Model 1: Affiliation	Model 2:Attendance	Model 3:Salience	Model 4:All Religious Characteristics
Religious affiliation
Conservative Protestant	.086^e,f			.077^e
Mainline Protestant	.074^e			.072^e
Black Protestant	.081^e			.076^e
Catholic	.086^e,f			.084^e
Other religious affiliation	.128^a,b,c,d,f			.122^a,b,c,d,f
No religious affiliation	.060^a,d,e			.076^e
Frequency of religious service attendance
Never		.068		.074
Once or twice a year		.073		.077
3–11 times a year		.079		.081
Once a month		.085		.084
2–3 times a month		.091		.087
Once a week		.098		.091
More than once a week		.106		.094
Religious salience
Not very important			.062	.070
Somewhat important			.079	.080
Very important			.099	.092

Notes. Predicted probabilities are average adjusted predictions, which set all covariates at their observed values.

The underlying logistic regression models control for age, race/ethnicity, nativity status, marital status, number of children, year of interview, gender, and educational attainment.

Significance tests are tests of difference in the average discrete change.

For religious affiliation, superscript letters denote differences among groups as described in the key below.

significantly different from conservative Protestant at p < .05.

significantly different from mainline Protestant at p < .05.

significantly different from Black Protestant at p < .05.

significantly different from Catholic at p < .05.

significantly different from other religious affiliation at p < .05.

significantly different from no religious affiliation at p < .05.

Table 6.

Predicted Probabilities from Logistic Regression Models Predicting a Husband and Wife Who Choose Not to Have Children Is Acceptable and as Good as (or Better Than) Any Other Way of Life (N = 4,782).

	Model 1:Affiliation	Model 2:Attendance	Model 3:Salience	Model 4:All Religious Characteristics
Religious affiliation
Protestant, born-again/evangelical	.704^b,e,g,h			.752^e,h
Protestant, not born-again/evangelical	.791^a,d,e,h			.792^e
Black Protestant	.780^e,h			.782^e
Catholic	.742^b,e,h			.756^e,h
Mormon/LDS	.382^{a,b,c,d,f,g,h}			.448^{a,b,c,d,f,g,h}
Jewish	.743^e,h			.728^e
Other religious affiliation	.768^a,e,h			.764^e
No religious affiliation	.862^{a,b,c,d,e,f,g}			.805^a,d,e
Frequency of religious service attendance
Never		.863		.821
Seldom		.832		.803
A few times a year		.796		.783
Once or twice a month		.755		.763
Once a week		.709		.741
More than once a week		.658		.717
Religious salience
Not important at all			.904	.854
Not too important			.857	.821
Somewhat important			.795	.784
Very important			.714	.741

Notes. Predicted probabilities are average adjusted predictions, which set all covariates at their observed values.

The underlying logistic regression models control for age, race/ethnicity, nativity status, marital status, number of children, gender, and educational attainment.

Significance tests are tests of difference in the average discrete change.

For religious affiliation, superscript letters denote differences among groups as described in the key below. For religious service attendance and religious salience, underlining reflects a significant average discrete change in the probability at p < .05 when increasing the variable from its observed value by one-unit, setting all covariates at their observed value.

significantly different from conservative Protestant at p < .05.

significantly different from mainline Protestant at p < .05.

significantly different from Black Protestant at p < .05.

significantly different from Catholic at p < .05.

significantly different from Mormon/LDS at p < .05.

significantly different from Jewish at p < .05.

significantly different from other religious affiliation at p < .05.

significantly different from no religious affiliation at p < .05.

The predicted probabilities are average adjusted predictions (AAPs) (Williams, 2012). AAPs calculate the predicted probability for every observation in the data setting all covariates to their observed value and setting the key independent variable at the value desired by the researcher. The predicted probabilities are then averaged across cases. Following the advice of Long and Mustillo (2018), our significance tests reflect the statistical significance of the average discrete change (ADC) in probability. For the categorical religious affiliation variable, this is simply the difference in the probability between each religious group. For the continuous attendance and salience variables, the ADC is the average change in the predicted probability setting all covariates at their observed value and increasing the key independent variable (either attendance or salience) from its observed value by one-unit. This is analogous to the interpretation of an unstandardized coefficient in an OLS model. The ADCs are produced using the user-written mchange command in Stata (Long & Freese, 2014). In the interest of space, only the predicted probabilities for the religion variables are displayed. Full models with logistic regression coefficients are available from the first author upon request.

Table 7 reports the statistically significant interaction effects between gender and the religious characteristics. All potential interactions revealing differences in the effect of religion across gender were tested for all outcomes, including differences in the effect of every religious affiliation. The tests of significance for the interaction effects, like for the main effects in previous tables, are tests of significant difference in the ADC for the religion variable across gender (Mize, 2019). We report the ADC separately for men and women (including tests of whether those ADCs are statistically significant), as well as the difference in the ADC across gender and whether that difference is statistically significant (i.e., a test of second difference).

Table 7.

Significant Interaction Effects between Gender and Religious Characteristics.

	ADC for Men	ADC for Women	Difference (Absolute Value)
Panel A. Bothered “a great deal” if childless
Black Protestant vs Other religious affiliation	−.022	−.155**	.133**
No religious affiliation vs. Other religious affiliation	−.065*	−.179***	.114**
Panel B. Agree childless can’t be happy
Mainline Protestant vs No religious affiliation	.032*	−.002	.035*
Catholic vs No religious affiliation	.037*	.016*	.021*
Other religious affiliation vs No religious affiliation	.087***	.051**	.036*
Frequency of religious service attendance	.010***	.003*	.007**
Panel C. Voluntary childlessness as good as other
Religious salience	−.068***	−.112***	.044*

Notes. *p < .05, **p < .01, ***p < .001 (two-tailed test).

Gender was interacted in separate models with religious affiliation, religious service attendance, and religious salience, including all control variables from previous tables, for all outcomes. Only significant interaction effects are displayed.

Average discrete change (ADC) is the difference in predicted probabilities across affiliation or for a one-unit increase in attendance or salience from its observed value, setting all covariates at their observed values.

The difference column shows the difference in the effect of religion across gender (i.e., second difference).

All possible religious affiliation comparisons were tested.

The NSFG data were analyzed using the svy commands in Stata in order to account for the complex sampling design and the unequal probability of selection into the sample. The SUSCFL data were weighted to account for unequal probability of selection into the sample.

Results

Table 3 presents AAPs for not intending to have a child among those with no children and who are not pregnant or do not have a pregnant partner, which has an observed probability of .229 (see Table 1). Model 1 reports the AAPs for each religious affiliation. Interestingly, there are no statistical differences across religious affiliations. The difference is between religiously affiliated respondents and those with no religious affiliation. If all respondents claimed no religious affiliation, there would be an average predicted probability of not intending to have a child of .304, which is statistically different from each religious affiliation, and represents more than a 50% increase in probability of intended childlessness vis-à-vis the three Protestant groups and Catholics. When it comes to childlessness intentions, what matters is having a religious affiliation of any kind (versus not).

Model 2 of Table 3 reports AAPs for the categories of frequency of religious service attendance. The probabilities in Model 2 are underlined, indicating that the ADC accompanying a one-unit increase in religious service attendance from its observed value is statistically significant (ADC = −.018, p < .001, results not shown). If all the NSFG respondents never attended religious services, their average predicted probability of intending to be childless would be .266. If they attended more than once a week, the average predicted probability would be only .162. Thus, we see a sizable effect of religious service attendance on one’s own intended childlessness. Similarly, for religious salience in Model 3, the ADC for a one-unit increase from its observed value is statistically significant (ADC = −.047, p < .001, results not shown). If all respondents were to report religion as not very important, the average predicted probability of intended childlessness would be .277. If all respondents were to report religion as very important in their daily life, that probability falls to .176.

Model 4 of Table 3 includes all three religious characteristics in an effort to identify which religious characteristics may matter most. The effects are diminished across all three variables, however, the distinction between the affiliated and non-affiliated remains, as does the effect of religious service attendance (ADC = −.007, p < .01, results not shown). The effect of religious salience, however, is not statistically significant in Model 4 once the other two religious characteristics are accounted for.

Table 4 reports predicted probabilities (AAPs) for those who would be bothered “a great deal” if they remained childless, among those who intended to have a child. Model 1 reports the AAPs for different religious affiliations. There are many differences among religious affiliations. If all NSFG respondents were conservative Protestants, the average predicted probability for the sample would be .422 (compared to the observed probability of .368 reported in Table 1). This is significantly higher than if all respondents were mainline Protestant, Black Protestant, Catholic, or no religion, but not “other religious affiliation.” If the entire sample were Catholic, the average predicted probability of a respondent being bothered a great deal by childlessness would be .384, which is significantly lower than conservative Protestant but significantly higher than no religion. Similarly, the AAP for mainline Protestant is .371, which like Catholic, is lower than conservative Protestant but higher than no religion. If the sample were entirely Black Protestant, the predicted probability of being bothered a great deal would be .384, significantly lower than if the sample were conservative Protestant or an “other” religious affiliation. A sample of nonreligious adults would produce the lowest predicted probability of being bothered a great deal if they remained childless at .308, significantly lower than conservative Protestant, mainline Protestant, Catholic, or “other” religious affiliation.

Model 2 of Table 4 reports the AAPs for categories of frequency of religious service attendance. The ADC for a one-unit increase in religious service attendance from its observed value is .011 (p < .001, results not shown). The predicted probabilities at either end of the attendance spectrum are thus substantially different. If the entire sample never attended religious services, the predicted probability of being bothered a great deal by childlessness would be .343, compared to .412 if the entire sample attended services more than once per week. Similarly, Model 3 shows a significant effect of religious salience on the probability of being bothered a great deal by childlessness among the sample who intends to have a child. The ADC for a one-unit increase in religious salience from its observed value is .060 (p < .001, results not shown). The AAP ranges from .311 if the entire sample reported religion was not very important to .429 if the entire sample reported religion was very important.

Model 4 includes all three religious characteristics. Differences across religious affiliations are diminished, suggesting they are a function of differential religiosity. In Model 4, Black Protestants stand out from conservative Protestants, Catholics, and “other” religions. If the entire sample were Black Protestant, only .316 of the sample would be bothered a great deal by childlessness, compared to .394, .379, and .400 for conservative Protestants, Catholics, and adherents to other religions, respectively.

Interestingly, the direction of the effect for religious service attendance changes when religious affiliation and religious salience are accounted for in Model 4. Net of these other religious characteristics, more frequent religious service attendance is associated with a lower probability of being bothered a great deal by childlessness. The ADC for a one-unit increase in attendance from its observed value is −.008 (p < .05, results not shown). If the entire sample never attended religious services, the predicted probability of being bothered a great deal by remaining childless would be .387, compared to .339 if the entire sample attended more than once per week.

Religious salience remains a strong, positive predictor of being bothered by remaining childless in the sample of respondents who intended to have a child, even after accounting for religious affiliation and religious service attendance. In Model 4, the ADC in the probability for a one-unit increase in salience from its observed value is .063 (p < .001, results not shown). If the entire sample reported religion was not very important, the predicted probability of being bothered a great deal would be .308, compared to .432 if the entire sample reported religion was very important to them.

Table 5 reports agreement with the statement, “People can’t really be happy unless they have children.” Generally speaking, few respondents agreed with this statement, with an observed probability of .082 (see Table 1). Nevertheless, we see differences across religious groups and religiosity in Table 5. In terms of religious affiliation, those with “other” religious affiliations stand out from every other group in agreeing, though this is a heterogeneous and thus fairly uninterpretable group. Conservative Protestants and Catholics also have above-average AAPs at .086, which distinguishes them from those with no religious affiliation (as well as the adherents to “other” religions). Mainline and Black Protestants have below-average AAPs at .074 and .081, respectively, but they differ significantly only from the “other” group. The religiously unaffiliated, with an AAP of .060, are statistically different from conservative Protestants, Catholics, and those with an “other” religious affiliation. Although these differences are small in absolute terms, the percentage increase in the probability is somewhat sizable. For example, the .026 increase in probability from no religious affiliation to conservative Protestant or Catholic (from .060 to .086) represents a 43% increase in the probability of agreeing people cannot be happy if they do not have children.

Model 2 of Table 5 reports AAPs for categories of frequency of religious service attendance. The ADC for a one-unit increase in attendance from its observed value is significant and positive (ADC = .006, p < .001, results not shown). As shown in Table 5, if the entire sample never attended religious services, the probability of reporting people cannot be happy without children would be .068, compared to .106 if the entire sample attended services more than once per week. Although this is a small absolute change in the probability (.038), it is a 56% increase in the probability. Model 3 reports AAPs for categories of religious salience. The ADC is significant and positive (ADC = .021, p < .001, results not shown). If all respondents reported religion was not very important, the average predicted probability of agreeing that people cannot be happy if they do not have children would be .062, compared to .099 if all respondents reported that religion was very important in their daily life. This represents a 60% increase in the probability.

Model 4 reveals that many of the religious effects seen in Models 1–3 disappear or are diminished when all three religious characteristics are included in the model. In Model 4, only those with an “other” religious affiliation stand out from all other groups. The ADC for the attendance (ADC = .003, p < .05, results not shown) and salience (ADC = .011, p < .05, results not shown) variables remains significant, though they are smaller than in Models 2 and 3. Both of these variables, however, appear to contribute to beliefs about people’s happiness in the context of childlessness.

Table 6 reports the AAPs for whether the respondent believes a husband and wife choosing to be childless is both acceptable and as good as (or better than) choosing to have children. Table 6 reveals a number of sizable effects across religious characteristics. In terms of religious affiliation, Model 1 reveals a significant difference between the religiously unaffiliated and every other group (much like the childbearing intentions outcome in Table 3). If the entire SUSCFL sample were unaffiliated, there would be a .862 probability of full acceptance of voluntary childlessness as a way of life. (The observed probability, in Table 2, is .766.) On the other end of the spectrum are Mormon respondents. If the sample were entirely Mormon, the average predicted probability of embracing voluntary childlessness would be just .382. This is a statistically lower AAP than any other group. Born-again or evangelical Protestants, Catholics, and Jews are also on the lower end of the spectrum with AAPs of .704, .742, and .743, respectively. The AAPs for born-again Protestants and Catholics are statistically lower than that of non-born-again Protestants (AAP = .791), and the AAPs for born-again Protestants, Catholics, and Jews are all statistically lower than the AAP for “other” religious affiliation (.768). The Black Protestant AAP of .780 is distinct only from Mormons (where it is higher) and those with no religious affiliation (where it is lower). Notably, however, full acceptance of voluntary childlessness is relatively likely (probabilities greater than .700) among all groups except Mormons.

There is a positive, statistically significant effect of religious service attendance on full acceptance of voluntary childlessness as a way of life in Model 2 (an observed probability of .766 per Table 2). The ADC in the probability of embracing voluntary childlessness for a one-unit increase in religious service attendance from its observed value is −.043 (p < .001, results not shown). If the entire sample never attended religious services, the average predicted probability would be .863, and if the entire sample attended services more than once a week the average predicted probability would be .658. There is also positive, statistically significant effect of religious salience in Model 3. Here, the ADC in the probability for a one-unit increase in salience from its observed values is −.082 (p < .001, results not shown). Here the AAPs range from .904 if the sample said religion was not important at all to .714 if the sample said religion was very important.

Model 4 includes all three religious characteristics, and like with Table 5, we see the size of effects diminished. The AAP for Mormons remains significantly lower than all other groups at .448, and born-again Protestants (.752) and Catholic (.756) AAPs remain significantly lower than the unaffiliated AAP (.805). The ADC in probability for religious service attendance (ADC = −.021, p < .001, results not shown) and religious salience (ADC = −.043, p < .001, results not shown) remain statistically significant and negative. Here we see differences in AAP ranging from .821 if all respondents never attended to .717 if all respondents attended more than once a week, and from .854 if all respondents reported religion was not important at all to .741 if they all reported religion was very important.

Table 7 presents the statistically significant gender–religion interactions for each outcome. There were no significant interactions for childlessness intentions. The only significant interactions for being bothered a great deal by remaining childless involved those with an “other” religious affiliation. Here we see the negative effect of being a Black Protestant or religiously unaffiliated is significantly stronger among women than among men, vis-à-vis adherents to an “other” religious tradition.

For agreeing that the childless cannot really be happy, we see differences in the effect of religious affiliation across gender for those with no religious affiliation. Specifically, the effect of being a mainline Protestant, Catholic, or adherent to an “other” religion, compared to being unaffiliated, is positive and statistically significant for men. The effect is not significant for mainline Protestant women, and the effect is weaker but still positive and significant for Catholic women and women from “other” religious affiliations. We also found a significant interaction by gender with religious service attendance. The effect of religious service attendance is positive and significant for both men and women, but it is stronger among men.

Lastly, we found one significant gender–religion interaction for attitudes toward voluntary childlessness. Religious salience was negatively and significantly associated with fully embracing voluntary childlessness among both men and women, but the negative effect was stronger among women.

Discussion

Using two large, national datasets, we examined the role of religion in shaping U.S. adults’ attitudes toward childlessness—both their own and others’. The results are clear: Religion is a strong correlate of all types of childlessness attitudes, including one’s own childlessness intentions, one’s potential response to unintended childlessness, assessing the possibility for people to be happy when childless, and fully embracing voluntary childlessness (of others) as a way of life. For all four attitudes, all three of our religion measures—affiliation, service attendance, and salience—are important factors. The evidence for gender differences in the effect of religion on these attitudes is less clear. With respect to assessing people’s happiness when childless, religion seems to matter more for men. But with embracing voluntary childlessness, religion seems to matter more for women. Religion, generally speaking, matters similarly for men and women when it comes to their own intentions and their feelings about their own unintended childlessness.

The findings for childlessness intentions have clear implications for childlessness as a family form in the United States. The vast majority (77%) of 18–44-year-olds who are childless still intend to have children. Even the vast majority of the nonreligious intend to have children, suggesting pronatalist attitudes remain widespread in the United States. The fact that religion is associated with being less likely to intend to be childless is no surprise given religion’s known association with fertility and older findings on religion and childlessness (Heaton et al., 1992). Having a religious affiliation, participating in a religious community, and valuing one’s faith are all predictive of having children and thus contribute to the perpetuation of a pronatalist culture. Notably, personal religious salience is not independently associated with childlessness intentions once affiliation and attendance are controlled. This suggests that it is the more public, organizational aspects of religion that matter for childlessness intentions. Religious schemas about fertility and co-religionists who serve as reference groups likely contribute to individuals’ desire to have children (Hackett, 2008). Yet these religious characteristics are less common now than they used to be (Chaves, 2017; Sherkat, 2014; Voas & Chaves, 2016). Thus, childlessness intentions are likely to increase in the population if religious commitment continues to decrease.

But religious effects go beyond childlessness intentions. Not only does religion diminish childlessness intentions, but it also increases the likelihood that an individual who wishes to have a child would be significantly bothered if they did not. The role of personal religious salience appears to be particularly pivotal here. After accounting for affiliation and attendance, we found even larger salience effects on being bothered by childlessness. The attendance association actually changes direction to negative once salience is accounted for. Although we can only speculate about what exactly is happening here, it appears that participation in a religious community can actually help reduce potential feelings of frustration with childlessness. We suspect this is because those who attend religious services anticipate support and alternative sources of intimacy from their religious community. Religion can in one way alleviate negative feelings about one’s childlessness, even as it exacerbates these negative feelings through its emphasis on family life.

Religion also matters for how people think about other people’s childlessness, or childlessness as a way of life. Whether people are religious or not, very few agree with the sentiment that it is impossible for childless people to truly be happy. Nevertheless, those who are conservative Protestant and Catholic (and from an “other” religion) are more likely to agree than the nonreligious (which appears to be a function of their religiosity in the case of conservative Protestants and Catholics), as are those who attend religious services more frequently and value religion more highly in their daily lives. There does seem to be a modest effect of religious norms and messages about fertility that lead some people to view childlessness as suboptimal in terms of individual happiness. But even the most religious individuals are unlikely to agree with this. There are two potential explanations for these low probabilities that are likely at work. First, religion in the United States is increasingly individualized (e.g., Bellah et al., 1985; Wuthnow 1988). Even as people may feel the need to apply religious teachings to their own lives, they are less inclined to speak about the choices of others. Second, this item in NSFG does not specify the cause of childlessness. Respondents are likely considering involuntarily childless individuals and are unwilling to relegate them to the realms of the unhappy based on something over which they had no control.

Indeed, we see less of an embrace of voluntary childlessness, especially among the more religious, although we should also note that the analysis of voluntary childlessness includes adults of all ages and not just 18–44-year-olds. Here we see differences among religious groups that are not merely a function of religiosity differences across groups. The religious traditions with the clearest pronatalist messages and cultures—born-again/evangelical Protestants, Catholics, and Mormons—are the least likely to embrace childlessness. Mormons stand out considerably from all other groups, even born-again/evangelical Protestants and Catholics. This is in keeping with Mormon distinctiveness on a whole host of family-related attitudes and behaviors, and is likely tied to theological differences as well as cultural and structural emphases on family life within the LDS church (Heaton, 1987). Cultural and structural emphases on family life likely extend to other traditions as well (Edgell, 2006; Wilcox, 2004), though to a lesser extent, and are reflected in the strong negative effects of religious service attendance and religious salience on a full embrace of voluntary childlessness. Much more so for religious Americans than nonreligious ones, childbearing remains normative and a piece of “the good life.” These findings should not be overstated, however. With the exception of Mormons, religious Americans are still highly likely to say voluntary childlessness is acceptable and as good as choosing to have children. For example, even if all Americans attended church more than once a week, we would still see a probability of .658 that a respondent would say voluntary childlessness is acceptable and good.

Evidence for gender differences in the effect of religion, as aforementioned, is mixed. We hypothesized that religion may matter more for men than for women because it is one of the few institutions that promote childbearing among men, while women are bombarded with pronatalist messages from many places. We find support for this when we look at agreement that people cannot really be happy if they do not have children. Here, mainline Protestantism, Catholicism, and “other” religions have a stronger positive effect on men’s agreement with this statement than women’s, as does religious service attendance. Given the lack of similar findings for the other outcomes, we hesitate to endorse our hypothesized explanation. Rather, we suspect that for this particular outcome, religious women are more inclined than religious men to empathize with those who are unable to have children and do not wish to say that they cannot be happy. Indeed, the pattern for voluntary childlessness attitudes is quite different. Here, religious salience matters much more for women than for men. Childbearing, as a choice, is more closely tied to women’s religious identity than men’s, which is the opposite of what we hypothesized.

Taken together, our findings suggest that religion continues to contribute to a pronatalist culture in the United States through its impact on a variety of attitudes regarding childlessness. Nevertheless, there are limitations to our research that merit discussion. Neither of our two datasets includes panel data and are thus subject to issues of reverse causation. Although we believe that we are theoretically justified in assuming religious commitment influences childlessness attitudes and not vice versa, it is conceivable that people with pro-family attitudes select into religious participation or more pro-family religious traditions and thus muddy the causal story. Selection may also impact our samples for the analysis of childlessness intentions and feelings about one’s own childlessness. Because religion is associated with fertility as well as fertility attitudes, religious people are more likely to select out of the samples that are restricted to the childless. This likely results in an underestimation or conservative estimate of the effect of religion.

Future research should incorporate data from panel studies and consider more direct tests of the mechanisms that link religion to childlessness attitudes. This study, however, provides evidence from recent survey data to suggest that religion remains a significant contributor to a variety of childlessness attitudes.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Jeremy E. Uecker

References

Bellah

R. N.

Madsen

Sullivan

W. M.

Swidler

Tipton

(1985). Habits of the heart: Individualism and commitment in American life. University of California Press.

Blackstone

(2014). Doing family without having kids. Sociology Compass, 8(1), 52–62.

Blackstone

Stewart

M. D.

(2012). Choosing to be childfree: Research on the decision not to parent. Sociology Compass, 6(9), 718–727.

Buis

M. L.

(2017). Logistic regression: When can we do what we think we can do? [Unpublished manuscript]. http://www.maartenbuis.nl/wp/odds_ratio_3.1.pdf

Catechism of the Catholic Church . (2011). http://www.usccb.org/beliefs-and-teachings/what-we-believe/catechism/catechism-of-the-catholic-church/

Chaves

(1994). Secularization as declining religious authority. Social Forces, 72(3), 749–74.

Chaves

(2017). American religion: Contemporary trends (2nd ed.). Princeton University Press.

Curran

(2018). Humanae Vitae: Fifty years later. Theological Studies, 79(3), 520–42.

Edgell

(2006). Religion and family in a changing society. Princeton University Press.

10.

Gillespie

(2001). Contextualizing voluntary childlessness within a postmodern model of reproduction: Implications for health and social needs. Critical Social Policy, 21(2), 139–159.

11.

Goldscheider

(1971). Population, modernization, and social structure. Little, Brown.

12.

Hackett

C. P.

(2008). Religion and fertility in the United States: The influence of affiliation, region, and congregation [Doctoral dissertation]. Princeton University.

13.

Hayford

S. R.

Morgan

S. P.

(2008). Religiosity and fertility in the United States: The role of fertility intentions. Social Forces, 86(3), 1163–1188.

14.

Hays

(1998). The cultural contradictions of motherhood. Yale University Press.

15.

Heaton

T. B.

(1986). How does religion influence fertility?: The case of Mormons. Journal for the Scientific Study of Religion, 25(2), 248–258.

16.

Heaton

T. B.

(1987). Four characteristics of the Mormon family: Contemporary research on chastity, conjugality, children, and chauvinism. Dialogue: A Journal of Mormon Thought, 20(2), 101–114.

17.

Heaton

T. B.

Jacobson

C. K.

X. N.

(1992). Religiosity of married couples and childlessness. Review of Religious Research, 33(3), 244–255.

18.

Heitlinger

(1991). Pronatalism and women’s equality policies. European Journal of Population/Revue Europeenne de Demographie, 7(4), 343–375.

19.

Jansen

N. A.

Saint Onge

J. M.

(2015). An internet forum analysis of stigma power perceptions among women seeking fertility treatment in the United States. Social Science & Medicine, 147, 184–189.

20.

Johnson-Hanks

J. A.

Bachrach

C. A.

Morgan

S. P.

Kohler

H. P.

(2011). Understanding family change and variation: Toward a theory of conjunctural action. Springer.

21.

Jordan

Revenson

T. A.

(1999). Gender differences in coping with infertility: A meta-analysis. Journal of Behavioral Medicine, 22(4), 341–358.

22.

Koropeckyj-Cox

Çopur

Romano

Cody-Rydzewski

(2018). University students’ perceptions of parents and childless or childfree couples. Journal of Family Issues, 39(1), 155–179.

23.

Koropeckyj-Cox

Pendell

(2007a). Attitudes about childlessness in the United States: Correlates of positive, neutral, and negative responses. Journal of Family Issues, 28(8), 1054–1082.

24.

Koropeckyj-Cox

Pendell

(2007b). The gender gap in attitudes about childlessness in the United States. Journal of Marriage and Family, 69(4), 899–915.

25.

Lehrer

E. L.

(1996). The role of the husband’s religious affiliation in the economic and demographic behavior of families. Journal for the Scientific Study of Religion, 35(2), 145–155.

26.

Long

J. S.

Freese

(2014). Regression models for categorical dependent variables using Stata, third edition. Stata Press.

27.

Long

J. S.

Mustillo

S. A.

(2018). Using predictions and marginal effects to compare groups in regression models for binary outcomes. Sociological Methods & Research. https://doi.org/10.1177/0049124118799374.

28.

Luhrmann

T. M.

(2012). When God talks back: Understanding the American Evangelical relationship with God. Knopf.

29.

Lynch

K. A.

(2006). Theoretical and analytical approaches to religious beliefs, values, and identities during the modern fertility transition. In Derosas

van Poppel

(Eds.), Religion and the decline of fertility in the Western world (pp. 21–39). Springer.

30.

Madsen

(2009). The Archipelago of faith: Religious individualism and faith community in America today. American Journal of Sociology, 114(5), 1263–1301.

31.

Martinez

Daniels

Chandra

(2012). Fertility of men and women aged 15–44 years in the United States: National Survey of Family Growth, 2006–2010. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics.

32.

Martinez

G. M.

Daniels

Febo-Vazquez

(2018). Fertility of men and women aged 15–44 in the United States: National Survey of Family Growth, 2011–2015. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics.

33.

McCutcheon

J. M.

(2018). Reviewing pronatalism: A summary and critical analysis of prior research examining attitudes towards women without children. Journal of Family Studies. https://doi.org/10.1080/13229400.2018.1426033

34.

McQuillan

Greil

A. L.

Shreffler

K. M.

Bedrous

A. V.

(2015). The importance of motherhood and fertility intentions among US women. Sociological Perspectives, 58(1), 20–35.

35.

McQuillan

Greil

A. L.

Shreffler

K. M.

Tichenor

(2008). The importance of motherhood among women in the contemporary United States. Gender & Society, 22(4), 477–496.

36.

McQuillan

Greil

A. L.

Shreffler

K. M.

Wonch-Hill

P. A.

Gentzler

K. C.

Hathcoat

J. D.

(2012). Does the reason matter? Variations in childlessness concerns among US women. Journal of Marriage and Family, 74(5), 1166–1181.

37.

McQuillan

(2004). When does religion influence fertility? Population and Development Review, 30(1), 25–56.

38.

Meyers

D. T.

(2001). The rush to motherhood: Pronatalist discourse and women’s autonomy. Signs: Journal of Women in Culture and Society, 26(3), 735–773.

39.

Mize

T. D.

(2019). Best practices for estimating, interpreting, and presenting nonlinear interaction effects. Sociological Science, 6, 81–117.

40.

Mood

(2010). Logistic regression: Why we cannot do what we think we can do, and what we can do about it. European Sociological Review, 26(1), 67–82.

41.

Mosher

W. D.

Williams

L. B.

Johnson

D. P.

(1992). Religion and fertility in the United States: New patterns. Demography, 29(2), 199–214.

42.

Park

(2002). Stigma management among the voluntarily childless. Sociological Perspectives, 45(1), 21–45.

43.

Park

(2005). Choosing childlessness: Weber’s typology of action and motives of the voluntarily childless. Sociological Inquiry, 75(3), 372–402.

44.

Parry

D. C.

(2005). Women’s leisure as resistance to pronatalist ideology. Journal of Leisure Research, 37(2), 133–151.

45.

Pearce

L. D.

(2010). Religion and the timing of first births in the United States. In Ellison

C. G.

Hummer

R. A.

(Eds.), Religion, families, and health: Population-based research in the United States (pp. 19–39). Rutgers University Press.

46.

Pearce

L. D.

Davis

S. N.

(2016). How early life religious exposure relates to the timing of first birth. Journal of Marriage and Family, 78(5), 1422–1438.

47.

Perry

S. L.

Schleifer

(2019). Are the faithful becoming less fruitful? The decline of conservative Protestant fertility and the growing importance of religious practice and belief in childbearing in the US. Social Science Research, 78, 137–155.

48.

Pew Research Center. (2015). U.S. Catholics Open to Non-Traditional Families. Pew Research Center.

49.

Russo

N. F.

(1976). The motherhood mandate. Journal of Social Issues, 32(3), 143–153.

50.

Schoen

Astone

N. M.

Kim

Y. J.

Nathanson

C. A.

Fields

J. M.

(1999). Do fertility intentions affect fertility behavior? Journal of Marriage and the Family, 61(3), 790–799.

51.

Sewell

W. H.

Jr. (2005). Logics of history: Social theory and social transformation. University of Chicago Press.

52.

Shain

(2019). Understanding the demographic challenge: Education, orthodoxy and the fertility of American Jews. Contemporary Jewry, 39(2), 273–292.

53.

Sherkat

D. E.

(2014). Changing faith: The dynamics and consequences of Americans’ shifting religious identities. New York University Press.

54.

Stark

Bainbridge

W. S.

(1985). The future of religion: Secularization, revival, and cult formation. University of California Press.

55.

Steensland

Robinson

L. D.

Wilcox

W. B.

Park

J. Z.

Regnerus

M. D.

Woodberry

R. D.

(2000). The measure of American religion: Toward improving the state of the art. Social Forces, 79(1), 291–318.

56.

Tark

(2003). Family-centered belief and practice in the Church of Jesus Christ of Latter-day Saints and the Unification Church. Peter Lang Publishing.

57.

Thornton

Young-DeMarco

(2001). Four decades of trends in attitudes toward family issues in the United States: The 1960s through the 1990s. Journal of Marriage and Family, 63(4), 1009–1037.

58.

Tichenor

McQuillan

Greil

A. L.

Bedrous

A. V.

Clark

Shreffler

K. M.

(2017). Variation in attitudes toward being a mother by race/ethnicity and education among women in the United States. Sociological Perspectives, 60(3), 600–619.

59.

Tichenor

McQuillan

Greil

A. L.

Contreras

Shreffler

K. M.

(2011). The importance of fatherhood to US married and cohabiting men. Fathering, 9(3), 232–251.

60.

Uecker

J. E.

Froese

(2019). Religious individualism and moral progressivism: How source of religious authority is related to attitudes about abortion, same-sex marriage, divorce, and premarital sex. Politics & Religion, 12(2), 283–316.

61.

United States Census Bureau. (2018). Historical Table 1: Percent childless and births in the past year per 1,000 women: Selected years, 1976–2018. https://www.census.gov/library/visualizations/time-series/demo/his-cps.html

62.

Veevers

J. E.

(1980). Childless by choice. Butterworth-Heinemann.

63.

Voas

Chaves

(2016). Is the United States a counterexample to the secularization thesis? American Journal of Sociology, 121(5), 1517–1556.

64.

Westoff

C. F.

Jones

E. F.

(1979). The end of “Catholic” fertility. Demography, 16(2), 209–17.

65.

Wilcox

W. B.

(2004). Soft patriarchs, new men: How Christianity shapes fathers and husbands. University of Chicago Press.

66.

Wilcox

W. B.

Wolfinger

N. H.

(2008). Living and loving “decent:” Religion and relationship quality among urban parents. Social Science Research, 37(3), 828–843.

67.

Williams

(2012). Using the margins command to estimate and interpret adjusted predictions and marginal effects. The Stata Journal, 12(2), 308–331.

68.

Wuthnow

(1988). The restructuring of American religion: Society and faith since World War II. Princeton University Press.

69.

Zhang

(2008). Religious affiliation, religiosity, and male and female fertility. Demographic Research, 18, 233–261.